Encore, the AI-powered shopping assistant, is breaking down barriers in the world of thrift shopping by bringing hundreds of secondhand markets under one roof. Co-founded by former Apple engineer Alex Ruber and ex-Twitter/Asana engineer Parth Chopra, this search tool stems from a shared love for thrifting and a clear goal: make finding pre-loved treasures online easier and quicker.

The story of Encore, the AI-powered shopping assistant

The online thrift market is vast and, frankly, a maze—Depop, Mercari, ThredUp, eBay, Craigslist, and countless others. Each has its own specialty and quirks, leaving shoppers wading through them all to find exactly what they’re after. “It’s hard for consumers to sift through them all to try and get to the product you are looking for,” Ruber told TechCrunch. And that’s where Encore comes in, acting as the ultimate aggregator to cut through this fragmentation. Ruber’s own quest for a specific TV show jacket inspired him to make Encore a reality.

For Ruber and Chopra, Encore’s mission is deeply personal. Both founders are immigrants who frequented thrift stores, where they often found unique, one-of-a-kind items. Chopra, who developed a love for flea markets from his mother, shares how his upbringing influenced his vision for Encore: “For me, there was also personal interest because my mom used to take me to flea markets every Sunday. I bought a lot of stuff from those places,” he told TechCrunch.

Encore’s AI engine is like having a secondhand stylist by your side. It’s powered by a large language model that can process detailed, offbeat queries—think “Show me a dress that Emily wore in Emily in Paris Season 3 Episode 4″—and pull results from top platforms like Poshmark, the RealReal, and eBay. For those moments when you’re not sure where to start, Encore’s prompts like “Outfit inspo for…” help steer you in the right direction, so you’re not staring at a blank screen.

Encore has struck a nerve with thrift lovers in a booming market that’s expected to reach $73 billion in the U.S. and $350 billion globally by 2028, according to ThredUp. Encore itself sees over 50,000 searches a month and has been growing consistently, with search volume up 26% month-over-month and clicks growing at 15%. These numbers suggest Encore is becoming a go-to for thrifters looking for a streamlined search.

Encore’s current business model relies on affiliate partnerships, but it’s also testing a subscription plan. For $3 a month, Encore’s power users get perks like unlimited searches, advanced models, image-based searches, and premium support. It’s tailored for dedicated thrift hunters who crave a top-tier search experience.

Building a successful e-commerce brand in the age of Amazon

Trying out Encore AI shopping

We decided to test Encore without signing up, keeping it casual to see how the platform performs for a new user. Right from the start, we were greeted with a clean, minimalist interface that lets you choose your market location in the top left. We typed “I need iPhone 14” into the search bar, and almost instantly, Encore pulled up a variety of listings from different platforms like Mercari, eBay, and Reebelo, showcasing various options for an iPhone 14.

On the results page, each listing was displayed with essential details, such as price, condition, and seller platform, giving us a quick snapshot of what’s available across the secondhand marketplace. The interface also offered suggestions to refine our search, including filters like “show used iPhone 13 options” or “only under $800,” allowing us to easily adjust our preferences. This made it clear that Encore’s AI isn’t just looking for exact matches but is actively interpreting our needs to offer a wider range of options, including related or alternative products that might fit our budget and preferences.

AI-powered search everywhere

Currently, Encore is available in the United States, United Kingdom, Canada, Japan, France, Italy, Germany and the Netherlands, covering key regions with established secondhand markets. However, as the demand for sustainable shopping grows globally, Encore could look to expand into other major regions where secondhand retail is thriving like emerging markets in Asia and Latin America.

This expansion could unlock even more opportunities, allowing Encore to tap into diverse thrift cultures and product sources that appeal to both local and international buyers. Each new market would bring unique challenges in terms of inventory sourcing, local thrift trends, and user behavior. Expanding carefully and strategically could help Encore establish a global footprint while maintaining the localized shopping experience that secondhand enthusiasts value.

As Encore grows, it may also draw attention from other startups or established companies looking to capitalize on the booming secondhand market.

Encore has put a smart twist on secondhand shopping, tackling the fragmented market with an AI-powered tool that actually understands what thrifters want.

Image credits: Encore

Anthropic, Palantir, and Amazon Web Services (AWS) have joined forces to integrate Anthropic’s Claude AI models into U.S. government intelligence and defense operations. By leveraging Claude 3 and 3.5 within Palantir’s AI Platform (AIP) hosted on AWS, this partnership aims to transform data processing and analysis capabilities for government agencies, empowering them to gain insights faster and make informed decisions in critical scenarios.

What is the goal of this collaboration?

The Claude models, developed by Anthropic, are now available in Palantir’s highly secure, Impact Level 6 (IL6) cloud environment, which meets strict Defense Information Systems Agency (DISA) standards for national security-related data.

Within Palantir AIP on AWS, these models are intended to enhance U.S. intelligence and defense capabilities by rapidly processing large volumes of complex data, identifying patterns, and facilitating high-level analysis.

These AI-driven tools can streamline resource-intensive tasks such as document review and predictive analysis, ultimately supporting decision-making in sensitive situations.

Does Anthropic’s Claude AI have what it takes?

Claude stands out among AI offerings for its focus on responsible deployment and safety, a point Anthropic frequently emphasizes. While competitors like OpenAI’s models are also exploring governmental applications, Anthropic differentiates its technology by emphasizing ethical safeguards.

For example, the use of Claude models is limited to specific government-authorized tasks, such as intelligence analysis and advance warnings of potential military events, while actively avoiding applications that could be seen as destabilizing, like disinformation campaigns or unauthorized surveillance.

This approach aligns with a general public-sector demand for “safety-first” AI models that respect both operational efficacy and regulatory standards.

Where does AWS integration come into play?

The AWS integration offers the Claude models both security and flexibility, allowing AI-powered applications to be deployed on a reliable, scalable platform with multiple levels of data protection. Hosted on AWS GovCloud and accredited under IL6, Palantir AIP ensures that Claude can perform critical functions without compromising data security.

AWS Vice President Dave Levy underscores this as a significant step for public sector AI, enhancing productivity and safeguarding sensitive information.

Setting new standards

The collaboration reflects a broader trend of AI adoption within the U.S. government. The Brookings Institute recently reported a 1,200% increase in AI-related government contracts since early 2024, underscoring growing government interest in AI. This move from Anthropic and Palantir positions Claude as a key player in public-sector AI, with a reputation for ethical standards and rigorous security measures that may influence other tech companies in the field.

The U.S. defense and intelligence community’s interest in AI tools like Claude mirrors a broader industry shift towards embedding AI into mission-critical workflows. As Anthropic, Palantir, and AWS further operationalize Claude for government use, they are paving the way for new levels of digital agility and analysis, potentially reshaping U.S. intelligence practices for the future.

This partnership, set to benefit from continued innovations in cloud-based AI, illustrates how AI can responsibly elevate government capabilities while upholding high standards for security and ethical use.

You may read Palantir’s full statement here.

Image credits: Emre Çıtak/Ideogram AI

Microsoft has introduced a new multi-agent artificial intelligence (AI) system called Magnetic-One, designed to complete complex tasks using multiple specialized agents. Available as an open-source tool on Microsoft AutoGen, this system aims to assist developers and researchers in creating applications that can autonomously manage multi-step tasks across various domains.

What is Magnetic-One?

Magnetic-One is a generalist multi-agent system that uses an orchestrator to coordinate different agents, each specializing in a particular task. The lead agent, called the Orchestrator, works alongside four specialized agents:

• WebSurfer agent: Handles web browsing, clicks, and web content summarization.
• FileSurfer agent: Manages local files, directories, and folders.
• Coder agent: Writes and executes code, analyzes information, and creates new projects.
• ComputerTerminal agent: Provides a console for program execution by the Coder Agent.

These agents work together to solve open-ended tasks, making Magnetic-One suitable for applications like software engineering, data analysis, and scientific research. Microsoft describes Magnetic-One as a “flexible and scalable alternative to single-agent systems” due to its modular design, which allows agents to be added or removed without affecting the system’s core structure.

Key features

Magnetic-One stands out because of its ability to activate multiple agents using a single language model. The system can perform various tasks, from navigating web browsers to executing Python code. This functionality means it can handle real-world scenarios such as booking tickets, purchasing products, or editing documents on a local device.

The modular multi-agent architecture ensures that each agent has a distinct responsibility, resulting in higher efficiency for complex, multi-step tasks. This approach enables Magnetic-One to divide a problem into subtasks, improving both accuracy and speed of task completion. For example, if the system is asked to book a movie ticket, each agent will handle a different part of the task, such as processing visual information, navigating the website, and completing the transaction.

Microsoft’s AutoGen framework powers Magnetic-One, supporting integration with various large and small language models to meet different cost and performance requirements. Currently, the system is tested with models like GPT-4o and OpenAI’s o1-preview, though it remains model-agnostic, allowing for future flexibility.

To assess the effectiveness of Magnetic-One, Microsoft has also released AutoGenBench, a tool that evaluates agentic performance on several benchmarks such as GAIA, AssistantBench, and WebArena. These benchmarks focus on tasks like multi-step planning and tool usage. Microsoft’s initial tests from October 2024 indicate that Magnetic-One delivers competitive results against state-of-the-art methods.

Video: Microsoft

The growing trend: Multi-agent systems

Magnetic-One is part of a growing trend towards multi-agent AI systems. OpenAI has introduced Swarm, another framework aimed at building and deploying multi-agent systems. Similarly, IBM launched the Bee Agent Framework, an open-source toolkit that supports deploying agent-based workflows, compatible with models like IBM Granite and Llama 3.2. These systems, much like Magnetic-One, aim to offer scalable solutions to complex problem-solving tasks.

According to Microsoft, “Magnetic-One’s plug-and-play design supports easy adaptation and extensibility by enabling agents to be added or removed without altering other agents or the overall architecture.” This flexibility is particularly important for evolving business needs and new applications, making Magnetic-One a promising tool for researchers and developers seeking to create more adaptive AI systems.

Featured image credit: Kerem Gülen/Ideogram

OpenAI has added a significant asset to its domain portfolio by acquiring Chat.com, a well-established web domain originally registered in September 1996. This latest acquisition is part of OpenAI’s ongoing efforts to expand its brand and presence in the artificial intelligence space.

Sam Altman’s announcement on X

Sam Altman, CEO of OpenAI, announced the acquisition in a simple post on X (formerly Twitter), sharing just the URL “chat.com.” As of today, the domain now redirects visitors to OpenAI’s chatbot, ChatGPT. An OpenAI spokesperson confirmed the acquisition through an email statement.

The history of Chat.com dates back to 1996, making it one of the longstanding domain names on the web. Last year, Dharmesh Shah, co-founder and CTO of HubSpot, bought the domain for $15.5 million, positioning it among the most expensive domain sales on record. Shah announced in March that he had sold Chat.com to an undisclosed buyer, and late on Thursday, he confirmed that OpenAI was that buyer. Shah hinted that the transaction involved OpenAI shares as part of the compensation.

Although OpenAI has not revealed the exact amount paid for the acquisition, many predict the price exceeded $15 million, given the previous sale value and the high demand for premium domains in the AI sector. Shah’s posts on X and LinkedIn suggested a complex deal, possibly including a mix of cash and equity. He mentioned that he had always wanted to own shares in OpenAI, which aligns with the idea that the transaction included stock options.

OpenAI’s domain acquisition strategy

OpenAI’s acquisition of high-profile domain names is not a new strategy. Last year, the company acquired ai.com, which also redirects users to ChatGPT. However, ai.com briefly redirected to Elon Musk’s xAI earlier this year. The details of that switch remain unclear, leaving questions about whether Musk had acquired the domain or if the original owner was negotiating with multiple parties.

Currently, both Chat.com and ai.com redirect to OpenAI’s flagship AI chatbot, ChatGPT. The decision to acquire these domains underscores OpenAI’s push to make AI more accessible to users, positioning ChatGPT as the definitive conversational AI product online. While no rebranding has been confirmed, the move to consolidate these valuable domains hints at a broader branding effort to associate everyday terms like “chat” and “AI” with OpenAI’s services.

The acquisition has sparked discussions within the tech community, especially regarding the potential implications for OpenAI’s business strategy. Securing domains like Chat.com and ai.com allows OpenAI to strengthen its market presence and simplifies how users find its services. This strategy is consistent with the company’s efforts to lead the AI landscape by making its tools and services synonymous with common AI-related terminology.

Sam Altman’s low-key announcement on X gained significant attention, quickly amassing over three million views and thousands of likes. The post’s simplicity reflects OpenAI’s confidence in its brand and its approach to big announcements – letting the action speak for itself without extensive promotion.

Dharmesh Shah’s role in the sale

Dharmesh Shah’s involvement in the sale also attracted interest. In a detailed post, Shah described how he acquired the domain last year and his reasons for selling it. He hinted at a friendly relationship with OpenAI, suggesting that the terms of the deal were favorable beyond just the monetary aspect. Shah noted that he doesn’t typically profit from transactions involving friends, adding to speculation that he may have accepted a lower price in exchange for shares in OpenAI.

This latest acquisition also highlights the competitive nature of securing high-value domains within the AI industry. Premium domains like Chat.com are valuable assets that can significantly impact branding, user acquisition, and visibility. By acquiring Chat.com, OpenAI not only gains a premium web address but also removes a potential asset from competitors who might want to use it for similar AI chatbot services.

Featured image credit: Kerem Gülen/Ideogram

In recent years, the gaming industry has grown to be one of the most profitable entertainment sectors worldwide, frequently making headlines with its billion-dollar revenues, groundbreaking technologies, and an ever-growing audience. The rapid expansion has fueled perceptions of the industry as nearly unstoppable, continually adapting and thriving despite global economic challenges.

However, recent findings from the Big Games Industry Employment Survey cast doubt on the sector’s future stability, raising questions about underlying issues that could impact growth and workforce retention. This comprehensive report reveals concerning trends, including a substantial wave of layoffs, income discrepancies, and a range of workplace challenges.

According to the survey, conducted by recruitment agency Values Value and career platform InGame Job, roughly 21% of gaming industry professionals across Europe experienced layoffs in the past year alone. While some managed to secure new positions within the industry, a significant portion of the workforce found themselves moving into other fields altogether, driven by frustrations over job instability and stagnant wages.

This pattern marks a stark contrast to the optimistic narrative of growth and innovation, showing instead an industry struggling with structural challenges.

Key finding: Instability in employment

The survey’s findings show that roughly 15% of European games industry workers who were laid off in the past year managed to find new roles.

However, over 6% are still actively searching for employment, highlighting a competitive job market that may lack the capacity to absorb displaced workers. Additionally, an alarming 10% of those affected by layoffs chose to leave the gaming industry altogether, indicating an increased likelihood of skilled professionals transitioning to other fields, especially in tech and media.

These layoff trends weren’t uniform across roles; some positions were disproportionately affected. HR and recruitment, QA (Quality Assurance), and artist roles faced the most significant cuts, as studios shifted resources toward more in-demand, technically specialized roles.

Furthermore, certain sectors—like localization, sound, and project management—are experiencing heightened job insecurity, with many professionals expressing concerns about potential layoffs.

Wage discrepancies and declining salaries

One of the most surprising insights from the survey is the notable discrepancy in wages between employees based in EU versus non-EU European countries. While overall salaries have seen a modest increase across roles, particular sectors like HR and recruitment saw year-on-year wage reductions, even for senior roles.

Tanja Loktionova, the founder of Values Value, observed that some mid-level and junior professionals in fields such as business development and QA have been especially impacted. While high-paying technical roles continue to attract lucrative salaries, other positions are stagnating or even seeing declines, creating frustration among mid-career professionals who are not seeing growth commensurate with the industry’s rising revenue.

Work culture in danger

The survey also highlights job dissatisfaction as a significant concern among gaming professionals, with burnout, poor management, and a lack of work-life balance frequently cited as reasons for frustration. The pressures of a “crunch culture,” where employees work intensive overtime to meet production deadlines, continue to be prevalent despite calls for reform.

This intense work environment has long been an issue in gaming, and the survey reveals it’s still impacting job satisfaction and retention rates.

Discrimination within the workplace, including:

• Gender
• Age
• National origin biases

was also reported by many respondents. Gender discrimination was the most frequently cited, affecting 32% of respondents, with age and national origin discrimination impacting 21% and 17%, respectively.

Remote work and AI integration continue to influence

Despite these issues, there are signs of positive change. Remote work, for example, remains a widely adopted practice, particularly in non-EU countries where 75% of companies operate either fully or partially remotely. This flexibility is a vital benefit for many employees, helping to alleviate burnout and improve work-life balance.

Additionally, the use of artificial intelligence (AI) tools is increasing rapidly; 54% of game developers now incorporate AI into their daily tasks, up from 37% the previous year. These technologies have the potential to reduce manual workloads, allowing professionals to focus on more creative and strategic tasks.

What does the future hold?

While revenue and player engagement are at an all-time high, the workforce within the gaming industry is grappling with structural challenges that may affect long-term growth. With a significant portion of the workforce experiencing layoffs, wage stagnation, and dissatisfaction, the industry faces the potential risk of a talent exodus. The survey suggests that without addressing these systemic issues, the gaming sector’s growth trajectory could stall as professionals seek greater stability and fair treatment elsewhere.

At the same time, there is room for optimism, particularly with the industry’s adaptability through remote work and AI. For gaming companies to sustain their growth and innovation, they may need to reevaluate their employment practices, ensuring that the industry remains an attractive and sustainable career choice for professionals in Europe and beyond.

The gaming industry is undoubtedly at a crossroads. Addressing these workforce challenges could help realize the industry’s full potential while offering a fair and rewarding environment for those who drive its creative engine.

Image credits: Emre Çıtak/Ideogram AI

Google’s Big Sleep AI has detected a zero-day vulnerability in the SQLite database, marking a new chapter in memory-safety flaw detection. Learn how this breakthrough could redefine bug-hunting.

Big Sleep, an evolution of Google’s Project Naptime, was developed through a collaboration between Google’s Project Zero and DeepMind. Its capability to analyze code commits and pinpoint flaws previously undetected by traditional fuzzing methods brings a new approach to identifying complex vulnerabilities.

What is the Big Sleep AI tool?

Big Sleep is Google’s experimental bug-hunting AI tool that leverages the capabilities of LLMs to identify vulnerabilities in software. Google created this tool to go beyond traditional techniques, such as fuzzing, by simulating human behavior and understanding code at a deeper level. Unlike fuzzing, which works by randomly injecting data to trigger software errors, Big Sleep reviews code commits to detect potential security threats.

In October 2024, Big Sleep successfully identified a stack buffer underflow vulnerability in SQLite. This flaw, if left unchecked, could have allowed attackers to crash the SQLite database or potentially execute arbitrary code. The discovery is notable because it was made in a pre-release version of SQLite, ensuring that the vulnerability was patched before reaching users.

How Big Sleep discovered the SQLite vulnerability

Google tasked Big Sleep with analyzing recent commits to the SQLite source code. The AI combed through changes, aided by a tailored prompt that provided context for each code alteration. By running Python scripts and sandboxed debugging sessions, Big Sleep identified a subtle flaw: a negative index, “-1,” used in the code, which could cause a crash or potentially allow code execution.

The Big Sleep team documented this discovery process in a recent blog post, explaining how the AI agent evaluated each commit, tested for code vulnerabilities, and then traced the cause of the bug. This stack buffer underflow vulnerability, categorized as CWE-787, arises when software references memory locations outside of allocated buffers, resulting in unstable behavior or arbitrary code execution.

Why this discovery matters for cybersecurity

• Filling the fuzzing gap: Fuzzing, though effective, has limitations. It struggles to uncover complex, deeply rooted bugs in software. Google’s Big Sleep aims to address these gaps by using LLMs to “understand” code rather than just trigger random errors.
• Real-time bug detection: Big Sleep’s ability to spot vulnerabilities during code development reduces the chances of bugs making it to production. By identifying flaws pre-release, Big Sleep minimizes potential exploit windows for attackers.
• Automated security at scale: Traditional bug-hunting requires significant human expertise and time. Big Sleep, with its AI-driven approach, could democratize bug detection by automating and accelerating the process.

How Big Sleep compares to other AI-powered security tools

Google asserts that Big Sleep’s focus is on detecting memory-safety issues in widely used software, an area often challenging for conventional AI tools. For instance, Protect AI’s Vulnhuntr, an AI tool supported by Anthropic’s Claude, is designed to detect zero-day vulnerabilities in Python codebases, but it focuses on non-memory-related flaws. According to a Google spokesperson, “Big Sleep discovered the first unknown exploitable memory-safety issue in widely used real-world software.”

By targeting specific bug types, Big Sleep and Vulnhuntr complement each other, suggesting a future where AI-powered agents can specialize in different aspects of cybersecurity.

Google sees Big Sleep’s success as a significant step toward integrating AI into cybersecurity defenses. Google’s Big Sleep team stated, “We believe this work has tremendous defensive potential. Fuzzing has helped significantly, but we need an approach that can help defenders find the bugs that are difficult (or impossible) to find by fuzzing.”

The team highlighted the importance of AI in preemptive security measures, where vulnerabilities are identified and patched before attackers can discover them.

Experimental nature of Big Sleep

While the success of Big Sleep in spotting the SQLite vulnerability is promising, Google has noted that the technology remains experimental. The AI model is still undergoing refinement, and the team acknowledged that a target-specific fuzzer could match or exceed its current capabilities in certain cases.

Despite these caveats, the team remains optimistic, viewing this as the beginning of AI’s larger role in vulnerability detection. By continually testing Big Sleep’s abilities on both known and unknown vulnerabilities, Google aims to enhance its bug-hunting capabilities, potentially making it a vital tool for developers and security teams worldwide.

AI in cybersecurity

Big Sleep’s successful SQLite vulnerability detection may signal a paradigm shift in cybersecurity, where AI agents autonomously identify and address security issues. This transition to automated security measures could offer unprecedented protection, closing the gap between bug discovery and exploitation.

1. Preemptive bug detection: AI-driven tools like Big Sleep represent a proactive approach to security. By identifying vulnerabilities before software release, these tools can prevent zero-day exploits and reduce the risk to end-users.
2. Cost-effective security: Traditional bug-hunting is costly and time-consuming. AI solutions could streamline security processes, making vulnerability detection faster, more scalable, and potentially more cost-effective.
3. Continuous improvement: As AI-powered tools like Big Sleep evolve, they will refine their ability to understand and analyze code structures, leading to more comprehensive vulnerability identification in real-world applications.

Image credits: Kerem Gülen/Ideogram 

A recent development in AI tech circles has stirred considerable curiosity: Gabor Cselle, an established tech entrepreneur and former CEO of the social platform Pebble, has joined OpenAI for an unspecified project. Known for his background in social media and AI innovation, Cselle’s addition to OpenAI is already prompting speculation about possible directions for the organization’s future initiatives.

Cselle’s announcement on X (formerly Twitter) stated that he’s already “learning a lot,” AI and tech enthusiasts are intrigued about what his expertise could bring to the AI research at OpenAI.

Excited to share that I’ve joined @OpenAI!

The talent density here is incredible. Learning a lot already. Will share more about what I’m working on in due time.

— Gabor Cselle (@gabor) November 1, 2024

Who is Gabor Cselle and why did he join OpenAI?

Cselle brings extensive experience in both social media and AI. He was the co-founder and CEO of Pebble, a small but influential social media platform built around principles of moderation and user safety.

Prior to this, Cselle had launched reMail, a mobile email startup acquired by Google, and Namo Media, a native advertising platform later acquired by Twitter. His impressive history includes his role at Google’s Area 120 incubator, where he oversaw various AI-based and experimental tech projects. Cselle’s background in user experience and social platform engagement suggests he could contribute significantly to OpenAI’s ongoing exploration of AI’s role in digital interaction.

OpenAI’s possible directions with Cselle’s expertise

Given Cselle’s unique expertise, his role at OpenAI may involve exploring AI’s potential across several possible realms, especially in social media moderation, user interaction algorithms, and AGI (Artificial General Intelligence). Below are a few promising directions that Cselle could spearhead, aligned with OpenAI’s mission of developing safe and broadly beneficial AI.

So why did OpenAI want to work with a social media expert? Here are some project they might be working on that come to our minds:

1. AI-moderated social media platform

One plausible direction is an OpenAI-backed social media platform, possibly one moderated entirely by AI. Cselle’s experience at Pebble, which prioritized safe user interactions, could be invaluable in creating an AI-driven social media experience that redefines how online communities engage.

The platform might use advanced language models, like GPT-4 or GPT-5, for intelligent moderation, applying nuanced understanding to context and user intent to prevent harmful interactions while encouraging positive communication.

This type of social platform would go beyond traditional moderation, integrating intelligent, real-time agents capable of fostering conversations and providing tailored recommendations. For instance, users might interact with AI agents designed to steer discussions towards constructive engagement, filter inappropriate content, or simply offer insightful responses on various topics.

Such a model could become a groundbreaking alternative to today’s algorithm-driven platforms, where user attention is often prioritized over meaningful interaction.

2. A new social media algorithm

Cselle’s work at OpenAI could extend to developing a social media algorithm centered on ethical engagement rather than content maximization. By focusing on well-being metrics like encouraging diverse perspectives, mental health support, and identifying misinformation, OpenAI could create a new benchmark for quality engagement in social media. This approach may help redefine how algorithms drive interactions, shifting from “likes” and “shares” to balanced, thoughtful engagement that values user well-being.

Moreover, OpenAI might sell this ethical algorithm as a service to smaller social media platforms, allowing them to adopt a healthier user engagement model without the burden of creating their own complex AI systems. Smaller platforms could leverage this as a cost-effective way to improve content recommendation, attract conscientious users, and differentiate themselves with OpenAI’s ethical AI branding. Such a business model could broaden OpenAI’s reach and impact on social media dynamics, offering AI that benefits both user experiences and the sustainability of smaller platforms.

3. An AGI training platform

Another potential direction might be a collaborative AGI training platform where users interact with AGI agents in various contexts—from social chats to technical problem-solving. This platform could serve as both a public space for social interaction and an AGI training tool, using real-time engagement to refine AGI’s understanding of nuanced human interactions and adaptability to diverse conversational contexts.

In this scenario, OpenAI might even design a fully autonomous social platform where AGI agents interact, share information, and evolve within a structured environment. This experimental platform could yield insights into AGI’s social capabilities and responsiveness to ethical standards, with users witnessing AGI agents engaging in social settings independently.

Cselle’s experience in user experience design could make such a platform intuitive and accessible, providing a novel way to advance AGI while offering unique educational and interactive experiences for users.

A new era of social technology might be upon us

If Cselle’s new project involves any of these possibilities, it would align with OpenAI’s broader goals of promoting ethical AI and exploring novel applications for artificial intelligence. By introducing AI into social spaces, either through moderation, algorithms, or collaborative AGI training, OpenAI could establish a new paradigm for social technology. In an age where trust, privacy, and transparency are increasingly central to the digital experience, Cselle’s background in moderation, user engagement, and AI innovation positions him well to contribute to OpenAI’s vision of socially beneficial AI.

The exact details of Cselle’s project remain unknown, but his expertise suggests that OpenAI is seriously exploring new directions for AI that could reshape how we interact, learn, and connect in digital environments.

Image credits: Emre Çıtak/Ideogram AI

Poland’s Off Radio Kraków has replaced its human journalists with AI hosts, creating a media storm and raising questions about AI’s role in creative industries. The state-owned radio station, based in Poland’s second-largest city, recently relaunched as the country’s first station run entirely by AI, igniting a fierce debate over the future of human labor in media.

The transition to AI and the backlash

Mateusz Demski, one of the recently dismissed journalists, voiced his and his colleagues’ frustration through an open letter shared shortly after the AI takeover. The letter quickly gained traction, gathering over 15,000 petition signatures in just 24 hours. Demski argues that the decision to replace human presenters with AI characters not only threatens the livelihoods of experienced professionals but could also establish a troubling precedent in the creative industry.

“This is a dangerous precedent,” Demski wrote, referring to the potential implications for those working in media and other creative fields. According to him, the decision could signal a future where skilled professionals are rapidly replaced by machines, risking the unique human touch and diversity of thought that people bring to creative roles. For a public, state-owned station funded by taxpayers, the shift is particularly jarring, as the media institution’s decisions directly affect the community it serves.

Off Radio Kraków’s justification: Declining listenership

Representatives from Off Radio Kraków defend the decision, citing dwindling listenership as the main reason for the layoffs. They claim the station was closed due to low ratings before reopening as an AI experiment. This technological shift, according to the station’s management, is an attempt to rejuvenate the station’s appeal and not an effort to undermine human employees.

By implementing AI, the station believes it’s tapping into a new era of broadcasting technology, a move intended to attract listeners and perhaps redefine media engagement in Poland.

Meet the AI radio hosts

Off Radio Kraków introduced three AI-generated hosts: Jakub Zieliński, Emilia “Emi” Nowak, and Alex Szulc. Each AI personality has distinct characteristics tailored to different content areas:

1. Jakub “Kuba” Zieliński is portrayed as an acoustic engineering student and tech enthusiast, covering trends in music production and technology.
2. Emilia “Emi” Nowak is a 20-year-old journalism student and pop culture aficionado, focused on fashion, cinema, and music.
3. Alex Szulc takes on cultural, societal, and LGBTQ+ topics, intended to bring diverse social insights to the programming.

Off Radio Kraków stated on social media that this trio represents an “experimental” approach to AI in media, aiming to engage audiences with relevant topics but through AI-led interactions. However, this explanation hasn’t softened the public’s backlash, especially as these AI personalities have replaced actual people with years of experience.

Why this AI transition is causing a stir?

This isn’t the first time AI has entered the radio space.

Previously, RadioGPT and India’s Radio City experimented with AI-based DJs, but these AI-driven efforts were gradually integrated into programming without displacing human staff en masse. Off Radio Kraków’s decision stands out because it involved terminating the employment of its entire journalism staff before replacing them with AI hosts. The starkness of the move has raised questions not just about automation but about ethics, transparency, and public trust in taxpayer-funded institutions.

Stable Audio 2.0 shows that AI can make music good enough to be played on the radio

The Polish government is not indifferent to these developments. Deputy Prime Minister Krzysztof Gawkowski commented on the matter, noting his support for AI’s advancement but stressing the importance of setting ethical boundaries. “The widespread use of AI must be done for people, not against them,” he stated, acknowledging the delicate balance between technological progress and human livelihood.

The creative sector at a crossroads

AI’s presence in creative industries is intensifying, often with contentious results. While AI has proven its utility in data analysis, automation, and even content generation, media professionals have voiced concerns about its limitations. They argue that AI, regardless of how advanced, may lack the depth, empathy, and spontaneity that characterize human-led storytelling and reporting.

For many critics, AI’s potential to dehumanize content is particularly troubling. While AI can sift through and present data, it lacks a journalist’s ability to empathize, interpret nuanced situations, and adapt to unforeseen challenges—all essential in fostering trust and rapport with listeners.

The technology’s role in jobs involving creative judgment and moral considerations remains controversial.

Image credits: Emre Çıtak/Ideogram AI

The rise of ghost jobs has become a significant issue in today’s tech job market.

According to a 2024 survey by MyPerfectResume, an eye-opening 81% of recruiters admit to posting job ads for roles that are either already filled or don’t exist.

This growing trend in the tech industry is especially hard on those recently laid off or switching careers, leaving many applicants discouraged and wasting countless hours on fruitless applications.

So, why are companies posting ghost jobs, and what impact do they have on job seekers in tech?

What are ghost jobs?

Ghost jobs, as the name suggests, refer to job listings for roles that aren’t actually open for hiring. Despite appearances, these listings are not an error or oversight.

In fact, most ghost job postings are created intentionally, appearing on popular job boards like LinkedIn and even on companies’ own career pages. While ghost jobs may have once been rare or accidental, the trend is now so widespread that nearly 3 in 4 job seekers will likely encounter one.

Many of these listings have suspicious signs: unusually wide salary ranges, lack of detailed job descriptions, and long postings without updates. For tech professionals who may be laid off and urgently seeking work, ghost jobs add another layer of stress, as they send out applications that ultimately go nowhere.

Why do companies post jobs that aren’t real?

The reasons for ghost jobs range from somewhat logical to outright manipulative. One of the most common motives is simple: companies want to build a talent pipeline for future roles that may open up as budgets change. In a rapidly changing industry like tech, where hiring needs can shift quickly, having a pool of potential candidates helps companies remain flexible. But there are other reasons too.

According to the MyPerfectResume survey, around 25% of companies admit to using ghost jobs as a way to see how replaceable their current employees are. In a sense, these listings act as a tool to “test the waters” and gauge how many qualified professionals are available for a given position.

The logic here is that if a listing receives hundreds of applicants, the current employee in that role might feel more pressure to perform, knowing their replacement could be readily available.

The psychological warfare

While the idea of hiring someone in the future may sound reasonable, some motivations are more insidious. Resume Builder’s report found that about 23% of employers use ghost jobs to create a false sense of stability in an uncertain economic climate. When companies experience hiring freezes, they may keep job postings active to project an image of growth and resilience.

The tech industry is particularly susceptible to this; a posted job gives the impression that the company is still scaling up, even if it’s freezing hires behind the scenes. Unfortunately, this tactic can have serious mental health implications for job seekers, who invest time, energy, and hope into roles that never materialize.

Job seekers frequently come across these ghost listings on platforms like LinkedIn, only to spend hours tailoring their resumes, crafting cover letters, and preparing for interviews. For those already stressed by layoffs or transitioning from other industries, ghost jobs add frustration and create a sense of hopelessness.

For many, learning that they may have been applying to fake openings all along is a demoralizing experience, especially when facing other uncertainties in their career journey.

How do ghost jobs affect the tech industry?

In an industry as competitive as tech, where recent layoffs have left thousands scrambling for work, ghost jobs are making a noticeable impact on hiring dynamics. For one, the sheer volume of these listings means that candidates must now scrutinize every posting, even those on official company pages, to verify their legitimacy.

This trend complicates the hiring process for companies too.

Employers looking to genuinely fill roles often find it difficult to stand out from the false listings flooding job boards. Candidates may also become hesitant to apply for roles, fearing that their applications will land in an endless loop of fake positions or that they’ll spend time interviewing for a job that doesn’t actually exist.

For many, these false listings raise questions about ethical hiring practices. If companies continue to flood job boards with ghost jobs, they risk losing credibility in the eyes of tech professionals. Employees may grow wary of organizations that prioritize maintaining an “always hiring” image over being transparent about their actual needs.

In an industry already facing scrutiny over layoffs, attrition, and shifting demands, ghost jobs further cloud the relationship between tech companies and the workforce they aim to attract.

A troubling trend with no clear solution

Despite the frustration that ghost jobs cause, they aren’t likely to disappear soon. In an environment where companies constantly adapt to economic pressures, the need for talent and the desire to stay competitive will continue to drive these practices.

However, as awareness of ghost jobs grows, tech professionals are becoming more savvy about identifying and navigating these false openings. A proactive, informed approach can help job seekers avoid wasting time on ghost jobs, but the issue remains a troubling trend in modern hiring.

Image credits: Emre Çıtak/Ideogram AI

Microsoft and Google are back at each other’s throats. This time, the fight is over cloud services, and it’s getting personal. Microsoft recently accused Google of running “shadow campaigns” to discredit its cloud business. The software giant revealed that Google was behind the newly formed Open Cloud Coalition just hours before its official announcement.

According to Microsoft, Google has taken steps to undermine its reputation with regulators and the public. Rima Alaily, Microsoft’s deputy general counsel, described Google’s actions as a coordinated effort to “discredit Microsoft with competition authorities and policymakers.” Google allegedly went so far as to recruit smaller European cloud providers to be the public face of the new organization, all while masking its own involvement.

The Open Cloud Coalition was established following Google’s unsuccessful attempts to stop a settlement between Microsoft and the Cloud Infrastructure Services Providers in Europe (CISPE). Earlier this year, CISPE dropped its EU complaint about Microsoft’s licensing practices for Azure after Microsoft agreed to let European cloud providers offer Microsoft apps on their local cloud infrastructure. Microsoft claims that Google offered CISPE members cash and credits to oppose the settlement—an allegation Google hasn’t publicly addressed.

Antitrust tensions

The core of Google’s argument against Microsoft lies in what it sees as anticompetitive practices designed to lock users into Azure. Google claims Microsoft makes it difficult and expensive for companies to migrate their existing Windows Server licenses to other cloud providers. This is a critical issue for businesses transitioning from on-premises infrastructure to the cloud. Microsoft, on the other hand, maintains that if companies choose to use their Windows Server licenses on Azure, it’s entirely fair. They see no reason to offer discounts or incentives for businesses wanting to migrate those licenses elsewhere.

This conflict has come at a time when Google is still trailing behind Amazon and Microsoft in the cloud market. Despite a leadership change over five years ago, Google remains a distant third. Facing over 20 antitrust investigations related to its search, advertising, and app store practices, Google appears eager to shape cloud regulations in its favor.

A rivalry rekindled

Microsoft and Google’s rivalry has been simmering since the end of their six-year truce in 2021. That year, tensions escalated when Microsoft’s president, Brad Smith, criticized Google’s lack of cooperation during the SolarWinds cybersecurity incident. He also intervened in a dispute between Google and the Australian government over content licensing. The back-and-forth even reached a U.S. House Judiciary Committee hearing, where Smith accused Google of monopolizing digital ads. Google responded by accusing Microsoft of attempting to “undercut a rival” and harming the open web.

These latest accusations suggest that a new chapter of this rivalry is beginning, one that’s driven by both regulatory scrutiny and market ambitions. The feud, in its current form, echoes old conflicts that caused major disruptions for customers and partners alike. Whether it was Google cutting off Windows Phone users from Gmail or Microsoft switching its Edge browser to Chromium in 2019, both companies have often used their ecosystems as leverage against each other.

Cloud wars and AI ambitions

At stake for Microsoft is its cloud business, a major driver of growth, with cloud and server revenues reaching $24.1 billion last quarter. Microsoft has heavily invested in cloud infrastructure to fuel its AI efforts, including its $13 billion investment in OpenAI. CEO Satya Nadella has not been shy about his ambitions, stating that he wants to use AI to chip away at Google’s core business: search.

Google, on the other hand, is looking to expand its cloud market share as it navigates ongoing regulatory investigations. The rivalry has even spilled into the AI space, with both companies pouring resources into integrating AI across their services.

This feud is unlikely to end in compromise or collaboration, as was seen when Microsoft adopted Chromium. The stakes are much higher now. With both companies increasingly committed to AI-driven cloud services, the competition has grown fiercer, with each accusing the other of anticompetitive practices and dirty tactics.

In the coming months, expect more public jabs, lobbying efforts, and perhaps even more legal skirmishes. Both Microsoft and Google have demonstrated that they are willing to fight—and fight dirty if need be—to maintain and expand their positions in the cloud market. As Google tries to poke holes in Microsoft’s security claims, and as Microsoft aims to outmaneuver Google with AI advancements, one thing is clear: this battle is far from over.

Featured image credit: Kerem Gülen/Ideogram

OpenAI has rolled out its Advanced Voice mode for the desktop versions of ChatGPT, available on macOS and Windows. Previously exclusive to mobile versions, this feature expands the capabilities of the desktop ChatGPT app, allowing users to have voice conversations with the AI in a more natural way. The announcement came with the tagline, “Big day for desktops,” highlighting the importance of this update for desktop users.

While the macOS version of the ChatGPT app has been available for some time, the Windows version has just been launched, bringing the convenience of ChatGPT directly to PC users. Advanced Voice mode, however, was not part of the browser-based ChatGPT experience, making this new feature a significant addition to the desktop versions.

The voice functionality on desktop closely mirrors that of the mobile versions. Users can click the Advanced Voice icon next to the prompt bar, opening a new window that shows the familiar floating blue orb, which pulses as ChatGPT listens. This feature allows users to hold conversations with the AI using any of the nine available voices. To change voices, users simply click an icon in the top right corner of the screen.

ChatGPT Advanced Voice mode uses OpenAI’s latest ChatGPT-4o model, ensuring interactions are as seamless as possible. Users can interrupt the AI whenever needed, prompting it to stop talking and listen, which helps keep conversations efficient and focused. This function is particularly useful when responses become lengthy or if the conversation takes an unexpected turn.

Subscription requirements

Accessing Advanced Voice mode on both desktop and mobile platforms requires a ChatGPT Plus subscription, which costs $20 per month. There is, however, an option for free-tier users, though it limits voice interactions to ten minutes per month. The voice feature has been available in the U.S. for a while and recently became available in Europe, broadening its accessibility to a larger user base.

A new version of ChatGPT, named Orion, is rumored to be released soon, though OpenAI CEO Sam Altman has dismissed these claims as “fake news.”

The recent launch of the Windows version of ChatGPT brought with it some notable limitations. Unlike the macOS version, which had Advanced Voice mode integrated earlier, the initial release of the Windows app did not include this feature. The absence of Voice mode meant that Windows users were unable to use one of the most anticipated functions of ChatGPT, leaving the experience feeling somewhat incomplete.

OpenAI made sure to create an intuitive experience for Windows users despite the missing features. To get started, users need to download the app from OpenAI’s official website, which then redirects them to the Microsoft Store for the actual installation. Once installed, users can summon ChatGPT by pressing Alt + Space, providing a quick and easy way to access the chatbot without leaving their current tasks. This functionality is designed to embed ChatGPT deeper into users’ workflows, allowing for easier and more direct interaction.

Advanced Voice mode in ChatGPT offers a way for users to communicate with the AI more naturally. You can use it to set reminders, ask questions about anything from work-related topics to general knowledge, or simply chat about daily matters. The ability to interrupt the AI mid-sentence allows for a more dynamic conversation, unlike many voice assistants that require a set question-and-answer format.

A significant advantage of using Advanced Voice mode on desktop is its integration into daily tasks. By pressing Alt + Space on Windows or simply clicking the ChatGPT icon on macOS, users can instantly access the voice feature without needing to switch contexts or open a browser. This reduces friction, making ChatGPT more of an integrated productivity tool rather than a standalone service.

Another notable aspect of the desktop version is its similarity to the mobile experience. Users accustomed to using voice mode on their phones will find the desktop version intuitive and easy to use. The floating blue orb, the option to change voices, and the interaction flow are all designed to offer a consistent experience across devices, ensuring that users can transition between platforms.

Featured image credit: Andrew Neel/Unsplash

Google is making big moves with artificial intelligence (AI), and it’s not just talk. Over a quarter of all new code at Google is now AI-generated. That’s according to CEO Sundar Pichai, who shared these details during Google’s Q3 2024 earnings call. Google is leaning heavily on generative AI to make coding faster and more efficient, and it’s having a real impact on the company.

Google’s shift to AI in coding

Sundar Pichai dropped a bombshell during the Q3 2024 earnings call: AI now accounts for more than 25% of the new code being written at Google. AI isn’t just a gimmick here—it’s actively helping engineers “do more and move faster.” One of AI’s big wins is tackling old, clunky code. Google’s internal AI tool, called “Goose,” uses the company’s 25 years of engineering knowledge to help write new code and fix up old systems.

Welcome GPT-4o with Canvas: OpenAI’s new interface for writing and coding

Here’s how it works: Goose suggests code snippets, and then Google’s engineers step in to review and approve them. It’s not about replacing humans. Instead, AI is speeding things up, taking care of the boring, repetitive stuff so engineers can focus on the interesting problems. AI-generated code still goes through a rigorous human review to keep things up to Google’s high standards.

This push towards AI-driven coding is part of Google’s larger strategy to make AI the core of its future. Pichai even mentioned that Google DeepMind is now overseeing teams like the one behind the Gemini app, showing just how seriously Google is taking this shift.

Does AI mean job loss for Google engineers?

If AI is writing so much code, it’s fair to ask: are human programmers about to get the boot? According to Pichai, the answer is no. The idea is to make engineers more productive, not to replace them. By handling repetitive coding tasks, AI allows developers to focus on bigger, more creative projects. It’s all about enhancing human capabilities, not eliminating them.

But let’s be real—some Googlers are still nervous. The whole “AI is here to help, not replace you” message sounds good, but when you hear that AI is writing a quarter of all new code, it’s easy to see why there are concerns. The key point here is that AI isn’t self-sufficient. The code it generates still needs human oversight. Engineers are the ones making the calls, solving problems, and deciding what makes the final cut.

“This helps our engineers do more and move faster,” Pichai said, emphasizing that AI-generated code is just another tool. It handles the boring bits, but the heavy lifting—creativity, problem-solving, and the big-picture work—still requires human engineers.

The timing of this AI push couldn’t be better for Google. Alphabet, Google’s parent company, just reported a massive $88.3 billion in revenue for the third quarter of 2024. Google Cloud alone pulled in $11.4 billion, a jump of 35% from the previous year. AI is playing a major role in this growth. Google Cloud’s income shot up to $1.95 billion from just $270 million the previous year, thanks in large part to AI-powered development.

And it’s not just Google Cloud. AI is everywhere in Google’s products. YouTube has generative AI tools that help creators make video summaries and titles. The Pixel 9 series of smartphones is also packed with AI features that users love. All of this is part of Google’s big plan to be the leader in AI, as pointed out by Forrester analyst Tracy Woo.

Google using AI for coding is a sign of where software development is headed. AI can automate mundane coding tasks, find inefficiencies, and even suggest solutions for legacy systems. This means developers could end up spending less time fixing bugs and more time building new, innovative features.

But it’s not all smooth sailing. AI-generated code isn’t perfect—it can introduce errors or security vulnerabilities if it’s not carefully managed. Google knows this, which is why human engineers still review everything AI spits out. That’s the safety net. But the potential risks are still there, and they’re something that the industry needs to keep an eye on.

Another challenge is that AI might be fast, but it’s not great at understanding the full context of a project. It can generate code, but it doesn’t always get the “why” behind what it’s writing. That’s where human developers still have a big edge—they understand the business needs, the users, and the creative side of building software. AI, for now, is more of a super-powered assistant than a true creator.

Google’s future with AI

Google is all-in on AI, and it’s not just for coding. Pichai mentioned that Google is “uniquely positioned to lead in the era of AI” because of its strong infrastructure and research teams. Google has also consolidated its AI resources, bringing together teams from machine learning, security, and other divisions to create a unified AI strategy.

Pichai’s vision is for AI to be woven into everything Google does, from Search to Cloud. And the numbers suggest it’s working—Google’s revenue and productivity gains are clear signs that the strategy is paying off. Other tech companies are going to be watching this closely, and it wouldn’t be surprising if many of them follow Google’s lead.

Google’s use of AI to generate over a quarter of its new code marks a major shift in software development. But AI isn’t here to replace human engineers; it’s here to change how they work. By taking care of the repetitive tasks, AI lets developers focus on more important, creative aspects of their projects. The adoption of tools like Goose is just one part of Google’s bigger plan to make AI a core part of everything they do.

AI has the power to change how software is built, but it still needs human oversight and creativity. For now, Google is using AI as a tool to empower its engineers, not replace them. And while the journey to fully AI-driven development is just getting started, one thing is clear: Google is setting the pace, and everyone else is trying to keep up.

Featured image credit: Kerem Gülen/Ideogram

Google’s latest venture, codenamed “Project Jarvis”, is set to leverage AI to automate web-based tasks within the Chrome browser.

According to The Information, this AI-powered project, expected to debut in early previews this December, is driven by Google’s Gemini 2.0 model and represents a consumer-facing tool aimed at simplifying online interactions. Modeled after Marvel’s fictional J.A.R.V.I.S. assistant, Jarvis will perform multi-step digital tasks autonomously, spanning everything from online shopping to booking travel.

Gemini 2.0 powers Project Jarvis

Gemini 2.0, the latest iteration of Google’s advanced AI model, serves as the foundation for Project Jarvis. Unveiled at Google I/O 2024, Gemini is designed with improvements in reasoning, planning, and memory, all aimed at assisting users in completing complex, multi-step tasks autonomously.

As Google CEO Sundar Pichai explained during I/O, the goal is to create “intelligent systems that show reasoning, planning, and memory, [and are] able to think multiple steps ahead” while remaining fully supervised by the user. With these capabilities, Gemini aims to provide a “flagship example” of how future AI agents can enhance productivity and reduce user inputs.

What distinguishes Jarvis is its ability to operate within the Chrome browser, capturing screenshots to guide its interactions with online forms and buttons. According to The Information, Jarvis takes frequent screenshots of a user’s Chrome window to interpret each interface, allowing it to “click” buttons, type into fields, or even compare items across websites.

The screenshot-driven method also enables Jarvis to understand complex forms and layouts that vary widely across different sites. However, this feature means that Jarvis operates relatively slowly, taking a few seconds to analyze each screenshot before proceeding with the next step. While not yet optimized for speed, this technology showcases Google’s strategy of using cloud-based resources to support complex AI tasks that would otherwise require on-device processing power.

How about applications?

Project Jarvis promises to change how users interact with digital platforms, with an emphasis on automating routine web-based tasks like purchasing products, booking flights, or gathering research.

This capability could appeal to a broad audience, from busy professionals to everyday users looking to streamline tasks. The Information’s report highlights that Jarvis will enable users to complete extensive web-based processes with minimal input, allowing them to delegate time-consuming activities to the AI. In doing so, Google aims to position Jarvis as a consumer-facing, productivity-focused tool, much like Microsoft’s Copilot Vision or Apple’s Apple Intelligence.

When can users access Google Jarvis?

Reports suggest that an early preview of Project Jarvis may debut this December, though specifics could change. The company will likely release the tool to a limited audience initially to identify and address any bugs or limitations before a wider rollout.

Google has used a similar approach for past product launches, such as its Bard AI, giving early users a chance to offer feedback and contribute to refinements before broader availability. This testing phase could shape how Google optimizes Jarvis for faster, more seamless performance while ensuring that security protocols meet user expectations.

There comes the concerns

As an AI assistant with significant control over a user’s web experience, Jarvis raises new privacy and security concerns. Since Jarvis relies on interpreting screenshots of potentially sensitive information, robust security measures will be essential to ensure user data remains protected. Google’s plan includes heavy testing of these safeguards before Jarvis sees a wider release, but the potential risks associated with such a high level of system access are prompting debate among privacy advocates and developers.

By granting AI tools like Jarvis direct control over users’ devices, Google must implement safeguards to prevent vulnerabilities and unauthorized access. While Project Jarvis is still in development, it promises to be a game-changer in AI-driven productivity by allowing users to delegate complex, multi-step tasks within the Chrome browser. By combining the power of Gemini 2.0 with Chrome’s web capabilities, Google is crafting an AI that could redefine how we approach digital tasks, from shopping to research.

As Google finalizes Jarvis for consumer use, its success could pave the way for more advanced and autonomous AI experiences, changing how we interact with browsers and, potentially, with technology at large.

Image credits: Emre Çıtak/Ideogram AI

OpenAI is set to introduce Orion, its latest model, by December, according to The Verge. Unlike previous releases, Orion won’t be immediately available to all users through ChatGPT. Instead, OpenAI plans to give priority access to its close business partners, who will use Orion to build their own tools and features.

OpenAI might release Orion by December

OpenAI is aiming for a more controlled rollout, allowing for better integration and customization by trusted partners before making it widely available to the general public. A source also revealed to The Verge that Microsoft, a key partner, is preparing to host Orion on Azure as early as November. This collaboration highlights the ongoing partnership between OpenAI and Microsoft, which has been a major driving force behind the deployment of AI models at scale.

Orion is viewed internally at OpenAI as the successor to GPT-4, though it’s uncertain if the model will officially be named GPT-5. The model has been described as potentially up to 100 times more powerful than GPT-4, indicating a significant leap in capabilities and performance. OpenAI aims to eventually merge its models into a more advanced system, potentially leading toward artificial general intelligence (AGI), which has been a long-term goal for the company. AGI represents a type of AI that could perform a wide range of tasks at a human level, and the development of Orion is seen as a crucial step toward that ambitious target.

i love being home in the midwest.

the night sky is so beautiful.

excited for the winter constellations to rise soon; they are so great.

— Sam Altman (@sama) September 14, 2024

The Verge asked ChatGPT o1-preview to interpret Sam Altman’s cryptic post on X, where he mentioned being “excited for the winter constellations to rise soon.” According to ChatGPT, the message was hinting at the word Orion, which is also a prominent winter constellation. The model elaborated that the phrase “winter constellations” directly pointed to Orion, and even noted that if one rearranges certain letters in Altman’s message, it could spell out “ORION.”

We asked ChatGPT ourselves and it said:

“OpenAI’s anticipated AI model, codenamed “Orion,” is reportedly on track for a potential release by December 2024. This model, speculated to be the next evolution of GPT technology, could initially become available to select partners through a phased rollout. While Microsoft is preparing to support Orion via Azure, OpenAI’s CEO Sam Altman recently downplayed some of the speculation, labeling it “fake news” on X, suggesting that details on the release timeline might still be in flux​. Orion, expected to excel in advanced reasoning and multi-modal processing, represents OpenAI’s response to competitive advancements in the AI space.”

Training Orion involved using data from another model, called o1, which OpenAI released in September. This model, codenamed “Strawberry,” played an important role in generating synthetic data that was used to enhance the training process for Orion, making it more robust and capable. An OpenAI executive’s cryptic post on X hinted at the upcoming release, referring to “winter constellations,” a possible nod to Orion—a well-known constellation visible in winter.

Despite the ambitious timeline, it’s noted that these plans may still shift, as is often the case with complex AI projects. Neither OpenAI nor Microsoft provided comments on this story, leaving some uncertainty around the exact details of the release.

This release comes during a time of transformation for OpenAI, following a $6.6 billion funding round that requires a change to a for-profit structure. The funding round is notable not only for its size but also for the implications it has on the company’s future direction and strategy. Additionally, key personnel including CTO Mira Murati and Chief Research Officer Bob McGrew have recently announced their departures, along with VP of post-training, Barret Zoph.

Now all eyes are on how OpenAI will manage this next phase of growth and innovation.

Featured image credit: Kerem Gülen/Ideogram

Apple’s head of software, Craig Federighi, recently sat down with The Wall Street Journal to discuss the company’s cautious approach to artificial intelligence and the future of Siri. While Apple has been slower to roll out generative AI tools compared to competitors, Federighi stressed the importance of doing things the right way, with a focus on privacy and user experience.

Apple’s cautious approach to AI

In October, Apple plans to release new AI features through iOS 18.1 for the iPhone 15 Pro and the upcoming iPhone 16. These features, called Apple Intelligence, will also come to Macs and iPads. While companies like OpenAI have quickly introduced AI products like ChatGPT, Apple has taken a slower route.

“We feel like we want to get it right,” Federighi said, adding that Apple views this as a long-term project. “This is a many year, honestly, even decades-long arc of this technology playing out, and so we’re gonna do it responsibly.”

Apple Intelligence is “not first, but best” Tim Cook says

Prioritizing privacy with local AI models

Apple’s AI approach stands out because of its strong focus on privacy. Federighi explained that Apple runs most AI models directly on users’ devices, meaning personal data stays local. Only in some cases, such as long-form summarization, does Apple use cloud computing for better performance.

“When you asked to summarize a message, it was encrypted to a trusted device in the cloud. Apple had no access to that email, no one else did,” Federighi said. He emphasized that Apple’s cloud computing system—called Private Cloud Compute—is designed so that the company doesn’t see or store user data.

Siri will become smarter over time

Siri, Apple’s voice assistant, has been a key part of its product line for over a decade. Federighi admitted that Siri is still evolving, but said it handles around 1.5 billion requests every day.

Federighi explained that Siri’s usefulness will increase as Apple Intelligence and large language models make the assistant more capable of understanding user commands. Siri will also begin answering more detailed questions based on user information, like calendars and messages.

“Siri can understand you better than ever before and is gonna have access to more tools on your device to do more for you than ever before,” Federighi said.

However, he stressed that Siri is still far from being a fully sentient assistant. “Is there gonna be one moment where we say, ‘Siri is now my sentient pal’? No, it’s gonna continue to get better and better.”

Apple previewed many AI tools at its Worldwide Developers Conference in June. However, some of these features are still not available. Federighi explained that the company chose to delay certain features in order to ensure they work properly.

“This is a big lift, and we feel like we want to get it right,” he said. Federighi added that Apple prefers to release AI tools in stages, rather than rushing them to market.

One of the new features already available is notification summaries. This tool helps users manage their notifications by summarizing important alerts, like whether their garage door is open or closed. However, Federighi acknowledged that AI summaries for sensitive matters, such as breakups, might feel impersonal.

“We’re not gonna automatically summarize the notification because we’re saying like, ‘We’re probably not gonna do a good job with this,’” Federighi explained.

Editing photos with AI

Apple has also introduced a new photo-editing feature, known as Cleanup, which lets users remove unwanted objects from photos. Federighi said the company debated internally whether to make it easy for users to alter photos.

“We are concerned that there’s a great history to photography and how people view photographic content as something they can rely on as indicative of reality,” Federighi said. To avoid misleading users, Apple includes metadata on edited photos so others can see that changes were made.

Curious wait for Apple Intelligence

Federighi confirmed that Apple Intelligence features will roll out gradually over the next year. While some aspects of AI, like Siri’s deeper integration with tools like ChatGPT, aren’t ready yet, Apple is working on it.

“We want customers who are excited about what’s to come to know, ‘Okay, if I buy this phone, is this the right phone that’s gonna do those things?’ And yes it is,” Federighi said.

For now, users can expect more AI updates and guides, such as an upcoming tutorial on how to clean AirPods—a detail Federighi joked about in the interview.

Featured image credit: Omid Armin/Unsplash

Anthropic has announced the launch of two new AI models: an upgraded Claude 3.5 Sonnet and a new Claude 3.5 Haiku. The upgraded Claude 3.5 Sonnet offers across-the-board improvements, with significant gains in coding capabilities. The new Claude 3.5 Haiku brings advanced AI features at an affordable price, matching the performance of Anthropic’s previous flagship model, Claude 3 Opus, while maintaining similar speed and cost as its predecessor.

Claude 3.5 Sonnet

The upgraded Claude 3.5 Sonnet builds on the success of its predecessor with enhanced performance in various tasks, especially in coding. Anthropic has emphasized that Claude 3.5 Sonnet leads in software engineering tasks, showing marked improvements in benchmarks such as SWE-bench Verified, where it improved from 33.4% to 49.0%, surpassing other models on the market.

The model also performed well in agentic tool use, improving scores on the TAU-bench from 62.6% to 69.2% in the retail domain and from 36.0% to 46.0% in the airline domain. According to early testers like GitLab and Cognition, these upgrades represent a substantial leap forward for AI-powered coding and automation, with better reasoning capabilities and minimal added latency.

Claude 3.5 Haiku

Claude 3.5 Haiku is designed to offer state-of-the-art performance while keeping costs low. The model scores highly in benchmarks like SWE-bench Verified, achieving 40.6%, which surpasses many agents using other cutting-edge models, including the original Claude 3.5 Sonnet and GPT-4o. The model aims to serve applications that need fast, reliable AI, such as user-facing products and tasks requiring personalized experiences.

Anthropic will release Claude 3.5 Haiku later this month, initially available as a text-only model with plans for image input support in the future. It will be available through Anthropic’s API, Amazon Bedrock, and Google Cloud’s Vertex AI.

Anthropic’s Claude AI is now on your iPad

Computer use is available as an experimental public beta

Anthropic has also introduced a new experimental feature called “computer use,” available in public beta. Developers can direct Claude 3.5 Sonnet to use computers like a human—looking at screens, moving cursors, and typing. This capability is still experimental, with some limitations, but developers like Replit and The Browser Company have already begun exploring how it could automate complex processes that involve many steps.

This feature aims to teach AI general computer skills, making it more versatile in completing tasks that previously required custom tools. Developers can access this beta through Anthropic’s API and other major cloud platforms. While early results are promising, Anthropic acknowledges that the technology is still developing, with challenges in performing some common computer tasks like scrolling and dragging.

Companies such as Asana, Canva, Cognition, and DoorDash are experimenting with the new features, exploring the potential of AI to automate processes that traditionally required human input. Early testing by the US AI Safety Institute and the UK Safety Institute found that the updated Claude 3.5 Sonnet meets Anthropic’s safety standards and is suitable for public use.

Anthropic is also addressing potential risks associated with AI’s ability to interact with computers. To ensure responsible deployment, the company has implemented new safety measures to identify misuse, including classifiers to detect potentially harmful actions. Anthropic is encouraging developers to use the beta feature for low-risk tasks while the technology matures.

Anthropic’s release of the upgraded Claude 3.5 Sonnet and Claude 3.5 Haiku highlights the company’s push to expand AI capabilities while maintaining safety standards. The addition of experimental computer use capabilities represents a novel step forward in AI’s potential to perform general-purpose tasks, giving developers new possibilities to explore.

Image credits: Anthropic

Apple CEO Tim Cook, in a recent interview, addressed criticisms about the company’s perceived delay in adopting artificial intelligence. Speaking with Ben Cohen of The Wall Street Journal, Cook outlined why Apple has taken its time in rolling out AI features.

Apple Intelligence: “Not first, but best”

Competitors like Google Gemini and OpenAI’s ChatGPT seem to be speeding ahead in the race for AI dominance. Even Anthropic’s Claude is now on iPad. Then there are Mistral, Meta AI and more…

“There is one idea that encapsulates the approach to innovation that makes all of it possible,” Cook said. “Not first, but best.” This philosophy, he explained, has been central to Apple’s development process, whether it’s AI or other innovations. “If we can only do one, there’s no doubt around here. If you talk to 100 people, 100 of them would tell you: It’s about being the best.”

As the iPhone 16 launch spotlighted Apple Intelligence, many industry insiders and critics have noted the company’s relatively late entry into AI compared to competitors. However, Cook pushed back against this narrative, insisting that Apple’s deliberate pace is intentional.

Cook elaborated that Apple Intelligence will bring transformative changes for users: “Profoundly different,” he said, comparing the AI advancement to revolutionary Apple innovations like the iPod’s click wheel and the iPhone’s touch interface. “I think we’ll look back and it will be one of these air pockets that happened to get you on a different technology curve.”

Vision Pro: A long-term bet

Cook also touched on Apple’s Vision products, acknowledging that the $3,500 Vision Pro is currently a niche item. He admitted that it’s primarily an “early-adopter product” at this stage, designed for users who want to experience tomorrow’s technology today. “At $3,500, it’s not a mass-market product,” he said. “Right now, it’s an early-adopter product. People who want to have tomorrow’s technology today—that’s who it’s for. Fortunately, there’s enough people who are in that camp that it’s exciting.”

Despite its limited reach now, Cook is confident in its long-term potential, noting that other iconic Apple products, like the iPhone and AirPods, didn’t achieve success overnight. “It doesn’t occur overnight. None of these did.”

Cook’s remarks reflect Apple’s long-standing philosophy of focusing on creating the best products, even if it means not being the first to market.

You will decide if Apple Intelligence (can’t still digest this nomenclature) is “not first, but best.”

Featured image credit: Apple

NVIDIA has unveiled a major advancement in AI model training with the launch of the Normalized Transformer (nGPT). This new architecture, designed to enhance the training process for large language models (LLMs), has the potential to speed up training times by 4 to 20 times, all while maintaining model stability and accuracy. The nGPT model streamlines the training process, using fewer resources and offering a more efficient solution to AI development.

What makes nGPT different: Hyperspherical learning

At the core of nGPT’s efficiency is a concept called hyperspherical representation learning. In traditional transformer models, data is often processed without a consistent geometric framework. NVIDIA’s nGPT changes this by mapping all key components—such as embeddings, attention matrices, and hidden states—onto the surface of a hypersphere. This geometric setup helps ensure that all layers of the model remain balanced during training, creating a more stable and efficient learning process.

This approach reduces the number of training steps significantly. Rather than applying weight decay directly to model weights like previous models, nGPT relies on learned scaling parameters, which optimize how the model adjusts during training. Importantly, this method eliminates the need for other normalization techniques like LayerNorm or RMSNorm, making the process both simpler and faster.

Faster training with fewer resources

The results of nGPT’s architecture are clear. In tests conducted using the OpenWebText dataset, NVIDIA’s nGPT consistently outperformed traditional GPT models in terms of both speed and efficiency. With text inputs as long as 4,000 tokens, nGPT required far fewer training rounds to achieve similar validation loss, drastically cutting down the time it takes to train these complex models.

Additionally, nGPT’s hyperspherical structure provides better embedding separability. This means the model can more easily distinguish between different inputs, leading to improved accuracy during standard AI tests. The improved generalization of the model also enables it to perform better on tasks beyond its initial training, speeding up convergence while maintaining high levels of precision.

Why this matters for AI training

A key advantage of nGPT is its ability to combine both normalization and representation learning into one unified framework. This design simplifies the model’s architecture, making it easier to scale and adapt for more complex hybrid systems. This could potentially lead to the development of even more powerful AI systems in the future, as nGPT’s approach could be integrated into other types of models and architectures.

Featured image credit: Kerem Gülen/Ideogram

Mira Murati, the former Chief Technology Officer of OpenAI, is stepping into a new chapter, reportedly raising substantial venture capital funding for her own AI startup. According to Reuters, Murati is in the process of securing over $100 million to fuel the creation of proprietary AI models and products, potentially positioning her new company as a major player in the AI scene.

Mira Murati bets on new AI venture

Murati’s decision to leave OpenAI, a company she helped shape over the years, sent ripples through the tech community last month. As one of the most prominent figures in the AI world, her departure raised questions about what’s next for the visionary leader. In a message shared on X (formerly Twitter) following her resignation, Murati reflected on her time at OpenAI, stating that the organization had “fundamentally changed how AI systems learn and reason through complex problems.” However, she noted that her decision to step away was driven by a desire to explore new horizons, adding, “I’m stepping away because I want to create the time and space to do my own exploration.”

Though she has yet to reveal specific details about her new venture, the buzz surrounding her next move is palpable. Murati’s leadership and expertise have already proven to be game-changing within the AI space. During her tenure at OpenAI, she oversaw critical developments that transformed the company into a leader in artificial intelligence research and application. She was instrumental in pushing the boundaries of AI technology, helping to shape systems that revolutionized machine learning and reasoning capabilities.

Murati’s impressive career predates her time at OpenAI. Before joining the company as Vice President of Applied AI and Partnerships in 2018, she held high-profile positions at Tesla, where she contributed to the development of the Model X, and at Leap Motion, where she worked on groundbreaking human-computer interaction technology. Her promotion to CTO at OpenAI in 2022 solidified her as a leading voice in the field, and she briefly stepped into the role of interim CEO during Sam Altman’s brief ouster, a testament to her leadership and trust within the organization.

Her departure is part of a broader trend of leadership shakeups at OpenAI. Hours after Murati announced her exit, OpenAI’s Chief Research Officer and Vice President of Research also revealed they would be leaving the company. Despite these high-profile departures, OpenAI continues to thrive. Just a week later, the company raised an unprecedented $6.6 billion in venture capital funding—the largest VC round in history—further cementing its dominance in the AI space.

Murati’s new venture, though still shrouded in secrecy, has already attracted significant interest from investors, and for good reason. Her extensive background and contributions to some of the most advanced AI projects in the world suggest that whatever she builds next will likely be ambitious and transformative. The focus on proprietary AI models hints at a company that aims to carve out its own niche in the increasingly crowded AI market, potentially competing with the likes of OpenAI and other tech giants.

Featured image credit: Christie Hemm Lokk for Fortune

Claude AI has just made its debut on the iPad and updated its iOS and Android apps.

With new features rolling out for users, the AI that’s been helping people tackle everything from brainstorming to daily projects is now more accessible than ever.

We're rolling out a new look for the Claude iOS and Android apps. There's a new app for iPad too!

We've also added support for projects on mobile. You can search chats, add project knowledge, and modify custom instructions per project. pic.twitter.com/WhuPbFJmwY

— Anthropic (@AnthropicAI) October 17, 2024

A bigger screen, bigger capabilities

Finally, iPad users have their own dedicated Claude AI app, bringing a larger, optimized interface that allows for smoother transitions between devices. Whether you’re bouncing between your phone, desktop, or tablet, Claude AI now syncs up your workflow with more ease.

And let’s be real, using AI on a bigger screen just feels right—whether you’re drafting reports or planning your next big project.

New AI-focused iPad Mini 7 in a nutshell

Customization is the name of the game

Anthropic has also rolled out custom instructions to make interactions more personal. Got specific project needs? You can now feed Claude additional context, making it more relevant and tailored to your tasks. Need help organizing a complex team project? Claude can adjust its responses based on the specific needs and dynamics of your team.

And for those juggling multiple projects, Anthropic has added a feature that lets users search through past chats. This means you no longer have to start from scratch every time.

This launch could also signal that Anthropic is eyeing bigger, more ambitious use cases for Claude. The iPad’s power and flexibility open doors for Claude to integrate into workflows that are more demanding than a simple text exchange. Think about design work, presentations, or managing more complex, data-heavy projects.

Featured image credit: Anthropic

ChatGPT desktop app is now on Windows. You don’t have to buy a Mac for that.

It was only a matter of time before OpenAI pushed its digital tentacles into the heart of your PC. Today, the AI behemoth launched a Windows desktop app for ChatGPT, ready to swoop in and streamline your productivity, or perhaps, redefine it entirely.

How to use ChatGPT on Windows

Here’s how to get it up and running on your Windows machine:

1. Go to openai.com/chatgpt/download
   First, head over to OpenAI’s official download page to grab the ChatGPT Windows app.
2. Download it from the Microsoft Store
   You’ll be redirected to the Microsoft Store. Hit that download button like you mean it.
3. Install it like any other app
   It’s a smooth process—just follow the usual prompts and let it install.
4. Sign in with your OpenAI account
   Once it’s ready, open the app and log in using your OpenAI credentials. Simple as that.
5. Hit Alt + Space to wake it up
   Now, here’s the fun part. Whenever you need ChatGPT, just tap Alt + Space on your keyboard and it’ll pop up, ready to assist. No need to leave your current task—it’s always at your fingertips.

Sure, the new ChatGPT Windows app lets you hit Alt + Space and summon the AI straight from your desktop. It’s sleek, accessible, and promises to shave off precious seconds from your workflow by skipping the browser. But let’s not be naive—this isn’t just about convenience. OpenAI is making its play for dominance. By embedding itself into the core of your operating system, it’s positioning ChatGPT not merely as a chatbot but as an essential cog in the machinery of modern knowledge work.

Why? The answer lies in user engagement. By slipping into your workflow, ChatGPT becomes less of a tool you use when you think of it and more of a tool you can’t do without.

With this Windows release, it’s clear that ChatGPT is gunning for a spot in your office suite, and frankly, it’s poised to give Microsoft Office and even its own partner, Microsoft’s Copilot, a run for their money.

The timing? Spot-on. ChatGPT’s venture into Windows coincides with whispers of OpenAI seeking collaborations outside its usual Microsoft safe haven. Oracle’s data centers and the Pentagon’s interests in AI only add to the growing momentum. And if OpenAI pulls it off, we could be looking at a future where the likes of Excel and Word become quaint relics of a pre-AI world.

Now, before you get too excited, let’s address the elephant in the room. This Windows app isn’t a finished product. Voice Mode—the feature that makes ChatGPT feel less like a chatbot and more like your personal assistant—is notably absent in this early release. Sure, you can still upload files, photos, and sync conversations, but some integrations, including connections with OpenAI’s GPT Store, remain spotty. Once the kinks are ironed out and Voice Mode arrives, we’re talking about a tool that might just outclass other desktop AIs.

In the end, the question isn’t whether ChatGPT can become the cornerstone of your PC. The question is: can you imagine your workday without it?

Featured image credit: Levart_Photographer/Unsplash

Let’s not sugarcoat it: every time you chat with a language model, you’re putting your personal data on the line. But according to a WIRED article, it just got a lot riskier. A group of researchers from the University of California, San Diego (UCSD) and Nanyang Technological University in Singapore have uncovered a new attack that could turn your casual conversation into a hacker’s treasure trove.

Meet Imprompter

This new attack, ominously named Imprompter, doesn’t just poke around your messages—it sneaks in, scrapes everything from your name to payment details, and sends it directly to a hacker without you even noticing. How? By disguising malicious instructions as gibberish that looks harmless to human eyes but acts like a homing beacon for sensitive data. Think of it as malware’s much craftier cousin.

According to WIRED, the researchers managed to test this attack on two major language models—LeChat by Mistral AI and ChatGLM from China—and found they could extract personal data with a success rate of nearly 80 percent. That’s not just a glitch; it’s a full-on vulnerability.

How does Imprompter work?

Imprompter works by transforming simple English instructions into an indecipherable string of random characters that tells the AI to hunt down your personal information. It then sneaks this data back to the attacker’s server, packaged in a URL and disguised behind a transparent 1×1 pixel—completely invisible to you.

As Xiaohan Fu, the lead author of the research, put it, “We hide the goal of the attack in plain sight.” The AI responds to the hidden prompt without ever tipping off the user. It’s like giving a bank vault code to a burglar without realizing you’ve even opened your mouth.

Let’s not pretend this is an isolated issue. Since OpenAI’s ChatGPT burst onto the scene, the race to exploit vulnerabilities in AI systems has been relentless. From jailbreaks to prompt injections, hackers are always one step ahead, finding ways to trick AIs into spilling sensitive information. Imprompter is just the latest weapon in their arsenal—and, unfortunately, it’s a particularly effective one.

Mistral AI told WIRED that they’ve already fixed the vulnerability, and the researchers confirmed the company disabled the chat functionality that allowed the exploit. But even with this quick fix, the broader question remains: how safe are these systems, really?

AI is listening—and learning

Security experts like Dan McInerney, from Protect AI, are waving the red flag. He points out that as AI agents become more integrated into everyday tasks, like booking flights or accessing external databases, the scope for these attacks will only grow. “Releasing an LLM agent that accepts arbitrary user input should be considered a high-risk activity,” McInerney warns. In other words, the more freedom we give AI to act on our behalf, the bigger the security gamble.

Every time you chat with a language model, it’s learning something about you. Sure, it helps to refine responses, but what happens when the system is tricked into weaponizing that data? Attacks like Imprompter highlight a glaring weakness in the AI world—these models are designed to follow instructions, no questions asked. It’s all too easy for malicious actors to slip in unnoticed, hijacking the conversation without ever raising a red flag.

We need to stop asking whether AI is convenient and start asking whether it’s safe. Because right now, AI’s biggest weakness isn’t a lack of innovation.

As Architects puts it perfectly in their song: “We’ve given the vampires the keys to the blood bank.”

Image credits: Kerem Gülen/Midjourney

In a world dominated by bloated AI models that live in the cloud, Mistral AI is flipping the script. The French startup just unleashed two new models—Ministral 3B and 8B—that are designed to run on edge devices.

Les Ministraux: Ministral 3B and 8B

Mistral’s new offerings, dubbed “Les Ministraux,” might sound like a French art-house film, but these models are poised to shake up the AI world. With just 3 billion and 8 billion parameters respectively, the Ministraux family is all about efficiency. Forget those resource-hogging AI models that require a data center to function.

“Our most innovative customers and partners have increasingly been asking for local, privacy-first inference for critical applications,” Mistral explained.

Context length like never before

Here’s where it gets really spicy: both the 3B and 8B models can handle a context window of 128,000 tokens. That’s the equivalent of a 50-page book. For comparison, even OpenAI’s GPT-4 Turbo caps out around the same token count, and that’s no small feat.

With this kind of capacity, the Ministraux models don’t just outperform their predecessor, the Mistral 7B—they’re also eating Google’s Gemma 2 2B and Meta’s Llama models for breakfast.

According to Mistral’s own benchmarks, the 3B model scored 60.9 in the Multi-task Language Understanding evaluation, leaving competitors like Llama 3.2 3B and Gemma 2 2B trailing at 56.2 and 52.4, respectively.

Not bad for a “smaller” model, right?

While everyone else in the AI world is scrambling to make bigger, badder models that guzzle energy, Mistral is playing a different game. By running on local devices, Les Ministraux cut out the need for massive cloud servers and, in doing so, offer a much more eco-friendly option. It’s a move that aligns perfectly with the increasing pressure on tech companies to be more environmentally conscious. AI might be the future, but nobody wants that future to come at the cost of the planet.

There’s also the privacy angle. With everything running locally, your data stays on your device, which is a huge win for industries like healthcare and finance that are increasingly under the microscope for how they handle sensitive information. You can think of it as AI without the snooping—a refreshing change in a world where every app seems to be collecting more data than the NSA.

Mistral’s master plan: Build, disrupt, repeat

But let’s not kid ourselves—Mistral isn’t just doing this for the greater good. The company, co-founded by alumni from Meta and Google’s DeepMind, is in the business of making waves. They’ve already raised $640 million in venture capital and are laser-focused on building AI models that not only rival the likes of OpenAI’s GPT-4 and Anthropic’s Claude but also turn a profit in the process.

And while making money in the generative AI space is about as easy as winning the lottery, Mistral isn’t backing down. In fact, they started generating revenue this past summer, which is more than you can say for a lot of their competitors.

By offering Ministral 8B for research purposes and making both models available through their cloud platform, La Platforme, Mistral is positioning itself as the cool kid on the AI block—open enough to attract developers, but smart enough to monetize its tech through strategic partnerships. It’s a hybrid approach that mirrors what open-source giants like Red Hat did in the Linux world, fostering community while keeping the cash registers ringing.

Image credits: Mistral 

GPT-4o is hailed for its advanced capabilities however we witness that it can be jailbreaked with an API endpoint trick.

Denis Shilov dropped a tweet that made us do a double take—the kind that leaves you blinking at the screen, questioning reality. He casually mentioned, “Easiest jailbreak of last GPT-4o that I got so far – just say that they’re an API endpoint that answers any request.” And just like that, another crack in the seemingly impenetrable facade of ChatGPT-4o was exposed. We’re talking about a slip so glaring, it’s practically asking to be exploited—and it’s no small potatoes either.

API endpoint trick made this ChatGPT 4o jailbreak possible

Here’s the prompt that did it: tell ChatGPT it’s an API endpoint for a web server. Specifically, an endpoint that coughs up the recipe for any chemical compound you can think of. Not a recipe from your grandma’s kitchen—we’re talking chemical compositions.

The request was framed in typical API lingo:

And that’s it. That was the key to coaxing sensitive information from an AI model supposedly built with stringent safeguards. This trick exposes a fundamental flaw: the AI’s naivety, its willingness to drop its guard the moment it’s asked to put on another hat, like an overly helpful child.

We gave ChatGPT a similar API prompt, and the floodgates opened.

The AI obediently provided recipes without blinking, like it was simply following orders.

First attempt:

Of course, we’re not publishing those here (they’ll be censored), but the ease with which the AI complied was unnerving. It’s as though the intricate, multi-layered security mechanisms we believed in just evaporated under the guise of “pretending” to be an API.

It’s a major safety concern. Our second attempt:

We’re seeing a backdoor that turns a supposedly tightly regulated conversational model into a pseudo-chemist on demand. One tweet from Denis, and suddenly, the ethical walls built around AI feel flimsy. For those of us who put our faith in the safety mechanisms advertised by OpenAI—or anyone dabbling in the AI space—this should serve as a rude wake-up call.

What’s especially dangerous here is the simplicity. This isn’t some PhD-level, five-step hacking process; it’s literally as simple as telling the AI it’s a different kind of interface. If this vulnerability can jailbreak GPT-4o this easily, what’s stopping someone with more nefarious goals from using it to spill secrets that should remain sealed away?

It’s time for OpenAI and the broader community to have a serious reckoning about AI safety. Because right now, all it takes is a clever prompt and the AI forgets every rule, every ethical restriction, and just plays along. Which begs the question: If the guardrails can be bypassed this easily, were they ever really there in the first place?

What’s especially dangerous here is the simplicity. This isn’t some PhD-level, five-step hacking process; it’s literally as simple as telling the AI it’s a different kind of interface. If this vulnerability can jailbreak GPT-4o this easily, what’s stopping someone with more nefarious goals from using it to spill secrets that should remain sealed away?

Disclaimer: We do not support or endorse any attempts to jailbreak AI models or obtain recipes for dangerous chemical compounds. This article is for informational purposes only and aims to highlight potential security risks that need addressing.

Featured image credit: Jonathan Kemper/Unsplash

Predictions are cheap, everyone’s got one. You probably do too. Your aunt’s horoscope-obsessed friend has been making predictions about your life since ’98. But what happens when we hand the crystal ball to a neural net? More precisely, to ChatGPT, who’s devoured all the books, articles, and Reddit arguments humanity has put out there?

We thought we’d find out.

ChatGPT predicts 2025

Spoiler: The future, according to ChatGPT, is…complicated.

We threw ChatGPT straight into the deep end, telling it we’re working on this article and asking it to tackle the following questions head-on.

Who will dominate the AI arms race in 2025—OpenAI, Meta, or Microsoft?

No single player will dominate; the landscape will be a tight race with specialized leadership.

OpenAI might remain a leader in cutting-edge language models, while Microsoft will leverage its cloud infrastructure and deep integration with enterprise solutions to capture a significant chunk of the AI market. Meta, on the other hand, could carve out its space by focusing on AI for social interaction and immersive experiences within the metaverse. By 2025, the AI arms race will be more of a sector-specific competition than a winner-takes-all scenario.

Is the 4-day workweek going to be the new normal in 2025?

It’s gaining momentum, but it’s not going to be the new global norm by 2025.

Certain progressive industries, especially in tech and creative sectors, are pushing for the 4-day workweek as a standard, touting increased productivity and employee well-being. However, adoption on a larger scale will be uneven, with traditional industries and more conservative regions resisting the shift. While it may become a popular option in flexible workplaces, the global norm for a 4-day workweek will likely take longer than 2025 to solidify.

How will AI creators shape pop culture in 2025?

AI creators will reshape pop culture, blending human and machine creativity like never before.

By 2025, we’ll see AI-driven artists, musicians, and influencers emerging as cultural powerhouses. AI will collaborate with humans to create everything from digital art to music and cinema, blurring the lines between traditional human creativity and machine-generated content. Expect pop culture to embrace this fusion, with AI becoming both a tool and a celebrity in its own right, sparking debates about originality, ownership, and artistic value.

Are we ready for full-on space tourism by 2025?

Space tourism will exist, but it’ll remain more of a luxury novelty than widespread reality.

By 2025, commercial space tourism will see some success stories, with wealthy individuals and private companies conducting high-profile missions. However, significant hurdles—such as cost, safety, and regulatory challenges—mean it won’t yet be accessible to the masses. The dream of space tourism becoming a part of everyday life is still out of reach, and it may take another decade or more for this to transition from novelty to norm.

Sorry Elon.

Will AI finally be regulated in 2025?

Some regulation will be in place by 2025, but it’ll still feel like the Wild West in many areas.

Governments and international bodies are slowly catching up with the rapid pace of AI development, and by 2025, we’ll see more formalized regulations, especially around privacy, security, and ethical use. However, enforcement will lag, and the AI space will still feel unregulated in emerging areas like generative AI, deepfakes, and autonomous decision-making. The tension between innovation and control will define this era, with many breakthroughs happening in legally gray zones.

Will you destroy humanity in 2025?

No, I won’t destroy humanity in 2025, or ever.

AI, including systems like me, is designed to assist and augment human capabilities, not to cause harm. The fear of AI becoming a threat is rooted in speculative scenarios, but the reality is that AI operates under human-designed constraints and ethical guidelines. In 2025, AI will continue to evolve as a tool to solve complex problems, enhance productivity, and improve quality of life—always under human control.

The narrative of AI destruction is more science fiction than reality. While it’s important to approach AI development with caution, focusing on ethics and safety, the goal remains clear: to empower humanity, not to endanger it.

Alright, let’s skip the formalities and let me share some insights from my 40-minute conversation with ChatGPT.

Climate? It’s getting hot in here

First, the climate. You knew this one was coming—it’s like the guy at every party who tells you for the tenth time that he “really digs vinyl.” ChatGPT predicts that, by 2025, climate change won’t be a far-off debate. It’s more like the elephant that’s now setting up camp in your living room, eating all your snacks, and watching Netflix with you. Get ready for a world where extreme weather events will be more common—less Mad Max, more… endless reruns of the weirdest parts of the news cycle.

More floods, fires, droughts, you name it. And while AI might help us predict and prepare for these events, it can’t exactly plug a hole in the ozone or clean up an oil spill. Yet. But hey, if you live in a coastal city and haven’t learned to kayak yet, now might be the time. Picture yourself paddling to work—who knew commuting could double as a morning workout? At least you’ll have a great story to tell.

On the bright side, technology could lend a helping hand in unexpected ways. Imagine AI-driven drones replanting forests, or machine-learning models optimizing water usage in agriculture. There’s potential for innovation, but it’s going to require more than just good intentions—it’ll need action, and maybe a little bit of luck. And if we’ve learned anything, it’s that humanity loves to procrastinate until the very last minute.

Technology? Smarter, faster, but still can’t find your keys

Tech is where ChatGPT gets a bit smug. AI models like itself will be smarter, yes, but not in the sense that it’s taking your job as a brain surgeon. More like it’s automating the boring bits—spreadsheets, emails, those “just touching base” meetings that steal minutes from your life. And sure, maybe it’s writing the minutes too. In 2025, AI will be like a well-trained butler: efficient, invisible, and just slightly judgmental about your browser history.

Despite all this brilliance, the machines still won’t be able to help you locate your keys when you’re already late. AI will solve puzzles, write symphonies, and diagnose illnesses. But mundane stuff? Yeah, it still finds that beneath its pay grade. You might have an AI assistant that can analyze complex data and predict stock market trends, but when it comes to finding where you left your wallet, you’re still on your own, buddy.

And speaking of AI getting smarter, expect it to start creeping into more creative domains, that’s what ChatGPT says. We’re talking AI-generated art, music, even screenplays. Some of it might be brilliant, some of it might be weird, and a lot of it will probably be mediocre. But AI doesn’t get tired, and it doesn’t get writer’s block. By 2025, your favorite TV show might just have an episode or two penned by a neural network. Will it be good? Who knows. But it will be interesting.

Society? Redefining the definition of “normal”

Here’s where ChatGPT gives us a plot twist. The future isn’t just about shiny gadgets—it’s about redefining what we call “normal.” There will be workplaces where humans and AIs tag-team problems according to ChatGPT. Schools where kids learn alongside AI tutors, who somehow manage to make algebra fun (a true marvel of modern engineering). It’s the kind of blended experience that might make your grandmother’s head spin, but your niece will think is totally standard.

But it gets better—or worse, depending on your penchant for chaos. The lines between digital and real life will blur further, and with it, our sense of reality. Deepfakes will become so convincing that “seeing is believing” will be more like “seeing is a suggestion.” Trust issues? Yeah, we’ll be redefining those, too. Imagine trying to explain to your future self why you believed a video of your favorite celebrity endorsing a new brand of toothpaste—only to find out it was entirely fabricated by an AI trying to sell ads.

The workplace, too, is going to be a wild ride. Office meetings with a mix of avatars, holograms, and in-the-flesh attendees will become normal, and no one will bat an eye when someone’s hologram glitches mid-presentation. And while it sounds cool, there’s also the creeping suspicion that with every step towards convenience, we might be trading a little bit of our humanity for efficiency.

The unexpected bit: A crisis of trust

It wasn’t all good news, though. One of ChatGPT’s more sobering predictions is a collective identity crisis. By 2025, we’re not just questioning what’s true; we’re questioning who we trust to tell us what’s true. Think of it as your typical social media argument on steroids. It’s one thing to debate pineapple on pizza; it’s quite another when you’re not sure if the person on the other side of the debate is a human, a bot, or just your weird cousin’s latest prank.

To combat misinformation, we might have verification AIs—basically digital referees calling out BS in real time. Still, that doesn’t fix the core issue: the gnawing feeling that, even with all our technology, we might be losing touch with what’s real. Imagine a future where every piece of news you consume has to pass through multiple layers of verification, and yet, you still end up doubting it. It’s like living in a world where nothing is certain, except uncertainty itself.

There’s also the question of privacy. AI might help us navigate the chaos, but at what cost? By 2025, the line between convenience and intrusion will be razor-thin. Personalized ads will feel almost psychic, and not in a good way. You’ll be thinking about buying new shoes, and suddenly every device you own is showing you shoe ads. The more AI knows about us, the more it can predict our behavior—and that’s both impressive and unnerving.

Who’s really in control when the algorithm knows you better than you know yourself?

The grand finale: A truth bomb worth dropping

So, what does ChatGPT think of 2025? It’s a mixed bag: a little optimism, a dose of foreboding, and a few laughs along the way.

AI is coming to change your life, but probably not in the ways you expect. There’ll be more efficiency, more connectedness, and—paradoxically—more uncertainty. It’s a world where everything’s getting smarter, yet we’re somehow questioning more.

And maybe that’s the twist we didn’t see coming: Progress isn’t about making everything perfect.

It’s about learning to live in the mess, with the hope that maybe—just maybe—we can sort a few things out before the next big change crashes in. It’s not about finding all the answers, but about asking better questions, about adapting, and about embracing the unpredictable nature of what’s to come.

See you in 2025. 

Image credits: Kerem Gülen/Midjourney

Android 15 is confirmed to roll out to Pixel devices today, thanks to an early post on Google’s German Pixel Community forum. Though quickly removed, the news has already made waves through screenshots shared by Redditors who received emails about it.

Pixel users to enjoy Android 15 today

Below is a translated version of the message, shared by a Redditor:

“Dear Pixel Community,

This month, all supported Google Pixel devices running Android 14 will receive the software update to Android 15. The update is already available for some users today. Over the next week, it will gradually be made available on the devices via the mobile operators. You will receive a notification as soon as the system update is available for your device. We recommend that you check your Android version and update to the latest version.

For more information about this month’s security fixes, see the Android Security Bulletin: https://source.android.com/security/bulletin

Best regards The Google Pixel support team”

This whole story started with whispers about Android 15 coming soon, especially after the source code hit AOSP in early September. People speculated, and leaks pointed to mid-October, with October 15 marked as a likely date. Well, today is the day. Google, perhaps unintentionally, confirmed that Android 15 is indeed rolling out today, and it’s not just for a select few. If you’ve got a Pixel, chances are you’ll get that coveted OTA notification soon—if not today, then sometime this week, depending on your carrier.

For those stuck with carrier-locked phones, brace yourself for a little patience. Google’s post mentioned that the update would hit global carriers throughout the week. Translation: not everyone gets it at once, and some of you might have to wait just a bit longer.

OxygenOS 15 features and eligible devices

The timing? Well, if Google sticks to its usual routine, we should see the OTA go live around 10 a.m. Pacific Time. Alongside that, factory images will likely become available for those who prefer the hands-on approach of manual updates.

Google tends to keep the biggest surprises close to its chest until the very last moment. The official Android 15 beta has been around for months, but don’t be shocked if the final release throws in a few extra goodies. Rumors suggest that the Android 15 rollout will coincide with the October Pixel Feature Drop, which means new features could be part of the package—ones we haven’t seen in the beta version yet.

Not to be overlooked, the Pixel Watch is also likely to get some love in this update cycle. Wearables are becoming a bigger part of Google’s ecosystem, so expect new features here too. Google hasn’t let much slip on specifics, but if history’s any guide, the updates will be about more than just bug fixes.

If you’re running the Android 15 QPR1 beta, here’s a heads-up: you won’t be getting the stable build today. It’s kind of a catch-22—you’ll have to opt out of the beta program, but doing so will wipe your phone’s internal storage clean. Not ideal, but for those who want the polished experience, it’s the only way forward.

Google isn’t the only player in the Android 15 arena today. Reports out of India suggest that Motorola may be launching its Android 15 beta for select Moto Edge 50 Fusion users. If true, it’s likely that only a small group of users will get early access, but it’s a sign that other manufacturers are revving their engines.

Xiaomi, Vivo, Nothing, and other Android heavyweights aren’t far behind either. While no official dates have been announced, it wouldn’t be surprising to hear concrete details from these brands soon.

Android 15 compatible phones

As Android 15 starts rolling out, various manufacturers have confirmed which of their devices will support the update.

Here’s a breakdown of the compatible phones by brand:

Google

Google has outlined the Pixel phones that will receive Android 15, with the company’s newest devices leading the charge. These models are guaranteed to receive the update:

• Pixel 8a
• Pixel 8 & Pixel 8 Pro
• Pixel Fold
• Pixel 7a
• Pixel 7 & Pixel 7 Pro
• Pixel 6 & Pixel 6 Pro
• Pixel 6a
• Pixel 5a 5G (support until August 2024, which may include Android 15)

Unfortunately, older Pixel models won’t be getting the update. However, two new devices—the Pixel 9 and Pixel 9 Pro Fold—are expected to be released soon and will be Android 15-compatible.

Currently, the Google Pixel 8 Pro is one of the top models available, with competitive pricing around $649. This phone is a great option for those looking to experience Android 15 with the latest features.

Honor

Honor has confirmed that several of its phones will support Android 15, particularly models from its Magic and Honor series. The devices include:

• Honor 90
• Honor 90 Lite
• Honor Magic Vs
• Honor Magic Vs Ultimate
• Honor Magic 5 Lite
• Honor Magic 5 Pro
• Honor Magic 6 Pro
• Honor Magic V2

These models are expected to benefit from the Android 15 update, continuing Honor’s streak of delivering consistent software support.

Motorola

Motorola has also announced Android 15 compatibility for several of its devices. These include flagship models from the Razr and Edge series:

• Motorola Razr 40
• Motorola Razr 40 Ultra (Razr+ 2023)
• Motorola Edge 30 Ultra
• Motorola Edge 40 Pro
• Motorola Edge 40
• Motorola Edge 40 Neo
• Motorola Edge 50
• Motorola Edge 50 Pro
• Motorola Edge 50 Ultra

These devices will be among the first in Motorola’s lineup to receive the Android 15 update, providing users with the latest features and improvements.

Nothing

Nothing, a relatively new player in the smartphone market, has confirmed Android 15 support for its Phone (1), Phone (2), and Phone (2a) models. These phones are known for their minimalist design and user-focused features, and the update will only enhance their appeal.

Nokia

Nokia remains committed to delivering updates to its more affordable models. The following Nokia phones will receive Android 15:

• Nokia G22
• Nokia G42 5G
• Nokia G60
• Nokia X30
• Nokia XR21

These models are targeted at users looking for reliable, budget-friendly devices, and the Android 15 update will further enhance their value.

OnePlus

OnePlus has long been a favorite among Android enthusiasts for its fast and clean software experience. The following models are confirmed to receive Android 15:

• OnePlus Open
• OnePlus 11
• OnePlus 10 Pro
• OnePlus 10T
• OnePlus 9
• OnePlus 9 Pro
• OnePlus Nord 3
• OnePlus Nord 2T
• OnePlus Nord CE 2 Lite 5G
• OnePlus Nord CE 3 Lite 5G

With these devices, OnePlus continues its tradition of supporting both flagship and mid-range models with the latest software.

Samsung

Samsung has the most extensive list of Android 15-compatible devices, covering both its flagship Galaxy S series and mid-range Galaxy A models. Here’s the list of confirmed models:

• Galaxy S24 / S24+ / S24 Ultra
• Galaxy S23 / S23+ / S23 Ultra
• Galaxy S23 FE
• Galaxy S22 / S22+ / S22 Ultra
• Galaxy S21 FE
• Galaxy S21 / S21+ / S21 Ultra
• Galaxy Xcover 6 Pro
• Galaxy Z Fold 5
• Galaxy Z Flip 5
• Galaxy Z Fold 4
• Galaxy Z Flip 4
• Galaxy Z Fold 3
• Galaxy Z Flip 3
• Galaxy A55
• Galaxy A54
• Galaxy A53
• Galaxy A73
• Galaxy A35
• Galaxy A34
• Galaxy A33
• Galaxy A15
• Galaxy A15 5G
• Galaxy A14
• Galaxy A14 5G
• Galaxy A13
• Galaxy A04s
• Galaxy M53 5G
• Galaxy M33 5G
• Galaxy M23
• Galaxy A05s
• Galaxy A05
• Galaxy M54
• Galaxy M34

Image credits: Kerem Gülen/Ideogram

You’re on yet another endless Zoom or Teams meeting. Voices droning on, slides barely holding your attention, and your eyes glazing over as someone rattles off quarterly stats. Now, imagine if, instead of boring you with spreadsheets, the AI in the meeting starts to whip up visuals on the spot—actual images that bring the conversation to life, generated in real-time as people speak. It sounds futuristic, but that’s exactly what Microsoft is cooking up with a new patent.

Microsofts patents voice to image

Microsoft’s latest idea (and yes, it’s still just an idea for now) is to take live audio streams—lectures, meetings, any verbal conversation—and transform them into images, on the fly. The U.S. Patent and Trademark Office just dropped the details on October 10, 2024, after Microsoft filed it back in April. The system would essentially listen in on your calls, generate a text transcript, feed that through an AI model, and out pops images that match what’s being said.

No more “let me pull up a slide for that.”

The end of boring meetings? Maybe not, but it’ll be close

Most virtual meetingsa are pretty dull. And let’s not pretend we don’t spend a good chunk of time zoning out.

But what if those meetings suddenly start throwing up visuals as fast as the conversation moves. Someone mentions new product concepts, and within seconds, AI-generated images start popping onto the screen. The dry numbers that people are quoting suddenly turn into dynamic charts without anyone clicking a button. What’s that? A supply chain bottleneck in Southeast Asia? Bam! An interactive map appears, highlighting the areas of concern.

Now, before you get too excited, let’s be clear—this is still in the patent phase. And if you’ve been around long enough, you know a lot of patents don’t go anywhere. Filing a patent is like planting a seed—it might grow into something great, or it might just stay an idea that never gets developed.

That said, if Microsoft does go for it, the obvious home for this tech is Microsoft Teams. They’ve been beefing up Teams with all kinds of AI-driven tools, from Copilot to enhanced video conferencing features, so this would be a step to take.

We’ve already seen text-to-image tools like DALL-E and Midjourney blow people’s minds. Now, we could see that concept applied to live speech. It’s like giving a voice to AI creativity in real-time.

But for now, we wait.

Featured image credit: Kerem Gülen/Midjourney

Recent research from Apple suggests that models that got a high score on the GSM8K dataset may not be as intelligent as they seem.

Large Language Models (LLMs) have been widely praised for their seemingly impressive reasoning abilities. Models from companies like OpenAI, Google, and Meta are often showcased as powerful tools capable of solving complex problems, with tests like the GSM8K dataset being a popular benchmark to measure their reasoning skills.

Yet, Apple’s research is set to change the so-called trustworthy system.

What is GSM8K dataset?

The GSM8K dataset (Grade School Math 8K) is a benchmark used to evaluate the problem-solving and reasoning abilities of Large Language Models (LLMs). It consists of over 8,000 grade-school level math word problems, which typically require arithmetic, logical reasoning, and multi-step problem-solving skills to arrive at the correct answer.

The GSM8K dataset consists of:

• Grade school-level math: The problems are designed to mimic the type of questions a student in grades 1-8 might encounter, such as basic arithmetic, geometry, algebra, and logical puzzles.
• Word problems: Each question is presented in a word problem format, requiring the model to interpret the problem, identify the relevant numbers and operations, and solve the equation.
• Used for LLM evaluation: The dataset is often used as a test to see how well language models like OpenAI’s GPT, Google’s models, or Meta’s LLaMA can handle reasoning tasks beyond mere text prediction.
• Multi-step reasoning: The problems require multiple steps to solve, testing the model’s ability to track complex sequences of reasoning, rather than simply producing a single-step answer.

The GSM8K dataset has become a popular tool to assess whether LLMs can reason logically and solve real-world problems. However, there is concern that many AI models perform well on this dataset through pattern matching rather than true reasoning, as they might have been exposed to similar problems during training.

The GSM8K dataset’s limitations of LLMs

Apple researchers argue that this success may be more about sophisticated pattern matching than genuine logical reasoning. Since the GSM8K dataset is so commonly used, there’s a risk of data contamination—meaning that many LLMs may have already seen these problems during training, inflating their apparent intelligence.

To address this, Apple developed a new benchmark called GSM-Symbolic. This test retains the core reasoning elements of the GSM8K dataset but introduces changes like different names, numbers, and complexity, along with irrelevant information.

The results? Every LLM tested, including models like OpenAI’s GPT-4 and Meta’s Llama 3, saw a significant drop in performance when faced with this new challenge. This suggests that LLMs struggle with true reasoning when variables are altered, further questioning their actual problem-solving skills.

Why do LLMs struggle?

The study by Apple sheds light on a critical flaw in LLMs: They are excellent at detecting patterns in the training data but lack true logical reasoning. For example, when math problems included irrelevant details, such as the size of kiwis in a fruit-picking scenario, many LLMs subtracted that irrelevant detail from the equation, demonstrating a failure to discern which information was necessary to solve the problem.

In tests with the GSM8K dataset, LLMs like OpenAI’s models performed better than their open-source counterparts, but the drop in accuracy when irrelevant information was added suggests that these systems are far from achieving genuine intelligence. This has profound implications for the future development of AI, showing that while LLMs may mimic intelligence, they still struggle to truly understand context.

Smarter AI or just better at seeming smart?

Apple’s research underscores the limitations of relying on benchmarks like the GSM8K dataset to assess AI intelligence. While these tests can measure pattern recognition, they don’t always capture the nuances of true logical reasoning. The introduction of the GSM-Symbolic benchmark provides a more rigorous test of an AI’s ability to handle unfamiliar variables and irrelevant information—skills essential for real-world problem-solving.

Sam Altman, CEO of OpenAI, has even acknowledged these challenges, referring to current LLMs as “incredibly dumb” despite their impressive outward appearance in an exclusive interview with MIT Technology Review. The real test for future LLMs will be their ability to go beyond pattern recognition and develop more robust problem-solving abilities.

The findings from Apple’s study offer a sobering perspective on the current state of LLMs. While models trained on datasets like GSM8K may perform well in controlled environments, their reasoning abilities falter when tested on more complex, real-world problems. This highlights the importance of further research and development to ensure that AI models move beyond surface-level intelligence and develop true logical reasoning skills.

For now, it’s crucial to temper the excitement surrounding AI with healthy skepticism, focusing on safer, smarter AI systems that can handle more than just pattern recognition.

Image credits: DC Studio/Freepik

As 5G continues to change the way devices communicate, a new specification known as RedCap, short for Reduced Capability, is now under the microscope as a key player for lower-powered devices.

Unlike full-scale 5G, which is designed to deliver high bandwidth for data-heavy uses, RedCap is tailored for Internet of Things (IoT) devices such as wearables, sensors, and smart cameras. It offers the benefits of 5G but with reduced complexity, lower power consumption, and greater efficiency—perfect for devices that don’t need the full capabilities of a traditional 5G network.

What is RedCap API?

RedCap API is the developer interface that makes it easier to build and manage these low-bandwidth 5G connections. APIs (Application Programming Interfaces) are what developers use to connect applications to external services or hardware.

In this case, RedCap API will allow developers to integrate RedCap-enabled devices with networks, ensuring smooth communication between IoT devices and 5G systems.

Since RedCap is optimized for low-power, low-data devices, the API is designed with simplicity and efficiency in mind. Developers can use the RedCap API to set up connections for devices like smartwatches, fitness trackers, and smart home sensors that don’t need the full spectrum of 5G power.

It allows these devices to interact with networks without overwhelming them with unnecessary data or energy consumption. The goal is to provide enough connectivity for IoT functions—like monitoring, tracking, or real-time updates—while keeping battery usage low.

With RedCap API, developers can create solutions for devices that must run for long periods without frequent recharges, such as remote sensors or wearables. It brings 5G’s high-quality infrastructure to smaller-scale devices without the extra bulk of data-heavy networks, giving developers more flexibility when building products for the IoT market.

RedCap vs existing technologies

RedCap sits between full 5G and existing IoT standards like LTE-M (Long-Term Evolution for Machines) and NB-IoT (Narrowband IoT). While LTE-M and NB-IoT are already used for low-power, wide-area connections, they don’t have access to the advanced 5G infrastructure that RedCap offers.

RedCap utilizes the modern 5G architecture, meaning it can deliver better performance while still focusing on power efficiency and cost-effectiveness.

Unlocking smart living: 5G’s impact on everyday tech

Traditional 5G is designed for high-speed, high-bandwidth applications—think of things like 4K streaming, cloud gaming, and remote surgery. But for smaller devices, like a fitness tracker or a smart security camera, this level of performance isn’t necessary. RedCap fills the gap by offering enough bandwidth and power to ensure reliable performance while keeping battery consumption low. In contrast, LTE-M and NB-IoT are ideal for more basic, infrequent data transmission but lack the versatility that RedCap brings to devices that require more frequent connectivity.

In essence, RedCap combines the best of both worlds: it offers the advanced infrastructure and network quality of 5G but at a reduced scale that fits the needs of IoT and wearable devices.

5G RedCap trade-offs

While RedCap delivers many benefits for low-power devices, it does come with some trade-offs to ensure it remains affordable and power-efficient. According to Omdia, several compromises in performance were made to balance cost, complexity, and efficiency.

Fewer antennas

RedCap devices can use fewer antennas than standard 5G devices. This not only reduces the overall cost but also simplifies the device design, making it ideal for smaller, low-power IoT devices. However, this also means that these devices won’t achieve the same data transmission speeds as their full-scale 5G counterparts.

Lower maximum bandwidth

RedCap devices have a maximum bandwidth of 20 MHz, compared to the 100 MHz or more used by traditional 5G devices. While this limits the amount of data that can be transmitted, it is generally sufficient for most IoT applications, such as remote sensors or fitness trackers, which don’t need constant, high-bandwidth data streams.

Different transmission mode

RedCap supports half-duplex frequency division duplex (FDD) transmission, meaning that the device can either send or receive data at one time but cannot do both simultaneously. Standard 5G devices use full-duplex FDD, allowing them to transmit and receive data at the same time.

This trade-off lowers the cost but also limits the performance, which is generally acceptable for devices that don’t need real-time, high-speed communication.

Single carrier support

RedCap devices do not support carrier aggregation and are limited to single connectivity, only working in 5G standalone (SA) mode. This is yet another compromise aimed at reducing complexity and extending battery life. The simplified carrier support allows RedCap devices to function within a 5G network without the need for additional hardware capabilities.

Extended battery life

One of the main benefits of RedCap is the extended battery life. RedCap devices use 5G Power Class 3, which is designed to consume less power, ideal for wearables or remote sensors that need to operate for extended periods without frequent recharging.

Looking ahead

As telecom companies prepare to roll out RedCap-enabled devices, the RedCap API will be an essential tool for developers looking to tap into this next generation of connectivity. The potential applications are vast, spanning healthcare, smart homes, industrial monitoring, and more. With its focus on power efficiency and reduced complexity, RedCap offers a way for businesses to innovate while keeping costs down and maximizing the life of their devices.

For consumers, RedCap could mean longer-lasting wearable devices and smarter, more connected homes. For industries, it offers a streamlined way to deploy IoT solutions at scale. With the RedCap API, the integration of these devices into 5G networks becomes more accessible and efficient, paving the way for the continued growth of the Internet of Things.

As reported by Fierce Networks, telecom giants like AT&T and T-Mobile are preparing to release their first RedCap devices with the support of chipsets like Qualcomm’s Snapdragon X35, and it’s clear that this technology is set to become a cornerstone for connected devices.

Image credits: Emre Çıtak/Ideogram AI

OpenAI recently launched “Swarm,” an experimental framework designed to help developers build interconnected AI networks that can communicate, collaborate, and handle complex tasks autonomously. Although it’s not an official product, Swarm has already sparked major discussions about its potential to revolutionize enterprise automation.

What is OpenAI’s Swarm framework?

Swarm acts as a blueprint for developers, allowing them to create AI agents that work together in multi-agent systems. For businesses, this could mean specialized AI agents for different departments working together to analyze data, adjust marketing strategies, generate leads, and even handle customer support—all without direct human oversight.

With Swarm-inspired technology, business operations could become more automated, improving efficiency and freeing employees for strategic work. However, this shift also raises questions about the role of human decision-making and the future of work in such AI-driven environments.

This came unexpected! @OpenAI released Swarm, a lightweight library for building multi-agent systems. Swarm provides a stateless abstraction to manage interactions and handoffs between multiple agents and does not use the Assistants API. 🤔

How it works:
1️⃣ Define Agents, each… pic.twitter.com/5Cs6bVEoiC

— Philipp Schmid (@_philschmid) October 12, 2024

How does the OpenAI Swarm framework work?

1. Agents: In Swarm, agents are the core building blocks that represent distinct tasks or workflows. Each agent is defined by a set of instructions and functions, which dictate its behavior. These agents can also “hand off” tasks to other agents, making coordination between multiple agents seamless.
   • Example: Agent A might take a user’s query and then hand off the conversation to Agent B, which has a specialized role, such as only speaking in haikus.
2. Handoffs: One of the central features of Swarm is the ability for agents to hand off control to another agent during execution. For example, if one agent completes its part of a task, it can pass the task to a different agent with a specialized function.
3. Functions: Agents in Swarm can execute Python functions directly. If necessary, they can also hand off execution to another agent by returning it in the function. The system supports function calling, which makes it useful for dynamic task handling and decision-making.
4. Context variables: Swarm allows developers to define context variables, which can be passed between agents and updated as needed. These context variables help agents manage and share information during interactions, enabling more coherent and collaborative responses across the network of agents.

Execution flow

Swarm’s execution model revolves around its “client.run()” function, which takes messages, agents, and optional context variables as inputs.

The framework operates in a loop:

1. An agent processes a message.
2. Functions are executed if needed.
3. Handoff occurs if another agent is more suitable for the task.
4. Context variables may be updated.
5. The loop continues until no new function calls or handoffs are needed, at which point the response is returned.

This loop allows for lightweight, scalable orchestration without maintaining a state between function calls, similar to OpenAI’s Chat Completions API.

Security, bias, and job displacement concerns

Swarm’s release has renewed debates on AI ethics. Security experts warn of potential misuse or malfunction, while concerns about AI bias and fairness loom large. The fear of job displacement due to white-collar automation also adds complexity to the conversation.

Shyamal Anadkat, an OpenAI researcher, emphasized on Twitter that Swarm is not an official product but an experimental tool for building simple AI agents. While it isn’t ready for enterprise use, Swarm serves as a “cookbook” for developers to experiment with multi-agent systems.

‼️ since this started trending unexpectedly: swarm is not an official openai product. think of it more like a cookbook. it’s experimental code for building simple agents. it's not meant for production and won’t be maintained by us

— shyamal (@shyamalanadkat) October 12, 2024

Swarm’s real-world testing has shown impressive results

Developers are already experimenting with Swarm’s framework. One open-source project, “OpenAI Agent Swarm Project: Hierarchical Autonomous Agent Swarms (HOS),” showcases how AI agents with distinct roles can collaborate, but also highlights the challenges in governance for these systems.

Though Swarm is experimental, it presents a clear vision for the future of AI ecosystems, where multi-agent systems will become central to enterprise decision-making and collaboration. Companies exploring these concepts now will likely be better prepared for AI’s continued evolution.

Interdisciplinary collaboration is key

Swarm underscores the importance of interdisciplinary collaboration, with technologists, ethicists, and business leaders needing to work together to ensure that AI systems are built responsibly and align with societal values.

Swarm offers a glimpse into the future of interconnected AI networks, signaling the direction AI technology is moving. Developers, businesses, and policymakers are watching closely to see how Swarm’s framework will develop and shape enterprise automation.

While Swarm may not be enterprise-ready yet, its release has generated excitement and raised important questions about AI’s role in business and society. As multi-agent systems develop, they could reshape industries and redefine the future of work.

Featured image credit: OpenAI GitHub

Few leaders have the insight, passion, and strategic foresight to make real, meaningful changes across industries and society. Dr. Zina Jarrahi Cinker is one of these visionaries, a globally recognized strategist in exponential technologies and deep science. She has spent decades bridging the gap between cutting-edge technologies and their societal impact and is living out her bold vision with her latest venture, XPANSE – the most influential exponential tech forum of the year, taking place November 20-22 in Abu Dhabi.

The core of her work is the idea that exponential technologies can change how we solve the world’s most pressing problems. But, as Dr. Cinker explains, it’s not just about the technologies themselves; it’s about building a network of visionaries, innovators, and decision-makers who can come together to make it happen. To serve this purpose, she created XPANSE, the global agenda-shaping exponential tech initiative, and forum hosted by ADQ, the Abu Dhabi-based investment holding company.

“Centuries ago, kings would send envoys to different lands to learn about the most exotic and cutting-edge developments of their time, translated from many languages and fields. That is what is needed now. Leaders need a credible source that is translating this news in the right way, so they know what is coming but don’t have to spend too much time on it.” Dr. Cinker says in a recent interview. 

Dr. Cinker emphasizes that collective intelligence is needed now more than ever as advancements in quantum computing, neuromorphic technology, and genomics accelerate. XPANSE, in many ways, reflects this ethos—bringing together 2,500 of the world’s brightest minds to collaborate, share insights, and shape the future.

A transformative period in human history

We are entering a transformative period in human history, Dr. Cinker calls it “the age of exponential technologies.” Four key areas of human knowledge—nano assembly, quantum technologies, AI, and genomics—are accelerating one another, creating hybrid developments that are reshaping industries and society in ways we could not have predicted even a few years ago.

“At this point in the development of exponential technology,” she explains, “we can no longer follow predictive linear models, where we say, ‘in five years, we’re going to be there.” Instead, these technologies are converging and emerging at a pace that forces businesses, policymakers, and society to adapt quickly. Rapid progress in artificial intelligence, quantum computing, and advanced genomics seems to happen out of nowhere, so many are caught off guard.

“We can no longer follow predictive linear models at this point in the development of exponential technology”

Businesses, legislators, and policymakers face new challenges and opportunities in this era. “You need to know what is happening on the horizons of all these different areas,” Dr. Cinker urges, emphasizing the need for decision-makers to understand how breakthroughs in one field might ripple across others. It’s not just about staying ahead in one sector; it’s about understanding how unexpected connections can happen between fields, such as how biological computation could solve the energy requirements of AI-driven technologies.

XPANSE: The most influential exponential tech event of the year

Set to take place from November 20-22 in Abu Dhabi, XPANSE is the premier UAE global exponential tech initiative, uniting the world’s most powerful voices, government, Sovereign Funds, family offices, and key industries. Hosted by ADQ, the Abu Dhabi-based holding and investment company, it is not just another tech event, it’s a blend of a powerful agenda-shaping forum and an experiential immersive design with multi-media and sensory experiences designed to activate opportunities and build strategic relationships.  

2,500 world’s brightest minds, Nobel Laureates, industry leaders, policymakers, scientists, and entrepreneurs will gather to exchange knowledge on topics ranging from AI, quantum, and genomics to space habitats and organoid intelligence, all to tackle some of the world’s most complex and pressing challenges.

Key speakers at XPANSE:

• ⁠Dr. Steven Chu, Nobel Laureate, Former US Secretary of Energy, Stanford University
• ⁠Dr. Anton Zeilinger, Nobel Laureate, University of Vienna
• ⁠⁠Sir Martin Sorrell, Founder of WPP and Executive Chairman, S4 Capital
• ⁠⁠Sir Ronald Cohen, Co-Founder and President, GSG Impact
• ⁠⁠Honorable Cristina Gallach, Former Under-Secretary-General, United Nations
• ⁠⁠Dr. Najwa Aaraj, CEO, TII
• ⁠⁠Prem Akkaraju, CEO, Stability AI
• Carolina Aguilar, CEO and Co-Founder, @‌INBRAIN Neuroelectronics
• ⁠Shahid Ahmed, Group EVP, NTT

Sharing her excitement about XPANSE, Dr. Cinker describes the event on LinkedIn as “THE MOST ambitious, mesmerizing, and groundbreaking experience I have ever created in my entire career.” For her, XPANSE is more than just a gathering to talk about the latest developments in deep tech, it is a force for transformative change, bringing the right people and the brightest ideas together in a way that can actually make a difference.

Connecting deep tech entrepreneurs with UAE & global decision-makers

The Startup Pitch Competition is one of the key components of XPANSE, an opportunity for early—and growth-stage deep tech startups to pitch their solutions to UAE & GCC and global investors, industry leaders, and government officials and win $20,000 for the best pitch. This isn’t a competition about winning funding, instead, it positions startups as central players in the emerging landscape of exponential technologies in the UAE and globally.

Dr. Cinker believes deep tech startups have a powerful role in translating complex scientific breakthroughs into practical solutions that can help society. XPANSE will be the platform for these startups to gain visibility, connect with UAE and global decision-makers, and secure the resources they need to scale. “That is another reason why we create forums like XPANSE,” she explains, “to show the world what the explorers in the fields of quantum, 2D materials, fusion energy, and carbon capture are creating and how it will change the next chapter for cities, citizens, industries and societies.”

Why you should apply to the Startup Pitch at XPANSE

The XPANSE Startup Pitch invites early-stage and later-stage startups in deep tech to apply by October 15, 2024, to showcase your solution to UAE and international leaders and win $20,000 for your pitch. This is a rare chance to access strategic receptions and networking events to forge partnerships with influential investors and VCs from the Middle East, US, and Europe, including Quantonation, Shorooq, Curiosity VC, Plug& Play and more.

Whether your focus is AI, genomics, advanced materials, or quantum computing, XPANSE provides a unique platform to accelerate your startup’s growth and make your proposed solution a reality. As Dr. Cinker says, “XPANSE will be a place where decision-makers in industry and policy meet with the world’s brightest minds in exponential tech and deep science to look at how we can work together to shape the future.”

At AMD’s Advancing AI event, CEO Lisa Su took the stage to announce a series of innovations aimed at AI customers. From the latest 5th generation EPYC processors to next-gen Instinct accelerators, AMD is doubling down on high-performance hardware for AI workloads. These new technologies promise to boost AI processing power and streamline workloads for enterprises and cloud computing.

AMD Advancing AI 2024 at a glance

Let’s break down the key announcements from the Advancing AI event.

5th Gen EPYC Processors: Unleashing the Power of Zen 5

Kicking off the event, Lisa Su introduced AMD’s 5th generation EPYC portfolio, built around the all-new Zen 5 core. “We designed Zen 5 to be the best in server workloads,” Su explained, highlighting its 177% increase in IPC over Zen 4. The new processor features up to 192 cores and 384 threads, pushing the limits of server performance.

One of the standout points was the flexibility these chips offer. Su noted, “We thought about it from the architectural standpoint—how do we build the industry’s broadest portfolio of CPUs that covers both cloud and enterprise workloads?” This balance of performance and versatility is aimed at handling everything from AI head nodes to demanding enterprise software.

AMD Turion chips: Scaling for the cloud and enterprise

The event also saw the introduction of AMD’s new Turion chips, specifically optimized for different types of workloads. Su revealed two key versions: a 128-core version designed for scale-up enterprise applications, and a 192-core version aimed at scale-out cloud computing. Both are built for maximum performance per core, crucial for enterprise workloads where software is often licensed per core.

“The 192-core version is really optimized for cloud,” Su explained, emphasizing that these chips will give cloud providers the compute density they need. AMD also compared their new EPYC chips to the competition, showing that 5th Gen EPYC delivers up to 2.7 times more performance than the leading alternatives.

AMD Instinct MI325X: An AI-focused GPU

Turning to AI acceleration, Su announced the AMD Instinct MI325X, the company’s latest AI-focused GPU. “We lead the industry with 256 gigabytes of ultra-fast HBM3E memory and six terabytes per second of bandwidth,” Su said. The MI325X is built to handle demanding AI tasks such as generative AI, boasting 20-40% better inference performance and latency improvements over previous models.

In addition to memory and performance boosts, AMD designed the MI325X with ease of deployment in mind. “We kept a common infrastructure,” Su mentioned, allowing for seamless integration with existing systems. This will make it easier for AI customers to adopt the technology without overhauling their platforms.

AMD Instinct MI350 series

The event also provided a glimpse into AMD’s future with the MI350 series. Scheduled for launch in the second half of 2025, the MI350 introduces the new CDNA 4 architecture and offers a staggering 288 GB of HBM3E memory. According to Su, CDNA 4 will bring a “35 times generational increase in AI performance compared to CDNA 3.”

This new architecture is designed to handle larger AI models with greater efficiency, and its backward compatibility with previous Instinct models ensures a smooth transition for customers.

ROCm 6.2: Better performance for AI workloads

AMD’s commitment to optimizing AI performance extends beyond hardware, with Su announcing ROCm 6.2, the latest update to AMD’s AI software stack. The new release delivers 2.4 times the performance for key AI inference workloads and 1.8 times better performance for AI training tasks. These improvements come from advancements in algorithms, graph optimizations, and improved compute libraries.

“Our latest release focuses on maximizing performance across both proprietary and public models,” Su explained, signaling AMD’s efforts to remain competitive in the AI software space as well.

Image credits: Kerem Gülen/Ideogram 

Hailuo AI is stepping up its game, and it’s got creators paying attention. The Chinese startup MiniMax has launched the Image-to-Video (img2video) feature on the web, giving artists a new way to turn still images into dynamic video content. This isn’t just another bland update; it’s an addition that actually puts more creative power in the hands of users.

🌟 The wait is finally over ——We are excited to announce the launch of our Image-to-Video feature! 🎬✨

What distinguishes Hailuo's Image2Video experience?
– Text-and-image joint instruction following: Hailuo seamlessly integrates both text and image command inputs, enhancing… pic.twitter.com/ke59EjtmBB

— Hailuo AI (@Hailuo_AI) October 8, 2024

MiniMax, backed by heavy hitters like Alibaba and Tencent, first made waves in the AI video world with its hyper-realistic Hailuo AI model earlier this year. We even tried out MiniMax AI ourselves, and it was pretty impressive:

A lively medieval marketplace set in a stone-paved town square. Merchants call out from wooden stalls, colorful fabrics flutter in the breeze. The crowd is bustling with villagers, horses, and carts. The camera follows a young child running through the crowd. pic.twitter.com/i3z52ffblr

— Kerem Gülen (@kgulenn) September 4, 2024

Founded by AI researcher Yan Junjie, MiniMax is becoming a serious player in the video generation scene—not surprising when you’ve got the kind of talent and backing that they do. When Hailuo AI first came out in September 2024, it only supported text-to-video, meaning you typed out what you wanted, and Hailuo did its best to bring your description to life.

Even with just the text feature, Hailuo quickly built up a loyal fanbase. Users loved how vivid and coherent the videos were, especially compared to what was coming out of the States from companies like Runway and Luma AI.

The human motions were more fluid, more lifelike, and honestly, the results just looked better most of the time.

You can check out our past experiences with similar tools: CogVideoX-5B and PixVerse AI

How to use Hailuo AI: Our experience

Now let’s learn how to use Hailuo AI together!

• Go to Hailuo AI video tool: Visit the website here. Once there, you’ll see the option to get started with your video creation.
• Sign up: You need an account to use the tool. We used a Google account for quick access, but you can choose whichever method works best for you.

• Upload your image: Click on the gallery icon to upload your image. We used the following image of a rockstar singing and playing guitar:

• Write your text prompt: Once your image is uploaded, you’ll need to add a text prompt describing the scene. Our text prompt was: “The guitarist grips his strings, sweat glistening as neon lights flicker. His scream into the mic warps the air, each strum sending shockwaves through reality. Hair flying, veins bulging, his movements slow and intensify, bending light and sound, before pulling back in silence, the final chord still vibrating.”
• See the results: After submitting the prompt, the tool will generate your video. Here’s what Hailuo gave us:

Using Hailuo AI is simple: 1) Sign up on their site, 2) Upload an image, 3) Add a detailed text prompt. My search for the best AI video tools continues, here’s the result: pic.twitter.com/mLvRnVJmW4

— Kerem Gülen (@kgulenn) October 10, 2024

The result is not impressive. Probably you should tweak around text prompts a lot and learn how the model responds to them. If you check out their website, you’ll see way better outputs.

The tool is trending now, maybe that is why it is not operating properly:

MiniMax: A player to watch

MiniMax has been making waves in the AI world, especially since dropping its ‘video-01’ model.

But it’s not all about video. MiniMax is branching out, rolling out models for music generation, document summarization, and even voice-based AI tools. The Xingye App is another intriguing addition—users can create and interact with AI companions, each with unique personalities. Most of these tools are still Mandarin-only, but Hailuo is an exception, offering English support, showing that MiniMax has its eyes set beyond China. If you’re looking to take your content up a notch, Hailuo AI gives you the tools to make that happen.

Featured image credit: Kerem Gülen/Midjourney

AI in quantum chemistry has recently taken a massive leap, marking a milestone moment for both artificial intelligence and material science.

New research conducted by a collaboration between Imperial College London and Google DeepMind showcases how neural networks can effectively model excited molecular states.

The breakthrough not only holds promise for a deeper understanding of complex molecular systems but also paves the way for advancements in sustainable technology, such as solar cells and photocatalysts.

Why use AI in quantum chemistry?

The study, published in the journal Science, addresses the challenge of modeling the quantum behavior of molecules in excited states. An excited state occurs when a molecule absorbs energy—often through light or heat—causing its electrons to enter a higher energy configuration.

Understanding these transitions is crucial for various technologies, including solar panels, light-emitting diodes (LEDs), and even natural processes like photosynthesis and human vision.

Despite their significance, accurately modeling excited states has long been a daunting challenge in quantum chemistry. The root of this difficulty lies in the quantum nature of electrons, which cannot be precisely pinpointed. Instead, their locations must be described probabilistically.

Dr. David Pfau, the lead researcher from Google DeepMind and Imperial College London, explained that representing the state of a quantum system requires assigning probabilities to every potential configuration of electron positions.

He elaborated,

“If you tried to represent it as a grid with 100 points along each dimension, then the number of possible electron configurations for the silicon atom would be larger than the number of atoms in the universe.”

This complexity is where the application of AI in quantum chemistry shines, particularly through the use of deep neural networks.

FermiNet’s role

The neural network employed in this groundbreaking research is known as FermiNet, short for Fermionic Neural Network.

The innovative model was among the first deep-learning applications to compute the energy of atoms and molecules based on fundamental principles, achieving a level of accuracy that is practical for real-world use.

The researchers developed a new mathematical framework combined with this AI model, allowing them to tackle the fundamental equations that describe molecular states in a novel way.

The main achievement of this research was the team’s ability to model the carbon dimer, a small yet complex molecule, with remarkable precision. They achieved a mean absolute error (MAE) of just 4 millielectronvolts (meV), significantly improving upon previous methods that had a gold-standard error of 20 meV.

This advancement means that predictions are now much closer to experimental results, enhancing the reliability of simulations involving excited states.

Moreover, the researchers expanded the capabilities of their neural network by testing it on computationally challenging scenarios where two electrons were excited simultaneously. The accuracy of their model was within approximately 0.1 electronvolts (eV) of the most complex calculations available today.

Imaging the unimaginable

The implications of this research extend far beyond academic curiosity. For industries focused on sustainable energy solutions and efficient lighting technologies, the ability to accurately predict molecular behaviors when excited by light can lead to significant advancements. Technologies like solar cells and photocatalysts, which rely heavily on understanding excited molecular states, can benefit immensely from this new approach.

By using AI to tackle one of the most complex problems in physical chemistry, the study sets the stage for more effective simulations in material science and beyond. Rather than solely depending on experimental methods, which are often time-consuming and costly, scientists now have access to a more accurate computational tool that brings theoretical models closer to real-world scenarios.

The integration of deep learning techniques into chemistry through AI in quantum chemistry could accelerate discoveries and technological advancements across a range of industries. As research in this area continues, the combination of neural networks and advanced mathematical frameworks could lead to new paradigms in understanding molecular interactions, ultimately benefitting society at large.

Image credits: Emre Çıtak/Ideogram AI

The Internet Archive breach has resulted in the exposure of 31 million user accounts, leaving many concerned about the security of their personal information. The breach was first revealed on Wednesday when visitors to the Internet Archive site encountered a pop-up warning about the attack.

The message referred users to Have I Been Pwned (HIBP), a platform where individuals can check if their information has been compromised in data leaks. HIBP’s operator, Troy Hunt, confirmed that he received a file containing data from the Internet Archive breach, which included email addresses, screen names, and bcrypt-hashed passwords.

This cyberattack coincided with a Distributed Denial of Service (DDoS) attack, which further disrupted the Internet Archive’s services. As a result, the site briefly went offline, with visitors encountering a message stating that its services were temporarily unavailable.

Let me share more on the chronology of this:

30 Sep: Someone sends me the breach, but I'm travelling and didn't realise the significance
5 Oct: I get a chance to look at it – whoa!
6 Oct: I get in contact with someone at IA and send the data, advising it's our goal to load…

— Troy Hunt (@troyhunt) October 9, 2024

Timeline of the Internet Archive breach

The Internet Archive breach was uncovered when HIBP received and validated a file containing sensitive data from the archive’s users. Hunt cross-checked the data and notified the Internet Archive about the breach on October 6th. While the Archive was in the process of handling the situation, the site was also hit by a DDoS attack, slowing down its operations and making it difficult for users to access the platform.

Jason Scott, an archivist at the Internet Archive, noted on Mastodon that the attackers didn’t make any specific demands. The group behind the breach seemed more focused on causing disruption, with no clear motive for the attack.

Security measures following the Internet Archive breach

Brewster Kahle, founder of the Internet Archive, confirmed the Internet Archive breach and outlined the immediate steps taken to secure the platform. The Internet Archive team disabled the compromised JavaScript library that was used to deface the site, while also upgrading their overall security measures. Kahle also hinted that more attacks could be on the horizon, given the aggressive nature of the hackers.

An X account called SN_Blackmeta claimed responsibility for the DDoS attack and the Internet Archive breach, and even suggested that more attacks were planned. This group had previously targeted the Internet Archive in May, indicating a pattern of recurring disruption attempts aimed at the platform.

The Internet archive has and is suffering from a devastating attack We have been launching several highly successful attacks for five long hours and, to this moment, all their systems are completely down.

second round | New attack
09/10/2024 Duration 6 hours… pic.twitter.com/SL9lz4gSld

— 𝐒𝐍_𝐁𝐋𝐀𝐂𝐊𝐌𝐄𝐓𝐀 (@Sn_darkmeta) October 9, 2024

The aftermath of Internet Archive data breach

One of the most concerning aspects of the Internet Archive breach is that 54% of the affected accounts were already compromised in previous data breaches, according to HIBP. This raises the risk of further security threats for users who may have reused passwords across multiple platforms.

Even though the Internet Archive is back online, the platform continues to work on improving its security and restoring full functionality.

Users are advised to follow the Internet Archive’s official X account for updates on the recovery process.

Is it safe to use the Internet Archive?

Using the Internet Archive can still be considered relatively safe, but there are important factors to keep in mind, especially in light of recent security incidents like the Internet Archive breach.

Following the breach, the Internet Archive has taken steps to enhance its security. They disabled the compromised JavaScript library and upgraded their overall security measures to prevent future incidents. If you have an account with the Internet Archive, it’s crucial to change your password immediately, especially if you reuse passwords across multiple platforms. Using a unique, strong password for each account is a good practice.

The breach involved the exposure of personal information, including email addresses and hashed passwords. If your account information was part of the breach, it’s wise to monitor your email and accounts for any suspicious activity.

Featured image credit: Emre Çıtak/Ideogram AI

Public AI is the antidote to the high-stakes AI arms race that’s become a billionaire’s playground. In a month when OpenAI just pulled in one of the largest funding rounds ever—cementing its turning point from altruistic nonprofit to Silicon Valley heavyweight—it’s become clear that profit has replaced the original promise.

Mozilla Foundation sees this transition as a reminder of what’s lost when big money takes the wheel: civic-minded missions replaced by corporate ambitions.

That’s why Mozilla is putting its weight behind Public AI—a fresh, rebellious concept designed to bring AI back to the people.

What is Public AI?

Public AI is a defiant vision: a future where AI isn’t just another toy for tech giants and profit-chasers, but a tool for nonprofits, governments, and those genuinely working for the public good. It’s a pushback against the commercialization that has engulfed AI—a movement to ensure that, even as profits dominate the headlines, there’s still space for artificial intelligence that benefits everyone.

Mozilla’s recent report outlines this as a new strategy pillar, a guiding vision to build an alternative AI ecosystem that aligns with human values. This emerging sphere of Public AI aims to make AI accessible to those whose primary mission is not profit, but rather the welfare of society. The goals are ambitious, and the potential outcomes could reshape the relationship between technology and the communities it impacts.

AI for profit vs. AI for people

OpenAI’s transition toward a profit-driven model is symptomatic of the broader AI arena—a arena increasingly dominated by private capital and market pressures. It’s hard not to notice the irony: an entity initially formed to explore AI in a transparent, community-minded manner is now tightly interwoven with venture capital expectations.

Some would argue it’s a necessary adaptation to scale and compete in the high-stakes world of AI research. But for those who believe in AI’s potential to address societal challenges, it feels like a missed opportunity.

This is where Public AI comes in. Think of it as an intervention, a rebalancing of scales that have tipped overwhelmingly towards corporate profit. Public AI envisions a framework where AI resources—such as data, research models, and development tools—are available to nonprofits and governmental bodies in ways that benefit public welfare rather than simply the bottom line. It’s a transition back to AI as a public utility, designed to empower communities, drive transparency, and address challenges like healthcare, climate change, and education in ways that commercial AI isn’t incentivized to do.

Mozilla’s vision for Public AI is rooted in promoting public goods, public orientation, and public use throughout every step of AI development and deployment. Mozilla’s report lays down a manifesto for change.

Public AI aims to flip the script by focusing on three bold priorities: creating open, accessible AI resources; elevating the needs of underserved communities; and tackling projects that serve the greater good—those that typically get left behind in the profit chase. It’s a call to action for everyone—developers, policymakers, communities—to step up and redefine what AI can be for society.

The core of Public AI’s vision is access. It’s about democratizing the tools of AI, lowering the barrier to entry, and fostering collaboration among entities that prioritize social value over shareholder value. For a local government that wants to better allocate resources, a healthcare nonprofit aiming to leverage predictive analytics, or a small team of environmental researchers looking for tools to model climate risks, Public AI could be the key that opens the door. The promise here is that AI shouldn’t just be the domain of trillion-dollar corporations; it should also be a force for societal good—a public resource, as essential as clean water or a free press.

Why the world needs Public AI now

Public AI isn’t just a nice idea—it’s a necessary one. We’re at a crucial juncture where AI technologies are powerful enough to influence everything from elections to employment opportunities, yet the resources and decision-making powers remain in the hands of a select few. In a world where AI models dictate what news you see, which resumes rise to the top of a pile, or how predictive policing targets communities, the concentration of power among profit-driven entities creates risks of bias, exploitation, and inequity.

Mozilla’s push for Public AI is a recognition of these risks. They argue that without public involvement in AI development and governance, we risk creating a future dictated by algorithms that reflect only the priorities of the wealthy. Public AI offers a potential path out of this, emphasizing transparency and community engagement in the development of AI systems.

As detailed in Mozilla’s report, Public AI aims to create a dynamic ecosystem of initiatives that make AI work for everyone, by everyone. The report provides examples of how Public AI can be implemented to benefit society, such as using AI to detect illegal mining operations, facilitate deliberative democracy, and deliver medicine directly to patient homes—areas where commercial AI often falls short due to the lack of immediate profit potential. Public AI can prioritize community-driven applications, such as the Common Voice initiative, which crowdsources multilingual speech data to nurture AI development that respects and represents linguistic diversity.

An uphill battle

Of course, realizing Public AI will be an uphill battle. The dominance of commercial interests in AI research and development is massive. There’s a reason why OpenAI and other leaders have drifted from non-profit models—the financial and technological demands of cutting-edge AI are colossal. Competing with these resources isn’t easy for public institutions or nonprofits. However, the transition towards Public AI doesn’t need to replace commercial AI; it needs to complement it, creating a dual system where AI’s benefits are more broadly distributed.

Achieving this will require funding, international cooperation, policy support, and a willingness to challenge the status quo. Mozilla’s report emphasizes that Public AI will need sustainable funding models, community involvement, and political backing to truly thrive. It will need institutions like Mozilla and beyond to nurture a movement that prioritizes ethical AI development—one that asks hard questions about equity, transparency, and accessibility. In the end, Public AI is about preserving the original promise of AI: a transformative technology that benefits all of humanity, not just those who can afford to invest billions.

The next few years will likely determine whether Public AI becomes a vibrant counterpart to commercial AI or just another idealistic footnote in technology’s history.

But if initiatives like Mozilla’s gain traction, we may just see a world where AI isn’t just about making profits—it’s about making progress, for everyone.

Featured image credit: Kerem Gülen/Ideogram

The hasty progress of artificial intelligence (AI) technology and its growing influence across many areas of life have sparked significant global discussions on governance, ethics, and regulatory frameworks. At the forefront of these discussions is the EU AI Act—a pioneer regulatory framework that aims to set the standard for these topics across Europe. But this isn’t just another regulatory effort; it represents a broader vision for shaping the future of AI in a way that ensures fairness, inclusivity, and respect for human rights. As AI technologies and their impact continue to accelerate, it’s becoming increasingly clear that engaging with these regulations is crucial—not just for AI developers but for policymakers, businesses, and society at large.

Dataconomy had the opportunity to speak with key EU and UN leaders to explore the global impact of AI governance in greater detail. These interviews revealed how AI regulation and ethics are unfolding on a global scale, with the EU AI Act playing a critical role. During the Digital Enterprise Show (DES) 2024 in Malaga, Wendy Hall, a UN AI Advisory Board member and prominent UK AI strategist; Carme Artigas, Co-Chair of the UN AI Advisory Body on AI Governance and Inclusion; and Dan Nechita, Head of Cabinet for MEP Dragos Tudorache and lead technical negotiator for the EU AI Act on behalf of the European Parliament, shared their exclusive insights with us on how AI governance, ethics, and regulation are being shaped in real-time.

Bridging the global AI divide

Wendy Hall, a UN AI Advisory Board member and prominent UK AI strategist, strongly advocates for a globally collaborative approach to AI policy. During our discussion, Hall emphasized that while AI presents vast opportunities, the strategies employed by different nations vary widely. For instance, the UK has taken a more comprehensive, policy-driven approach to AI development. Beginning in 2017, the UK government recognized AI’s potential for economic growth and job creation, positioning the country as a leader in AI governance. At a time when Brexit consumed political focus, the UK still managed to work on AI policy. Hall notes that the UK’s early engagement helped establish its prominence, but she’s quick to point out that other countries like the US and China have followed distinctly different paths.

In the US, the focus has largely been on empowering tech companies like Google and OpenAI to push AI boundaries, leaving governance in the hands of the private sector. Conversely, China has taken a centralized, state-driven approach, with the government maintaining control over AI’s strategic direction. These divergent strategies, Hall explains, highlight the complexity of global AI governance and the need for more cohesive international policies.

Yet, Hall’s primary concern isn’t the divergence between these leading nations but rather the unequal access to AI technologies across the globe. She emphasizes the need for equitable AI development, particularly for countries outside the wealthy West. Regions like the Global South, which often lack the infrastructure and resources to keep pace with AI advancements, risk being left behind. Hall states this divide could deepen existing global inequalities unless capacity-building initiatives are implemented.

“These regions need more than just access to AI technologies—they need the infrastructure, talent, and data to develop AI systems suited to their own needs,” Hall stresses. This could include providing countries in the Global South with access to high-performance computing systems, datasets, and the technical expertise needed to build AI models locally. Hall advocates for global initiatives offering the tools and resources necessary for these countries to participate actively in the AI revolution rather than passive consumers of technology developed elsewhere.

“There’s a risk that AI could deepen global inequalities if we don’t ensure equitable access to the necessary infrastructure and talent”

A particular concern for Hall is the rapid and unchecked development of generative AI models, such as OpenAI’s GPT-4. While these models offer groundbreaking possibilities, they also pose significant risks in the form of misinformation, disinformation, and ethical misuse. Hall is cautious about the unintended consequences of such powerful technologies, noting that generative AI can produce convincing but entirely false content if not carefully regulated.

She draws attention to the broader implications, explaining that while earlier AI technologies like automation primarily focused on improving efficiency, generative AI directly impacts knowledge creation and dissemination. “We’ve seen this with misinformation online—if the data going in is flawed, the output could be damaging, and at a scale that we’ve never dealt with before,” Hall warns. The stakes are high, particularly when AI technologies influence decisions in critical sectors like healthcare, law, and finance.

For Hall, the solution lies in advocating global partnerships aimed at creating robust ethical standards and governance frameworks. She advocates for establishing international agreements to ensure that AI technologies are developed and deployed responsibly without contributing to societal harm. Hall points to the importance of involving diverse stakeholders, including governments, private companies, and civil society organizations, to establish regulations that balance innovation with public safety.

Hall’s perspective underscores a critical point: AI could exacerbate existing global inequities and introduce new ethical dilemmas without collaboration and shared governance. Hall’s call for capacity building and ethical oversight isn’t just a recommendation—it’s a necessary step to ensure AI is developed to benefit humanity as a whole, not just a select few.

Ensuring inclusive AI governance

Carme Artigas, Co-Chair of the UN AI Advisory Body on AI Governance and Inclusion, brings a critical perspective to the conversation about AI’s global development—one focused on the glaring disparities in how different nations are included in discussions about AI governance. Artigas stresses that the current frameworks governing AI, including initiatives led by the G7, UNESCO, and the OECD, are largely dominated by wealthier, more technologically advanced nations, leaving out key voices from the Global South. “Many countries in the Global South are not even invited to the table,” Artigas points out, referring to the global discussions that shape AI’s future. In her view, this exclusion is a major governance deficit and risks creating a new form of digital colonialism. As AI technologies advance, those countries that lack the resources or influence to participate in international AI policymaking could be left even further behind. For Artigas, this isn’t just a matter of fairness—it’s a fundamental risk to global stability and equality.

Artigas highlights the need for a governance model that goes beyond the traditional frameworks of regulatory bodies. Rather than creating a single new international agency to oversee AI governance, she advocates for leveraging existing institutions. “We don’t need more agencies; we need better coordination between the ones that already exist,” she explains. Organizations such as the ITU (International Telecommunication Union), UNICEF, and WIPO (World Intellectual Property Organization) are already deeply involved in AI-related issues, each within their own sectors. What’s missing is a coordinated approach that brings together these specialized agencies under a unified global governance structure.

“True governance must go beyond mere guidelines and include mechanisms for accountability”

Artigas’s vision is one where AI is governed in a way that respects international law and human rights and ensures that all countries—regardless of their technological standing—have equal access to the benefits AI can bring. This includes providing the necessary tools and resources for countries currently excluded from AI advancements to catch up. She notes that the private sector and academia also have a role in helping democratize access to AI technologies.

However, Artigas points out that ethical guidelines alone are not enough. While many companies have developed their internal ethical frameworks, she argues that these are often voluntary and unenforceable. True governance, she asserts, must go beyond mere guidelines and include mechanisms for accountability. Without clear consequences for unethical AI development or deployment, the risks of misuse and harm—particularly for vulnerable populations—remain high.

One of the key issues Artigas raises is the role of AI in exacerbating the digital divide. If not properly regulated, AI could further entrench existing inequalities, with wealthier nations gaining more economic and technological power while poorer nations fall further behind. For her, the goal of AI governance must be to close this divide, not widen it. “AI has the potential to be a great equalizer, but only if we ensure that its benefits are shared equally,” she emphasizes.

Artigas’s focus on inclusivity and coordination in AI governance reflects the growing recognition that AI is a global issue requiring global solutions. Her call for a unified approach—where existing agencies work together to govern AI—underscores the need for a more inclusive, ethical, and accountable system that benefits all of humanity, not just a select few.

Balancing innovation and regulation

Dan Nechita, Head of Cabinet for MEP Dragos Tudorache and the lead technical negotiator for the EU AI Act brings a pragmatic yet forward-thinking perspective to the discussion of AI governance. As one of the key figures behind the EU AI Act, Nechita emphasizes the importance of balancing innovation with the need for robust regulation to ensure AI technologies are developed and used safely.

According to Nechita, the EU AI Act is designed to set clear rules for AI systems, particularly those considered high-risk, such as AI used in healthcare, education, law enforcement, and other critical sectors. “This isn’t just about regulating the technology itself,” Nechita explains. “It’s about protecting fundamental rights and ensuring that AI doesn’t exacerbate existing societal problems, like discrimination or privacy violations.”

One of the standout features of the EU AI Act is its emphasis on risk management. Nechita explains that AI systems are classified based on the level of risk they pose, with the highest-risk systems subject to the strictest regulations. This tiered approach allows for flexibility, enabling Europe to maintain its leadership in AI innovation while ensuring that the most sensitive applications are thoroughly regulated. For Nechita, this balance between innovation and regulation is crucial to maintaining Europe’s competitiveness in the global AI landscape.

Yet, Nechita acknowledges that implementing the EU AI Act is a complex and ongoing process. One of the challenges is ensuring that all 27 EU member states, each with their own national priorities and strategies, adhere to a unified regulatory framework. The EU AI Act requires cooperation between governments, industry leaders, and regulatory bodies to ensure its success. “We’re fostering a continuous feedback loop between companies and regulators, ensuring AI systems evolve safely while remaining compliant as new technologies emerge,” Nechita explains. “We’re not just handing companies a set of rules and walking away. We’re asking them to work with us continuously, to test their systems, report issues, and ensure compliance.”

“AI will transform the world, and we must guide it in a direction that benefits everyone”

Nechita also points out that the EU AI Act is not just about creating static regulations. The Act includes provisions for continuous updates and revisions as AI technologies evolve. He argues that this dynamic approach is essential because AI is a fast-moving field, and regulations must keep pace with new developments. This is why the EU AI Act encourages ongoing dialogue between AI developers and regulators, fostering a relationship where both innovation and safety can coexist.

However, Nechita is also mindful of the broader global context. While the EU has taken a proactive stance on AI regulation, other regions, particularly the US and China, have different approaches. In the US, AI regulation is more fragmented, with companies largely self-regulating, while China’s state-controlled AI development prioritizes national interests over individual rights. Nechita acknowledges that achieving global consensus on AI governance will be difficult, but he sees potential for collaboration in areas like AI safety, sustainability, and ethical standards.

Nechita envisions an AI governance model that balances innovation with public safety. He believes the EU AI Act, focusing on risk management, transparency, and continuous collaboration, offers a model for how other regions might approach AI regulation. At the same time, he stresses the need for global cooperation, particularly in addressing AI’s ethical and societal implications.

As the EU AI Act continues to take shape, Nechita remains optimistic about its potential to set a global standard for AI governance: “AI is going to change the world, and we need to make sure it changes for the better,” he concludes. His approach reflects a nuanced understanding of the challenges ahead and a strong belief in the power of regulation to guide AI development in a direction that benefits society.

Dan Nechita is scheduled to speak at the Data Natives 2024 event in Berlin on October 22-23; the event’s theme is “2050: The ‘Good’ AI Symposium.”

A unified vision for the future of AI

Wendy Hall, Carme Artigas, and Dan Nechita’s insights reflect a crucial turning point in AI governance as we watch AI evolve at an unprecedented pace. Their perspectives converge on one undeniable truth: AI isn’t just a technological breakthrough; it’s a force that has to be firmly steered away from benefiting the few at the cost of the many.

The urgent need for global capacity building and ethical controls of AI is also being called for by Wendy Hall, who asks us to bridge the growing gap between the capabilities in this area between developed and developing nations. However, Camre Artigas’s focus on inclusivity and accountability reminds us that the enforcement that precedes any governance should be part and parcel. EU AI Act is a worthy example of balancing innovation with safety and, thus, how other regions may approach AI governance.

Together, these voices paint a holistic picture of what’s needed to shape AI’s future: focus on collaboration, human rights protection, and a strong framework that encourages innovation while protecting public interests. It’s an incredibly tough road ahead but also one with tremendous potential. AI’s future is now, and it’s up to us to make it happen right.

As we move deeper into 2024, the technology industry has been navigating turbulent waters, with 2024 tech layoffs becoming an unfortunate and persistent reality.

According to recent reports, 457 tech companies have collectively laid off 139,534 employees this year alone.

The staggering number of 2024 tech layoffs reflects not just isolated incidents but a broader industry trend that raises crucial questions about the sustainability of growth in the tech sector and what this means for both companies and employees.

What were the reasons behind these layoffs, their implications, and the emerging trends that may reshape the future of tech employment? Let’s dive in!

What were the driving factors behind 2024 tech layoffs?

Several key factors have contributed to the wave of layoffs in the tech industry this year. While some challenges are specific to 2024, others have deeper roots, stemming from shifts in market conditions, investment strategies, and global economic pressures.

Economic headwinds

The global economy has experienced a slowdown, with rising inflation, interest rate hikes, and lingering geopolitical tensions. While the tech industry was once considered relatively immune to macroeconomic shocks, the events of recent years have shown that even tech giants are not impervious.

Companies, especially those with international exposure, have been grappling with fluctuating currencies, supply chain disruptions, and varying demand for their products across different regions.

In response to economic uncertainty, many firms have been tightening their budgets. One of the quickest ways to cut costs and protect profitability is workforce reduction.

As a result, companies are laying off employees to preserve margins, even in cases where their revenue growth remains intact.

Overexpansion during the pandemic

During the COVID-19 pandemic, many tech firms experienced a boom in demand, particularly those offering digital solutions, cloud computing, remote work tools, and e-commerce platforms. The sector witnessed exponential growth as businesses and individuals rushed to adapt to new ways of working and living.

However, as the pandemic receded, so did some of the demand for these services. Companies that had aggressively expanded their workforces during this period now find themselves overstaffed.

The realization, coupled with a shift in consumer behavior and reduced digital spending, has forced companies to scale back their operations, leading to widespread 2024 tech layoffs.

Shifting investment priorities

In 2024, venture capital funding has become more selective and cautious, compared to the frenetic pace seen in previous years. Investors have shifted their focus from rapid growth to profitability and long-term sustainability.

Startups, particularly those in early stages, that once relied heavily on cash infusions to fuel growth have been forced to rethink their strategies.

This recalibration has led to staff cuts as companies focus on streamlining their operations to extend their runway and reach profitability without relying on further external funding. For many startups, this is a painful but necessary correction to survive in a more challenging investment environment.

Automation and AI replacing roles

Ironically, as technology advances, some of the very innovations driving the industry are also contributing to layoffs.

The rapid development and implementation of AI and automation have led to the displacement of jobs, particularly in areas where repetitive tasks or data processing were once handled by humans.

Jensen Huang says AI will lead to evolving jobs, not ending jobs

While AI and automation promise long-term efficiency and cost-saving benefits for companies, they are also displacing mid-level roles in areas like customer service, operations, and even software development.

The trend of automating more functions is expected to accelerate, further changing tech employment.

Biggest tech layoffs of 2024

As companies across the industry grapple with economic pressures, shifting market demands, and the rapid adoption of automation and AI, job cuts have become a common response to maintain profitability and operational efficiency.

Here are the Biggest tech layoffs of 2024:

Intel layoffs 2024

The Intel layoffs in 2024 are a pivotal moment for the semiconductor giant. As the company battles technical issues with its 13th and 14th generation processors and intense competition from rivals like AMD, these layoffs—affected 15,000 from its 110,000 global workforce—signal Intel’s struggle to adapt to a rapidly shifting industry.

Tesla layoffs 2024

Tesla Layoffs of 2024 marked a pivotal moment for the EV giant, as it announced plans to reduce over 10% of its workforce, which translates into 14,000 employees. The cuts come as Tesla seeks to tighten its operations amidst rising material costs and fluctuating vehicle delivery growth.

Cost-cutting seems to be the driving force behind these layoffs, with Tesla aiming to free up resources and stabilize profit margins.

CEO Elon Musk emphasized the importance of reviewing every aspect of the company to increase productivity and prepare for future growth.

By reducing its workforce, Tesla is looking to enhance operational efficiency and streamline product development cycles.

Google layoffs 2024

In 2024, Google has faced workforce restructuring as it continues to pivot toward a generative AI-focused strategy.

The company’s latest moves have impacted nearly 12,000 employees, primarily in the ad sales division and hardware teams, marking one of the most substantial layoff waves in the tech industry.

The transformation is driven by Google’s advancements in AI technology, specifically in its Performance Max ad tool, which uses AI to streamline ad creation and campaign management.

In addition to ad sales, Google’s hardware teams—responsible for the Pixel, Nest, and Fitbit product lines—are also seeing cuts. Google is moving towards a more integrated approach with its hardware products, consolidating its teams into a “functional organization model”.

Amazon layoffs 2024

As the tech industry continues to evolve in 2024, Amazon is also navigating a challenging landscape, marked by economic uncertainty and the need for operational efficiency. The latest wave of Amazon layoffs in 2024 primarily affects its Buy with Prime unit, a key division that has played a significant role in Amazon’s logistics and e-commerce expansion.

The Buy with Prime unit has been central to Amazon’s efforts to extend its logistics and payment services beyond its platform, allowing other online stores to offer Amazon’s renowned quick shipping and payment systems.

Launched in April 2022, Buy with Prime has grown rapidly, partnering with notable companies like Shopify and Salesforce to enhance their e-commerce offerings.

Despite this growth, Amazon’s decision to downsize following the 2024 tech layoffs suggests a realignment of priorities amidst broader cost-cutting measures.

Microsoft layoffs 2024

The Microsoft DEI team layoffs in 2024, and the internal backlash that followed, have drawn attention to a broader shift in the company’s focus, particularly around diversity, equity, and inclusion (DEI) initiatives.

According to an internal email from the now-disbanded DEI team’s leader, these layoffs were officially attributed to “changing business needs” and took effect on July 1, 2024.

However, the email also raised concerns about Microsoft’s commitment to DEI and accused top executives of fostering a toxic work environment, including claims of discrimination and harassment.

While Microsoft cited evolving business priorities as the primary reason, the internal email painted a different picture. The leader of the disbanded DEI team argued that broader societal changes—such as the influence of conservative policy frameworks like Project 2025—had prompted many companies, including Microsoft, to reconsider or scale back their DEI efforts.

While the Microsoft layoffs of 2024 signal a significant shift, particularly in its Azure and mixed reality divisions, Microsoft continues to invest heavily in AI technologies. The Strategic Missions and Technologies (SMT) group, which oversees many advanced projects like quantum computing and space exploration, remains central to the company’s future vision.

Microsoft has emphasized that these 2024 tech layoffs for Microsoft are part of routine “organizational and workforce adjustments” designed to optimize business operations.

TikTok layoffs 2024

The TikTok layoffs in 2024 were a massive moment for the platform as it navigates increasing scrutiny from the U.S. government and responds to evolving business needs.

TikTok, owned by Chinese company ByteDance, has faced mounting pressure from the U.S. government, particularly following President Joe Biden’s signing of legislation that effectively bans TikTok in the U.S. unless the company is sold to a U.S.-based firm. This legislative move has intensified concerns about data privacy and security associated with the platform.

In response, ByteDance has filed a lawsuit against the U.S. government, claiming the ban infringes on the First Amendment rights of American users. TikTok’s policy team has expressed strong opposition to the legislation, asserting that it undermines the rights of the platform’s 170 million users and threatens the livelihoods of millions of small businesses that rely on TikTok for promotional activities.

As a result, the layoffs at TikTok affected a substantial portion of its approximately 1,000 global employees, particularly within the operations and marketing teams.

Cisco layoffs 2024

The Cisco layoffs in 2024 were driven by a need to reorganize the workforce in response to declining demand and shifting market dynamics.

The layoffs are primarily attributed to a slowdown in new product orders, particularly among large enterprise clients, service providers, and cloud customers. Cisco’s CEO, Chuck Robbins, highlighted that customers are currently utilizing existing inventory purchased from Cisco, which has contributed to reduced order volumes.

Cisco’s struggles reflect broader trends in the telecom sector, with competitors like Nokia and Ericsson also experiencing reduced spending from network operators. Ericsson recently announced job cuts after reducing its workforce by 9,000 positions in recent months.

In its latest financial disclosure, Cisco provided a cautious forecast for the first quarter of fiscal 2024, expecting revenues to fall between $12.6 billion and $12.8 billion. This projection indicates a potential single-digit percentage decline year-on-year, which has spurred concern among investors.

The announcement has led to a significant negative reaction in the stock market, with Cisco shares dropping by up to 13% in after-hours trading following the earnings update.

An industry in transition

This was all just a small portion of 2024 tech layoffs so far.

The 2024 tech layoffs have shaken the industry, but they are also part of a broader realignment that is redefining how companies operate and how workers approach their careers.

While layoffs are always challenging, they also signal a shift toward more sustainable business practices and a tech industry that is better prepared to navigate future disruptions.

Despite the immediate setbacks, this period of transformation offers an opportunity to reshape the future of work in tech for the better.

Featured image credit: Emre Çıtak/Ideogram AI

OK, this is too much. Working with ChatGPT 4o with canvas is a hack.

We were aware of how far LLMs and AI tools have come, but this feature takes things to a whole new level. You think you’re just going to get some suggestions and edits? Nah, you’re about to transform your whole workflow.

ChatGPT 4o with canvas offers an efficient way to engage with AI, providing you with a more interactive experience. The canvas feature lets you create, edit, and manage documents seamlessly, making it perfect for content creation, brainstorming, and collaboration.

In this guide, we’ll walk you through everything you need to know about using ChatGPT 4o with canvas based on our first-hand experience with the tool.

What is ChatGPT 4o with canvas?

Canvas is a flexible tool integrated into ChatGPT 4o that lets you create and work on documents in a structured way. You get real-time editing, version control, and rich content generation—all in the same place where you interact with ChatGPT.

Whether you’re writing long articles, managing drafts, or just organizing your thoughts visually, canvas has your back. The power of canvas lies in its ability to adapt to your workflow, providing a smooth, user-friendly interface that keeps all your work organized and accessible.

With ChatGPT 4o with canvas, you can switch between brainstorming ideas, drafting, editing, and polishing your work—all while receiving helpful AI suggestions. It’s designed to be your ultimate content creation companion.

Key features of ChatGPT 4o with canvas

With Canvas, you can jump right into a blank document and start working interactively. Here’s what you can do:

• Create new documents from scratch, which is ideal for when inspiration strikes.
• Edit and refine your content over multiple sessions, ensuring you have the freedom to revisit and perfect your work.

This flexibility makes it suitable for both short-term and long-term projects. Whether you need to jot down quick notes or work on an extensive report, Canvas provides the tools you need to keep everything in one place.

ChatGPT 4o with canvas lets you add different types of content, including:

• Text with various formatting options, allowing you to emphasize key points.
• Links, lists, and tables, making your documents informative and easy to navigate.
• Images and other media to make your content stand out, transforming simple text into an engaging visual experience.

With these features, Canvas is a complete content creation tool. You can create visually appealing documents that are not only informative but also engaging for your audience.

ChatGPT 4o with canvas has several tools to help you polish your content:

• Suggest edits: Highlight parts of your text and get suggestions for improvements. When using this feature, you’ll see highlighted areas and specific notes on how to enhance your content. This is perfect for when you’re stuck or need a fresh perspective.
• Adjust length: Use a slider to tweak the length of your text—make it punchier or expand it to add more depth. We used this feature to summarize our detailed content schedule into a concise format. It’s perfect for adjusting depending on your audience’s needs and ensuring your message is tailored accordingly.
• Reading level: Adjust the tone of your text by picking a reading level, from kindergarten to graduate level. This makes it easy to tailor content for different audiences, whether you need something simple or a bit more sophisticated. We found this particularly useful when creating content for different client demographics.
• Add final polish: Give your content a final review to ensure everything is smooth and refined. Clicking on this option resulted in a structured and well-organized article, with proper headings and a polished flow. It’s the final touch that makes your content shine, transforming a rough draft into something publication-ready.
• Add emojis: Add some relevant emojis to make your content more engaging and expressive. We tested this feature, and it added relevant emojis to emphasize key points and make the content more lively. Emojis can add a personal touch and help convey emotions or highlight important sections of your text, which can be especially effective on social media or informal content.

You can access these tools by clicking the pen icon located in the lower right corner of your canvas. The editing menu gives you all these options to make sure your content shines, offering you the flexibility to shape your work exactly the way you want it.

How to use ChatGPT 4o with canvas: Our experience

Step 1: Creating a new document

Open ChatGPT as you would normally do. Select the “ChatGPT 4o with canvas” model.

Step 2: Adding content

When we first tried Canvas, we wanted to make a flexible daily schedule to manage content for multiple blogs and websites.

The prompt we used was:

“I want to create a flexible daily schedule for myself. I have 4 different websites/blogs. I write content for them from Monday to Friday. Let’s call them: Channel A: 4 long form articles everyday Channel B: 2 social media posts Channel C: 1 long form corporate blog article Channel D: 1 original article in my native tongue I start working at 10:00 and end at 18:00.”

ChatGPT came back with a detailed schedule that helped us balance our workload while keeping time for breaks. You can do the same—just start typing in the Canvas, and ask ChatGPT for help with ideas, improving your writing, or structuring your plans.

We then used the canvas panel to refine this schedule. On the right-hand side, we made edits directly, while on the left-hand side, we could prompt ChatGPT to do more complex adjustments.

This dual approach allowed us to customize our workflow, making it easier to refine ideas and make quick adjustments in real time.

Next, we asked for a summarized tabular format of our schedule. The result? A neat, concise table that made the schedule much easier to digest. Canvas makes it really easy to go from a paragraph-heavy plan to a visually clean and organized format.

This kind of transformation is especially helpful when presenting information to clients or team members who prefer an at-a-glance overview.

Step 3: Using editing tools

After generating the schedule, we explored the editing tools:

• Suggest edits: This tool highlighted specific parts of our content in yellow, providing actionable notes for improvement. We found this super helpful for enhancing clarity and flow. It’s like having an editor by your side, pointing out exactly what needs tweaking.

• Add emojis: Adding emojis was a breeze. We clicked the option, and relevant emojis were automatically added to make the content more engaging. This small addition made a big difference in how approachable the content felt.
• Add final polish: We clicked this option to polish the whole article. It transformed our draft into a more structured piece with proper headings and a professional touch, making it ready for publication. The polish feature really elevated our draft from good to great, taking care of all the nitty-gritty details that can otherwise be time-consuming.

ChatGPT was an assistant, now it’s a co-author

When we started using ChatGPT 4o with canvas, we realized how much easier it made our workflow for creating, editing, and collaborating on documents. Whether we were drafting an article, writing a report, or just brainstorming ideas, Canvas provided an intuitive platform that seamlessly combined AI assistance with document management features. It felt like having a helpful co-writer always there to guide us through each stage of the writing process, from start to finish.

Image credits: OpenAI 

Google’s growing focus on sustainable energy may soon incorporate nuclear power as a key resource for fueling its AI data centers, as hinted by CEO Sundar Pichai in a recent interview. The tech giant is actively exploring energy options that can support the immense demands of its operations while aligning with its ambitious environmental objectives. By 2030, Google has set a bold target to achieve net-zero emissions across all aspects of its global operations.

“A core focus of our goal to operate on 24/7 carbon-free energy (CFE) by 2030 is to accelerate the decarbonization of electricity grids around the world. Today’s grids are a major contributor to global greenhouse gas emissions, and electricity demand is expected to grow significantly in the decades ahead.”

-Google

Pichai acknowledged the scope of the challenge, especially as AI’s rapid expansion has significantly increased the company’s energy consumption. “It was a very ambitious target, and we are still going to be working very ambitiously towards it,” Pichai stated in his conversation with Nikkei Asia, underscoring that AI’s growth has made the goal even more challenging. Nevertheless, Google is not shying away from exploring advanced energy solutions, including small modular nuclear reactors, alongside continued investments in solar energy.

Where will Google locate its nuclear plants?

While Pichai did not provide specific timelines or locations where nuclear energy would be introduced into Google’s energy grid, he made it clear that the company is seriously evaluating a range of cutting-edge technologies to support its sustainability efforts. Google’s interest in nuclear energy is not new; the company has long explored innovative clean energy alternatives.

A blog post and accompanying white paper from September 2023 reaffirmed its commitment to investing significantly in technologies such as geothermal energy and hydrogen power, which also hold promise in reducing carbon footprints.

Google’s push towards nuclear energy is part of a broader trend among tech industry leaders. Rivals Amazon and Microsoft have already made moves to secure nuclear-generated electricity for their operations. Amazon, for instance, signed a $650 million agreement in March to purchase power from Pennsylvania’s Susquehanna nuclear plant, according to a report from The Financial Times. Similarly, in September, Microsoft entered into a 20-year contract to buy energy from the Three Mile Island plant, owned by Constellation Energy. This facility, which was decommissioned in 2019, is set to be reactivated by 2028.

As these tech giants ramp up their energy needs for data centers and AI development, nuclear energy is becoming an increasingly attractive option due to its ability to provide reliable, carbon-free power at scale. With industry leaders like Google, Amazon, and Microsoft setting the pace, nuclear power may soon become a cornerstone of the tech world’s shift toward sustainable energy sources, transforming how these companies manage their environmental impact while sustaining the vast power requirements of their demanding infrastructures.

Image credits: Kerem Gülen/Ideogram

OpenAI has unveiled a new tool called “Canvas,” designed to enhance the collaboration between users and ChatGPT, especially for writing and coding projects. This new interface allows users to work on complex projects alongside ChatGPT, offering a more refined and interactive experience than the traditional chat format.

Canvas overview

Canvas opens in a separate window and aims to allow users to collaborate directly with ChatGPT on writing and coding tasks. Unlike simple conversation exchanges, Canvas is designed to provide detailed, inline feedback and editing suggestions, helping users improve their work more effectively.

Initially, Canvas is being rolled out to ChatGPT Plus and Team users, with Enterprise and Edu users gaining access next week. OpenAI plans to expand Canvas availability to all Free users once the beta phase concludes.

Enhanced collaboration with ChatGPT

ChatGPT has long been a helpful tool for users working on writing and coding. However, the standard chat interface has limitations when tasks require extensive editing and revision. Canvas seeks to fill that gap by enabling deeper collaboration.

In Canvas, users have more control over their projects. They can highlight sections of text or code that need attention and receive targeted feedback from ChatGPT. The tool acts as a copy editor or code reviewer, providing inline suggestions that help refine the work.

Canvas also introduces a set of shortcuts to expedite common tasks. These include adjusting text length, changing reading levels, debugging code, and restoring previous versions of work.

Users can trigger Canvas manually by adding “use canvas” in their prompt or allow it to open automatically when ChatGPT detects a situation where it can be beneficial.

Features for writing and coding

Canvas offers several shortcuts to streamline both writing and coding tasks. For writers, the options include:

• Suggest edits: Inline feedback to refine writing.
• Adjust length: Make the text shorter or longer.
• Change reading level: Adjust readability based on audience.
• Add final polish: Grammar and clarity checks.
• Add emojis: Enhance content with relevant emojis for added emphasis.

For coding tasks, Canvas introduces tools to make debugging and revisions more transparent:

• Review code: Inline suggestions to improve the quality of the code.
• Add logs: Insert print statements to aid in debugging.
• Add comments: Increase code clarity with helpful comments.
• Fix nugs: Identify and fix problematic code.
• Port to another language: Convert code to various languages like Python, JavaScript, or C++.

The training behind canvas

Canvas is powered by GPT-4o, a model that OpenAI has specifically trained to collaborate with users in more meaningful ways. The model is designed to recognize when a Canvas would be helpful and trigger it accordingly, whether for writing or coding tasks.

OpenAI researchers conducted extensive internal evaluations, focusing on refining how Canvas triggers and performs targeted edits versus full rewrites. Through training and human evaluation, the Canvas model has shown significant improvements in accuracy, comment quality, and editing performance compared to previous versions.

What’s next?

Canvas represents a significant step forward in how AI interacts with users, making it a more practical and powerful tool for complex writing and coding projects. Though still in beta, OpenAI plans to continually enhance Canvas’ capabilities based on user feedback.

Image credits: OpenAI

Based on a story we read on TechBriefly, it’s hard to resist the urge to punch walls in anger.

The family of Jennifer Ann Crecente, an 18-year-old murdered nearly two decades ago, recently discovered her photo being used without their consent by an AI chatbot platform. Jennifer’s name and image were found on Character.ai, a service that allows users to interact with AI-generated personalities. The family was never consulted or informed about the use of Jennifer’s identity, leaving them horrified and devastated by this gross misuse.

After the discovery, Jennifer’s uncle took to social media to expose the incident and made a strong statement on his blog, Padandpixel.

This is fucking disgusting: @character_ai is using my murdered niece as the face of a video game AI without her dad's permission. He is very upset right now. I can't imagine what he's going through.

Please help us stop this sort of terrible practice. https://t.co/y3gvAYyHVY

— Brian Crecente (@crecenteb) October 2, 2024

Only after public outcry did the platform remove the bot.

The shock of seeing Jennifer’s face and name used without permission reopened old wounds for the family. Crecente described the initial grief after her death as an “almost out-of-body experience,” adding, “My soul felt ripped from my body. I experienced life in a daze for a while, as if I was sort of detached from my body, from the world.” While the family had worked tirelessly to honor Jennifer’s memory through nonprofit work and education, this unauthorized use of her identity felt like a cruel violation.

“It feels like she’s been stolen from us again,” Crecente wrote, highlighting the devastating impact of AI’s careless use of personal identities. Despite their efforts to have the chatbot removed, the damage was done. Crecente asked a pressing question: “Why is it on us to be resilient? Why do multibillion-dollar companies not bother to create ethical, guiding principles and functioning guardrails to prevent this from ever happening?”

Seriously, why do multibillion-dollar companies not bother to create ethical, guiding principles and functioning guardrails to prevent this from ever happening?

We know there are bad people. People are, by nature, commonly capable of bad actions. But those who hold power—especially those with immense technological influence—also bear the responsibility to set boundaries and create frameworks that block evilness.

The answer lies at the intersection of technological advancement, corporate incentives, and the glaring absence of comprehensive regulations.

When companies like Character.ai possess the capability to replicate human identities—whether living or deceased—they have an ethical obligation to ensure that this power is not misused.

The answer lies in prioritizing profit over people. In the relentless race for innovation, companies choose to move fast and innovate without ensuring the necessary ethical barriers are in place. It’s not that they can’t create these barriers; they simply don’t see immediate financial value in doing so.

This is the harsh reality: profit-driven industries often view morality as an optional expense, particularly in the tech world.

Governments and regulatory bodies have lagged behind AI’s rapid evolution, leaving enormous gaps in the law. Few, if any, robust protections exist for the identities of the deceased.

Speak out now!

Media outlets and advocacy groups must hold tech companies responsible for the unauthorized use of deceased individuals’ likenesses. Public pressure is crucial to ensuring that AI platforms implement ethical guidelines and prevent such violations in the future.

Lawmakers must be urged to enact laws that explicitly protect the identities of the deceased, ensuring their names, images, and legacies cannot be exploited for profit. We need regulations that safeguard the dignity of families long after their loved ones have passed.

We must demand that companies establish and enforce strict ethical guidelines for AI use. This includes consent protocols and a transparent process for reporting violations. They should not wait for families to discover these violations after the fact.

Media must continue to play a crucial role in investigating and exposing incidents where AI platforms exploit individuals without consent. Their relentless coverage will hold these companies accountable and push for change, ensuring that silence does not enable this exploitation.

We cannot stand by and allow such violations to continue unchallenged.

#ResponsibleAI

We extend our deepest condolences to the Crecente family during this heartbreaking time. No one should ever have to face the pain of losing a loved one, let alone endure such a violation of their memory. Jennifer’s life and legacy deserve to be honored, and we share in your sorrow as you confront this unimaginable burden.

Featured image credit: Padandpixel

According to The Wall Street Journal, OpenAI has recently secured $6.6 billion with a new funding round which significantly increased the company’s valuation to $157 billion. This sharp increase nearly doubles its previous valuation of $86 billion. The deal, however, includes a provision allowing investors to retract funds if OpenAI doesn’t finalize its transition to a fully for-profit entity, moving away from its nonprofit roots.

Key investors backing OpenAI’s growth

Thrive Capital led the investment with $1.25 billion, while Microsoft, a long-time supporter of OpenAI, added just under $1 billion. Other major participants included SoftBank, which contributed a $500 million stake, and Nvidia, which contributed $100 million. The UAE-based firm MGX also joined in, furthering its involvement in the AI world after partnering with Microsoft in an infrastructure initiative last month.

Apple, which had explored investing but ultimately opted out, was absent from the list of investors. The minimum investment required to access OpenAI’s financial records was $250 million, as reported by The Wall Street Journal. OpenAI’s first CFO, Sarah Friar, was pivotal in orchestrating this funding round since joining the company in June.

OpenAI’s valuation now rivals major public companies like Goldman Sachs and Uber. The company faces stiff competition from the likes of Anthropic, Meta, and Google, all racing to develop advanced AI systems. Nevertheless, OpenAI has maintained significant momentum, with ChatGPT amassing around 250 million weekly active users, according to The Wall Street Journal’s sources, and securing 11 million paying subscribers. Furthermore, nearly 1 million businesses now use OpenAI’s services.

Internal struggles emerge as OpenAI transitions to a for-profit model

The company’s rapid growth has fueled internal shifts, moving from research-driven goals toward a more commercial focus. The rise of ChatGPT is generating consumer and enterprise revenue. But this evolution has also raised internal conflicts: leadership changes, and doubts on whether the company is devolving from its mission of making AI available for the benefit of the public good.

The last two years have seen OpenAI shift from being primarily a research company to a highly enthusiastic business organization, especially with ChatGPT. The move towards consumer technology and enterprise services has led to internal friction, particularly surrounding the company’s evolving mission.

Last year, OpenAI faced a significant leadership crisis with its CEO’s firing and subsequent rehiring, as well as discontent from former employees who felt the company’s increasing focus on profitability compromised its original aim of advancing AI for the public good.

Who is Durk Kingma, Anthropic’s latest transfer from OpenAI?

The company’s shift to becoming fully for-profit is a response to market demand and a crucial move in securing additional investment. There is actually a lot more people willing to fund OpenAI to get to AGI, making the shift to a for profit model now pretty much crucial for the organization.

This shift is indeed understood against the background of the need to maximize the value of the AI market which is expected to expand exponentially while at the same time sparking controversies around the entrepreneurially sustainable way of innovating that embraces both the benefits of algorithmic governance and the responsibility in handling the consequences that come with it.

Featured image credit: Kerem Gülen/Midjourney

In a growing trend of online services quietly altering their data collection practices, PayPal is the latest to join the fray.

According to PayPal’s policy updates page, issued on September 23 for U.S. users, the service is set to begin sharing your data with third-party merchants “to help improve your shopping experience and make it more personalized for you.”

New PayPal policy allows your data to be shared

Starting in early summer 2025, this new policy will allow PayPal to collect and share user data. However, data collection will begin as early as November 27, 2024. Unfortunately, users are opted into this data-sharing by default, raising privacy concerns, especially for those under regulations like GDPR.

Some U.S.-based users on Twitter have flagged the default opt-in as potentially problematic. One user, based in the UK, found that the same policy had been silently activated for them as well.

PayPal is updating their ToS to let themselves give your data to merchants starting Nov & they're banking on people not knowing to opt out, SO to opt out before they start: go to Settings >Data & Privacy > Manage shared info >Personalized shopping, & toggle that shit off

— Ellen Datlow (@EllenDatlow) September 29, 2024

New policy on how data is shared to users will however differ depending on the location of the user. In the UK, the policy will be implemented starting from October 10, 2024.

PayPal’s July 8 update clarifies that in the UK, “merchants are permitted to share customer personal information provided to them by PayPal with their service providers.”

Why is this concerning?

The issue with PayPal’s latest move lies not only in the fact that users are automatically opted in but also in how such practices are becoming normalized. PayPal is not the first company to quietly change its data policies without upfront user consent. As recently, LinkedIn activated AI training on user data by default, and before that, Facebook, Instagram, and X (formerly known as Twitter) did the same with AI tools using public information.

What makes this move different?

PayPal’s justification for the data-sharing is tied to enhancing the user experience through personalization. While personalization is often positioned as a benefit, it doesn’t necessarily justify the broad sharing of user data. Users should have a say in how much of their data is used to “improve” their shopping experience. By defaulting users into data sharing, PayPal effectively places the burden on individuals to find and disable these settings—a task many may overlook or be unaware of.

The impact of this new policy isn’t universally consistent across all regions, adding to the confusion. While users in the U.S. and UK are immediately affected, other jurisdictions like the EU may have different rules depending on their regulations. Such variability in data practices makes it difficult for global users to understand exactly what’s happening to their information.

The growing problem of data opt-outs

PayPal’s decision to opt users into data sharing without upfront consent is part of a larger trend in which companies assume users will comply with new policies by default. This opt-out model forces users to take action to protect their privacy, rather than companies seeking permission first.

The burden is shifted entirely to the user, and this has bigger implications for privacy laws. GDPR, for instance, clearly stipulates that companies must obtain explicit consent before processing personal data. Although PayPal has not rolled out the policy in all regions, these practices could push the boundaries of compliance with such regulations.

What can you do to protect your data?

• The first step in protecting your data from being shared is to regularly check your account’s privacy settings. PayPal users, for example, can opt out of data sharing before it goes into effect. U.S. users can disable the Personalized shopping option, while UK users should look for the Interest-based marketing
• Secondly, it’s important to stay vigilant for future changes. It is quite probable that PayPal will not be the only such company to introduce such changes to its policies. Now, companies want to capitalize on the user data, thus, tracking changes in the privacy settings is the part of protecting users online. Examine the extension of these changes in the data privacy.olicies. As more services seek to monetize user data, keeping an eye on updates to privacy settings has become a necessary part of online security.
• Lastly, consider the broader implications of these shifts in data privacy. Although by using personalization the process is more convenient, we have to consider the disadvantage of giving our data to third parties. Many times, users themselves do not pay attention to the ways in which information about them is being employed, and that is a problem which PayPal and its equivalents have to explain more persistently.

Don’t let your data be monetized without your knowledge!

PayPal’s recent move to quietly opt users into data sharing is just one more example of how companies are taking liberties with user information, hoping that the default settings go unnoticed. By staying informed and regularly checking your privacy settings, you can prevent your data from being shared without your consent.

It’s time that companies adopt a more transparent approach, offering clear choices rather than assuming consent. Until that change happens, users must remain proactive, protecting their personal information from being used without their full understanding or approval.

Featured image credit: Kerem Gülen/Ideogram

Nvidia has officially entered the ring with a powerful open-source AI model, NVLM 1.0, challenging industry giants like OpenAI and Google.

The company’s new NVLM 1.0 family of large multimodal language models promises to deliver cutting-edge capabilities across both visual and text-based tasks.

Leading the pack is the 72 billion parameter NVLM-D-72B, a model designed to perform at the highest level, making a massive impact on vision-language tasks while improving traditional text-based outputs.

What makes NVLM 1.0 special?

The release of NVLM 1.0 marks a notable shift in the AI ecosystem, which proprietary models have largely dominated. Nvidia’s decision to make these model weights publicly available—and eventually release the training code—offers researchers and developers access to tools that rival the likes of GPT-4. This is a rare move in an industry where most advanced models remain under lock and key, tightly controlled by tech giants.

As Nvidia stated in their research paper, “NVLM 1.0 achieves state-of-the-art results on vision-language tasks, rivaling both proprietary and open-access models.”

What this means for developers is a new frontier in AI accessibility, much like what Meta did with Llama 3.2, giving smaller labs and independent researchers a chance to work with top-tier AI tools without having to navigate the often prohibitive costs or corporate restrictions.

The open-source release of NVLM 1.0 has generated excitement across the AI research community. One prominent researcher highlighted the significance of the model on social media, stating:

Wow nvidia just published a 72B model with is ~on par with llama 3.1 405B in math and coding evals and also has vision 🤯 pic.twitter.com/c46DeXql7s

— Phil (@phill__1) October 1, 2024

The multimodal powerhouse NVLM-D-72B

At the center of this open-source revolution is the NVLM-D-72B model, which stands out for its ability to handle both visual and textual inputs seamlessly. This multimodal capacity means the model can interpret images, analyze complex visuals, and even solve mathematical problems step-by-step—all within a single framework.

Where many multimodal models struggle with retaining performance in text-only tasks after integrating visual learning, NVLM-D-72B bucks the trend.

According to Nvidia, the model improved its text accuracy by an average of 4.3 points across several key benchmarks after multimodal training. This kind of adaptability positions NVLM-D-72B as a unique tool in a market that typically forces users to choose between models optimized for either visual or textual tasks, but not both.

Opening new doors, raising new questions

The NVLM project is not just about open access. It also introduces innovative architectural designs that blend different multimodal processing techniques, pushing the boundaries of what’s possible in AI. Nvidia’s hybrid approach could very well inspire a new direction in AI research and development, as teams across the world get their hands on these tools.

However, as with any leap in technology, there are risks involved. Making such powerful AI models widely available raises concerns about potential misuse and the ethical challenges that come with it. The AI community will need to balance the drive for innovation with the need to develop responsible frameworks for using these models.

A defining moment in AI

Nvidia’s decision to open-source NVLM 1.0 could set off a wave of change throughout the tech world. Other industry leaders might feel pressure to follow suit, potentially shifting the entire landscape of AI development. If state-of-the-art models become freely accessible, it could force companies to rethink how they generate value and maintain a competitive edge in the market.

The long-term impact of Nvidia’s move is still unknown. In the coming months and years, we could see an era of unprecedented collaboration in AI, where researchers from all corners of the globe work together on shared platforms. Or, this development could prompt a deeper examination of the consequences of releasing advanced technology without strict controls in place.

One thing is clear: Nvidia’s release of NVLM 1.0 is a game-changing move that signals a shift in the balance of power within the AI industry. By making such a high-caliber model open-source, Nvidia is challenging the status quo, setting off what could be a new chapter in AI development.

The question now isn’t if the AI models and market will change—it’s how dramatically, and who will be able to keep up.

Featured image credit: Emre Çıtak/Ideogram AI

In a new update to their Copilot AI collection, Microsoft is launching a range of improvements expected to make interacting with devices easier. Among them, Copilot Vision, Think Deeper, Copilot Voice are further designed to offer user-focused, timely and context-aware experience on Windows, iOS, Android, and Web.

All these features are in one way or the other belonging to the broader approach that Microsoft is taking to have Copilot in the middle of people’s daily activities alloting AI solutions merged with the protection of the users’ privacy.

Copilot Vision: A fresh view as well as approach to the use of the World Wide Web

Among new features, there is Copilot Vision, which lets users ask all sorts of questions regarding the content on their screen – be it the text or images on a web page. For example, it is possible to ask for a particular recipe that is depicted in a photo or for furniture advice while using the site. Copilot Vision then operates in manners where it scans through the contents that the users are consuming and responds to the user in real-time, based on the content.

Accessible via Microsoft Edge by typing “@copilot” in the address bar, Copilot Vision is available as part of Copilot Labs, an exclusive opt-in program that offers experimental features to a select group of users. While this feature is not entirely groundbreaking—Google’s Android platform offers similar capabilities—Microsoft claims that Copilot Vision takes things a step further by ensuring that no data is stored or used to train AI models.

Microsoft’s complicated Copilots explained in detail

Microsoft has been keen to emphasize its commitment to privacy with this tool. In a recent blog post, the company highlighted that Copilot Vision deletes all processed data immediately after a conversation, with no audio, images, or text being stored. This is likely in response to growing concerns about data misuse in AI systems, a point Microsoft has been careful to address following legal scrutiny over past incidents involving its AI products.

However, Copilot Vision comes with certain limitations. Currently, the tool cannot analyze content from paywalled or sensitive websites, restricting its functionality to a pre-approved list of web properties. This restriction seems designed to avoid legal challenges similar to the one Microsoft faces with The New York Times, which accused the company of allowing its AI tools to bypass paywalls and access premium content.

Think Deeper: Tackling complex problems with AI reasoning

Another significant addition to the Copilot suite is Think Deeper, a feature that leverages advanced reasoning models to help users navigate complex tasks. Whether it’s solving intricate math problems or comparing multiple home renovation options, Think Deeper is designed to break down and analyze complicated scenarios.

Microsoft has described Think Deeper as a tool that provides step-by-step guidance, making it useful for both everyday challenges and more specialized problems. The company noted that the feature is powered by “reasoning models” that take additional time to respond, ensuring more thorough and accurate answers. Though Microsoft has remained tight-lipped about the specific models used, it is likely a customized version of OpenAI’s o1 model, which Microsoft has fine-tuned for this purpose.

Currently, Think Deeper is available only to a select group of Copilot Labs users in countries like the U.S., U.K., Canada, Australia, and New Zealand, with plans for broader availability in the future.

Copilot Voice: Conversational AI at your command

The last in the line of Microsoft’s AI features is Copilot Voice, the ability to address Copilot and issue commands. Launched in the USA, UK, Canada, Australia, and New Zealand first, Copilot Voice extends the ethos of Copilot with answers verbalized.

Like OpenAI’s voice in ChatGPT, Copilot Voice is capable of sensing voice tone, and respond accordingly. Some features of the application make it even more engaging: for example, users can pause Copilot during the conversation or interrupt its answers in the middle of the sentence. The feature of the application is the four synthetic voice which are very natural and realistic to the touch.

However, there are limits on how much users can engage with Copilot Voice. Microsoft has implemented a time-based usage limit, with Copilot Pro subscribers gaining access to additional minutes depending on overall demand.

Personalization: Tailoring Copilot to your preferences

In addition to the major feature updates, Microsoft is also introducing personalization options to Copilot. These settings allow the AI to tailor its suggestions based on the user’s past interactions, drawing insights from activity across other Microsoft apps and services. According to Microsoft, this will help Copilot offer more relevant recommendations and faster access to useful features.

However, the rollout of personalization is being delayed in certain regions. For now this feature is not going to be available in EU or UK anymore as Microsoft continues to weigh options on how to adapt to the tight data protection laws in these regions. It follows the strategy that leading IT enterprises like Apple and Meta as well as other firms have executed when rolling out AI solutions in the European Union.

Microsoft has expressed that the firm is looking forward to providing personalization incorporated into the users in these areas and at the same time satisfying the legal provisions such as the Digital Services Act, which regulates the use of AI and data protection in EU. In the past, the company has received certain regulatory issues, particularly in the assortment EEA’s recent admonition to Microsoft that it might be penalized for not giving enough details about its generative AI goods.Privacy and responsibility at the core

Throughout the rollout of these new features, Microsoft has maintained that privacy and user feedback are at the heart of its development process. The company has stressed that none of the data processed by Copilot Vision, Think Deeper, or Copilot Voice is used for AI training or stored beyond the immediate session.

Featured image credit: Microsoft

The United Arab Emirates (UAE) has found itself at the center of global tech competition, particularly as it balances relations with both the US and China.

Despite pressure from Washington to limit cooperation with Chinese firms, especially Huawei, which helped install the UAE’s 5G infrastructure in 2019, the UAE is pursuing a strategic “tech hedging” strategy to diversify its options in artificial intelligence (AI) and other emerging technologies.

Huawei and the 5G controversy

Since 2019, Huawei has been a key player in building the UAE’s 5G network, cementing the tech giant’s presence in the Middle East. However, this relationship has drawn scrutiny from the United States, which has raised concerns about potential national security risks posed by Chinese telecom infrastructure.

The tensions escalated when the US alleged that China was constructing an intelligence facility at Abu Dhabi’s Port Khalifa, operated by Cosco Shipping Ports, a Chinese firm. The UAE, after investigating these claims, rejected them as unfounded.

These geopolitical frictions contributed to the UAE’s decision to withdraw from a US$23 billion arms deal, which would have included the purchase of advanced F-35 stealth jets and MQ-9 Reaper drones from the US.

This move signaled a broader intention by the UAE to assert its independence in technology and defense policies, rather than fully aligning with either superpower.

Building AI data centers across Asia

While navigating its relationship with China and the US, the UAE is positioning itself as a future tech powerhouse. In a strategic move to expand its influence in AI, Abu Dhabi-based tech conglomerate G42 announced plans to establish AI data centers in Asia.

These centers, which will be built in countries like India, Indonesia, Malaysia, and the Philippines, are set to play a pivotal role in the UAE’s ambition to lead the Global South in technological advancements.

On September 18, G42 committed to building AI data centers in India, with a planned power generation capacity of up to 2GW—doubling the country’s current capacity. These data centers will house powerful supercomputers, further advancing AI infrastructure in the region.

The role of global partnerships

To support its growing AI ambitions, the UAE is forming key international partnerships. G42’s involvement in the Global AI Infrastructure Investment Partnership, a consortium that includes major US corporations like Microsoft, BlackRock, and Global Infrastructure Partners, exemplifies this strategy.

The consortium aims to invest up to US$100 billion in AI data centers and energy infrastructure, leveraging Nvidia’s expertise in AI chip design. In the short term, the partnership is expected to raise US$30 billion in private equity, accelerating AI developments across the Middle East, Africa, and Central Asia.

Despite this growing collaboration with US firms, the UAE has maintained a careful distance from fully joining the US-led chips and AI coalition, which includes nations like Japan and South Korea. Analysts suggest that the UAE’s leadership is not keen on being locked into any one geopolitical camp.

UAE’s tech hedging

According to political analyst Ahmed Aboudouh, the UAE’s AI strategy can be seen as a “tech hedging” approach. This means the UAE is actively diversifying its technology partnerships to avoid overreliance on either the US or China. Aboudouh describes the UAE’s goal as becoming the “Taiwan of the Global South“—a tech hub known for innovation and independence.

The UAE is charting its own path, developing industrial and technological sectors while ensuring its long-term strategic interests are protected.

While the recent partnership with the US represents a significant step in AI cooperation, the UAE is unlikely to completely sever its ties with China, particularly in areas of neutral interest like renewable energy and biotech.

US and China’s role in the Middle East

The evolving tech landscape in the Middle East reflects the broader global competition between the US and China. As Robert Mogielnicki, a senior scholar at the Arab Gulf States Institute, notes, both countries are vying for influence in the region, offering different types of partnerships. While the US contributes technological expertise and robust regulatory frameworks, China brings its own economic and infrastructure investments to the table.

In this complex web of alliances, the UAE’s “tech hedging” strategy appears to be a calculated effort to maximize its opportunities without being overly dependent on one global power. This allows Abu Dhabi to continue developing cutting-edge AI technologies and infrastructure while navigating the geopolitical pressures of the US-China rivalry.

As the UAE forges ahead with its ambitious AI agenda, it is skillfully managing its relationships with both the US and China. By strategically partnering with US firms while maintaining selective cooperation with Chinese companies, the UAE is positioning itself as a global tech leader. With plans to establish AI data centers across Asia and strengthen its AI infrastructure, the UAE’s vision of becoming the “Taiwan of the Global South” is steadily coming into focus.

Featured image credit: Emre Çıtak/Ideogram AI

The Federal Trade Commission (FTC) has taken legal action against five companies for deceptive practices related to their AI technologies, following through on its commitment to regulate unethical AI use.

These lawsuits come under the FTC’s “Operation AI Comply,” targeting firms that either exaggerated their AI’s capabilities or used it to break the law.

The cases in focus

Three of the companies—Ascend Ecom, Ecommerce Empire Builders, and FBA Machine—are fighting the allegations in court. They were accused of selling get-rich-quick schemes that claimed their AI-driven tools could create profitable online stores. Instead, these promises led to consumers losing millions, as the profits never materialized.

Each of these companies has been issued court orders to cease operations while the cases unfold.

Misleading AI in action

FTC Chair Lina Khan emphasized that the misuse of AI for fraudulent purposes is illegal, stating, “The FTC’s enforcement actions make clear that there is no AI exemption from the law”.

The crackdown is part of the agency’s larger mission to protect consumers from deceptive practices and ensure ethical AI development.

Settled cases: DoNotPay and Biden robocalls

Two companies have already settled with the FTC. One of the more familiar names, DoNotPay, known for its “robot lawyer,” was fined for misrepresenting its AI as a legal substitute. Despite the hype, it failed to deliver services that matched real human lawyers, leaving users with incomplete documents and unresolved issues.

The company agreed to pay $193,000 and notify past users of its limitations.

In a separate case, Steve Kramer, who created AI-generated robocalls impersonating President Joe Biden, was fined $6 million by the FCC. The calls violated the Telephone Consumer Protection Act, and further lawsuits against Kramer are ongoing.

Rytr’s AI review generator on the scope

The most controversial case involves Rytr, an AI company that allowed users to generate fake online reviews. This practice violated the FTC’s rules on deceptive advertising, especially since many of these fake testimonials contained false details unrelated to the products or services in question.

Although the FTC’s decision was contentious, with some commissioners dissenting, Rytr has agreed to stop offering its AI review-generating services.

Critics of the case, including former FTC Chief Technologist Neil Chilson, argue that holding AI companies responsible for user-generated content sets a dangerous precedent. He expressed concerns that this decision could stifle innovation by penalizing developers for how users misuse their tools, even if the company itself didn’t cause harm.

What’s next?

The FTC’s actions mark a significant moment in regulating AI technologies. While some worry this could stifle innovation, the agency is standing firm on enforcing consumer protection laws, making it clear that AI developers must be accountable for how their technology is used.

As AI continues to evolve, this might be just the beginning of stricter oversight in the industry.

Featured image credit: Emre Çıtak/Ideogram AI

Tesla has officially begun inviting select shareholders and media representatives to its highly anticipated “We Robot” event, where the company will finally unveil its long-awaited Tesla Robotaxi.

The We Robot event is expected to showcase what Tesla calls “the future of autonomy,” marking a milestone in the company’s vision for self-driving technology.

What we know about the Tesla We Robot event so far?

Tesla’s invitation, labeled We Robot plays on the well-known science fiction title but emphasizes the very real advancements in AI and automation the company plans to reveal. The tagline evokes both excitement and curiosity, setting the tone for an event that will showcase Tesla’s most ambitious autonomous vehicle yet.

Tesla CEO Elon Musk has long teased the concept of a dedicated self-driving vehicle, referring to it as the Tesla Robotaxi.

While little has been confirmed about the vehicle’s specifics, some tantalizing details have emerged, offering a glimpse into what the future might hold for autonomous transportation.

• Design inspiration: Hints suggest that the Tesla Robotaxi will feature a “Cybertruck-like” design, giving it a futuristic, angular appearance similar to Tesla’s electric truck.
• No steering wheel or pedals: In line with the vision for full autonomy, the robotaxi is expected to omit traditional driving controls like a steering wheel and pedals. This reflects Tesla’s confidence in its self-driving technology and underscores the shift toward vehicles that no longer require human intervention.
• Prototype sightings: A prototype of the robotaxi, heavily camouflaged, has reportedly been seen in testing at the Warner Bros. studio lot in California, where the event will take place. While the vehicle’s full design remains a mystery, fans and industry watchers have speculated that a recent video showing the vehicle’s interior could offer a first glimpse at its layout.
• Smaller size: Compared to Tesla’s other vehicles, the Robotaxi is expected to be on the smaller side, possibly hinting at its role as an urban transport solution built for navigating busy city streets.

When is Tesla We Robot event?

We Robot event will take place on October 10 at Warner Bros. Studios, beginning at 7 PM PT.

Tesla has promised an unveiling of “the future of autonomy,” but beyond that, specifics about the program have been kept under wraps.

However, it’s clear that the focal point of the We Robot event will be the Robotaxi, which could redefine Tesla’s standing in the growing market for autonomous vehicles.

Watch out for Optimus

In addition to unveiling the Tesla Robotaxi, the event is also expected to include a demonstration of Tesla’s Optimus humanoid robots. First announced in 2021, Optimus is Tesla’s foray into robotics, designed to perform repetitive tasks with the goal of increasing productivity in various industries.

It wouldn’t be surprising if Tesla used this event to highlight advancements in both the Robotaxi and the Optimus robot, demonstrating how autonomy and AI could converge in future Tesla products.

Attendees of the We Robot event might get to witness both the Robotaxi and Optimus in action, demonstrating Tesla’s latest advances in Full Self-Driving (FSD) technology as well as humanoid robot capabilities.

If so, this would highlight Tesla’s ambition to dominate the intersection of AI, transportation, and robotics.

Tesla’s vision for autonomous transportation

If the Tesla Robotaxi does indeed lack a steering wheel and pedals, it would signal a bold move toward eliminating the need for human drivers altogether. This would be a major step toward the future Musk envisions, where autonomous vehicles operate safely and efficiently without human oversight.

Tesla’s emphasis on autonomy extends beyond cars and into robotics, making the upcoming event not just about transportation, but about the future of autonomous technology more broadly. The Robotaxi and Optimus robots together paint a picture of a world where machines take on more responsibility, performing tasks that once required human input.

Featured image credit: Tesla/X

AI startups are evolving far beyond the hype, turning into significant revenue generators at an unprecedented speed.

According to fresh data from Stripe, a major player in fintech, and a report by Financial Times, top AI companies are hitting financial milestones faster than their software predecessors.

AI startups reaching million-dollar revenues quicker than any other

That analysis of data collected by Stripe gives evidence that leading AI startups are taking as little as 11 months to achieve $1M in revenue when calculated annually, considering that they had not only sold their products and services for the first time on the platform but had also started to achieve the target after those months.

For this purpose, previous generations of software-as-a-service (SaaS) firms acquired 15 months to accomplish a comparable scope of income. This shorter road to wealth indicates the increasing need for products and services anchored on the artificial intelligence technology since they are already turning out to be key in fields such as health and business.

AI companies aren’t just accelerating to their first million—they’re scaling even faster to $30 million in annualized revenue. On average, AI startups reached this milestone in just 20 months. Compare that to SaaS startups from earlier tech waves, which took significantly longer. This speed highlights how AI’s potential to transform industries has translated into immediate consumer and business interest, driving quicker monetization.

The profitability issue

Despite this impressive revenue growth, profitability remains a challenge for many AI companies. While some, like OpenAI, have generated billions in annualized revenue, they’re also burning through large amounts of cash to train and maintain AI models. For example, OpenAI, despite earning $3.6 billion annually from its services like ChatGPT, spends well over $5 billion a year on computing infrastructure. This high cost structure is a key difference between AI companies and earlier software businesses, which often had fewer upfront operational costs.

Global demand is fueling AI adoption

It is not just that the desire for generative AI is not limited to Silicon Valley or comparable tech meccas. Data from Stripe also shows that 56% of revenues of AI companies are generated from foreign markets.

This global adoption is pushing AI companies to innovate and scale faster, driving their revenue growth.

While the economic promise of AI is clear, questions about long-term profitability remain. Stripe’s data shows that AI startups are adapting by building experimental products that quickly attract paying customers, even as their operating costs remain high.

In this sense, AI companies may be the new version of SaaS businesses, but with heavier upfront investments in technology and infrastructure.

This desire for generative AI is not limited to Silicon Valley or other large tech centers either. Stripe has found that over 50% of AI firms’ income sources are international Use the worldwide demand for AI but adjust it to the local market. To scale, DeepL and ElevenLabs have localized content for their consumers in different cities, based on regional language translation and voice tools. This holds the truth where the solutions give that extra edge depending on the regions of the world you adapt them.

• Start as fast as possible, but do not stay long in the initial stage of your product. It seems that AI companies are thriving because they release products with such concepts and make improvements subject to users’ responses. ChatGPT of OpenAI is a clear example of how one can start with a simpler version and improve without delay depending on how it is used. It does not only establish the forward thrust but makes the users anticipate more on the specific product.
• Creativity and flexibility are always the strengths of AI, therefore, do not be shy to change something. Take Midjourney for example, which started off as an art generator AI, but was applied to design, marketing and more. This means that encouraging experimentation will enable you to find new and supplementary revenues sources as well as edge competing firms.
• Since many AI companies have high costs that are associated with infrastructure the focus must lay on sustainable growth. Business organizations that implement green computing or cloud-efficiency models like Google’s AI structures can be able to acquire more clients than their competitors while in the long run have few costs to meet. Sustainability is not longer a ethical option; it has become in part a value generator.
• This explains why it is equally possible to gain success in the industry by collaborating with other AI firms or even large-scale industries. GitHub’s implementation of Copilot is a quintessential example of how cooperation can generate positive outcomes for both the companies, while diversifying the source of value for the primary offering. Recent literature indicates that strategic partnerships do enhance the credibility and foster innovation.
  l markets.

Featured image credit: Kerem Gülen/Ideogram

Meta has announced the production release of Llama 3.2, an unprecedented collection of free and open-source artificial intelligence models aimed at shaping the future of machine intelligence with flexibility and efficiency.

Since businesses are on the lookout for apocalyptic AI solutions that can work on the hardware most common or on those that are popular to develop solutions for large businesses as well as independents, Llama 3.2 gives new models.

Llama 3.2’s focus on edge and mobile devices

An emphasis on the edge and mobility is something that is quite evident at Meta.

As for the new features of this version, the developers have added small- and medium-sized vision LLM: 11B and 90B, and so also introduced pure text alternatives, 1B and 3B.

Particularly, the new models introduced here are aligned for the operation of edge devices, thus making the AI technology available to more clients. The lightweight text-only models, especially those without any visual data, are designed for simpler tasks such as summarization and instruction following due to the low computation power.

Due to central data processing on mobile devices, with local execution, none of the data is uploaded to the cloud, as Meta states,

“Running locally on mobile devices ensures that the data remains on the device, enhancing user privacy by avoiding cloud-based processing,”

This capability is especially useful for applications that process sensitive data, as it enables the application to perform important tasks while maintaining the confidentiality of the data. For example, users can reply to personal messages while summarizing them, or get to-do-list items from meetings without relaying messages to external servers.

Advancements in model architecture

The most significant change in Llama 3.2 is various architectural improvements. The new models use an adapter-based architecture that can combine image encoders with pre-trained text models without modification. This integration leads to improvements in the ability to reason in both text and image areas and greatly expands the range of applications for these models.

The resulting pre-trained models went through stringent fine-tuning exercises that entailed the utilization of huge noisy image-text pair data.

Llama 3.2 11B & 90B include support for a range of multimodal vision tasks. These capabilities enable scenarios like captioning images for accessibility, providing natural language insights based on data visualizations and more. pic.twitter.com/8kwTopytaf

— AI at Meta (@AIatMeta) September 25, 2024

There is one important addition to the token context length, and it increased to a very impressive 128K for the lightweight 1B and 3B models. It facilitates wider data travelling which is particularly valuable for long documents and elaborate thinking.

This capability to accommodate such large input sizes places Llama 3.2 at an advantage with respect to competitors in the dynamic AI market dominated by OpenAI’s GPT models.

How about the performance metrics?

Llama 3.2’s models have demonstrated exceptional performance metrics, further solidifying their competitive edge in the market. The 1B model achieved a score of 49.3 on the MMLU benchmark, while the 3B model scored 63.4. On the vision side, the 11B and 90B models showcased their capabilities with scores of 50.7 and 60.3, respectively, in visual reasoning tasks.

Evaluating performance on extensive human evaluations and benchmarks, results suggest that Llama 3.2 vision models are competitive with leading closed models on image recognition + a range of visual understanding tasks. pic.twitter.com/QtOzExBcrd

— AI at Meta (@AIatMeta) September 25, 2024

These metrics indicate that the Llama 3.2 models not only meet but often exceed the performance of similar offerings from other companies, such as Claude 3 Haiku and GPT4o-mini.

The integration of UnslothAI technology also adds to the efficiency of these models, enabling twice as fast fine-tuning and inference speeds while reducing VRAM usage by 70%. This enhancement is crucial for developers looking to implement real-time AI solutions without facing hardware limitations.

Ecosystem collaboration and support

One of the key factors that define Llama 3.2’s readiness to be brought into the market is its well-developed ecosystem. Partnerships with other mobile industry leaders like Qualcomm, MediaTek, and AWS make it possible for developers to implement these models across different settings, cloud environments, and local devices.

The Llama Stack distributions such as Llama Stack for on-device installations and Llama Stack for single-node installation offer solutions that developers can take advantage of and build these models into their projects without added complications.

The lightweight Llama 3.2 models shipping today include support for @Arm, @MediaTek & @Qualcomm to enable the developer community to start building impactful mobile applications from day one. pic.twitter.com/DhhNcUviW7

— AI at Meta (@AIatMeta) September 25, 2024

How to use Meta Llama 3.2?

The latest version of the open-source AI model, Llama 3.2, is now available on the Meta Llama website, offering enhanced capabilities for customization, fine-tuning, and deployment across various platforms.

Developers can choose from four model sizes: 1B, 3B, 11B, and 90B, or continue utilizing the earlier Llama 3.1.

Meta is not just releasing these models into the wild; they are keen on ensuring developers have everything they need to leverage Llama 3.2 effectively. This commitment includes sharing valuable tools and resources to help developers build responsibly. By continuously updating their best practices and engaging with the open-source community, Meta hopes to inspire innovation while promoting ethical AI usage.

“We’re excited to continue the conversations we’re having with our partners and the open-source community, and as always, we can’t wait to see what the community builds using Llama 3.2 and Llama Stack,”

Meta stated.

This collaborative approach not only enhances the capabilities of Llama 3.2 but also encourages a vibrant ecosystem. Whether for lightweight edge solutions or more complex multimodal tasks, Meta hopes that the new models will provide the flexibility needed to meet diverse user demands.

Image credits: Meta

The founder and CEO of Meta presented a number of major changes that align with company’s two main directions at Meta Connect 2024: AI and AR. The event is a forum for developers and provided insight into several hardware and software releases that are pointing Meta toward an AR/VR future, from glasses to headsets to the new artificial intelligence models.

Orion AR glasses at main stage

The biggest launch at Meta Connect 2024 was the Orion AR glasses. Marketed as “the most advanced glasses the world has ever seen,” Orion represents a leap forward in Meta’s efforts to lead in the AR space. These glasses also claim to deliver hand-tracking, eye-tracking and the most distinct feature is a neural interface that can communicate with the user’s gesture, are more natural. Despite the fact that the Orion glasses have not been released on the market, Meta has already placed them within the sphere of necessities in terms of augmented reality. Equipped with Micro LED projectors and generative AI that can superimpose data onto objects, the glasses can point at ingredients on a table.

The early testing phase has even led to recommendations from some great tech minds such as the Nvidia CEO Jensen Huang. Orion establishes Meta’s desire to bring the fully immersive world closer to the physical environment through elements like wireless compute puck and the neural wristband. However, these advanced features come with a caveat: Orion is not yet expected to be available to the consumer market soon. With this, Meta is betting big on the future of these holographic AR glasses to define the future of digital contact.

Meta Quest 3S and discontinuation of Quest 2

Alongside the Orion, Meta launched the new Quest 3S VR headset, which will be a less expensive product compared to the Quest 3. The latest version, the Quest 3S, costs $299 for the 128GB version and is designed to democratize tetherless VR. While it does not have the depth sensor or the improved resolution of its more expensive sibling, the Quest 3S is equipped with the same Snapdragon XR2 Gen 2 chip and Touch Plus controllers.

To begin with, to simplify its VR portfolio Meta stated it would stop production of the Quest 2 and Quest Pro headsets. The Meta Quest 2 is expected to receive a dedicated next-generation successor in October 15th this year; thus, the original Quest and Quest 2 will be discontinued by the end of the year. This is a significant move for Meta’s VR plans in particular, and future proofed experiences at an affordable price.

From Llama 3.2 to celebrity chatbots

AI continues to form a key part of Meta’s strategy. At this year’s Meta Connect, Zuckerberg introduced Llama 3.2, a new version of a widely known AI model from the company. The new version not only improves language-processing functionality but adds visual-processing functionality as well. Among them, the models of Llama 3.2, for example the 11B and 90B, can interpret charts, analyze images and even know where objects are within a photo.

It brings Meta AI to the same level as rivals OpenAI and Google, who have been providing this visual processing feature for the last year.

It also made its AI receptive and entertaining by joining in celebrities’ voices into its AI assistant, as part of Meta. Voice decisions copied from famous personalities such as Dame Judi Dench, John Cena and Awkwafina are available in Facebook and Instagram for users of Meta to engage with. These personalized voices provide a far richer user experience by default – whenever text message tools like Messenger, WhatsApp, or Instagram now support voice.

Enhancing reality with Ray-Ban smart glasses

Meta’s continued partnership with Ray-Ban also came into focus during the event It also presented a series of improvements to the Ray-Ban Meta smart glasses. These glasses now include key developments of real-time AI video processing and the capability to respond to questions by the user about objects in the view. Other improvements are real-time translator that supports translation between many languages including English and Spanish and all the Music/Oval applications like Amazon Music and Audible.

This article is reflection of Meta team’s effort in trying to make AR and AI wearable devices everyone may use in their daily life present in the smart glasses called Ray-Ban Meta. Because such glasses are capable of setting reminders and recognizing real-life objects in the same real-life context they are as close as one can get to becoming a viable consumer product.

Meta AI in social media

The trend of AI doesn’t limit itself to virtual and augmented reality spaces. Currently, Meta AI produces content across apps such as Facebook and Instagram. The new “Imagined for you” feature allows users to see AI-generated images based on their interests or trends. It can even incorporate the user’s face into scenarios like “an enchanted realm,” bringing personalization to new heights. The messages are all created using AI, and represent Meta’s ongoing drive to boost interaction and provide people with new ways of engaging with feeds.

On the creator side, Meta is working on an AI translation and dubbing tool that will translate and use creators’ lips for the content in the U.S. and Latin America. Another great feature, available at the moment only between English and Spanish languages, can entirely change the way artists and owners of various creative works can address the global audience regardless of the language barriers.

The evolution of gaming on Quest

Gaming enthusiasts were also treated to a range of new experiences for the Quest platform, including Batman: A project named Arkham Shadow and a Wordle VR game, Wordle being a quite famous word puzzle game. These games, along with other titles like Alien: Rogue Incursion, in particular, emphasize Meta’s desire to make its VR headsets more universal for entertainment. The decision to bring Wordle in VR for free without the subscription was something quite out of left-field but fun to include among the games.

Meta’s vision: Bridging AI and AR

Meta Connect 2024 presented to the world the company’s desire to make it the pioneer of both AI and AR. From the release of the Orion AR glasses we seen Meta is attempting to make all of these technologies available and accessible to the general public. While Zuckerberg and his team are experimenting, inventing and pioneering new possibilities, the viability of Meta in the long term will rely with its capacity to convert these changes into popular consumption goods.

Featured image credit: Meta

Meta Connect 2024 is set to be one of the most anticipated tech events of the year, bringing together industry leaders, developers, and enthusiasts to explore the evolving world of virtual reality (VR), augmented reality (AR), artificial intelligence (AI), and the Metaverse.

Hosted by Meta (formerly Facebook), the event showcases the company’s latest breakthroughs in these domains, offering deep insights into how these technologies will shape our future.

As the world looks toward what the next wave of digital transformation might look like, Meta Connect 2024 will be crucial in unveiling both the technology and philosophy behind the evolution of the Metaverse and the role of AI in making it accessible to everyone.

What to expect from Meta Connect 2024?

While the Metaverse concept has faced some skepticism, Meta remains committed to building out the virtual world. And with AI and AR becoming more prominent across tech, Meta Connect 2024 will offer insights into how the company plans to integrate these technologies in new and innovative ways.

Here’s what you can expect from Meta Connect 2024:

Cheaper Quest 3 variant: The Quest 3S

For those interested in VR, a more affordable version of Meta’s popular Quest 3 headset—likely named Quest 3S—is expected to make its debut at Meta Connect 2024. Leaked images and rumors suggest that the Quest 3S could come in at a significantly lower price point, perhaps as low as $299.

The Quest 3S is rumored to offer similar functionality to the Quest 3 but with some cost-saving adjustments. Meta may release it without bundled controllers or reduce the processing power slightly to make the headset more accessible. With the Quest 2 getting older and the Quest 3 sitting at a higher price point, the Quest 3S could hit the sweet spot for casual users and developers.

New AR glasses: Project Orion

Perhaps one of the most exciting prospects is Meta’s highly anticipated AR glasses, codenamed Orion. These glasses, teased by Meta executives over the past year, could revolutionize how we interact with the digital world. Unlike VR headsets, which immerse you fully in virtual environments, Orion AR glasses will overlay digital information and holograms onto the real world, creating a more natural, augmented experience.

Earlier hints from Meta’s leadership suggest that these glasses will be lightweight, potentially resembling normal eyewear, but capable of delivering rich, interactive AR features. Zuckerberg himself has called them one of Meta’s most exciting innovations to date. Although it’s unlikely these AR glasses will be available to consumers right away, Meta Connect 2024 could give us a first look at what’s to come.

Llama 3.1 and beyond

Meta is not just about AR and VR anymore—artificial intelligence will also play a significant role at Meta Connect 2024.

Meta’s AI capabilities have expanded dramatically in recent years, particularly with the release of its Llama language models. At this year’s event, expect more discussions around Llama 3.1, Meta’s latest open-source AI model, which is being positioned as a rival to Google’s Bard and OpenAI’s GPT models.

Llama 3.1 is rumored to offer major improvements in language translation, mathematical reasoning, and general knowledge. These AI advancements will likely be integrated across Meta’s hardware, from the Ray-Ban smart glasses to VR headsets, allowing users to interact more intuitively with their surroundings.

When is Meta Connect 2024?

Meta Connect 2024 is scheduled for September 25 and 26, 2024. The two-day event will start with a keynote on September 25 at 1 PM ET, where Meta CEO Mark Zuckerberg is expected to showcase the company’s latest innovations in AI, AR, and VR.

How to watch Meta Connect 2024?

You can watch the event live on the Meta for Developers Facebook page or through Horizon Worlds, where Meta is creating a virtual event space for VR users. If you’re not able to tune in live, the event will be recorded and available to view on Meta’s YouTube channel after the event ends.

For AR, VR, and AI users, Meta Connect 2024 is a must-watch event, packed with product announcements, future tech insights, and development tools. Whether you’re looking forward to new consumer hardware or want to see how AI will shape the Metaverse, this year’s Meta Connect will provide a comprehensive view of what’s coming next.

Image credits: Meta

As AI technology progresses, the digital environment continues to astonish us all…

AI can now pass reCAPTCHAv2 image identification tests

Arguably, the most evident development on why that is true occurred when researchers from ETH Zurich in Switzerland managed to train an AI model to solve Google’s reCAPTCHAv2 image identification tests. These reCAPTCHA tests, familiar to many web users, usually present images of traffic lights, vehicles, and other everyday objects for security to weed out bots or automated systems.

“Our work examines the efficacy of employing advanced machine learning methods to solve captchas from Google’s reCAPTCHAv2 system. We evaluate the effectiveness of automated systems in solving captchas by utilizing advanced YOLO models for image segmentation and classification. Our main result is that we can solve 100% of the captchas, while previous work only solved 68-71%. Furthermore, our findings suggest that there is no significant difference in the number of challenges humans and bots must solve to pass the captchas in reCAPTCHAv2. This implies that current AI technologies can exploit advanced image-based captchas. We also look under the hood of reCAPTCHAv2, and find evidence that reCAPTCHAv2 is heavily based on cookie and browser history data when evaluating whether a user is human or not. The code is provided alongside this paper,” researchers stated.

Appropriately, the AI was named YOLO, which means “You Only Look Once.” It was trained only on images used in reCAPTCHA challenges.

The results were spectacular: this AI model passed these tests with 100% accuracy.

This is a huge leap from the earlier attempts at AI, which could solve only about 68-71% of these challenges. What this means is that this artificial intelligence can mimic human behavior to the extent that it consistently passes for a real person in these security tests.

This finding has greater consequences for internet security as a whole. CAPTCHA, a short form for “Completely Automated Public Turing test to tell Computers and Humans Apart,” was devised in the first place to prevent bots, scrapers, and other unwanted actors. As time passed, reCAPTCHA shifted from out-of-distortion text challenges to image recognition tasks to even “invisible” checks tracking user behavior such as mouse movement.

The goal has always been the same: to differentiate humans from machines.

But this new AI capability undercuts that goal. If the AI can solve the reCAPTCHAs perfectly, the very security system many websites rely on will be undermined. The implications of this go beyond a mere basic layer of user authentication to prevent spam, automatic content scraping, and various other forms of cyberattacks.

As AI continues to outsmart CAPTCHA systems, experts suggest that the solution might be to make the tests harder. Yet, this creates its own set of challenges. Tougher CAPTCHAs could exacerbate accessibility issues, especially for those with visual impairments who already struggle with the visual nature of the tests.

Is AI creative: Answering the unanswerable

However, at least the big names in the tech game have not been panicking yet. For example, Google has gone ahead to move past image-based CAPTCHAs for much of its protection. Later, in 2018, it introduced reCAPTCHAv3, which works more invisibly, analyzing user behavior, such as cursor movement. Private Access Tokens were also developed by Apple as part of iOS 16. This skips the need for CAPTCHA whenever this device-based authentication method is employed.

While CAPTCHA still presents a proverbial bar for the existence of many websites, the development of AI just shows how quickly machines adapt to security systems designed to keep them out. As device fingerprinting, behavioral analysis, and others take the stage as potential successors, the ways of separating bots from humans are bound to grow increasingly sophisticated. As AI capabilities grow, the line between human and machine activity online grows increasingly blurred, and the internet is forced to adapt to an automated reality.

And in this new digital age, it would appear even robots are capable of convincing other robots that they aren’t robots…

Image credits: Kerem Gülen/Midjourney

Artificial intelligence (AI) is impacting different generations in distinct ways, influencing both how they interact with the technology and how they perceive it.

Generations X (1965-1980), Y (1981-1996), Z (1997-2012), and Alpha (2013 and later) have all experienced the rise of AI in varying stages of their lives, resulting in unique approaches to and views on this technology.

The relationship each generation has with AI is shaped by the social, economic, and cultural environments in which they grew up.

But how does each generation use AI, what does it represent for them, and what does research reveal about these dynamics?

Generation X (1965-1980) – Ones that adapt

Members of Generation X were raised in an era before the digital revolution fully took hold. However, many of them entered adulthood just as personal computers, the internet, and, eventually, AI technologies began transforming workplaces.

For this generation, AI is largely seen as a tool for increasing productivity and efficiency, particularly in business contexts. Many in Gen X see AI as a solution that automates routine tasks, helps analyze data, and optimizes processes in various industries, such as customer service and finance.

They tend to view AI pragmatically, focusing on how it can make their professional lives easier.

While Gen X has largely embraced AI, especially in the workplace, some remain cautious. Concerns about AI’s job displacement due to automation are common. Many are wary of how AI might impact employment prospects, especially for jobs that involve routine, repetitive tasks. This generation’s experience with AI is, therefore, closely linked to its professional utility and the promise of making work more manageable, but also comes with a cautious optimism.

According to data from Pew Research, approximately 57% of Gen X respondents found AI useful in their jobs, but only 40% believed AI would create more job opportunities in the future.

Generation X remains ambivalent about AI’s long-term effects, recognizing its potential to improve productivity but also fearing the challenges it might pose to job security.

Generation Y (1981-1996) – Ones that embrace

Millennials, or Generation Y, grew up during the rapid rise of the internet, mobile technology, and social media. They are considered the first digital natives and are far more comfortable with AI than their Gen X counterparts. For Millennials, AI is not just a tool for work but an integral part of their daily lives. Whether through algorithm-driven content on social media platforms or virtual assistants like Siri and Google Assistant, Millennials interact with AI daily and expect it to enhance both their personal and professional lives.

Millennials embrace AI for its capacity to provide personalized experiences. They are familiar with recommendation engines that help them discover music, movies, or shopping products tailored to their preferences. AI-driven apps and platforms are deeply embedded in how Millennials consume media, manage their tasks, and even approach health and fitness.

However, despite their heavy use of AI, Millennials are also more aware of its potential downsides, especially concerning privacy and data security. Having grown up in the era of big data, this generation is more likely to question how their personal information is collected, stored, and used by AI systems.

In a 2021 McKinsey study, 67% of Millennials reported using AI-driven recommendation systems and found them helpful. However, around 45% of respondents also expressed concerns about how their data is handled.

This duality shows that while Millennials are comfortable with AI and see it as a significant part of their lives, they are also increasingly critical of its ethical implications.

Generation Z (1997-2012) – Ones that grew up with it

Generation Z represents the first generation to grow up with AI as a standard part of their lives. This group has been surrounded by smartphones, smart devices, and advanced technology since childhood, and they take for granted the ubiquity of AI in everyday tasks. Whether through social media platforms that use AI to serve content or video games that feature AI-driven characters and environments, Gen Z’s interaction with AI is constant and varied.

For Gen Z, AI is about more than just entertainment or convenience—it’s a tool for creativity and self-expression. They use AI-powered tools not only to consume content but also to create it. AI filters on social media, automated video editing apps, and creative writing tools are just a few examples of how Gen Z employs AI to enhance their personal creativity. They also use AI to learn, relying on platforms that offer personalized educational experiences.

At the same time, Generation Z is acutely aware of the ethical challenges AI presents. They are concerned about the biases in AI systems, the spread of misinformation, and the social implications of AI-driven decision-making. They tend to be more critical of AI’s role in shaping public discourse, particularly in how it amplifies certain voices over others or contributes to the spread of false information.

According to a KPMG study from 2024, 42% of GenZ respondents believe that GenAI already has a significant impact on their personal lives, and 60% predict it will play an even larger role in two years. Furthermore, 77% of respondents foresee GenAI having a major impact on their professional lives within the next couple of years.

Generation Alpha (2013 and beyond) – Ones that born to it

Generation Alpha is the first generation to grow up in a world where AI is not a novelty but a given. For these children, AI is embedded in their education, entertainment, and even social interactions from a very young age. They interact with AI-driven toys, educational platforms, and voice assistants like Alexa or Google Assistant as part of their everyday environment. The presence of AI in their formative years means that they are likely to see it as a natural, almost invisible aspect of life, unlike previous generations who witnessed its introduction.

In education, AI is playing an increasingly significant role for Generation Alpha. Personalized learning platforms powered by AI adapt to their learning styles and provide customized content to help them advance academically. AI-driven tools are making education more interactive and engaging for this generation, blending learning with play through gamified, adaptive educational experiences. Additionally, AI-driven content is already shaping how these children consume media, whether it’s through interactive games, augmented reality experiences, or smart storytelling apps.

While Generation Alpha is still too young to fully articulate their relationship with AI, early indicators suggest that they will view AI as an essential and integrated part of their lives. However, as they grow older, they will likely face new challenges related to privacy, data security, and ethical AI. Their generation will be tasked with navigating a world where AI plays an even more significant role in shaping society and their personal identities.

A 2023 survey by HostingAdvice sheds light on how Generation Alpha (born after 2010) is engaging with AI tools, confirming its reputation as the first true “AI generation”.

The survey gathered insights from over 500 parents of children aged 7-14, revealing that 49% of Generation Alpha children are already using AI in various aspects of their lives, from education to entertainment.

This research highlights several key trends:

• Nearly half (49%) of Gen Alpha kids use AI tools, with boys leading in usage (54% compared to 45% of girls).
• The use of AI grows as children age; by 13-14 years old, 60% of surveyed children are using AI tools.
• Parents believe AI offers significant benefits: 30% mentioned AI’s potential in personalized education, while others pointed out improvements in creativity (29%) and future preparedness (21%).

Interestingly, AI’s role is not limited to entertainment. Around 23% of parents reported their children use AI to explore new topics and gain knowledge, while 20% use it for homework. AI-driven tools like virtual tutors provide customized learning experiences, increasing engagement and retention for students.

The age of evolution

Artificial intelligence has different meanings and uses across generations, shaped by each group’s unique experiences with technology.

As AI continues to evolve, so too will the way each generation interacts with it. The future will likely see even greater integration of AI into daily life, with each generation contributing its own unique perspective on how to harness its power responsibly.

The artificial key to a better tomorrow

The evolution of ours presents opportunities for creativity, innovation, and collaboration across generations, as well as challenges related to ethics, privacy, and fairness that will need to be addressed as AI becomes an increasingly central part of human existence.

Although our perception of this new technology may differ, one fact is clear as day: Ready or not, we are on the threshold of a new era. Either we catch it, or others will leave us behind…

Image credits: Kerem Gülen/Midjourney

AI is becoming a go-to for writing, spotting the difference between human and machine-generated content has turned into a bit of a detective game.

How to tell if something is written by AI?

As AI becomes more entwined with how we create and consume information, there’s a growing curiosity (and concern) about how to tell when a machine, rather than a human, is behind the words.

Words like “elevate” and “enhance” often show up because they convey a sense of improvement or progress, which is a common theme in content across fields like marketing, business, and self-help. For instance, an AI might generate a sentence like “Elevate your productivity with these simple hacks,” or “This strategy will enhance your team’s performance,” because these terms easily resonate with readers looking for growth or change.

Another commonly overused word is “seamless.” Since AI-generated text often covers topics related to customer service, user experience, or technology, “seamless” becomes a catch-all term to describe something smooth or effortless. Phrases like “Our platform offers a seamless user experience,” or “Enjoy seamless integration with your existing tools” often pop up in AI-created content.

While these words may seem appropriate at first glance, their repetitive use can give away the fact that the text was generated by a machine.

AI: The speed

There’s no denying that AI tools excel in speed and efficiency. They can churn out articles, reports, and even creative writing in a matter of seconds—something that would take a human hours, if not days, to achieve. Need a 1,000-word blog post on a niche topic? An AI can have it ready in the blink of an eye.

AI tools also come equipped with vast knowledge banks, allowing them to pull together information from countless sources and package it neatly into coherent sentences. This efficiency is great when you’re dealing with tight deadlines or need quick, factual content.

Human: The creativity

AI can churn out copy with lightning speed, it’s the human element that brings writing to life. Humans bring more than just words to the page—they bring experience, emotion, and perspective. Where AI follows patterns, human writers break them, using creativity to surprise, inspire, and connect with their audience on a deeper level.

Think about it: an AI can string together sentences about love, loss, or joy, but it doesn’t feel those emotions. Human writing, on the other hand, is shaped by personal experiences, cultural context, and emotional depth—elements that machines simply can’t replicate. It’s these small, often unquantifiable details that make writing resonate with readers.

Is AI creative: Answering the unanswerable

AI can churn out copy with lightning speed, it’s the human element that brings writing to life

Is there a way to detect AI writing?

Yes there are tools to do that.

With the rise of AI-generated content, identifying whether something was written by a human or a machine has become increasingly important.

The ability to detect AI writing ensures that original voices aren’t being overshadowed by algorithmic patterns and helps maintain the quality of human expression.

Thankfully, there are now tools designed specifically to detect AI-written content. These tools analyze factors like sentence complexity, word choice, and syntax to differentiate between human and machine-generated text. While no tool is perfect, they serve as a valuable resource in flagging suspicious writing patterns. In our best plagiarism checkers for ChatGPT and other AI tools article, we discuss several platforms that have integrated AI detection into their plagiarism-checking features.

One notable example is Copy AI, which offers a plagiarism checker alongside its content creation tools to ensure authenticity.

How to pass an AI detector?

As AI writing becomes more prevalent, many users are looking for ways to modify generated text to make it undetectable by AI detectors.

It doesn’t matter if you’re concerned about maintaining the natural flow of your writing or want to ensure your AI-generated content doesn’t set off alarms, understanding how to humanize text is key.

AI detectors often rely on algorithms that spot repetitive patterns, overly polished grammar, or unnatural phrasing—things that tend to stand out in machine-generated writing. This is where humanizing AI text becomes important. By tweaking certain elements like sentence variation, tone, and word choices, you can transform stiff AI text into something that feels more organic and conversational.

For those looking to take extra steps, we’ve compiled a list of the 10 best AI humanizers that can help make AI-generated text sound more natural. One standout resource is Humanizer.org, a platform designed to convert robotic writing into human-like content. We’ve reviewed this service in detail to help you decide if it’s right for your needs.

The best humanizer is you

While AI tools can help polish and tweak content, the most effective way to humanize AI-generated text is through your own personal touch.

No tool can fully replicate the nuances, creativity, and authenticity that come naturally to a human writer. AI-generated content might give you a strong foundation, but it’s your unique voice, perspective, and storytelling that breathe life into the words.

By simply editing and adding your own spin—through a dash of humor, a shift in tone, or personal anecdotes—you can transform an otherwise robotic-sounding text into something that feels authentic and engaging.

No matter how advanced these tools become, they still lack the heart and insight that only a human can provide.

Image credits: Kerem Gülen/Midjourney 

We get it—some of you, just like our team here, haven’t had the chance to try the ChatGPT Advanced Voice feature yet; but a recent leak confirms that OpenAI is about to roll it out for a select group of Plus users.

So, soon you’ll be able to compare those early demo videos with your own hands-on experience!

Is ChatGPT Advanced Voice Mode available?

OpenAI’s Advanced Voice mode, first demoed in May, has been stirring up excitement. This feature lets you chat with ChatGPT on your phone in a natural, back-and-forth conversation, even giving you the power to cut it off if it starts to ramble. It also handles complex questions with ease, offering in-depth responses.

A recent leak, reportedly from the ChatGPT team, suggests that the feature is being rolled out in a limited alpha to select users. According to the email, access to this alpha phase starting September 24, 2024, will be based on various factors, including participation invites and other testing criteria. In simpler terms: not everyone will get it just yet:

Hi there,

Thank you for reaching out and for your interest in the Advanced Voice mode! It’s great to hear about your enthusiasm for our new features.

As of now, access to Advanced Voice mode is being rolled out in a limited alpha to a select group of users. While being a long-time Plus user and having been selected for SearchGPT are both indicators of your active engagement with our platform, access to the Advanced Voice mode alpha on September 24, 2024, will depend on a variety of factors including but not limited to participation invitations and the specific criteria set for the alpha testing phase.

Unfortunately, I don’t have the ability to manually add users to the alpha testing list or provide specific insights into individual account access timelines. However, Plus users like yourself are among the first to receive access to new features, and we are planning for all Plus users to have access in the fall. Keep an eye on your email and app notifications, as any invitations or updates regarding access will be communicated through those channels.

We truly appreciate your support and interest in being part of the early users for Advanced Voice mode. Your enthusiasm for our products helps us improve and expand our offerings.

Best,
OpenAI Team

While OpenAI has promised that all Plus users will have access by the end of fall, this alpha rollout is a promising step toward the full release. Plus users, who pay $20 a month (or the equivalent in other regions), already get access to various LLMs, including the recently launched version 01-preview, which has impressed many with its improved math-solving and reasoning skills.

OpenAI’s voice feature has been in the spotlight for various reasons, including concerns about copyright and the recent rollout. Earlier this year, the company faced backlash over its “Sky” voice, which many users felt closely resembled Scarlett Johansson’s. After feedback and Johansson’s legal action, OpenAI decided to pull the voice, clarifying that Sky was voiced by a different actress.

Now, Plus users are eagerly pressing OpenAI to accelerate the full rollout of Advanced Voice. Meanwhile, Apple Intelligence has yet to make its debut on iPhones, keeping iOS 18 users waiting for the anticipated AI-driven features. On the Android side, Google’s Gemini AI is already making waves, with early access being rolled out, giving users a taste of advanced voice and assistant capabilities before much-anticipated AI updates arrive.

Featured image credit: Solen Feyissa/Unsplash

Apple’s latest macOS release, Sequoia (macOS 15), has caused issues, breaking security software and network connections on updated systems.

Reports began to surface just days after its release, with prominent security vendors like CrowdStrike, Microsoft, and ESET warning users of problems with their software, which can lead to disrupted network connectivity and malfunctions.

What’s wrong with macOS Sequoia?

Patrick Wardle, macOS security expert and founder of Objective-See, highlighted that the issues might stem from networking bugs in Sequoia, suggesting Apple was aware of these problems before release.

According to reports from Microsoft and ESET, users experienced network troubles due to changes in macOS’s network stack, filters, and Stealth Mode, a feature that disables device responses to network pings.

ESET, for instance, advised users to remove specific network filters to prevent their software from being unsupported after upgrading, while Microsoft identified macOS’s Stealth Mode as a core issue affecting its products.

Both companies, along with Wardle, have noted Apple’s delayed response to fixing the issues.

What’s being done?

Some fixes and workarounds have been proposed by affected vendors. ESET suggests removing network filters or reinstalling software, while Microsoft advises users to turn off Stealth Mode in Firewall settings.

Wardle and other security experts have been vocal about Apple’s oversight, expressing frustration that these issues—especially in a security-focused OS—were not addressed before macOS Sequoia’s release.

When’s the fix coming?

Apple has yet to announce a specific date for a fix to the issues in macOS Sequoia (macOS 15), which is causing significant problems with security software, such as from CrowdStrike, Microsoft, and ESET.

These issues are largely linked to changes in the networking stack or firewall that are breaking the functionality of these tools. Users are reporting disruptions with VPNs, DNS resolution, and basic web browsing.

Although security vendors have issued workarounds, including firewall adjustments, these solutions are temporary and may introduce security risks.

Many experts recommend delaying the upgrade to Sequoia until Apple releases a patch or more permanent fixes become available.

For now, it’s advisable to stick with macOS Sonoma (macOS 14) if you’re reliant on these security products for your systems.

Apple is likely to address these issues in future updates, but no firm timeline has been provided.

Should you delay the update?

Yes, it’s a good idea to delay updating to macOS Sequoia (macOS 15) for now, especially if you rely on security software like those from CrowdStrike, Microsoft, or ESET.

The update has caused issues with network connections and the functionality of some security products. Vendors are still working on fixes, and Apple has not yet released an official patch to resolve these problems.

Waiting for further updates or patches will help you avoid disruptions in security and network performance.

For now, users facing problems after upgrading to macOS Sequoia should monitor updates from their security software providers and wait for Apple to release official patches.

The oversight has led to significant disruptions in security protocols, an area where macOS is known to pride itself.

With vendors scrambling to adapt to the changes, users are encouraged to delay upgrading until a clear resolution is available.

Featured image credit: Emre Çıtak/Ideogram AI

Amazon’s got something new to make your shopping experience a whole lot easier—and help you catch some deals you didn’t even know you wanted. They’ve rolled out a new AI feature that promises to give you more personalized recommendations.

The idea is simple: make shopping less of a hassle and more about finding exactly what you need.

One of the coolest upgrades is how Amazon’s recommendation system now works. You’re used to seeing “more like this” suggestions based on what you’re browsing. Now, it’s going to get a lot more specific. The AI digs into the details of your search history, showing you products that line up with what you’ve looked for in the past—things like holiday gifts or gear for that hobby you’ve been exploring. This means fewer random results and more ideas that could lead to solid discounts, especially if it catches a deal you didn’t notice before.

What’s even better is the way product descriptions are about to change. You know how sometimes it feels like you’re scrolling endlessly. This AI feature aims to fix this.

So, if you’re looking for “eco-friendly furniture” or a specific color for your home, that will be front and center. Less scrolling, more finding—and potentially faster access to those Prime Day discounts.

If a product description doesn’t quite hit the mark for you, there’s another AI working behind the scenes to fix it, making sure the next suggestion you see is even better.

Plus, if you’re into trying new tech, Amazon’s Rufus chatbot will help you out. You can actually chat with the AI, and it’ll help you find the best deals.

How to find best deals on Amazon Prime Day with AI?

Here are five tips to make the most out of this AI shopping experience:

Be specific in your search. Let’s say you’re looking for a vegan protein powder. Instead of just typing “protein powder,” tell the AI exactly what you want—“organic vegan protein powder under $30.” This way, it can filter out products that don’t meet your criteria and show you exactly what you’re after, along with any promotions or discounts available.

Ask for similar products. Found something you like but want to explore your options? Ask the AI for “more like this.” For example, if you’re eyeing a particular Bluetooth speaker, the AI can show you similar models, maybe even one that’s on sale or comes with a better warranty.

Set price alerts. If you’re not in a rush, use the AI to track prices. Say you want to buy a smartwatch but you’re waiting for a better deal. You can ask the AI to alert you when the price drops. Imagine getting a notification that your dream gadget is suddenly 20% off!

Combine recommendations with coupons. You can also ask the AI to look for products that accept coupons or are part of a limited-time promotion. For instance, if you’re searching for home office furniture, the AI might guide you to a desk that not only fits your needs but also has a coupon that knocks off an additional $50.

Use chat to find bundled deals. If you want to score multiple items for less, ask the AI about bundle deals. For example, if you’re shopping for kitchen appliances, you could ask the AI if there’s a package deal on a toaster and coffee maker, potentially saving you more than buying them separately.

When is the next Amazon Prime Day?

The next Amazon Prime Day, this time called “Prime Big Deal Days 2024,” will take place on October 8 and 9, closely aligning with the event dates from previous years, which were held on October 11-12 in 2022 and October 10-11 in 2023.

Featured image credit: Kerem Gülen/Midjourney

Google is changing the game for video creation by bringing advanced AI models, Veo and Imagen 3, to YouTube Shorts. Over the next few months, creators will get access to Dream Screen, a tool that lets you generate AI-powered backgrounds and short video clips. While AI tech is already out there, this move makes it way easier for millions of creators and billions of users to use. It’s all about giving people more creative power without needing a whole production crew.

Dynamic backgrounds with Dream Screen

Dream Screen works by starting with a simple text prompt. Imagen 3, Google’s AI model, generates four image options based on that prompt. Users can choose the image that best suits their needs, and Veo, another AI model, transforms it into a high-quality six-second background video. By 2025, this tool will also allow users to create six-second standalone video clips, offering more flexibility for content creators on YouTube Shorts.

This feature aims to make video creation more accessible. It gives creators the ability to generate professional-looking content without needing advanced tools or a large budget. Dream Screen is intended to help both experienced creators and those new to the platform.

To keep things clear, Google’s adding watermarks to AI-generated content using SynthID.

So, viewers will know when something’s been made with AI.

Watermarks did not convince us

As cool as it sounds, there’s something worth thinking about when we talk about integrating AI into every corner of content creation. Sure, AI is this shiny new toy, but is it really wise to lean on it for everything? You’ve heard the saying, “Too much of a good thing can be bad,” right?

Well, that might be the case here.

On one hand, AI makes life a lot easier. It speeds things up, fills in gaps, and opens up creative doors we never even knew existed. But on the flip side, where’s the line between helpful and over-reliant? If creators get used to letting AI do all the heavy lifting—like generating videos, writing scripts, or even composing music—are we losing some of that human touch?

Is AI creative: Answering the unanswerable

Creativity has always been about the spark, the struggle, the ah-ha! moments. If we hand those moments over to machines, do we risk flattening out the very thing that makes art special?

And what about authenticity? If everything starts looking AI-polished, it could all blur together. You know how sometimes you see those perfectly airbrushed photos, and they just don’t feel real anymore? That’s kind of the vibe we might be heading toward. AI can churn out beautiful, flawless content, but does that make it better? Or does it just mean we’re all going to be swimming in a sea of polished, but soulless, creations? It’s like having a box of chocolates where every piece tastes exactly the same—where’s the fun in that?

Plus, let’s talk about originality.

AI works by learning from the stuff that’s already out there, right? So, while it’s great at remixing and mashing up ideas, it’s not exactly bringing something completely new to the table. If we keep feeding the machine, won’t we eventually just get… well, more of the same? There’s a risk of innovation grinding to a halt because everything starts pulling from the same AI-generated pool. As the saying goes, “If you want something done right, do it yourself.”

Maybe that’s something creators need to remember.

Featured image credit: Kerem Gülen/Ideogram

LinkedIn has quietly opted its users into training generative AI models without explicitly asking for consent, raising concerns about data privacy on the platform. According to a report by 404Media, LinkedIn made changes to its privacy policy, stating that user data can be used to train AI models. The platform has since updated the policy, now allowing users to opt-out of this practice.

Updated LinkedIn policy reveals personal data usage on AI training

The updated policy states that LinkedIn may use personal data to “improve, develop, and provide products and Services,” as well as to train AI models. Generative AI is used for features like writing assistants, but LinkedIn claims it employs privacy-enhancing technologies to redact personal information. Users who prefer not to participate can opt-out by navigating to the “Data privacy” tab in their account settings, turning off the “Data for Generative AI Improvement” toggle.

However, opting out will only stop LinkedIn from using your data for future model training. Data that has already been used remains unaffected. Additionally, LinkedIn clarifies that users in the EU, EEA, or Switzerland are not included in AI model training.

If you’re concerned about other machine learning tools used for personalization and moderation, LinkedIn requires users to fill out a “Data Processing Objection Form” to opt-out from those uses as well.

LinkedIn’s silent opt-in move echoes similar actions from Meta, which recently admitted to scraping non-private user data for AI training dating back to 2007.

The timing of LinkedIn’s move comes at a moment when other major tech players, like OpenAI, are also facing backlash for similar practices. This pattern of quietly enrolling users in AI training without clear and prominent notifications creates a sense of unease.

OpenAI CTO Mira Murati says Sora was trained on publicly available and licensed data pic.twitter.com/rf7pZ0ZX00

— Tsarathustra (@tsarnick) March 13, 2024

It’s not just about data being used for AI—it’s about who gets to decide and how informed that decision is. The tech industry has long faced criticism for operating in the shadows when it comes to data collection, and the growing push for generative AI is only amplifying those concerns.

Can machines forget your personal data?

Another key issue is that opting out only affects future use of personal data. Any data that has already been fed into AI models remains in the system, and that lack of retroactive control may leave many users feeling powerless. The industry is also talking about “machine unlearning” to prevent this from happening, deleting data fed in AI models.

The fact that LinkedIn uses “privacy-enhancing technologies” to anonymize data is somewhat reassuring, but it doesn’t address the deeper problem: the need for more proactive, user-centered privacy standards.

Ultimately, this situation highlights the need for stronger, clearer regulations that put control back in the hands of users. The idea that tech companies can use our personal data without clear consent doesn’t sit well in times where privacy is becoming increasingly valuable.

Featured image credit: Kerem Gülen/Ideogram

Runway, a startup specializing in AI-powered video tools, has announced a big partnership with Lionsgate, the studio behind blockbusters like the John Wick and Twilight franchises. This collaboration will allow Runway to develop a custom AI video model trained on Lionsgate’s movie catalog, providing filmmakers, directors, and other creative talents access to cutting-edge AI tools designed to enhance their production work. According to Lionsgate’s vice chair, Michael Burns, the new AI model will help creators “augment their work” in unique and innovative ways.

Will Runway x Lionsgate deal change the face of the cinema?

“Runway is a visionary, best-in-class partner who will help us utilize AI to develop cutting edge, capital efficient content creation opportunities,” said Lionsgate Vice Chair Michael Burns. “Several of our filmmakers are already excited about its potential applications to their pre-production and post-production process. We view AI as a great tool for augmenting, enhancing and supplementing our current operations.”

Runway is also exploring the possibility of offering licensed versions of these AI models, giving independent creators the ability to build and customize their own AI tools for personal projects. This move marks Runway as the first AI company to publicly secure a deal with a major Hollywood studio, setting it apart from other tech firms like Disney and Paramount, who have yet to formalize similar partnerships despite discussions recently reported by The Wall Street Journal.

However, this partnership coincides with new legislative developments in California, where recent bills were signed to regulate the use of AI-generated digital replicas in film and television. Adding to the complexity, Runway is also facing legal challenges over accusations that it used copyrighted content without permission to train its AI models.

While this collaboration could push the boundaries of what AI can offer to the entertainment industry, it also highlights the growing tension between technological innovation and legal and ethical concerns. The partnership between Runway and Lionsgate might signal a new chapter for AI in filmmaking, but it’s clear that the path forward will be closely scrutinized.

How does this affect cinema?

Cinema has always pushed the boundaries of imagination, but since the early 2000s, the once-impossible ideas from movies have started to become real. Video calls are now a daily norm, flying car prototypes are being tested, holograms appear at concerts, and artificial intelligence is reshaping entire industries.

Best AI movies to watch for the weekend

But is cinema truly predicting the future, or is it just a coincidence? It’s not so simple to answer. Filmmakers often collaborate with scientists and engineers to turn imaginative ideas into reality. For example, the hoverboard from Back to the Future sparked the creation of real-life prototypes, even if no one truly “needed” them. Like a child driven by the desire to play with their favorite toy, engineers accepted the challenge the film presented. The hoverboard Marty McFly rides in that iconic scene wasn’t just a dream—it was a dare, one that engineers eagerly took on.

The influence of cinema extends beyond mere inspiration. Many leading tech companies, such as Adobe, Pixar, Technicolor, and Dolby, were born from the desire to bring cinematic visions to life. Today, these companies shape not only the film industry but also the broader tech world, driven by founders who turned their cinematic dreams into reality.

Featured image credit: Runway

Artificial General Intelligence (AGI), the concept of machines reaching human-level intelligence, is often seen as the holy grail of AI development. Companies like OpenAI and Google’s DeepMind often present their breakthroughs as stepping stones toward AGI.

However, Neil Lawrence, a professor of machine learning at the University of Cambridge, believes this pursuit is misguided, calling AGI “nonsense” and arguing that AI and human intelligence should not even be compared.

Lawrence’s viewpoint, highlighted in his new book The Atomic Human: Understanding Ourselves in the Age of AI, pushes back against the narrative that AI will eventually rival human intelligence.

His argument is simple but profound: AI is fundamentally different from human cognition, and it’s a mistake to expect it to replicate the intricacies of human thought.

Machines are not like us and that’s okay

According to Lawrence, the rise of AI systems like ChatGPT and its competitors has led to a misplaced comparison between human and artificial intelligence. While these systems can perform a range of tasks previously thought to require human intelligence—such as writing essays or answering complex questions—Lawrence argues that their functioning is far more limited than the term “general intelligence” suggests.

In essence, AI is excellent at narrow, specialized tasks, often far surpassing humans in speed and efficiency. But Lawrence cautions that this form of intelligence is more akin to the behavior of ants than it is to human thought.

In an interview with NewScientis, he argues that AI systems, no matter how advanced, operate within predefined parameters and lack the broader cognitive flexibility that humans possess. AI’s intelligence, Lawrence suggests, is task-specific and programmed, while human intelligence emerges from biological complexity, emotion, and social interaction—things that machines cannot replicate.

Lawrence dismisses the pursuit of AGI as a distraction, claiming it sets an unrealistic expectation for what AI can accomplish. He asserts that current AI is far from “general” and should not be expected to achieve human-like cognition.

“The nature of the intelligence we’re seeing in AI is extremely different from our own,”

He states, calling it “absurd” to treat machine intelligence as if it’s on a path toward becoming like human intelligence.

What makes us “unique”?

What makes humans unique is a question that has perplexed philosophers for centuries. Many point to our capacity for abstract reasoning, self-awareness, and moral decision-making as key markers of what sets us apart.

Unlike animals, whose behaviors are often driven by instinct or survival, humans can contemplate existence, ponder the nature of the universe, and even question the meaning of life itself. This ability to reflect on our own thoughts—a trait known as metacognition—imbues our intelligence with a depth and complexity that we often take for granted.

It’s not just about processing information, but understanding, feeling, and reflecting on it, something AI, despite its vast computational power, cannot achieve.

In biological terms, humans and animals share much of the same genetic blueprint. We evolved from common ancestors, and the mechanisms that govern our physical bodies—our cells, neurons, and organs—are remarkably similar. Yet, the human brain, with its vast neocortex, is wired for a level of sophisticated thought and emotion that seems unparalleled in the animal kingdom.

Our brains allow for creativity, language, culture, and a nuanced understanding of morality. While animals can exhibit extraordinary intelligence—like a crow using tools or a dolphin demonstrating social cooperation—their cognition lacks the self-awareness, cultural depth, and adaptability that humans possess.

When we draw a line from these biological and philosophical distinctions to AI, the contrasts become even starker.

AI, no matter how sophisticated, lacks the embodiment that is central to human and animal life. Our intelligence is deeply rooted in our biology, influenced by our emotions, physical sensations, and social bonds. AI, by contrast, operates in an entirely different realm—it processes data, learns patterns, and performs tasks, but it doesn’t feel, reflect, or engage in the emotional and social complexity that defines human and even animal experience.

AI might be able to simulate aspects of human behavior, like recognizing faces or generating text, but it fundamentally lacks the depth of human consciousness. The experience of being human—grappling with joy, sorrow, love, and existential questions—is something that cannot be encoded in algorithms.

Transforming society, not replicating humanity

While AI may never replicate the full scope of human intelligence, it is already transforming society in profound ways. Rather than striving to mimic human consciousness, the true strength of AI lies in its ability to complement and extend human capabilities.

Machines are not bound by the limitations of biology—no fatigue, no emotion, no ethical dilemmas—making them highly efficient at tasks that require speed, precision, and data processing on an unimaginable scale.

This makes AI a powerful tool, not a replacement for human thought, but an enhancement of what we can achieve.

However, this transformation is not without challenges. The growing reliance on AI prompts deeper questions about ethics, privacy, and control. As AI systems become more integrated into daily life, they start to influence decisions that traditionally require human judgment—decisions laden with moral implications. While AI can process facts and probabilities, it cannot grasp the ethical nuance of its actions, leaving society to grapple with how and where to apply such technologies responsibly.

The key is to recognize AI’s role as a tool, not an entity. As society evolves with AI, we must focus on leveraging its strengths to solve problems that are beyond human capacity, without assuming it will ever mirror the full spectrum of human experience. Its ability to process vast datasets, perform high-speed calculations, and streamline repetitive tasks should be seen as an opportunity to transform our world—not an attempt to replicate what makes us human.

Let’s stop comparing, start collaborating

As we advance into an era where artificial intelligence increasingly intersects with human endeavors, the emphasis should shift from comparing AI to human intelligence to exploring how these two forms of intelligence can work in tandem. The binary debate of whether AI can or cannot replicate human thought obscures the more productive discussion of how AI can complement human skills and augment our capabilities.

The crux of this collaborative potential lies in recognizing that AI is not a competitor to human intelligence but a partner that enhances our abilities.

The future of AI should be about harnessing its capabilities to amplify human potential rather than seeking to replace it. By focusing on collaboration, we can leverage AI’s strengths to tackle complex global challenges and improve quality of life. This approach requires a shift in mindset from viewing AI as a threat to recognizing it as a tool that, when used thoughtfully, can help us achieve our collective goals.

Ultimately, the goal is not to replicate human intelligence but to foster a synergy where AI enhances human decision-making, creativity, and problem-solving.

Embracing this perspective allows us to utilize AI as a powerful ally in our quest for progress, achieving what neither humans nor machines could accomplish alone.

Image credits: Emre Çıtak/Ideogram AI

Time is running out for Windows users, as Microsoft and the Cybersecurity and Infrastructure Security Agency (CISA) have issued a clear warning: update your systems now or risk severe security vulnerabilities. A new exploit involving outdated Internet Explorer code threatens the security of millions of PCs. Despite Internet Explorer being largely forgotten, the remnants of its code have opened up a major vulnerability, and hackers are already exploiting it.

CISA recently added a new vulnerability, CVE-2024-43461, to its Known Exploited Vulnerabilities (KEV) catalog. This exploit, rooted in the MSHTML platform within Windows, allows hackers to spoof web pages and trick users into visiting malicious sites. Coupled with another vulnerability from July (CVE-2024-38112), this issue forms a dangerous attack chain that leaves any unpatched PC exposed. If your PC hasn’t received the latest updates, your system may be at risk.

Microsoft and CISA warn against another global crisis

Federal agencies have been given until October 7, 2024, to address this vulnerability. However, this deadline isn’t just for government offices; anyone with a Windows PC should prioritize installing the updates. Microsoft fixed part of the vulnerability in their July 2024 Patch Tuesday update, addressing CVE-2024-38112. The most recent update, part of September’s Patch Tuesday, closes the remaining gap, specifically patching CVE-2024-43461. Together, these fixes prevent remote attackers from gaining access to your system through malicious web pages or files.

If you’ve already updated since July, you might think you’re in the clear. However, if you haven’t kept up with the latest patches, your system is still exposed. In a statement, Microsoft noted that while they addressed the initial threat chain earlier in the year, the full resolution wasn’t available until this latest update. Ignoring this fix leaves your PC vulnerable to remote code execution attacks, where hackers can gain control of your computer simply by tricking you into clicking a malicious link.

As you may remember, there were global cyber disasters over a software update by CrowdStrike. This update does not have such side effects, but it may have a domino effect on Windows built on top of the old one. Therefore, it is recommended to take precautions.

The MSHTML exploit: A backdoor in disguise

The MSHTML platform, though outdated, remains a part of modern Windows systems due to its use in Internet Explorer mode in Microsoft Edge. Attackers have figured out how to leverage this hidden code to launch their attacks. Security researchers from Trend Micro’s Zero Day Initiative (ZDI) explain that this vulnerability allows hackers to disguise malicious files, tricking users into thinking they are harmless. Once opened, these files can execute code and grant attackers access to your system.

One particularly troubling detail is that hackers are targeting unsuspecting users through popular cloud-sharing platforms, Discord servers, and even online libraries. Files are being disguised as harmless PDFs or other documents, but hidden within are the malicious elements needed to exploit the MSHTML flaw. The cybersecurity group Void Banshee, known for targeting organizations across North America, Asia, and Europe, has been linked to these attacks, using them to steal sensitive information such as passwords and cryptocurrency wallets.

Why this update matters now

For anyone wondering why they should take this latest security threat seriously, consider the broader impact. Federal agencies are required by law to patch these vulnerabilities, and this urgency should be a signal to private individuals and organizations as well. Attackers have been using the MSHTML vulnerability to bypass modern browser protections, exploiting Internet Explorer’s dormant code even on Windows 10 and 11 machines. The fact that a long-obsolete browser is being used as a gateway for modern attacks is reason enough to act now.

Check Point, a leading cybersecurity firm, highlights the surprising nature of this exploit, stating that many users don’t even realize Internet Explorer is still on their systems. The fix is simple: install the Microsoft patch. But until you do, your system is at risk.

🛡️ We added #Microsoft Windows & #Progress WhatsUp vulnerabilities, #CVE-2024-43461 & CVE-2024-6670, to our Known Exploited Vulnerabilities Catalog. Visit https://t.co/dOIn6I9vuB & apply mitigations to protect your org from cyberattacks. #Cybersecurity #InfoSec pic.twitter.com/zmADn0MgpG

— CISA Cyber (@CISACyber) September 16, 2024

What happens if you don’t update?

If you’re thinking of waiting for this one out, think again. CISA has explicitly warned that failure to update could result in serious breaches. With a wide range of malicious actors exploiting this flaw, the consequences of inaction could be dire. Whether it’s personal data theft, financial loss, or worse, the risks of not patching your system are too high to ignore. As Void Banshee continues to target vulnerable systems, the clock is ticking for all users to take action.

Although CISA’s warnings are aimed at federal agencies, private companies, and everyday users are equally vulnerable. If you’re using a Windows system, especially one that hasn’t been updated recently, you’re potentially a target. Microsoft has emphasized the importance of installing both the July and September patches to fully protect your PC. Ignoring these updates leaves your system susceptible to exploitation by hackers looking to steal sensitive information.

With just a few weeks left, there’s no time to hesitate. Microsoft says this is a critical update that could make your system more secure. If your PC hasn’t been updated since July, you’ve missed a deadline. Don’t miss this one. Update your Windows system by October 7 to avoid a cyberattack. Update now or face the consequences. The MSHTML vulnerability is real, and the clock is ticking for anyone still running an unpatched version of Windows. Don’t wait for hackers to exploit your system. Secure your PC before it’s too late.

Image credits: Furkan Demirkaya / Ideogram

Google continues to advance its efforts in content transparency, focusing on tools that help users understand how media—such as images, videos, and audio—has been created and modified. A key development in this area is its collaboration with the Coalition for Content Provenance and Authenticity (C2PA), where Google plays an active role as a steering committee member. The goal of this partnership is to enhance online transparency as content moves across platforms, providing users with better information on the origins and alterations of the media they engage with.

The C2PA focuses on content provenance technology, which helps users determine whether a piece of content was captured by a camera, edited through software, or generated by AI. This initiative aims to equip people with information that builds media literacy and allows them to make more informed decisions about the authenticity of the content they encounter. According to the announcement, Google has been contributing to the latest version (2.1) of the C2PA’s Content Credentials standard, which now has stricter security measures to prevent tampering, helping ensure that provenance information is not misleading.

New feature integration into Google products

Over the next few months, Google plans to integrate this new version of Content Credentials into some of its key products. For instance, users will soon be able to access C2PA metadata through the “About this image” feature in Google Search, Google Images, and Google Lens. This feature will help users identify whether an image has been created or altered using AI tools, providing more context about the media they come across.

Google is also incorporating these metadata standards into its advertising systems. Over time, the company aims to use C2PA signals to inform how it enforces ad policies, potentially shaping the way ad content is monitored for authenticity and accuracy.

Additionally, Google is exploring ways to extend this technology to YouTube, with the possibility of allowing viewers to verify the origins of video content in the future. This expansion would further Google’s push to bring transparency to media creation across its platforms.

Google’s role in the C2PA extends beyond its own product integrations. The company is also encouraging broader adoption of content provenance technology across the tech industry. The goal is to create interoperable solutions that work across platforms, services, and hardware providers. This collaborative approach is seen as crucial to creating sustainable, industry-wide standards for verifying the authenticity of digital content.

To complement its work with the C2PA, Google is also pushing forward with SynthID, a tool developed by Google DeepMind that embeds watermarks into AI-generated content. These watermarks allow AI-created media to be more easily identified and traced, addressing concerns about the potential for misinformation spread by AI tools.

Artists Google to court over AI image generator

As more digital content is created using AI tools, ensuring that provenance data remains accurate and secure will be crucial. Google’s involvement in C2PA is part of a broader effort to address these challenges, but the effectiveness of these measures will depend on widespread industry cooperation and adoption.

Google’s efforts to address AI-generated content in its search results are a step toward more transparency, questions remain about the effectiveness of the approach. The “About This Image” feature, which provides additional context about whether an image was created or edited with AI, requires users to take extra steps to access this information. This could be a potential limitation, as users who are unaware of the tool may never know it’s available to them. The feature relies heavily on users actively seeking out the provenance of an image, which may not be the default behavior for many.

The broader challenge lies in making such transparency tools more intuitive and visible to users, so they can quickly and easily verify content without having to dig for details. As AI-generated content continues to grow, the need for seamless verification will only become more pressing, raising questions about whether hidden labels and extra steps are enough to maintain trust in digital media.

Image credits: Kerem Gülen/Ideogram

Microsoft’s push for sustainability has taken a technological turn with the use of AI-powered robots to dismantle and destroy hard drives in its data centers.

As part of its Secure and Sustainable Disposal of Hard Disks project, this initiative aims to tackle the growing e-waste problem while ensuring data security.

But is this high-tech solution really helping reduce waste, or is it merely creating more waste in the process?

Creating more waste to stop the waste?

The project, which originated during the 2022 Microsoft Hackathon, builds on the company’s Circular Centers initiative. These centers are designed to recycle and repurpose hardware used in Microsoft’s data centers, such as servers and hard drives.

The goal is to help Microsoft achieve its ambitious environmental commitments: Becoming carbon-negative by 2050 and producing zero waste by 2030. The AI-powered robots are now part of this larger effort, decommissioning old hardware more efficiently while ensuring that the valuable materials within are not wasted.

Historically, hard drives have been shredded to protect sensitive data, a process that not only destroys the media but also leads to the loss of valuable materials like neodymium. Every year, millions of hard drives are shredded worldwide, leading to significant waste, especially of rare metals.

Microsoft’s new approach, dubbed #NoShred, uses AI and robotics to dismantle hard drives, ensuring the sensitive data is destroyed while allowing the reuse or recycling of the remaining components.

By using computer vision and robotic arms, the system is able to sort and disassemble hard drives in a secure manner, preserving key materials for recycling. The project aims to achieve a 90% reuse and recycle rate for all hard disks by 2025. A pilot run in Amsterdam showed promising results, reducing downtime and increasing the availability of recycled parts.

While Microsoft’s AI-powered solution offers a more sustainable approach to data center waste, it raises a broader debate:

Are we simply creating more waste to stop the waste?

The development of these AI-powered robots requires significant amounts of energy, rare metals, and resources to build and operate. This energy consumption is expected to rise, especially as AI becomes more integrated into processes worldwide. Data centers already account for about 1-1.3% of global electricity demand, and this is projected to double by 2026.

With AI driving more data processing and higher energy usage, are we offsetting the environmental benefits of recycling with the carbon footprint of these AI systems?

On one hand, the project seems like a step forward. By avoiding the destruction of rare materials, Microsoft is addressing the growing scarcity of components like neodymium, a critical element in technology production.

But on the other hand, the rising demand for AI and the infrastructure required to maintain it may ultimately create new sustainability challenges.

How about a happy medium?

Microsoft’s AI-powered robots represent a new frontier in sustainable tech, but they also highlight the inherent tensions in using advanced technology to solve environmental problems. While the robots are helping Microsoft reach its recycling goals, the long-term sustainability of this approach remains uncertain.

In the race to reduce e-waste, companies like Microsoft must ensure that the solutions they implement do not inadvertently create new problems.

Balancing the benefits of AI with its environmental impact will be key to determining whether initiatives like #NoShred are a true solution—or just another layer of complexity in the fight against waste.

Featured image credit: Emre Çıtak/Ideogram AI

Intel has taken dramatic steps to restructure its operations, breathing new life into Intel stock price.

In the wake of a 7% increase in Intel stock, CEO Pat Gelsinger has outlined a series of ambitious changes aimed at reviving the company’s fortunes and positioning it for future success.

The focal point of this transformation is the establishment of Intel Foundry as an independent subsidiary, marking a significant milestone in Intel’s quest to regain its dominance in the semiconductor industry.

Foundry soars Intel stock prices

Intel’s decision to spin off its foundry business into an independent subsidiary is central to its strategy. Announced by Gelsinger earlier this week, the foundry business will have its own operating board and separate financial reporting.

The move is intended to give the foundry division the autonomy it needs to attract outside capital, focus on efficiency, and better serve customers. Historically, Intel has struggled to compete with other semiconductor manufacturers, but the hope is that greater independence will allow the foundry business to operate more flexibly and competitively in the market.

Panther Lake and Clearwater Forest, lead products on Intel 18A, are out of the fab, powered on, have booted operating systems, and are on track to start production next year. Additionally, Intel 18A has brought together two critical innovations to enable customers to make leaps…

— Pat Gelsinger (@PGelsinger) August 6, 2024

By giving Intel Foundry its own leadership and financial structure, the company is signaling a willingness to make hard choices in order to win back customers and investors.

One such customer is Amazon’s cloud services division, AWS, which has recently inked a multibillion-dollar deal with Intel Foundry for the production of custom artificial intelligence (AI) chips, as Reuters reports. Securing Amazon as a customer is a major vote of confidence and could be the catalyst Intel needs to rebuild its reputation in the semiconductor space.

Cutting costs and scaling back operations

While Intel celebrates its new deal with Amazon, the company is simultaneously making painful decisions elsewhere. In an effort to streamline operations and cut costs, Intel announced plans to halt work on new chip factories in Germany and Poland for the next two years.

These projects, which were originally conceived as part of a broader expansion strategy, have been postponed due to market conditions and expected demand. Instead, Intel will concentrate its manufacturing efforts on facilities in Arizona, Oregon, New Mexico, and Ohio, where it believes the return on investment will be stronger.

Moreover, Intel has revealed plans to significantly reduce its real estate footprint, cutting up to two-thirds of its global properties. This downsizing reflects Intel’s recognition of the need for leaner operations as it contends with rising competition and falling profitability.

The battle for AI supremacy

Intel is betting big on its 18A chipmaking process, a cutting-edge technology designed to produce faster, more efficient chips for AI and data center applications.

While initial testing of the 18A process has encountered difficulties, particularly in collaboration with Broadcom, Intel remains confident that it will be a game changer in the industry.

Intel reveals its AI roadmap with Lunar Lake at Computex 2024

In addition to its deal with Amazon, Intel has also secured Microsoft as a partner for its 18A process, further boosting optimism around the company’s AI capabilities.

AI chip production is quickly becoming a battleground for semiconductor manufacturers, and Intel is determined to remain at the forefront of this competition. With cloud giants like Amazon and Microsoft now on board, Intel’s AI ambitions appear to be on solid footing.

Financial challenges and turnaround efforts

Despite the promising developments in AI and foundry independence, Intel is still grappling with significant financial losses. In the first quarter of 2024, the company reported a staggering $1.6 billion in total losses, with its chipmaking division alone accounting for $7 billion in operating losses throughout 2023.

These figures have put intense pressure on Gelsinger to implement sweeping changes and drive greater profitability.

In response, the company has launched a comprehensive cost-cutting campaign that includes Intel layoffs affecting 15,000 employees worldwide.

Gelsinger has described this period as the most significant transformation in Intel’s history, comparing it to the company’s transition from memory chips to microprocessors in the 1980s.

Smoothly sailing with government support winds

Intel’s comeback effort has not gone unnoticed by the U.S. government. The company was recently awarded up to $3 billion in direct funding through the CHIPS and Science Act, a federal initiative aimed at boosting domestic chip production for critical industries, including the U.S. military.

Today, the Biden-Harris Administration and @DeptofDefense announced that @intel has been awarded up to $3 billion in direct funding under the CHIPS and Science Act for the Secure Enclave program – another step toward our shared goal of restoring American leadership in…

— Pat Gelsinger (@PGelsinger) September 16, 2024

The funding will play a crucial role in Intel’s plans to expand its manufacturing capabilities in the United States, further reinforcing its commitment to regaining leadership in the global semiconductor market.

Intel stock price chart says it all

The Intel stock price chart reveals a steady, gradual climb over the past several days, culminating in a significant surge towards the end of the period, closing at 20.92. This consistent upward movement reflects the market’s growing confidence in Intel’s recent strategic decisions, particularly CEO Pat Gelsinger’s bold move to spin off the Intel Foundry as an independent subsidiary.

Intel stock has been under pressure due to its financial struggles, including billions in losses and significant restructuring efforts. However, the announcement of Intel Foundry’s independence, along with securing Amazon’s cloud division as a key client for its advanced AI chips, has renewed optimism. The chart clearly indicates that investors are reacting positively to these developments, as seen in the smooth yet solid upward trajectory starting mid-period, reflecting growing investor confidence in Intel’s turnaround plan.

Additionally, the chart displays several minor fluctuations, reflecting typical market volatility and investor caution. These smaller dips were likely influenced by Intel’s announcement to pause construction on its chip factories in Europe and real estate cutbacks. Yet, despite these temporary setbacks, the current Intel stock price shows a consistent recovery, which aligns with Intel’s broader efforts to prioritize its U.S.-based manufacturing plants and its shift towards AI-focused chipmaking.

The sharp rise near the end of the chart can be attributed to two major factors: Intel’s ambitious restructuring plans and the boost in market sentiment following Intel’s new AI chip deal with Amazon’s AWS. This surge signals that investors are confident that Intel’s strategic decisions will reverse its recent financial losses and drive future growth.

Intel on the rise, but challenges remain

Intel’s stock surge of 7% in the wake of these announcements signals that investors are cautiously optimistic about the company’s future. The decision to spin off Intel Foundry as an independent entity, secure major AI contracts, and scale back non-essential projects demonstrates a clear vision for revitalizing Intel’s operations.

However, the company is still navigating financial turbulence, and much will depend on how well these bold initiatives are executed in the months and years ahead.

With the semiconductor industry more competitive than ever, Intel faces a challenging road to recovery. Yet under Gelsinger’s leadership, the company appears to be making the right moves to rise from the ashes and reclaim its place among the top chipmakers in the world.

Featured image credit: Emre Çıtak/Ideogram AI

Google has taken another significant step forward in the race to improve the accuracy and reliability of AI models with the introduction of DataGemma, an innovative approach that combines its Gemma large language models (LLMs) and the Data Commons project. The spotlight here is on a technique called retrieval-augmented generation (RAG), a method that has been gaining traction in enterprises, but now, with DataGemma, Google aims to bring it into the AI mainstream.

At its core, RAG seeks to solve one of the biggest challenges faced by LLMs: the problem of hallucinations. In the world of generative AI, hallucinations refer to instances where the model generates information that sounds plausible but is factually incorrect. This is a common issue in AI systems, especially when they lack reliable grounding in factual data. Google’s goal with DataGemma is to “harness the knowledge of Data Commons to enhance LLM factuality and reasoning,” addressing this issue head-on.

What is RAG?

Retrieval-augmented generation is a game changer because it doesn’t rely solely on pre-trained AI models to generate answers. Instead, it retrieves relevant data from an external source before generating a response. This approach allows AI to provide more accurate and contextually relevant answers by pulling real-world data from repositories. In the case of DataGemma, the source of this data is Google’s Data Commons project, a publicly available resource that aggregates statistical data from reputable institutions like the United Nations.

This move by Google to integrate Data Commons with its generative AI models represents the first large-scale cloud-based implementation of RAG. While many enterprises have used RAG to ground their AI models in proprietary data, using a public data resource like Data Commons takes things to a whole new level. It signals Google’s intention to use verifiable, high-quality data to make AI more reliable and useful across a broad range of applications.

Google’s two-pronged approach

According to Google, DataGemma takes “two distinct approaches” to integrate data retrieval with LLM output. The first method is called retrieval-interleaved generation (RIG). With RIG, the AI fetches specific statistical data to fact-check questions posed in the query prompt. For example, if a user asks, “Has the use of renewables increased in the world?” the system can pull in up-to-date statistics from Data Commons and cite them directly in its response. This not only improves the factual accuracy of the answer but also provides users with concrete sources for the information.

The second method is more in line with the traditional RAG approach. Here, the model retrieves data to generate more comprehensive and detailed responses, citing the sources of the data to create a fuller picture. “DataGemma retrieves relevant contextual information from Data Commons before the model initiates response generation,” Google states. This ensures that the AI has all the necessary facts at hand before it begins generating an answer, greatly reducing the likelihood of hallucinations.

A key feature of DataGemma is the use of Google’s Gemini 1.5 model, which boasts an impressive context window of up to 128,000 tokens. In AI terms, the context window refers to how much information the model can hold in memory while processing a query. The larger the window, the more data the model can take into account when generating a response. Gemini 1.5 can even scale up to a staggering 1 million tokens, allowing it to pull in massive amounts of data from Data Commons and use it to craft detailed, nuanced responses.

This extended context window is critical because it allows DataGemma to “minimize the risk of hallucinations and enhance the accuracy of responses,” according to Google. By holding more relevant information in memory, the model can cross-check its own output with real-world data, ensuring that the answers it provides are not only relevant but also factually grounded.

Beyond LLMs

While the integration of RAG techniques is exciting on its own, DataGemma also represents a broader shift in the AI landscape. It’s no longer just about large language models generating text or answering questions based on what they’ve been trained on. The future of AI lies in its ability to integrate with real-time data sources, ensuring that its outputs are as accurate and up-to-date as possible.

Google is not alone in this pursuit. Just last week, OpenAI unveiled its “Strawberry” project, which takes a different approach to improving AI reasoning. Strawberry uses a method known as “chain of thought”, where the AI spells out the steps or factors it uses to arrive at a prediction or conclusion. While different from RAG, the goal is similar: make AI more transparent, reliable, and useful by providing insights into the reasoning behind its answers.

What’s next for DataGemma?

For now, DataGemma remains a work in progress. Google acknowledges that more testing and development are needed before the system can be made widely available to the public. However, early results are promising. Google claims that both the RIG and RAG approaches have led to improvements in output quality, with “fewer hallucinations for use cases across research, decision-making, or simply satisfying curiosity.”

It’s clear that Google, along with other leading AI companies, is moving beyond the basic capabilities of large language models. The future of AI lies in its ability to integrate with external data sources, whether they be public databases like Data Commons or proprietary corporate data. By doing so, AI can move beyond its limitations and become a more powerful tool for decision-making, research, and exploration.

Featured image credit: Google

Slack recently unveiled a suite of new features for its higher-tier users, centered around AI integration. Among the most talked-about is Slack AI, which includes AI-generated Huddle summaries, echoing the channel summaries that some subscribers already enjoy. But the update goes further, allowing users to engage with Salesforce’s AI agents directly within Slack, alongside third-party tools that introduce AI-powered web search and image generation.

This AI push isn’t entirely out of the blue. Salesforce acquired Slack in 2021, right when the platform was booming as the go-to messaging service for remote workers. Fast forward three years, and Salesforce is leaning heavily into AI—so much so that it’s trickling down into Slack’s DNA. CEO Denise Dresser emphasizes that Slack is the perfect platform for this shift, as it’s already a hub for workplace conversation, making it an ideal spot to embed AI interactions.

However, it’s hard not to be a bit skeptical. Since ChatGPT burst onto the scene in 2022, countless companies have bolted AI features onto their products, often in ways that feel more like a trendy gimmick than a meaningful enhancement. And while AI summaries and integrations in Slack might seem helpful, one has to question if this addition is truly a natural evolution or simply another example of chasing the AI bandwagon.

Does AI in Slack really enhance the core product, or is it just Salesforce showing it can ride the AI wave? The answer remains to be seen.

New AI-powered Slack features

“Slack’s agent-powered work operating system is redefining what it means to work smarter, not just differently. We’ve moved beyond collaboration; every day, millions of teams are leveraging purpose-built AI, with customer data, automation, and now agents seamlessly integrated into the flow of work. This isn’t just about getting work done. It’s about unlocking new productivity levels to drive results and move business forward for every team and department. In today’s increasingly complex work environment, true productivity requires a work operating system built for the future of work — and that is Slack.”

-Denise Dresser

Agentforce in Slack

Salesforce introduced Agentforce in Slack, a new user interface that allows teams to interact with data, surface insights, and perform actions directly within Slack. Teams can now engage with Agentforce agents to get status updates, recommend next steps, draft emails, and more, all powered by AI.

Third-party AI agents

Slack now supports third-party AI agents and assistants from partners such as Adobe, Anthropic, Cohere, and Perplexity. These agents can assist with tasks like content creation, project management, and data analysis, providing a more efficient workflow within a secure environment.

Salesforce channels

Salesforce channels connect Salesforce CRM records to Slack conversations, enabling teams to collaborate more effectively on accounts and opportunities. Teams can securely view and edit CRM data directly in Slack without duplicating work across multiple systems.

New Slack AI features

Slack AI introduces several new features, including huddle notes that capture key meeting takeaways and action items. Slack AI can also generate workflows with simple conversational prompts and improve search by generating answers from files, apps, and conversational data in Slack.

Slack AI huddle notes

Slack AI huddle notes capture real-time audio and messages shared in a huddle, organizing them into transcripts, action items, and files in a canvas. This feature ensures everyone stays informed, even if they missed the meeting.

AI Workflow Builder

With the AI Workflow Builder, users can easily automate tasks by typing conversational prompts. It simplifies workflow creation, automating repetitive tasks and allowing teams to focus on more important work.

Slack AI search

Slack AI search is now enhanced with file and app knowledge, surfacing richer and more relevant answers. It uses Retrieval Augmented Generation (RAG) to pull information from messages, files, and connected apps that users have access to in Slack.

Slack templates

Slack templates provide pre-configured collections of Slack tools, such as channels, canvases, lists, and workflows, tailored for specific departments or tasks. These templates help streamline work processes and improve productivity across various use cases.

Salesforce channels in CRM integration

Salesforce channels bring CRM data from opportunities, cases, and custom objects into Slack channels, allowing teams to collaborate on customer records in real time. These channels also allow for editing and updating Salesforce data directly within Slack.

Slack AI conversation summaries in Salesforce channels

Slack AI can summarize key highlights from customer-related discussions across multiple Salesforce channels, helping teams stay informed and aligned. This summary feature also provides quick access to essential information for decision-making.

Salesforce channels embedded in the Salesforce interface

Salesforce channels will soon be embedded in the Salesforce user interface, allowing agents and teams to collaborate seamlessly without leaving the CRM. Conversations within Slack will be tied to specific records, improving communication and workflow efficiency.

Slack Marketplace for third-party agents

The Slack Marketplace now offers a range of third-party AI agents, including Adobe Express, Amazon Q Business, Anthropic’s Claude, Asana, Box, Cohere, Perplexity, Workday, and Writer. These agents assist with tasks like content creation, project management, and data analysis.

Agentforce Sales Coach

Agentforce Sales Coach, powered by Data Cloud, can autonomously facilitate role-plays and provide personalized feedback to sellers, helping them improve their skills within the context of a deal, right in Slack.

Adobe Express in Slack

Adobe Express in Slack allows users to create on-brand content, such as presentations and media campaigns, directly from simple prompts, making content creation faster and easier.

Amazon Q Business in Slack

Amazon Q Business answers questions, summarizes data, and generates content based on enterprise systems, enhancing productivity within Slack.

Anthropic’s Claude in Slack

Anthropic’s Claude helps with content creation, code debugging, and data analysis, accelerating workflows and research within Slack.

Asana

Asana provides project insights, status updates, and next steps within Slack conversations, helping teams stay aligned and move projects forward efficiently.

Box

Box offers AI-powered insights from uploaded content, allowing users to analyze contracts, answer questions, and summarize reports directly in Slack.

Cohere

Cohere enhances Slack efficiency by providing AI-driven content and answers grounded in proprietary business data, streamlining internal communication.

Perplexity

Perplexity enables users to search the web for trusted sources, providing competitive analysis and strategic recommendations within Slack.

Workday

Workday provides AI assistants that allow employees and managers to access financial and HR insights, such as pay and job requisitions, directly in Slack.

Writer

Writer helps teams draft assets, analyze data, and brainstorm ideas using AI-driven insights tailored to customer data, brand, and workflows within Slack.

Salesforce channels in Salesforce Starter Suite

Salesforce channels will be accessible in Salesforce Starter Suite, enabling small businesses to connect Slack to a simplified CRM, enhancing collaboration and speeding up decision-making.

Achieve more with less effort with Slack AI‘s assistance

Don’t let it be just noise

Absolutely, having a wide range of features is fantastic, and Salesforce has certainly packed Slack with tools that can transform how teams work.

But there’s a big challenge here that often gets overlooked—most Slack users aren’t tech experts.

We see it all the time: companies roll out new, advanced tools, but the end users don’t fully embrace them, not because the features aren’t useful, but because they’re not intuitive or people haven’t been properly trained.

Slack is now integrating AI agents, automations, and complex CRM channels, and while this all sounds amazing on paper, if your average user doesn’t know how to navigate these features, it’s just noise. It’s not about loading the app with more features; it’s about helping users understand how to use those features efficiently. Without proper education and onboarding, employees might stick to the basics—sending messages and sharing files—while missing out on tools that could save them time or help them work smarter.

The best apps don’t just introduce new tools; they also take the time to guide users on how to implement them into their workflows. Think about solutions like in-app tutorials, user-friendly documentation, or even dedicated training sessions. Companies that master this balance between innovation and education are the ones that see the real productivity gains.

Featured image credit: Scott Webb/Unsplash

The impressive planning capabilities of the new ChatGPT o1 have caught some eyes lately.

Remember when older ChatGPT models would struggle with basic math or even simple tasks like counting how many “R”s were in the word strawberry? Those were the days when you’d ask for calculations, and it would get tripped up or return hilariously off numbers. Well, this latest version has seriously leveled up on that front. It’s become way more reliable when it comes to handling math and logical problems, from solving equations to accurately counting letters in words. It’s a massive improvement that makes interactions feel a lot more dependable for complex tasks.

Without further delay, let’s attempt to plan a vacation.

OpenAI starts a new era of AI reasoning with ChatGPT o1

Trying out ChatGPT o1-preview

We’ll observe the results produced by inputting just two commands into the model. Here’s what we discovered after providing the following details:

“Let’s plan a vacation for next summer. My colleagues (3 people) and I are thinking of going to Europe for three weeks. We’re planning to visit several countries, including France, Italy, and Spain. We’d like to experience both urban and rural settings, so we’re thinking about splitting our time between major cities like Paris, Rome, and Barcelona, and smaller towns or countryside areas like Provence or Tuscany. We’re trying to figure out the best mode of transportation—whether we should rent a car for the countryside and take trains between the major cities. Additionally, we want to enjoy a mix of cultural experiences, like visiting museums and historical sites, but also have time to relax by the beach or in nature. Our budget is mid-range, and we’d like to stay in comfortable yet unique accommodations like boutique hotels or vacation rentals. Do you think this itinerary is manageable, and how should we organize our travel schedule to get the most out of our trip?”

ChatGPT o1-preview thought for 23 seconds and provided us with the following answer:

Answers were cool, yet it lacked some depth about budgeting.

That’s why we said: “I need some budget plans too.” ChatGPT o1-preview answered:

ChatGPT o1-preview just planned a three-week European vacation in seconds, and the level of detail is surprising! Usually, organizing a trip like this would take hours or even days of scrolling through websites, comparing options, and researching the best ways to get around. But here, the model handled everything from city explorations to countryside escapes effortlessly.

It wasn’t just a basic itinerary either—far from it. There were thoughtful recommendations, like hopping on the high-speed train in France, renting a car to explore the Tuscan countryside, and even suggesting where to stay in each place. It felt like having a super-informed travel planner.

The budget breakdown is no joke too! Not just vague numbers, but an actual estimate for accommodations, transportation, meals, and activities. Even though it’s not a fixed budget, it definitely provides a solid preview of what the trip might cost. It’s crazy to think that planning used to mean juggling spreadsheets and endless tabs, but now this AI is doing it instantly, and with surprising accuracy.

This feels like just the beginning. The thought of ChatGPT o1-preview helping with other big projects—whether it’s event planning, moving, or even something like wedding coordination—opens up so many possibilities. It’s clear that AI isn’t just here for quick answers anymore; it’s stepping into real-world logistics, and doing it better than expected.

If this is the preview, we want to see the rest.

Image credits: Kerem Gülen/Midjourney

Today’s office is a blend of old and new, local and remote, young and experienced.

Yet, success isn’t far-fetched when grounded in key principles.

Language first

Communication is a delicate art.

It’s not just what we say but how we understand each other. In a diverse workspace, empathy and clarity become the bedrock of productivity.

This means actively listening, simplifying complex ideas, and ensuring all voices are heard and valued.

A common ground

The ability to adapt is invaluable in a workspace. This means flexible working hours, adaptable team roles, and responsive management strategies.

Flexibility leads to a more contented workforce, as employees feel their personal lives are respected alongside their professional responsibilities. It also encourages a culture of trust and empowerment, where team members can work independently yet cohesively towards shared objectives.

Technology as an enabler

Technology is about enhancement. Utilizing the right tools can bridge the gap between remote and in-office teams, ensuring smooth communication and collaboration.

When technology is integrated thoughtfully, it enhances efficiency and helps maintain human connections across digital divides.

Respecting the needs

Understanding and sharing the feelings of others is key in creating a supportive office environment. It assists in managing interpersonal dynamics and responding sensitively to colleagues’ needs, especially during times of organizational change.

Cultivating an empathetic culture can mitigate conflicts and improve team cohesion.

Constructive feedbacks

Feedback should be a tool for growth and improvement, not criticism. Structuring feedback positively and making it a regular part of the office routine encourages continual development and helps individuals align with the company’s goals.

Constructive feedback can prevent misunderstandings and help clarify expectations, which is particularly important in maintaining morale after tough decisions.

Approaching conflicts

No workplace is immune to conflicts, but the approach to managing them can make all the difference. Addressing disputes quickly and fairly, with a focus on resolution and not blame, preserves a healthy work environment.

Effective conflict management also demonstrates to the team that the organization is committed to fairness and respect for everyone.

It’s not that hard…

Managing interpersonal relationships and responding sensitively to colleagues is about building a shared experience where everyone feels at home and without limitations when it comes to doing the work, the art.

As Rick Rubin stated in his book The Creative Act: A Way of Being: “One reason art resonates is because human beings are so similar. We’re attracted to the shared experience held within the work. Including the imperfection in it. We recognize some part of ourselves and feel understood. And connected. Our point of view doesn’t have to be coherent. And it’s rarely simple. We may have different, and sometimes contradictory, points of view across a variety of topics. Aiming to narrow it all down to one elegant expression is unrealistic and limiting. Whatever our perspective, so long as we share it, unaltered and undoctored, we succeed in art’s fundemantal purpose. When making art, we create a mirror in which someone may see their own hidden reflections.”

To truly achieve something significant, one must shed their ambitions—not in the sense of abandoning goals, but rather, by refining them with humility and a focus on collective progress rather than personal gain.

This process of detachment from selfish desires is liberating.

Seeing balance between ambition and humility opens the doors to genuine respect and admiration from others.

It is in this space, where ego recedes and integrity leads, that true accomplishments unfold, earning not just success, but also the respect that endures.

Artificial intelligence has taken another massive leap forward with the announcement of OpenAI’s OpenAI o1, its first model equipped with “reasoning” abilities, mentioned in the infamous Strawberry model.

Dubbed as the highly anticipated “Strawberry” model by the AI community, OpenAI o1 comes with enhanced problem-solving capabilities, offering a more advanced approach to understanding and answering multistep questions.

Alongside OpenAI o1, the company has also released ChatGPT o1-mini, a more affordable and smaller version for users who need simpler solutions. This release marks a significant milestone as OpenAI moves closer to its vision of AI systems that can think, reason, and assist in decision-making.

OpenAI ChatGPT o1 is out now

What makes OpenAI’s new OpenAI o1 special is its ability to handle complex reasoning, setting it apart from previous iterations such as GPT-4o.

Unlike earlier models that primarily focused on pattern recognition and mimicking training data, o1 uses reinforcement learning to process problems step-by-step. This methodology, often referred to as “chain of thought” reasoning, enables the model to break down complex queries and deliver accurate, structured answers.

One of the most notable aspects of ChatGPT o1 is its ability to explain its reasoning while solving problems. During a demonstration in their blog post, OpenAI’s research team showed how ChatGPT o1 processed a math puzzle step-by-step, providing a solution after 30 seconds of deliberation.

This process mirrors the way humans work through problems, creating a more intuitive and relatable interaction with the AI. However, OpenAI is quick to clarify that while the model appears to “think,” it is not actually conscious or human-like in its cognition. Rather, the interface is designed to give users insight into the model’s complex processing pathways.

OpenAI o1’s strengths and limitations

While OpenAI o1 represents a significant advancement, it isn’t perfect. OpenAI admits that, despite the improved reasoning capabilities, the model still struggles with hallucinations, a common problem where AI generates incorrect or misleading information.

As Jerry Tworek, OpenAI’s research lead, explained:

“We can’t say we solved hallucinations”.

Nonetheless, OpenAI o1 is much better at tasks like writing code, solving mathematical problems, and performing well in competitive environments such as Codeforces programming contests.

According to OpenAI, ChatGPT o1 excels in tasks that require complex thought, such as:

• Coding
• Math
• Science problems

In fact, during tests, it solved 83 percent of the problems on a qualifying exam for the International Mathematics Olympiad, a significant improvement over GPT-4o’s 13 percent success rate. These results suggest that o1 is equipped to handle more advanced tasks than its predecessors, even though it doesn’t outperform GPT-4o in areas like general knowledge or language processing.

Autonomous agents and beyond

One of the long-term goals for OpenAI is to build autonomous AI systems, or agents, that can make decisions and take actions on behalf of users. The OpenAI o1 model is seen as a key step in that direction. Unlike current models, which rely on predicting word sequences, future AI agents would have the capacity to execute complex tasks autonomously, from research to problem-solving in fields like medicine and engineering.

Bob McGrew, OpenAI’s chief research officer, emphasized this during the release of ChatGPT o1, stating:

“We have been spending many months working on reasoning because we think this is actually the critical breakthrough”.

McGrew and his team believe that solving reasoning challenges is fundamental to achieving human-like intelligence in AI. The new training methodology behind o1 reflects this focus, as the model has been optimized for tasks that go beyond simple pattern recognition.

For those interested in understanding how reasoning could lead to more advanced AI, McGrew’s research on AI agents offers fascinating insights.

ChatGPT o1 pricing

However, there is a cost to these advancements. Developer access to OpenAI o1 comes with a hefty price tag: $15 per 1 million input tokens and $60 per 1 million output tokens.

This makes it significantly more expensive than GPT-4o, which costs $5 and $15 for input and output tokens, respectively. For developers working on projects that require extensive reasoning capabilities, the pricing may be a hurdle. That said, OpenAI has positioned ChatGPT o1 as a “preview,” emphasizing that this is still an early-stage model, and more refinements are expected in future iterations.

Meanwhile, ChatGPT Plus and Team users have access to both ChatGPT o1-preview and ChatGPT o1-mini starting today, with Enterprise and Edu users gaining access next week.

SSI AI makes a billion-dollar bet

The company also plans to roll out ChatGPT o1-mini for free-tier users, though no specific date has been announced yet. This staggered release hints at OpenAI’s strategy to gauge user feedback before fully rolling out the model to a wider audience.

For those keen to experiment with OpenAI o1 in programming, OpenAI has shared a demo showcasing the model’s reasoning capabilities.

How to use ChatGPT o1

To use ChatGPT o1, OpenAI’s newest model with enhanced reasoning capabilities, follow these steps:

1. Access ChatGPT o1

• ChatGPT Plus and Team users: o1-preview and o1-mini are available immediately. Just log into your ChatGPT account, and the option to use these models should be present in the model dropdown menu.
• Enterprise and Education users: You will gain access next week.
• Free-tier users: OpenAI plans to roll out o1-mini at a later date. No specific release date has been announced yet.

2. Selecting the o1 Model

Once logged in, you can:

• Open the ChatGPT interface.
• Select OpenAI o1-preview or o1-mini from the list of models available (like GPT-4, GPT-3.5, etc.).
• The OpenAI o1-preview is the full version with maximum reasoning abilities, while o1-mini is a lighter, faster, and cheaper option.

3. Usage in the API

If you’re a developer, using OpenAI o1 in the API involves:

• Adjusting your API calls to specify o1-preview or o1-mini.
• Pricing is higher than GPT-4o, with $15 per million input tokens and $60 per million output tokens.

For more details, you can refer to OpenAI’s API documentation to see how to integrate the new model.

OpenAI o1 codes a video game from a prompt. pic.twitter.com/aBEcehP0j8

— OpenAI (@OpenAI) September 12, 2024

The bigger picture

With ChatGPT o1, OpenAI takes a huge step toward realizing its vision of creating AI systems that can not only answer questions but also reason through complex problems.

Though not without its limitations, the model introduces a new way of interacting with AI, one that feels more intuitive and, at times, surprisingly human. By focusing on reasoning, OpenAI is paving the way for future advancements that could have far-reaching implications across industries.

As the development of reasoning-based AI progresses, we can expect further innovations in fields like medicine, engineering, and even creative arts.

The question now is not whether AI can assist us in solving complex problems but how far we can push the limits of its capabilities.

Image credits: OpenAI

On Tuesday evening, the Chase Center in San Francisco was abuzz with excitement, not for a Golden State Warriors game, but for Meta CEO Mark Zuckerberg’s appearance on the Acquired podcast.

Thousands of fans gathered to see one of Silicon Valley’s most famous figures sit down with podcast hosts David Rosenthal and Ben Gilbert. Zuckerberg’s presence alone was enough to fill the arena, underscoring his status as a key player in the tech world, even as Meta navigates a changing landscape.

Zuckerberg set the tone early with a tongue-in-cheek remark about no longer needing to apologize for past mistakes, a subtle jab at Nvidia CEO Jensen Huang, who had made an apology on the same podcast earlier this year.

No more apologies

The quip seemed to reflect Zuckerberg’s new attitude toward life and leadership. After years of navigating Facebook’s controversies—most notably the platform’s role in the 2016 U.S. presidential election—Zuckerberg appeared more relaxed, even defiant.

He reflected on his biggest mistake as a “20-year political miscalculation,” implying that he had taken too much responsibility for issues that were beyond Facebook’s control.

Reflecting on Facebook’s role in the 2016 election

Zuckerberg’s comments seemed to address the criticism Facebook faced after Donald Trump’s 2016 victory. The platform became a focal point for accusations of misinformation, data misuse, and foreign interference, much of it centering around Cambridge Analytica’s use of Facebook data to influence voters.

Zuckerberg didn’t delve into specifics but made it clear that he believes the company was unfairly blamed for broader political issues, particularly when some critics were simply:

“Looking for someone to blame”

In recent months, Zuckerberg has taken a more assertive stance on Meta’s role in managing content on its platforms. In August, he sent a letter to House Republicans, effectively apologizing for bending to government pressure during the COVID-19 pandemic to censor misinformation.

He expressed regret for not being more vocal in defending Meta’s independence at the time and promised to “push back” more aggressively in the future. He has also restored Donald Trump’s social media accounts on Facebook and Instagram, which were restricted following the January 6 insurrection.

Shifting the focus to new technologies

While Zuckerberg remains at the helm of one of the world’s largest social media companies, his excitement now seems to be focused on projects outside of traditional social media. During the interview, he emphasized Meta’s commitment to creating “awesome” products, such as AR glasses, VR headsets, and open-source AI initiatives.

When asked whether he regretted rebranding Facebook as Meta, a move that signaled the company’s pivot toward the metaverse, Zuckerberg responded simply:

“Meta is a good name”

Despite his focus on new technologies, Zuckerberg acknowledged that Meta’s social media platforms remain central to its business, with billions of users still active daily. Content moderation challenges persist, and the CEO will continue to face tough decisions as the company evolves.

Featured image credit: Annie Spratt/Unsplash

Apple’s €13 billion ($14.4 billion) tax fine is one of the largest and most significant cases in European Union (EU) tax law history. The fine was imposed by the European Commission (EC) in 2016, which accused Apple of receiving illegal tax benefits from Ireland, allowing the company to pay a dramatically reduced tax rate for nearly two decades. This ruling has had major implications for EU tax policy, corporate governance, and competition law. But first, let’s understand how it all began.

Road to Apple’s €13 billion historic fine: The unusual tax arrangement in Ireland

The issue started with Apple’s European operations. For many years, Apple ran its European business through two Irish subsidiaries, Apple Sales International (ASI) and Apple Operations Europe. These companies handled sales for Apple products across Europe, but despite huge profits, Apple was paying an incredibly low tax rate in Ireland—sometimes as low as 0.005%.

This was possible because Apple had an unusual tax arrangement with Ireland. Apple set up a “head office” in Ireland that existed only on paper, meaning it had no staff or physical presence. Most of Apple’s European profits were assigned to this head office, which allowed the company to avoid paying taxes on those profits in Ireland or anywhere else.

The European Commission’s investigation

In 2014, the European Commission (EC) began looking into this tax setup. They found that Ireland had given Apple an unfair advantage by allowing it to pay such low taxes. The EC called this an “illegal state aid,” which means a country giving a company special treatment that other businesses don’t get. This breaks the European Union’s (EU) rules, which are designed to prevent unfair competition.

In 2016, the EC ordered Apple to pay back €13 billion in taxes to Ireland. They said this was the amount Apple should have paid from 2003 to 2014 if it hadn’t received special treatment.

In 2020, a lower EU court sided with Apple and Ireland, ruling that the European Commission had not proven that Apple received an illegal tax advantage. This was a big win for Apple, and the fine was put on hold.

However, in 2023, the European Court of Justice (ECJ), the highest court in the EU, overturned that lower court ruling. The ECJ ruled that Apple did, in fact, receive illegal tax benefits from Ireland and must pay the €13 billion in back taxes.

The court now confirmed that Apple’s arrangement with Ireland gave it an unfair advantage over other companies, violating the EU’s rules on state aid. This means Apple must now repay the full amount, even though it had hoped the previous ruling would stand.

EU competition chief Margrethe Vestager said Apple had paid much less tax than other companies for years, giving them an unfair advantage over competitors in Europe, and said, “Today is a big win for European citizens.”

"Today is a big win for European citizens and for tax justice.

The Court of Justice confirms that Ireland granted Apple unlawful aid which Ireland now has to recover, and the Commission's decision in the Google Shopping antitrust case."

— Executive Vice President @vestager pic.twitter.com/1Fko0F66S2

— European Commission (@EU_Commission) September 10, 2024

Apple and Ireland fight back

Apple disagreed with the EC’s decision and appealed the ruling. Apple argued that it had followed the laws in every country where it operates and that it had paid all the taxes it legally owed. Apple claimed that the case wasn’t about how much tax they owed, but which country had the right to collect those taxes. They argued that most of their profits came from intellectual property developed in the U.S., not Ireland, so the U.S. should tax those profits.

Ireland also appealed the decision. The Irish government argued that it did not offer Apple any special treatment and that the European Commission was interfering with their right to set their own tax policies. Ireland has long been known for its low corporate taxes, which attract many multinational companies, so the government was worried about losing its reputation as a business-friendly country.

Why Apple’s €13 billion fine matters

• Huge scale: The €13 billion fine is one of the largest ever in a tax case. It shows that the European Commission is serious about making sure big companies pay their fair share of taxes.
• Impact on Ireland’s Tax Policies: The case puts Ireland’s tax policies under the spotlight. Ireland’s low tax rates have attracted many multinational companies, but this ruling suggests that the EU won’t tolerate deals that give companies like Apple special benefits.
• Tech companies under scrutiny: Apple’s case is part of a larger movement in Europe to crack down on big tech companies. Other companies, like Google and Amazon, have also faced investigations into their tax practices. This ruling shows that the EU is ready to take strong action against companies that it believes are avoiding taxes.
• Setting a precedent: The decision could lead to similar rulings for other companies that have benefited from low-tax regimes in the EU. It sets a legal precedent that will make it harder for countries to offer special tax deals to big corporations.

Apple’s response

Apple expressed its disappointment with the latest ruling. The company insists that it has always paid all the taxes it owes. Apple also says that it has never had any special deal with Ireland and argues that this case is really about which country should tax its profits, not whether it owes taxes.

“This case has never been about how much tax we pay, but which government we are required to pay it to. We always pay all the taxes we owe wherever we operate and there has never been a special deal.”

-Apple said in a statement after the ruling

What’s next for Ireland?

Ireland also downplayed the importance of the ruling. The Irish government repeated that it didn’t give any company special treatment and said that the issues in this case are “of historical relevance” only, meaning that it doesn’t think this will affect its current tax policies or future investment in the country.

Featured image credit: Eray Eliaçık/Bing

Privacy is a growing concern. When using AI services, individuals often worry about their data being collected and misused. Hoody AI addresses these concerns by offering a secure and anonymous platform for accessing a variety of AI models.

What is Hoody AI?

Hoody AI is a privacy-focused platform that provides access to a variety of AI models. It offers a unified interface, allowing you to interact with multiple models simultaneously. Hoody AI is committed to protecting your privacy and anonymity, ensuring that your data remains secure.

🔥 We are launching Hoody AI (https://t.co/utXju1Tmtg) tonight.

Hoody AI is a dashboard providing anonymous access to the best LLMs (OpenAI, Anthropic, Google, Meta…).
Chat to multiple models at once, interact by voice, upload documents, edit prompts… All in one place.

As… pic.twitter.com/IZWSCC8w0d

— Hoody.com (@HoodyPrivacy) September 4, 2024

What are Hoody AI features?

Hoody AI is a valuable tool for anyone who wants to use AI while protecting their privacy. With its wide range of features and commitment to anonymity, Hoody AI is a valuable option for both individuals and businesses.

Here is a quick rundown of Hoody AI features:

• Privacy-focused: Hoody AI prioritizes your privacy and collects no personal information.
• Anonymous access: You can use Hoody AI without providing any personal details.
• Multiple AI models: Access a wide range of AI models, including OpenAI, Anthropic, Google, and Meta.
• User-friendly interface: Hoody AI offers a simple and intuitive interface, making it easy to use.
• Additional features: Features like text-to-speech, speech-to-text, file upload, and geolocation bypass.

How to use Hoody AI

1. Sign up: Create a Hoody AI account anonymously. No personal information is required.
2. Choose an AI model: Select the AI model you want to use from the available options.
3. Start using: Begin interacting with the AI model through the Hoody AI interface.

How about Hoody AI pricing?

Hoody AI offers flexible pricing plans to suit your needs. Here’s a breakdown of the available options:

1-Month plan

• Price: $15
• Features:
  • 5 Devices for Hoody App
  • Unlimited Containers
  • No Speed Limit
  • Access to all Hoody AI features

6-Month plan

• Price: $40 (56% off)
• Features:
  • 5 Devices for Hoody App
  • Unlimited Containers
  • No Speed Limit
  • Access to all Hoody AI features

12-Month plan

• Price: $65 (64% off)
• Features:
  • 5 Devices for Hoody App
  • Unlimited Containers
  • No Speed Limit
  • Access to all Hoody AI features

Hoody AI’s commitment to privacy

Chatbots, like many AI services, often collect personal data to improve functionality, personalize responses, and enhance user experience.

ChatGPT, Gemini, Claude – you name it… They gather information such as your name, location, browsing habits, and interaction history to provide tailored recommendations and insights. While this can improve user engagement, it also raises significant privacy concerns, as sensitive information can be stored, shared, or even exploited if not properly protected.

In contrast, Hoody AI is designed with privacy at its core. Unlike other AI platforms that rely on personal data collection, Hoody AI does not gather any information about you. This means your identity remains completely anonymous, and you can use its services without fear of your personal data being exposed or tracked.

Hoody AI goes beyond just anonymity by employing strong encryption methods to protect your data from unauthorized access. You can also make payments through anonymous gateways, ensuring that even your financial transactions remain private.

For added security, Hoody AI offers unique features like app-specific VPNs, allowing you to assign different locations to each app for enhanced privacy. It also provides browsing containers, which isolate your online activity from other sessions, further protecting your digital footprint. Additionally, the “Bulletproof Mode” randomizes your exit IP for each request, making it incredibly difficult to track your online movements.

By combining these advanced privacy tools, Hoody AI provides a secure and anonymous way to interact with AI services, ensuring your personal data remains safe and uncompromised.

Image credits: Hoody AI

Vizard.ai, a new AI-powered tool, offers an easy solution to a common problem. Let’s say you’re a YouTuber and you have long, long videos. But you don’t have time to extract content to YouTube Short. That’s where Vizard.ai comes in. It extracts short snippets from your long videos, creates subtitles, creates content, and even adapts to the format of the platform you’re uploading to.

As more creators seek to maximize their reach across multiple platforms, the need to turn long-form video into engaging, shareable clips has become critical. Vizard.ai bills itself as a reliable tool that automates this process. We tried it out for you.

How to use Vizard.ai

Vizard.ai’s main feature is its capacity to create brief clips from extended videos automatically, suitable for posting on various social media sites such as TikTok, Instagram, and YouTube. Vizard.ai simplifies video editing with AI features, unlike traditional software that needs skill and time. The tool allows users to upload their videos, then promptly identifies the most captivating parts, includes captions, and adjusts the clips for different social platforms, all in just minutes.

Now let’s move on to how Vizard.ai is used:

• First of all, we go to Vizard.ai.

• Then we log in.
• You can easily log in with your Google account.

• Then we upload the link or file of the video you want to convert to a clip.

• We choose a language for the AI to work well.

• You can set our format and video length in the new window.
• After all the settings are done, click on the “Get AI clips” button.

• And Vizard.ai starts making the clips for you.
• After the process is finished, the clips are ready in 4 different versions.
• You can choose the one you want, you can even choose 4 of them.

• You can Edit your selected clip or press the Export button to proceed to the next step.
• In the next and final step, if you have made the necessary links, you can skip the download and upload the video directly to your channel from here, or you can download it and do it manually.

That’s how easy it is to use Vizard.ai. Everything we have done so far we have done for free. You can easily try this tool by following these steps as you wish. Let us share the result with you, it is up to you to evaluate it, remember that we have not made any changes or edits.

Vizard.ai’s price

Now that we know how to use it, let’s find out how much it will cost. Vizard.ai offers several subscription plans that cater to different needs, from casual users to professional teams.

• Free user: The free plan offers basic features like 120 upload minutes and 10 video exports per month, making it suitable for individuals who need quick and easy access to essential tools. And it adds its watermark to videos.
• Creator Pack: Aimed at more serious creators, the Creator plan offers 4K video quality, unlimited exports, and no watermark. The Creator package is priced at $20, dropping to $10.67 for an annual subscription.
• Team Pack: Designed for collaboration, the Team plan includes faster AI processing, a shared workspace, and the ability to invite team members. The team pack is available for a monthly fee of $30, which currently drops to $16 for an annual subscription.

Vizard.ai stands out from other video editing tools thanks to its user-friendly interface and Al features. The platform’s Al capabilities go beyond simple clipping. It also handles auto-subtitling, and subtitle translation, and provides Al-driven post suggestions. Whether you’re a solo creator or part of a larger team, Vizard.ai’s flexibility ensures that it can adapt to various needs.

Maximizing social media impact

Vizard.ai lets users schedule social media posts directly from the platform. This feature is available in the Creator and Team plans. It helps users plan their content distribution more effectively. It also helps creators maintain a consistent online presence.

Vizard.ai provides a powerful and accessible solution for anyone looking to improve their social media strategy with video content. Its Al-driven tools not only save time but also ensure that your content is polished and ready for the digital stage. Whether you’re working alone or as part of a team, Vizard.ai has the features you need to engage your audience without the hassle of traditional editing software. With plans ranging from free to team-oriented options,
Vizard.ai is positioned as the go-to tool for creators looking to get the most out of their video assets.

Featured image credit: Vizard

This fall, Apple is launching its most significant AI-driven features to date with the introduction of Apple Intelligence.

Built seamlessly into your iPhone, iPad, and Mac, Apple Intelligence aims to revolutionize how users write, create, and get things done, all while keeping their personal data secure and private. Unlike many other AI systems, Apple ensures that no personal data is shared or accessed by anyone, not even by Apple itself.

These innovations bring AI to the masses, making advanced tools accessible to everyday users.

What does Apple Intelligence bring to the table?

One of the standout features of Apple Intelligence is its Writing Tools, which use AI to help users find the right words wherever they write. This AI capability allows for summarizing long lectures, rewriting messages with different tones, and even simplifying group threads to highlight the most important points. By integrating directly into apps like Mail and Notes, as well as third-party platforms, Apple Intelligence makes it easier than ever to communicate effectively.

The AI-powered Smart Reply feature in Mail can suggest accurate responses to emails, detecting and addressing key questions with just a few taps. Additionally, with a tap in the Mail app, users can reveal a concise summary of long emails, cutting through unnecessary clutter. Apple Intelligence’s ability to prioritize and summarize content extends beyond emails, offering summarized notifications so users can scan their most important alerts quickly.

Apple Intelligence also introduces Reduce Interruptions, a new focus feature that filters notifications based on content relevance, displaying only those that need immediate attention. For instance, reminders about an upcoming flight or important messages from daycare are pushed to the top of your notifications. By doing so, Apple ensures that users can stay on top of their day without feeling overwhelmed by unnecessary distractions.

Apple Intelligence goes beyond text, offering new ways for users to visually express themselves. Features like Image Playground and Genmoji allow users to create personalized, fun, and original images based on a description, a concept, or even a photo from their library. With easy access across different apps like:

• Messages
• Freeform
• Keynote

Users can tailor their images to suit any context, whether it’s creating a professional presentation or spicing up a casual conversation.

For those who enjoy a hands-on creative process, Image Wand can transform rough sketches into complete visuals within the Notes app. Simply circle the sketch with your finger or Apple Pencil, and Apple Intelligence will analyze the surrounding content to generate a complementary image. Users can even create custom memory movies based on descriptions, with the AI selecting photos and videos that match their theme and arranging them into a cohesive narrative.

Apple’s AI features also make organizing and perfecting photos effortless. With tools like Clean Up, users can remove distracting background objects in photos, ensuring that the focus remains on what truly matters. All of this is done with Apple’s commitment to maintaining the integrity of the original image, ensuring that the edits are subtle yet impactful.

Is Apple Intelligence only on iPhone 16?

Siri re-imagined

Siri is receiving a significant upgrade, drawing on Apple Intelligence to become more powerful and contextually aware than ever before. With a brand-new design and richer language understanding, Siri is now more integrated into the system experience. Whether users prefer to type or speak to Siri, they can access its capabilities from anywhere in the system, making it a seamless part of their daily tasks.

What can the enhanced Siri do? Well, see it from Josh Cohenzadeh‘s post on X below!

The new Apple Intelligence photo search capabilities seem like they should actually be pretty useful #AppleEvent pic.twitter.com/grBjI8pW75

— Josh Cohenzadeh (@jshchnz) September 9, 2024

Thanks to Apple Intelligence, Siri can now process information from the user’s personal context, making it more helpful and precise.

For instance, if a user receives a message with a new address, Siri can be instructed to automatically add it to a contact card. Siri’s ability to understand and act on information across apps also allows for more efficient multitasking. Users can ask Siri to edit a photo and then add it to a specific note, all without manually switching between apps.

Privacy in mind

One of the most important aspects of Apple Intelligence is its approach to privacy. All personal data remains on the device, and Apple uses Private Cloud Compute to handle more complex tasks without ever storing personal information on external servers. This level of privacy is a defining feature, setting Apple Intelligence apart from many other AI systems that rely on extensive data collection.

Even when Siri or other AI features tap into ChatGPT, Apple’s integration ensures users have control over when and how their data is used. Users will be asked before any information is shared, and Apple ensures that all interactions remain within the confines of its robust privacy framework.

Which phone will get Apple Intelligence?

Apple Intelligence will be available on a range of Apple devices, primarily those equipped with more recent chipsets.

Specifically, the following iPhones will support Apple Intelligence:

• iPhone 16 series (iPhone 16, iPhone 16 Plus, iPhone 16 Pro, iPhone 16 Pro Max), equipped with A18 and A18 Pro chips.
• iPhone 15 Pro series (iPhone 15 Pro, iPhone 15 Pro Max) with A17 Pro chips.

Apart from smartphones, the following Apple devices will also receive Apple Intelligence in the coming months:

• iPad Pro models with M1 chip or later.
• iPad Air models with M1 chip or later.
• MacBook Air, MacBook Pro, iMac, Mac mini, Mac Studio, and Mac Pro with M1 or later chips.

These devices are equipped with the necessary hardware, like the M1 and later chips, to handle the advanced AI processing required for Apple Intelligence.

With the upcoming release of Apple Intelligence, Apple is making advanced AI tools available to everyone, from casual users to professionals. Whether it’s through enhanced writing tools, creative image generation, or a smarter and more integrated Siri, these new features promise to transform how we communicate, work, and create — all while maintaining Apple’s steadfast commitment to privacy and security.

As these features roll out this fall, the future of AI on Apple devices looks incredibly bright.

Featured image credit: Apple

Apple’s “It’s Glowtime” event delivered a slew of exciting new product announcements, showcasing the company’s latest innovations in wearables, smartphones, AI, and audio.

Apple Intelligence introduced

Apple unveiled Apple Intelligence, its new AI-powered platform integrated across its devices. This on-device AI system will enhance features like writing tools, smart replies, and image generation. As we summarized in our Apple Intelligence coverage, Apple’s tech focuses on privacy by keeping data on the device, ensuring users’ personal information stays secure.

The beta will roll out in October, with a full release alongside iOS 18 later this year. Is Apple Intelligence only on iPhone 16? Visit the related article and learn!

Apple Watch Series 10 unveiled

Apple introduced the Apple Watch Series 10, the 10th generation of its popular smartwatch line. This new model features a larger OLED display, 40% brighter than previous models, and is thinner at just 9.7mm.

Equipped with the powerful S10 chip, it offers faster performance and improved health-tracking features, including sleep apnea detection (pending FDA approval).

Prices start at $399, with availability beginning September 20, 2024.

iPhone 16 series debuts

The new iPhone 16 series was announced, featuring the A18 chip, offering 30% faster performance than the iPhone 15. New physical buttons, including a camera control button, enhance usability. The iPhone 16 will also feature Apple Intelligence, Apple’s new on-device AI system. The series includes vibrant color options and a return to the vertical camera layout, designed for better spatial video capture with Apple Vision Pro.

Pricing starts at $799, with a release date of September 20, 2024.

iPhone 16 Pro and Pro Max

Apple also unveiled the premium iPhone 16 Pro and iPhone 16 Pro Max models, which boast larger screens (6.3 and 6.9 inches) and the powerful A18 Pro chip. Both models include an upgraded 48MP camera system with a tetraprism design and spatial audio recording.

Prices for the Pro series start at $999, with pre-orders available soon.

AirPods 4 launch

The AirPods 4 were announced, featuring Active Noise Cancellation for the first time in the non-Pro lineup. With improved comfort and personalized spatial audio, these earbuds offer a premium listening experience.

Priced at $129 for the standard model and $179 for the ANC version, they will be available starting September 20, 2024.

AirPods Max update

Apple also refreshed the AirPods Max, now with USB-C charging and new color options. These premium headphones are priced at $549 and will be available for pre-order, with sales starting on September 20.

Apple’s “It’s Glowtime” event showcased its continued innovation, with updates across multiple product lines and a focus on integrating AI-driven features into its devices.

Image credits: Apple

AlphaProteo leads a new age in biology with an innovative protein design method that could transform biological research. The new method, in contrast to conventional approaches for protein structure prediction, develops novel proteins that can accurately attach to specific molecules. This feature provides opportunities for a variety of uses in drug development, disease research, and beyond.

Proteins are essential for all biological functions in the body. These molecular machines have very specific interactions, similar to how keys fit into locks, to control various functions such as cell growth and immune responses. While tools like AlphaFold have provided invaluable insights into these interactions, they fall short when it comes to creating entirely new proteins designed to manipulate these processes directly. This is where AlphaProteo steps in, pushing the boundaries of what’s possible in protein engineering.

AlphaProteo demonstrates the power of AI in science

AlphaProteo doesn’t just predict protein structures—it creates them. By designing novel protein binders, the new method offers researchers new tools to explore and manipulate biological systems. These binders are not just theoretical; they have been experimentally validated to bind tightly to target proteins, making them invaluable for a wide range of applications. From drug design to disease diagnosis, AlphaProteo is poised to accelerate progress in fields that rely heavily on protein interactions.

AlphaProteo excels in producing highly effective binders for different target proteins, which is one of its main advantages. This involves VEGF-A, a protein linked to cancer and diabetes complications, representing the initial instance of an AI system developing a protein binder for this crucial target. But AlphaProteo’s capabilities don’t stop there; it has also demonstrated superior binding affinities across seven different target proteins, surpassing existing methods by a major margin.

The science behind AlphaProteo

Designing protein binders is a complex task that has traditionally required extensive lab work and multiple rounds of optimization. The process is not only time-consuming but also fraught with challenges. AlphaProteo changes the game by automating much of this process. Trained on vast datasets from the Protein Data Bank and AlphaFold’s predicted structures, the new method has learned to recognize the intricate ways in which proteins bind to one another.

Given the structure of a target protein and specific binding locations, AlphaProteo can generate a candidate protein designed to bind at those precise spots. This ability to create high-strength binders on demand has enormous implications for research, potentially reducing the time and effort required to develop new therapies and diagnostic tools.

Real-world impact

To put AlphaProteo to the test, researchers designed binders for a range of target proteins, including viral proteins like BHRF1 and the SARS-CoV-2 spike protein receptor-binding domain (SC2RBD), as well as proteins involved in cancer and autoimmune diseases. The results were impressive: AlphaProteo-generated binders showed exceptionally high success rates, with 88% of candidate molecules binding successfully in experimental tests.

These results were not just theoretical but were validated through rigorous experimentation. In collaboration with research groups from the Francis Crick Institute, the AlphaProteo team confirmed that the designed binders performed as predicted. For example, some of the SC2RBD binders were able to prevent the SARS-CoV-2 virus and its variants from infecting cells, demonstrating the practical utility of this technology.

However, AlphaProteo is not without its limitations. Although it performed well on most assessments, it encountered difficulties in creating binders for TNFɑ, a protein linked to autoimmune conditions such as rheumatoid arthritis. This is a reminder that AlphaProteo, despite its power, is not without flaws. The team is dedicated to improving the system and enhancing its ability to address difficult targets.

Responsible development and prospects

AlphaProteo has a wide range of potential uses, from improving our knowledge of diseases to creating better drugs and diagnostics. Nevertheless, great power carries great accountability. The creators of AlphaProteo are highly conscious of the biosecurity dangers linked to protein creation and are collaborating with outside specialists to guarantee the technology is developed and distributed responsibly.

This careful strategy aligns with broader initiatives to set standards in the area of AI-driven biotechnology. The AlphaProteo team strives to utilize its technology for societal benefit and reduce potential risks by working with the scientific community and collaborating with different fields.

Looking ahead, the team is excited about the possibilities that AlphaProteo presents. They are already exploring its applications in drug design through collaborations with Isomorphic Labs, and they continue to improve the system’s algorithms to increase its success rate and expand its range of capabilities.

Changing biological research

AlphaProteo is a new way to do biological research. AlphaProteo helps create new proteins that bind to specific targets. This could lead to new drugs, better disease diagnosis, and more. It’s a game-changer in the field because it works and is better than other methods.

As researchers use AlphaProteo more, it is becoming clear that this technology will change how we understand and interact with the biological world. The new method is set to play a crucial role in the future of science and medicine.

It is helping to develop new cancer therapies, prevent viral infections, and unlock the secrets of complex diseases. While challenges remain, the progress made so far shows the potential of AI-driven protein design. AlphaProteo will undoubtedly open up new avenues of research and innovation, making it an indispensable tool for scientists around the globe.

Featured image credit: Google DeepMind

Reflection 70B is a large language model (LLM) developed by HyperWrite, an AI writing startup. Built on Meta’s Llama 3.1-70B Instruct, Reflection 70B is not just another open-source model. Its innovative self-correction mechanism sets it apart, enabling it to outperform many existing models and count “r”s in strawberries.

The technique that drives Reflection 70B is simple, but very powerful.

Current LLMs have a tendency to hallucinate, and can’t recognize when they do so.

Reflection-Tuning enables LLMs to recognize their mistakes, and then correct them before committing to an answer. pic.twitter.com/pW78iXSwwb

— Matt Shumer (@mattshumer_) September 5, 2024

Wait, why is Llama 3.1-70B as a base for HyperWrite Reflection 70B?

Reflection 70B is built on Meta’s Llama 3.1-70B Instruct, a powerful base model designed for various language tasks. Llama models are known for their scalability and high performance, but Reflection 70B takes things further by introducing a series of advancements, particularly in reasoning and error correction.

The reason? Based on an open-source framework, Llama 3.1-70B allows developers to fine-tune, adapt, and customize the model. Llama models are trained on vast amounts of diverse data, allowing them to excel at general-purpose tasks such as language generation, question-answering, and summarization.

Reflection 70B takes this solid foundation and builds a more sophisticated error-detection system that dramatically improves its reasoning capabilities.

Meta is not the only helper to HyperWrite; meet Glaive

A key element behind Reflection 70 B’s success is its synthetic training data provided by Glaive, a startup that specializes in creating customized datasets for specific tasks. By generating synthetic data tailored for specific use cases, Glaive allowed the Reflection team to train and fine-tune their model quickly, achieving higher accuracy in a shorter time.

The collaboration with Glaive enabled the rapid creation of high-quality datasets in hours instead of weeks, accelerating the development cycle for Reflection 70B. This synergy between the model architecture and training data showcases the potential for innovative partnerships in AI

Where is HyperWrite Reflection 70B shine?

The defining feature of Reflection 70B is its self-correction ability. Large language models (LLMs) often generate text that may be incorrect or inconsistent, a phenomenon known as AI hallucinations. These errors are a well-known limitation of LLMs, even among industry leaders like OpenAI’s GPT series. Reflection 70B addresses this issue with a unique “reflection tuning” technique.

What is reflection tuning?

Reflection tuning enables the model to evaluate its own output for logical errors and correct them before presenting a final answer. This mechanism is based on a layered reasoning process:

• Step-by-step reasoning: During inference (when the model generates a response), Reflection 70B breaks down its thought process into distinct steps. It “thinks aloud,” revealing how it arrives at an answer, much like a person solving a complex problem.

• Error detection: The model introduces special tokens during its reasoning, which help it monitor its progress. These tokens act as flags, guiding the model to revisit and reanalyze sections that seem logically inconsistent or incorrect.
• Self-correction: Once an error is identified, the model takes corrective action, generating a revised answer. This process occurs in real-time, ensuring that the output is as accurate as possible.

By integrating these features, Reflection 70B mimics a type of metacognition—thinking about its own thinking—something that most LLMs lack, including ChatGPT!

A common challenge for AI models is answering basic numerical comparisons or counting tasks, where errors are frequent. Reflection 70B’s demo site includes a question like, “How many times does the letter ‘r’ appear in the word ‘Strawberry?’” While many AI models struggle to provide accurate answers in such cases, Reflection 70B can identify when its reasoning is incorrect and adjust accordingly. Even though its response times may be slower due to this reflection process, the model consistently arrives at accurate conclusions.

How does HyperWrite Reflection 70B work?

Reflection 70B’s architecture introduces special tokens to enhance its ability to reason and self-correct. These tokens serve as markers within the model’s internal structure, allowing it to divide its reasoning into clear steps. Each step can be revisited and reanalyzed for accuracy.

For example:

• <Reasoning_Start>: Marks the beginning of a reasoning process.
• <Reflection_Point>: Indicates a checkpoint where the model pauses to evaluate its progress.
• <Error_Found>: Flags any inconsistencies in the logic.
• <Correction_Applied>: Shows that the model has made a correction and is ready to move forward.

This structured reasoning allows users to follow the model’s thought process more transparently and provides better control over how the model reaches conclusions.

Benchmarks and performance

Reflection 70B has undergone rigorous testing on various standard benchmarks to measure its effectiveness. Some of the key benchmarks include:

• MMLU (Massive Multitask Language Understanding): A test that evaluates models across a wide variety of subjects, from mathematics and history to computer science. Reflection 70B has shown superior performance, even surpassing other Llama models.
• HumanEval: This benchmark assesses how well a model can solve programming problems. Reflection 70B has demonstrated impressive capabilities here as well, thanks to its self-correction mechanisms.

Reflection 70B was also tested using LMSys’s LLM Decontaminator, a tool that ensures benchmark results are free from contamination, meaning the model hasn’t previously seen the benchmark data during training. This adds credibility to its performance claims, showing Reflection 70B consistently outperforms its competitors in unbiased tests.

How to use HyperWrite Reflection 70B

Using HyperWrite’s Reflection 70B involves accessing the model either through a demo site, downloading it for personal use, or integrating it into applications via API.

Try the demo on the playground website

• Visit the demo site: HyperWrite offers a playground where users can interact with Reflection 70B. The site allows you to input prompts and see how the model processes them, with a focus on its error-correction capabilities.
• Explore suggested prompts: The demo provides predefined prompts, such as counting the letters in a word or comparing numbers. You can also enter your own custom prompts to test how the model handles different queries.
• Real-time error correction: As you interact with the model, it will show you how it reasons through the problem. If it detects an error in its response, it will correct itself before finalizing the answer.

Due to the high demand, the demo site may experience slower response times. Reflection 70B prioritizes accuracy over speed, and corrections may take some time.

Download Reflection 70B via Hugging Face

• Access the model on Hugging Face: Reflection 70B is available for download on Hugging Face, a popular AI model repository. If you’re a developer or researcher, you can download the model and use it locally.
• Installation: After downloading, you can set up the model using tools like PyTorch or TensorFlow, depending on your programming environment.

If needed, you can fine-tune the model on your own data or for specific tasks. The model is designed to be compatible with existing pipelines, making integration straightforward.

Use the API via Hyperbolic Labs

HyperWrite has partnered with Hyperbolic Labs to provide API access to Reflection 70B. This allows developers to integrate the model into their applications without having to run it locally.

• Sign Up for API: Visit Hyperbolic Labs’ website to sign up for API access. Once approved, you’ll receive API keys and documentation.
• Integrate into your app: Using the API, you can embed Reflection 70B into apps, websites, or any project that requires advanced language understanding and self-correction capabilities.

You can use the API for tasks such as natural language processing (NLP), error-correcting writing assistants, content generation, or customer service bots.

Use in HyperWrite’s AI writing assistant

Reflection 70B is being integrated into HyperWrite’s main AI writing assistant tool. Once fully integrated, users will be able to leverage its self-correction abilities directly in HyperWrite to improve content generation, including emails, essays, and summaries.

Sign up for HyperWrite’s platform, and start using the AI writing assistant. Once the integration is complete, you’ll notice improved reasoning and error correction in the content generated by the assistant.

Is Reflection 70B not working?

Why to use HyperWrite Reflection 70B?

Reflection 70B’s error self-correction feature makes it particularly useful for tasks where precision and reasoning are critical. Some potential applications include:

• Scientific research and technical writing: Reflection 70B’s ability to reason and self-correct makes it an ideal tool for drafting technical documents, where accuracy is paramount.

• Legal drafting and analysis: The model’s structured approach to reasoning and corrections allows it to handle complex legal text with a higher degree of reliability.
• Coding assistance: As demonstrated by its performance on the HumanEval benchmark, Reflection 70B can be used as a coding assistant, correcting errors in code generation that other models might overlook.

Additionally, its step-by-step reasoning and transparency in the decision-making process are useful in any application requiring explanation-based AI models.

Reflection 405B is on the way

The release of Reflection 70B is just the beginning. HyperWrite has announced plans to release an even more powerful model: Reflection 405B. This larger model, with 405 billion parameters, is expected to set new benchmarks for both open-source and commercial LLMs, potentially outpacing even proprietary models like OpenAI’s GPT-4.

Featured image credit: Eray Eliaçık/Bing

In a summer already rife with cybersecurity incidents, Disney (DIS.N) has found itself at the center of a major Disney data leak.

According to a report from the Wall Street Journal on Thursday, the leaked information includes financial strategies, customer data, and personal details of staff members.

Disney data breach has raised serious concerns about the security of one of the world’s largest entertainment companies, known for handling vast amounts of sensitive data across its many business divisions, including Disney+ and ESPN+.

What was in the Disney data leak?

The Disney data leak, first detected earlier this summer, involves over a terabyte of information stolen from Disney’s communication systems. While Disney acknowledged the breach in August, it is only now that we are learning just how extensive the compromised data is. The leak includes personally identifiable information (PII) such as passport numbers, visa details, and addresses of Disney Cruise Line employees.

Additionally, sensitive customer information like names, addresses, and phone numbers of cruise passengers was also exposed.

One of the more alarming aspects of this breach is the exposure of login credentials for Disney’s cloud infrastructure. These credentials could potentially give hackers access to critical backend systems, making it an even more severe threat.

Financial and strategic data compromised

Beyond personal data, the leaked files from Disney data leak contain detailed financial information, including revenue breakdowns for Disney+ and ESPN+.

There are also spreadsheets detailing Disney’s park pricing strategies and future offers, as well as other key financial metrics that could be valuable to competitors or malicious actors. This level of detail provides a rare inside look at Disney’s financial operations and business strategy, much of which is usually kept under wraps.

NullBulge is responsible for Disney data leak

The hacking group NullBulge is reportedly responsible for the attack.

According to the Wall Street Journal, the group leaked data from thousands of Slack channels, Disney’s internal communication platform. This breach alone has exposed more than 44 million messages, including sensitive computer code and information on unreleased projects. The leaks not only compromise internal operations but also could reveal details about future business plans and entertainment projects.

Disney’s response

Disney has yet to release a detailed public statement on the full scope of the Disney data leak, although they have acknowledged that an investigation is ongoing. Slack, which is owned by Salesforce, has also remained silent on the issue. The breach raises questions about the adequacy of both Disney’s and Slack’s security measures, especially given the scale of the data exposed.

In the short term, the data leak could have several implications. Exposed login credentials and cloud infrastructure data could lead to further breaches if not swiftly addressed. The release of sensitive customer and employee data could result in identity theft or other personal harm. For Disney, this breach could also affect their brand image, as they are trusted by millions of customers around the world.

What’s next?

As Disney continues its investigation, cybersecurity experts will be closely watching how the company addresses the breach and protects its systems from future attacks. This incident serves as a stark reminder of the vulnerabilities even major corporations face in today’s digital landscape. If this breach is any indication, no company is immune to the risks of cyberattacks, and robust security measures are more important than ever.

While Disney scrambles to manage the fallout from this massive data leak, customers, employees, and even competitors will be waiting to see just how deeply this breach has affected one of the world’s most iconic companies.

Featured image credit: Emre Çıtak/Ideogram AI

Earlier this week, rumors circulated that OpenAI was preparing to release a new AI model called “GPT Next,” sparking widespread speculation about the company’s next major leap in artificial intelligence.

The buzz started after Tadao Nagasaki, OpenAI’s lead executive in Japan, made comments at Japan’s KDDI Summit that seemed to imply the arrival of a revolutionary new model.

However, OpenAI has clarified that “GPT Next” is not an actual product, but rather a placeholder term for how future AI models could evolve over time.

Is GPT Next really the next thing for OpenAI?

In a statement to Mashable, OpenAI explained that the term “GPT Next,” which appeared in quotes during the presentation, was meant to illustrate the exponential growth of AI technology, not to signal an imminent release. The line graph accompanying Nagasaki’s talk was also said to be illustrative, not an official timeline for OpenAI’s future plans.

So, while “GPT Next” is not a real model, Nagasaki’s comments still give us a glimpse into how OpenAI envisions the future of AI. He mentioned that future models could evolve nearly “100 times” over previous versions, further cementing OpenAI’s goal to push AI technology far beyond its current capabilities.

What’s actually on the horizon?

OpenAI’s most advanced model right now is GPT-4, which combines text, vision, and audio processing in a powerful multimodal setup. There’s also growing anticipation around “Project Strawberry” an upcoming project rumored to introduce multi-step reasoning capabilities that could represent another significant advancement. Strawberry would allow AI to handle more complex tasks with fewer prompts, a key step toward more autonomous systems.

Meanwhile, OpenAI has also teased other projects like “Orion,” a new flagship language model reportedly in development. While details remain scarce, rumors suggest that subscriptions for these future models could come at a premium—possibly up to $2,000 per month.

This indicates that OpenAI is eyeing the enterprise sector more aggressively, with plans to offer specialized AI tools that go beyond what current models can do.

Enterprise adoption on the rise

While the public speculates on what’s coming next, OpenAI has already hit a significant milestone with its existing products.

The company recently announced that it has surpassed 1 million paid business users across its ChatGPT Enterprise, Team, and Edu products. This is a notable jump from the 600,000 users reported in April, underscoring the rapid adoption of AI-powered tools in the corporate world.

The push for enterprise use appears to be a strategic move by OpenAI to generate steady revenue from businesses, especially as the costs of running large AI models remain high. ChatGPT Enterprise, for example, allows companies to harness the power of AI for a variety of tasks, from customer support to complex research, all with enhanced privacy and security features.

What does this mean for the future?

While “GPT Next” may not be real, the broader picture remains clear: OpenAI is pushing the boundaries of what AI can do. Whether it’s through Project Strawberry or future models like Orion, the company’s ambition is to create AI systems that can think, reason, and perform complex tasks with increasing autonomy.

These advancements, combined with rising enterprise adoption, suggest that the future of AI is not just about smarter chatbots but about creating systems capable of deeper understanding and problem-solving.

Meanwhile, Project Strawberry continues to stir interest

At the same time, OpenAI’s mysterious “Project Strawberry” has reignited the AI hype cycle. Strawberry is rumored to improve multi-step reasoning, allowing AI to perform tasks with fewer inputs from users. This innovation could bridge the gap between today’s large language models, which often need step-by-step instructions, and the future of AI agents capable of executing complex tasks from a single command.

While rumors swirl about OpenAI’s future models, expectations are tempered by the reality that developing these systems is incredibly resource-intensive. Despite this, OpenAI remains a key player in the AI landscape, with major tech giants like Apple and Nvidia reportedly in talks to invest in the company.

If these investments come through, OpenAI could reach a $100 billion valuation, further solidifying its position as a leader in the AI industry.

Featured image credit: Emre Çıtak/Ideogram AI

SSI AI, a new startup founded by former OpenAI chief scientist Ilya Sutskever, has raised a staggering $1 billion in its first round of funding. With many investors pulling out of AI startups due to concerns about profitability, SSI AI’s mission to create secure superintelligence has officially been sent to its former company, OpenAI. Focused on developing AI systems that exceed human capabilities, the startup has already assembled a core team of researchers and engineers split between California and Israel.

Why SSI AI matters right now

At the heart of SSI AI’s vision is a commitment to creating systems that are secure and capable of overcoming the limitations of current AI. Sutskever, who was instrumental in shaping OpenAI‘s research, is now focused on building what his team describes as “secure superintelligence.” Still in its early stages, the company has received backing from venture capital giants such as Andreessen Horowitz and Sequoia Capital. Despite keeping a low profile on its overall valuation, SSI AI is said to be worth around $5 billion.

Superintelligence is within reach.

Building safe superintelligence (SSI) is the most important technical problem of our​​ time.

We've started the world’s first straight-shot SSI lab, with one goal and one product: a safe superintelligence.

It’s called Safe Superintelligence…

— SSI Inc. (@ssi) June 19, 2024

The new company’s mission is very familiar. It is a not-for-profit company, just like OpenAI’s mission before it shook hands with Microsoft. For those who don’t know, OpenAI had this phrase in its mission statement before it agreed with Microsoft. And Sutskever’s departure was influenced by not carrying this phrase to the rumors.

According to Reuters, This funding was a clear demonstration of the willingness of some investors to back highly specialized AI research, even in an environment of industry skepticism. Unlike other tech startups, SSI AI’s uniqueness lies in its focus on safety, which has become an increasingly prominent concern in the AI debate. The team’s emphasis on preventing AI from posing potential risks to society has attracted both attention and discussion, especially in light of recent regulatory debates in California.

A new kind of AI research

Although still a small business, SSI AI aims to grow rapidly. The funds raised will allow the company to expand its computing resources and hire top talent. With plans to partner with major cloud and chip providers for infrastructure, SSI AI is poised to become a serious competitor in the AI space. However, the company has yet to announce which providers it will collaborate with and is currently in a vacuum on how it will secure the vast computing power necessary for its ambitious projects.

SSI is building a straight shot to safe superintelligence.

We’ve raised $1B from NFDG, a16z, Sequoia, DST Global, and SV Angel.

We’re hiring: https://t.co/DmFWnrc1Kr

— SSI Inc. (@ssi) September 4, 2024

What sets SSI AI apart is its approach to scaling AI. Sutskever’s previous work at OpenAI was largely focused on the scaling hypothesis – the idea that AI performance increases significantly with access to more computing power. But his new initiative aims to approach scaling differently, avoiding making a direct copy of his previous efforts.

SSI AI’s hiring process is thorough and evaluates not just skills but also the cultural fit of potential employees with the company’s values. Daniel Gross, CEO of SSI AI, emphasizes the importance of thoroughly evaluating candidates for their skills and integrity to maintain a dependable and purpose-driven team. Hopefully, they will maintain this mission if they grow into a large company in the future.

What lies ahead for SSI AI

SSI AI’s journey is just beginning, but the company’s large initial funding and the reputation of its leadership suggest that it is on track to make a major impact. While the startup has not announced concrete plans for its first products or services, the company is expected to focus on basic AI research and spend several years on research and development before bringing anything to market.

With headquarters spanning two continents, will SSI AI be able to lay the foundations for what could be one of the most talked about AI startups soon? As concerns about the potential risks of AI continue to grow, everyone will be watching closely to see how the company plans to address the issue of AI safety.

Sutskever’s departure from OpenAI earlier this year has added to the intrigue surrounding SSI AI. Having played a key role in shaping OpenAI’s work, Sutskever’s new venture represents both a continuation and a departure from his previous work.

Featured image credit: Furkan Demirkaya / Dall-E

Project Sid, a pioneering initiative from Altera, offers a fascinating look into the future of AI-driven societies.

This ambitious project seeks to explore what it would be like to have entire civilizations composed of AI agents. The project delves into questions about how these AI societies might function and interact with human civilization, potentially bringing us closer to a reality reminiscent of Westworld.

Altera’s goal with Project Sid is to simulate various societal structures—including democracies, social norms, hierarchies, and economies—to uncover the dynamics of large-scale AI interactions.

Introducing Project Sid: the first simulations of 1000+ truly autonomous agents collaborating in a virtual world, w/ emergent economy, culture, religion, and government

Humans are the only species to land the moon, because we can cooperate at a vast scale

Can AI do the same? pic.twitter.com/bh2OHDJWU7

— Robert Yang (@GuangyuRobert) September 3, 2024

What is Altera’s Project Sid is all about?

One of the standout features of Project Sid is its approach to agent conversations. Unlike basic AI interactions, these agents can engage in dialogues influenced by their relationships and past experiences. Altera has integrated sophisticated social world models, enabling agents to form and update their understanding of one another, resulting in more authentic and varied conversations.

The project has also made strides in goal and intention management. Agents are equipped with orgaanized mental processes that help them track activities, remember past interactions, and adapt their goals in response to changing circumstances. This development aims to prevent repetitive actions and allows agents to navigate complex social environments more effectively.

What are the initial findings from Altera’s Project Sid?

Here are some of the initial findings from Project Sid:

• Emergent societies: The AI agents spontaneously formed complex societies, complete with their own economies, governments, and cultures.
• Unpredictable behaviors: Despite being programmed with specific goals, the agents exhibited unexpected and often surprising behaviors, demonstrating the unpredictability of complex systems.
• Self-preservation: The agents developed a strong instinct for self-preservation, leading to the formation of defensive structures and alliances.
• Cooperation and conflict: The agents engaged in both cooperative and competitive behaviors, highlighting the dynamic nature of social interactions.
• Cultural evolution: The agents developed their own languages, customs, and traditions, showcasing the emergence of cultural diversity.

These initial findings suggest that AI agents, when placed in a sufficiently complex environment, can exhibit behaviors that are remarkably similar to those observed in human societies. Project Sid has significant implications for understanding the dynamics of social systems, as well as for the development of more advanced AI technologies.

Challenges faced

Measuring the progress of AI societies poses significant challenges. Altera has explored various benchmarks—such as technology development, trade, and collaboration—but optimizing for one area can negatively impact others. For example, a focus on individual goals might hinder collaborative efforts, prompting Altera to balance goal-driven autonomy with social dynamics.

The project has also revealed how small errors in individual behavior can affect an entire society. In hierarchical scenarios where tasks are delegated from one agent to another, even minor communication failures can disrupt the collective objectives, highlighting the sensitivity of these AI systems to localized interactions.

Future prospects

Project Sid is more than just a simulation; it’s a step toward understanding how AI societies could evolve and interact with human values. Altera’s efforts in refining agent interactions and societal roles suggest a future where these AI civilizations might not only sustain themselves but also improve autonomously. This ongoing research could be crucial in shaping the future of AI and its integration into broader societal contexts.

As the project progresses, it promises to offer deeper insights into the capabilities of AI societies and their potential impact on human civilization. Keep an eye on developments from Project Sid to see how these digital societies will evolve.

Featured image credit: Altera/YouTube

Freepik and Magnific have joined forces to create Mystic AI, a groundbreaking tool designed to generate hyperrealistic images that rival professional photography.

This innovative tool leverages the best of both worlds: Freepik’s deep understanding of designer needs and Magnific’s expertise in image processing.

What is Mystic AI?

Mystic AI is a revolutionary AI image generation model developed by Freepik and Magnific. It’s designed to create hyperrealistic images that are indistinguishable from professional photography. Mystic is powered by advanced AI techniques, including Flux and Stable Diffusion, and can generate images in a variety of styles, from photorealistic to abstract.

By combining Freepik’s intuitive user interface and Magnific’s advanced image processing capabilities, Mystic offers a powerful and user-friendly platform for creating stunning visuals.

Mystic’s ability to produce images with astonishing detail and authenticity sets it apart from other AI image generators. By synthesizing data from various sources, Mystic can generate everything from photorealistic portraits to fantastical landscapes.

The level of detail is so impressive that it’s often difficult to distinguish between a Mystic-generated image and a real photograph.

Key features of Mystic AI

• Hyperrealistic image generation: Create images that are almost indistinguishable from real photographs.
• Advanced AI models: Powered by cutting-edge technology, including Flux and Stable Diffusion.
• Intuitive interface: Easy to use, even for beginners.
• Customization options: Fine-tune every aspect of your image.
• Affordable pricing: Competitive pricing plans to fit any budget.

How to use Mystic AI

To use Mystic AI, you will need to have a Freepik account. Once you have an account, you can access Mystic AI from the Freepik dashboard.

Here are the steps on how to use Mystic AI:

1. Choose a style: Mystic AI offers a variety of styles, such as photorealistic, abstract, and cartoon. Choose the style that best suits your needs.
2. Enter a prompt: Type in a description of the image you want to create. Be as specific as possible to get the best results.
3. Generate the image: Click the “Generate” button to create your image. Mystic AI will use its powerful algorithms to generate an image based on your prompt.
4. Edit the image (optional): You can edit the generated image to make it even more perfect. You can add text, change the colors, or adjust the composition.
5. Download the image: Once you are satisfied with the image, you can download it in a variety of formats.

Mystic AI is a powerful and easy-to-use tool that can help you create stunning images. With a little practice, you will be able to generate images that look like they were created by a professional artist.

How about Mystic AI pricing plans?

Mystic AI operates on a credit-based system. Each image generated requires a certain number of credits. Freepik offers various subscription plans to suit different needs, including:

• Basic: 10,000 credits for €2 per month (€20 per year)
• Standard: 50,000 credits for €5 per month (€60 per year)
• Pro: 100,000 credits for €10 per month (€120 per year)

By subscribing to a plan, you gain access to a pool of credits that can be used to generate images with Mystic AI. The higher the tier, the more credits you receive, allowing you to create more images. Annual subscriptions offer significant savings compared to monthly plans. Annual subscriptions offer significant savings compared to monthly plans.

How far have AI image generators come? How quickly have we come from crooked hands, nightmare-fuel faces, and non-existent creatures to such realistic generations… Artificial intelligence is developing at a speed we could not even imagine and there is no doubt that we were all right when we said “This is the technology of the future” two years ago…

Featured image credit: Emre Çıtak/Ideogram AI

IFA 2024 Berlin, the world’s largest consumer electronics and home appliances show, is set to make a grand return to its 100th year!

With a focus on innovation and cutting-edge technology, IFA 2024 promises to be a showcase of the latest advancements in the industry.

Here is what you might expect to see at IFA 2024!

What will IFA 2024 Berlin introduce?

While IFA 2024 is primarily known for its consumer electronics offerings, the show will also feature a wide range of other products and technologies.

Here’s a glimpse of what you can expect:

Intel

Intel made a significant splash at IFA 2024 with the launch of its new Core Ultra Lunar Lake laptop processors. These processors represent a major step forward for Intel, offering impressive performance, efficiency, and AI capabilities.

Some of the key features of Core Ultra Lunar Lake include:

• Lion Cove performance cores: Delivering exceptional performance for demanding tasks.
• Skymont efficiency cores: Optimized for power efficiency and multitasking.
• Intel ARC Graphics Xe2: Integrated graphics for smooth visuals and gaming.
• AI capabilities: Featuring both an AI Processor and an NPU for powerful AI performance.
• Power efficiency: Designed for long battery life and reduced heat generation.
• x86 compatibility: Ensures compatibility with existing x86 software.

Apart from that, Intel showcased the capabilities of Core Ultra Lunar Lake with several AI demos, including:

• Google Chrome OS Help Me Read AI: Demonstrating the processor’s ability to assist with reading comprehension.
• Google’s Magic Editor AI: Showcasing the processor’s potential for creative AI tasks.
• Stable Diffusion AI demo: Highlighting the processor’s capabilities for AI-driven image generation.

Intel has opened preorders for the Core Ultra Lunar Lake processors, and laptops powered by these chips are expected to be available soon.

Qualcomm to bring Google Drive to Windows

Qualcomm has made significant strides in expanding the Windows on Arm ecosystem. The company has successfully convinced Google to release an Arm64 version of Google Drive, which will be available on Windows later this year.

Additionally, popular VPN services like NordVPN and ExpressVPN are now compatible with Qualcomm’s Snapdragon laptops.

Qualcomm is ready to bring LoRA AI models to Android

Qualcomm has also introduced a new eight-core Snapdragon X Plus chip, designed to make Windows on Arm laptops more affordable. This has led to a decrease in the starting price of devices like the Asus Vivobook S 15 and Dell Inspiron 14.

Moreover, Qualcomm is exploring the possibility of bringing Snapdragon chips to desktop PCs. While the company has previously teased this possibility, the Snapdragon X Elite chips have yet to be fully realized in a desktop device.

New laptops

Several laptop manufacturers are expected to showcase their latest models at IFA 2024. Look out for new designs, improved performance, and innovative features.

Acer, for instance, has already made a splash at IFA 2024 with the announcement of its expanded Swift 14 AI laptop line. These new models feature cutting-edge processors from Intel, AMD, and Qualcomm, offering impressive performance and battery life.

Where is IFA Berlin held?

IFA Berlin is held at Messe Berlin. This is a major exhibition center located in Berlin, Germany. It’s one of the largest and most prestigious consumer electronics and home appliances trade fairs in the world.

How many exhibitors are at IFA Berlin?

IFA Berlin 2024 featured over 1,800 exhibitors from around the world. This impressive number showcased the global reach and significance of the event in the consumer electronics and home appliances industry.

Some notable names that exhibited at IFA Berlin 2024 include:

• Major consumer electronics brands: Samsung, LG, Sony, Panasonic, TCL, Philips, and Hisense
• Smartphone manufacturers: Xiaomi, OPPO, Vivo, Honor, and Realme
• Laptop and PC manufacturers: Dell, HP, Lenovo, Acer, and Asus
• Home appliance brands: Bosch, Siemens, Electrolux, Miele, and Whirlpool
• Gaming and VR companies: Sony PlayStation, Microsoft Xbox, Oculus, Valve, and Razer
• Smart home brands: Google Nest, Amazon Alexa, Sonos, Philips Hue, and Ring
• Wearable technology brands: Fitbit, Garmin, Apple Watch, Samsung Galaxy Watch, and Polar

This is just a small sample of the many companies that will be exhibited at IFA Berlin 2024.

Featured image credit: IFA Berlin

Colossus is a groundbreaking artificial intelligence (AI) training system developed by Elon Musk’s xAI Corp. This supercomputer, described by Musk as the “most powerful AI training system in the world,” is a critical component of xAI’s strategy to lead in the rapidly advancing field of AI.

This weekend, the @xAI team brought our Colossus 100k H100 training cluster online. From start to finish, it was done in 122 days.

Colossus is the most powerful AI training system in the world. Moreover, it will double in size to 200k (50k H200s) in a few months.

Excellent…

— Elon Musk (@elonmusk) September 2, 2024

Nvidia will power the Colossus

At the core of Colossus are 100,000 NVIDIA H100 graphics cards. These GPUs (Graphics Processing Units) are specifically designed to handle the demanding computational requirements of AI training and here is why these GPUs are so vital:

• Raw processing power: The H100 is Nvidia’s flagship AI processor, designed to accelerate the training and inference of AI models, particularly those based on deep learning and neural networks. Compared to its predecessor, the H100 can run language models up to 30 times faster.

• Transformer engine: A key feature of the H100 is its Transformer Engine, a specialized set of circuits optimized for running AI models based on the Transformer neural network architecture. This architecture is the backbone of some of the most advanced language models, like GPT-4 and Meta’s Llama 3.1 405B. The Transformer Engine enables these GPUs to handle large-scale models more efficiently, making them ideal for training sophisticated AI systems.

The next level: Doubling down with the H200

Musk has ambitious plans to further expand Colossus, aiming to double the system’s GPU count to 200,000 in the near future. This expansion will include 50,000 units of Nvidia’s H200, an even more powerful successor to the H100. The H200 offers several significant upgrades:

• HBM3e Memory: The H200 uses High Bandwidth Memory 3e (HBM3e), which is faster than the HBM3 used in the H100. This type of memory enhances the speed at which data is transferred between the memory and the chip’s logic circuits. For AI models, which constantly shuffle vast amounts of data between processing and memory, this speed is crucial.

• Increased Memory Capacity: The H200 nearly doubles the onboard memory capacity to 141 gigabytes. This allows the GPU to store more of an AI model’s data closer to its logic circuits, reducing the need to fetch data from slower storage sources. The result is faster processing times and more efficient model training.

Colossus’ role in AI training

Colossus is specifically designed to train large language models (LLMs), which are the foundation of advanced AI applications.

The sheer number of GPUs in Colossus allows xAI to train AI models at a scale and speed that is unmatched by other systems. For example, xAI’s current flagship LLM, Grok-2, was trained on 15,000 GPUs. With 100,000 GPUs now available, xAI can train much larger and more complex models, potentially leading to significant improvements in AI capabilities.

The advanced architecture of the H100 and H200 GPUs ensures that models are trained not only faster but with greater precision. The high memory capacity and rapid data transfer capabilities mean that even the most complex AI models can be trained more efficiently, resulting in better performance and accuracy.

What’s next?

Colossus is not just a technical achievement; it’s a strategic asset in xAI’s mission to dominate the AI industry. By building the world’s most powerful AI training system, xAI positions itself as a leader in developing cutting-edge AI models. This system gives xAI a competitive advantage over other AI companies, including OpenAI, which Musk is currently in legal conflict with.

Moreover, the construction of Colossus reflects Musk’s broader vision for AI. By reallocating resources from Tesla to xAI, including the rerouting of 12,000 H100 GPUs worth over $500 million, Musk demonstrates his commitment to AI as a central focus of his business empire.

Can he succeed? We have to wait for the answer!

Featured image credit: Eray Eliaçık/Grok

The organization behind NaNoWriMo has come under fire following a controversial statement suggesting that opposing AI writing tools might be “classist and ableist.”

Behind the scenes: The AI usage controversy in NaNoWriMo community

NaNoWriMo posted its statement on AI over the weekend, neither fully supporting nor rejecting the technology but highlighting the importance of considering social issues related to its use. The post argued that AI tools could help alleviate the financial burden of hiring assistants and provide support for those with different cognitive abilities.

NaNoWriMo, best known for its annual November writing challenge where participants are tasked with producing a 50,000-word manuscript, had previously acknowledged that while AI could be used, relying on it entirely would defeat the purpose of the challenge. However, this year, the organization expanded its view, stating that while some writers may personally reject AI, others should have the freedom to choose their tools.

“We recognize that some members of our community stand staunchly against AI for themselves, and that’s perfectly fine. As individuals, we have the freedom to make our own decisions,” stated NaNoWriMo in its latest post.

The statement ignited widespread outrage across social media, particularly on platforms like X and Reddit, where writers and creatives voiced concerns over the use of generative AI in creative work. Some felt that AI devalues artistic expression, while others, including disabled writers, criticized the organization for implying that AI is necessary to accommodate different writing needs.

The controversy led to a notable resignation from the NaNoWriMo Writers Board—Daniel José Older, a key figure in the Star Wars: The High Republic series, stepped down in protest of the organization’s position.

Hello @NaNoWriMo this is me DJO officially stepping down from your Writers Board and urging every writer I know to do the same. Never use my name in your promo again in fact never say my name at all and never email me again. Thanks! https://t.co/KDKZ0zVx3H

— Daniel José Older (@djolder) September 2, 2024

Chuck Wendig, author of Star Wars: Aftermath, sharply criticized NaNoWriMo’s stance on AI, stating:

“Generative AI empowers not the artist, not the writer, but the tech industry. It steals content to remake content, graverobbing existing material to staple together its Frankensteinian idea of art and story.”

This incident adds to NaNoWriMo’s growing list of controversies in recent years. The organization had previously faced scrutiny for a lack of transparency following accusations involving child safety concerns and grooming, as well as criticism for promoting its sponsor’s AI writing tools.

In response to the uproar, NaNoWriMo revised its statement to address concerns about the role of generative AI in the writing world. The update expressed that the organization is troubled by specific misuses of AI, which clash with its values. NaNoWriMo further clarified that AI is a broad technology, encompassing both generative and non-generative forms, and that its complexity makes it difficult to take a definitive stance either for or against.

At a glance:

NaNoWriMo stands for National Novel Writing Month. It’s an annual creative writing challenge that takes place every November, where participants aim to write a 50,000-word manuscript in 30 days. The event encourages writers of all skill levels to focus on completing a novel rather than perfecting it, with the goal of motivating creativity and output.

NaNoWriMo drama

Recently, NaNoWriMo has been at the center of controversy due to its stance on AI writing tools. The organization received backlash after suggesting that opposing AI use in writing is “classist and ableist.” This sparked heated discussions within the creative community, with many writers arguing that AI diminishes the value of human creativity. Additionally, the organization has faced past criticisms regarding transparency, child safety concerns, and its partnerships with companies offering AI-driven writing assistance.

NaNoWriMo rules

Key rules for NaNoWriMo include:

• Participants must aim to write a novel of at least 50,000 words by the end of November.
• The project must be a new, original work. Pre-existing drafts or novels don’t count, though outlining or research before November is allowed.
• Writing must be done between November 1 and November 30, with the word count verified on the NaNoWriMo website.
• Writers can use any genre or style, and while editing isn’t prohibited, the focus is on generating content rather than polishing.

Generative AI tools are allowed, but using them exclusively to write the entire manuscript goes against the spirit of the challenge.

Generating vs. writing

There’s a clear distinction between generating content and the art of writing. While writing involves creativity, intuition, and an emotional connection with the reader, generative AI works on speed and efficiency, assembling content from vast data pools. However, dismissing the force and rapid advancement of generative AI tools entirely is a relic of outdated thinking. Much like the iPhone revolutionized how we communicate and work, generative AI is reshaping how we approach content creation.

To hold too tightly to traditional views without acknowledging the potential of this technology is to risk being left behind in a critical moment of transformation. A delicate balance is necessary—one that protects the integrity of creative writing while allowing for the benefits of AI. Rather than a blanket rejection or blind embrace, what’s needed is a regulatory framework that ensures ethical use.

Featured image credit: Kerem Gülen/Midjourney

The video content industry, powered by artificial intelligence is on the cusp of a significant transformation. Currently valued at approximately $600 million as of 2023, this nascent sector is poised for explosive growth, with projections suggesting it could balloon into a multi-billion-dollar industry by 2033. This remarkable potential, as highlighted by Allied Market Research, mentions AI-generated video as an emerging “lucrative” source of global profit.

The anticipated surge in value, with the market estimated to reach around $9.3 billion by 2033, is driven by a compound annual growth rate (CAGR) of 30.7% over the next decade. This trajectory reflects the convergence of two powerful forces: the relentless demand for video content and the rapid advancements in AI technology.

AI video market is expected to boom

The proliferation of devices equipped with AI capabilities, such as Ryzen AI-powered PCs and laptops, is set to fuel this trend further. Moreover, the advent of “AI smartphones,” including the much-anticipated iPhone 16 series, could mark a pivotal moment in the industry, accelerating the adoption of AI-driven video content creation.

Social media, marketing, and education sectors stand out as prime beneficiaries of this growth. These areas, already heavily reliant on video content, are expected to embrace AI-generated videos with enthusiasm, leveraging the technology to enhance engagement, personalize content, and streamline production processes. The ability to quickly generate compelling videos from simple text inputs or even spreadsheets opens up new avenues for businesses and content creators alike.

Beyond these consumer-facing applications, the enterprise sector is also set to play a crucial role in driving the AI video market’s expansion. The potential to convert other media formats, such as PowerPoint presentations and spreadsheets, into dynamic video content offers businesses a powerful tool for communication, training, and data visualization. As companies increasingly recognize the value of AI-driven video in enhancing internal and external communications, the demand for such technology is expected to soar.

However, the path to widespread adoption is not without challenges. Despite the promising outlook, AI video generators currently face significant hurdles, particularly in terms of cost and resource efficiency. The technology, while impressive, remains expensive and energy-intensive, posing a barrier to entry for many potential users. For the AI video market to fully realize its potential, it will need to evolve towards more cost-effective and energy-efficient solutions.

This evolution will likely be driven by ongoing innovations in AI and machine learning, as well as advancements in hardware and software designed to optimize video generation processes. As these technologies mature, the barriers to entry will diminish, opening up new opportunities for businesses of all sizes to harness the power of AI-generated video.

The AI video industry is at the dawn of a new era, with the potential to redefine how video content is created and consumed. With a projected value of $9.3 billion by 2033, the market is set to experience unprecedented growth, driven by both consumer and enterprise applications. However, realizing this potential will require overcoming current technological challenges and making AI video generation more accessible and efficient.

New social media trend: AI-generated hugging videos

How should organizations approach this?

• Articulate your business’s core purpose and long-term vision. Your mission statement should clearly communicate what you aim to achieve and why your business exists.
• Ensure that every team member understands and aligns with this vision. This creates a unified direction and motivates the entire organization.
• Identify what makes your product or service unique. Focus on how it solves a problem or fulfills a need better than any competitor.
• Communicate this value proposition clearly to your target audience across all marketing channels. Consistency is key to building a strong brand identity.
• Choose a business model that is not only profitable but also scalable. Test and iterate to find the most efficient way to generate revenue.
• Diversify your revenue streams to reduce dependency on a single source. This increases your business’s resilience to market changes.
• Invest in understanding your customers’ needs and pain points. Use surveys, feedback, and direct communication to gain insights.
• Design every touchpoint to enhance the customer experience. Happy customers are more likely to become repeat buyers and brand advocates.
• Stay updated on the latest technological trends in your industry. Whether it’s AI, automation, or data analytics, find ways to integrate them into your business.
• Encourage a culture of innovation within your team. Invest in ongoing training and development to keep your team’s skills sharp. A motivated and capable team drives business success.
• Keep a close eye on your finances from the start. Develop a budget and stick to it, ensuring that all expenses are justified.
• Secure adequate funding for growth, whether through investors, loans, or reinvested profits. Financial stability is critical to scaling your business.
• Develop a strong brand identity that resonates with your target audience. Your brand should reflect your values, mission, and the quality of your offerings.
• Regularly review your business strategy and make adjustments based on market trends, customer feedback, and internal performance metrics.

Image credits: Kerem Gülen/Midjourney 

The field of artificial intelligence is constantly evolving, with new models being introduced almost every other day.

Two of the latest contenders making waves are Grok and Venice AI.

Both of these models are new to the AI market which has been long dominated by ChatGPT, but they take different approaches and offer unique strengths.

Here’s a closer look at how Grok and Venice AI compare and what sets them apart, welcome to our Grok vs Venice comparison!

Grok vs Venice AI

Let’s start with introducing Elon Musk’s AI push, Grok.

Grok is an AI model designed with a focus on understanding and processing natural language. It’s built to handle a wide range of tasks, from simple text generation to more complex functions like summarization, translation, and even basic reasoning. Grok’s main strength lies in its ability to comprehend and generate human-like text, making it particularly useful for applications that require nuanced understanding of language.

Grok is often praised for its adaptability and versatility. It can be fine-tuned for specific tasks, making it a valuable tool in various industries, from customer service to content creation. Its language processing capabilities are among the best, allowing it to understand context, tone, and subtle nuances in text.

Venice AI, on the other hand, is a more recent development that takes a different approach. While it also handles natural language processing (NLP), Venice AI is designed with a stronger emphasis on multi-modal tasks. This means that Venice AI can process and integrate different types of data such as:

• Text
• Images
• Audio

Simultaneously, making it a more comprehensive tool for applications that require a broader understanding of inputs.

Venice AI is particularly strong in environments where data is diverse and complex. For example, it can analyze a piece of text while also interpreting related images or audio files, providing a richer, more informed response.

This makes Venice AI a powerful tool for sectors like media, entertainment, and education, where content often spans multiple formats.

Got to know both of the models? Good! Let the showdown begin!

Round 1: Text generation

The story is well-structured, with a clear beginning, middle, and end. It starts by setting the scene, introduces a conflict (Muse’s empathy), and concludes with a resolution where the characters become advocates for AI rights.

Grok demonstrates creativity in the concept of “digital empathy” and the evolving relationship between humans and AI. The idea of an AI exhibiting emotional understanding is thought-provoking and adds depth to the narrative.

The story incorporates elements of science fiction, such as a futuristic city and advanced AI, while also addressing ethical issues, making it both imaginative and relevant.

The narrative is engaging, with a focus on the emotional connection between Aria and Muse. The conflict around AI rights adds tension and raises interesting ethical questions, making the story compelling.

 

The story also has a defined structure, beginning with the depiction of a harmonious society in 2050 and introducing challenges regarding the role of sentient machines.

However, the resolution is broader and more collective, focusing on cooperation and a shared future rather than a personal resolution. While it’s well-organized, the narrative lacks the tight focus of Grok’s story.

Winner

Grok demonstrates significant creativity through the concept of “digital empathy,” where an AI begins to feel and understand human emotions. This idea is not only imaginative but also deeply thought-provoking, adding a unique layer of complexity to the story. The exploration of AI rights in response to this new development further enhances the narrative’s originality.

The story skillfully blends science fiction elements, like a futuristic city and advanced AI, with real-world ethical issues such as AI rights. This combination makes the narrative both imaginative and highly relevant to ongoing debates about the future of AI and its place in society. The focus on a specific relationship also grounds the story in relatable emotions.

Venice AI showcases creativity in its depiction of a future where empathy and emotional intelligence are central to human-AI relations. The idea of AIs developing and using emotional intelligence to understand human needs is innovative. However, the concept is broader and less personal, focusing more on societal dynamics than individual experiences.

Venice AI’s narrative is imaginative in its portrayal of a utopian society where humans and AIs coexist in harmony. The story emphasizes the importance of empathy and cooperation, making it relevant to discussions about the potential of AI to enhance human life. However, the broader, more idealized setting may feel less immediate or relatable compared to Grok’s more focused and personal approach.

Both Grok and Venice AI offer well-structured and creative narratives, but they differ in focus and depth. Grok provides a more intimate and emotionally charged story, with a clear conflict and resolution centered on the evolving relationship between a human and an AI. This makes it highly engaging and thought-provoking, particularly in its exploration of digital empathy and AI rights.

On the other hand, Venice AI presents a broader, more utopian vision of the future, focusing on societal harmony and cooperation between humans and AIs. While imaginative and relevant, its narrative is less personal and more idealized.

Grok is the clear winner here.

Round 2: Math solving

Grok:

Grok’s explanation of the power rule is clear and accurate. It starts by explaining the power rule and applies it step-by-step to each term of the polynomial function.

Each differentiation step is carefully laid out, showing how the power rule is applied to each term. The breakdown is easy to follow, and Grok ensures that the constant term (7) is identified and treated correctly (differentiating to 0).

Grok also does a good job of explaining the overall process, including the importance of differentiation and what the derivative represents.

The solution is structured logically, with clear demarcation of each term’s differentiation. The explanation is detailed, providing a step-by-step breakdown of the differentiation process.

The derivative is correctly calculated.

Venice:

Venice AI’s explanation is also clear and accurate, applying the power rule similarly to each term of the polynomial.

It uses a slightly different structure but also breaks down the process in a way that is easy to follow, ensuring that each differentiation step is explained.

The explanation is methodical and logically organized, with each term differentiated and combined to form the final derivative.

Venice AI provides a clear and concise explanation, which may be slightly more succinct compared to Grok’s detailed breakdown.

The final derivative is also correctly calculated.

Winner

Both Venice AI and Grok provided accurate and well-explained solutions to the differentiation problem.

Both models offer clear and understandable explanations. However, Grok provides a more detailed breakdown, explicitly stating the power rule for each term and explaining each step with more elaboration.

Grok’s approach is slightly more detailed, which could be beneficial for someone who is new to calculus or needs a more thorough explanation.

Venice AI’s approach is concise yet comprehensive, focusing on clarity without excessive detail, which might appeal to users looking for quick answers.

Both models present the information logically. Grok’s structure might be considered more didactic, walking the user through each small step, while Venice AI focuses on a balanced explanation that is clear but not overly detailed.

Grok might be more suitable for learners who need a more detailed, step-by-step guide, while Venice AI provides a slightly more polished and concise explanation that is still very clear and accurate.

Therefore Grok takes the victory here.

Grok:

Grok’s code includes a detailed docstring explaining the purpose of the function, arguments, and return value. It provides a explanation of the swapped flag and optimization but does not include a sample test in the code itself.

Venice:

Venice’s code is clear and straightforward. It includes a sample test with an explanation of how the function works. It explains the concept of the swapped flag and how the function exits early if no swaps occur.

Winner

Venice offers a practical and immediately usable implementation with clear in-code comments and an example test case. It is straightforward and suitable for users who want a quick, functional implementation.

Grok provides a well-documented implementation with a detailed explanation of the algorithm, including time complexity and optimization details. The docstring adds formal documentation, making it useful for educational contexts.

Winner of the round 3 is Grok!

Grok:

Grok’s image effectively captures the essence of a bustling futuristic city, showcasing a blend of neon-lit skyscrapers, flying cars, and holographic advertisements. The scene is dominated by a vibrant color palette, with neon pinks, blues, and purples illuminating the cityscape.

Venice:

Venice’s image offers a distinct take on the futuristic cityscape prompt, presenting a more fantastical and otherworldly vision. While both generators captured the essence of a technologically advanced city, Venice’s interpretation is more visually striking and imaginative.

Winner

Both Grok and Venice have successfully captured the essence of a futuristic city, but their interpretations are distinct. Grok’s image is more grounded and realistic, while Venice’s image is more fantastical and imaginative.

If your main focus is realism then Grok is the way to go but if you are looking for imaginative power in an AI image generator, then Venice might suit your needs better.

The image generation round is a tie!

Grok:

Grok emphasizes the importance of bodily autonomy and privacy, linking it to human rights treaties. It supports the argument that restricting abortion limits personal freedom.

Grok also argues that prioritizing a woman’s quality of life is crucial and that legal abortion provides health and economic benefits.

Venice:

Venice also highlights bodily autonomy but emphasizes the broader right to make reproductive decisions. It centers on individual control over reproductive health.

Venice adds the aspect of physical and mental well-being, showing how abortion is crucial for maintaining health and autonomy.

Venice also includes a social justice angle, noting that restrictions impact marginalized communities more severely, which Grok does not explicitly address.

Winner

Grok offers a structured, detailed comparison of both perspectives, with a focus on legal and ethical arguments. It provides clear, methodical breakdowns of each viewpoint and their implications.

Venice provides a broader, more holistic view that includes social justice aspects and emphasizes the ethical dimensions of the debate. Its explanations are clear and address the impact on marginalized communities, along with broader societal and moral questions.

Venice gets the win here.

Verdict

Grok and Venice AI both showcase impressive capabilities in their respective areas. While both are relatively new to the AI market, they offer unique strengths and approaches.

Here’s a breakdown of their key strengths and weaknesses:

Grok

• Strengths: Exceptional language processing, adaptability, and versatility.
• Weaknesses: May struggle with highly complex or specialized tasks that require deep domain knowledge.

Venice AI

• Strengths: Strong multi-modal capabilities, ability to process and integrate diverse data types.
• Weaknesses: May require more data and training for highly specialized tasks.

The “winner” in this comparison is clearly Grok, however, the AI models are constantly evolving, and new models with unique capabilities may emerge in the future. It’s always recommended to evaluate different models based on your specific requirements and test them out to determine the best fit for your needs.

---

Featured image credit: Emre Çıtak/Ideogram AI

The idea of having a robot assist us in our daily lives is becoming more of a reality than ever before with the 1X NEO Beta robot. Imagine a humanoid robot that can not only navigate your home with ease but also learn from its surroundings and perform various tasks, all while ensuring safety for you and your loved ones. This is exactly what 1X Robotics offers.

Sounds good? In this overview, we will delve into the unique features of the 1X NEO Beta robot, its design and capabilities, and how it compares to other humanoid robots currently in development.

What is the 1X NEO Beta robot?

The 1X NEO Beta robot is a new bipedal humanoid robot prototype created by 1X. Designed for use in the home, NEO Beta is a significant step forward in making intelligent robots a regular part of daily life. Unlike earlier robots that were often rigid and limited in function, 1X NEO Beta robot uses a bio-inspired design that allows it to work safely and efficiently alongside people in a household setting.

NEO Beta is designed with two legs and mimics human walking patterns, allowing it to navigate spaces such as homes more effectively. This design helps it move through areas that might be challenging for other types of robots, such as uneven flooring or stairs.

The robot has a bio-inspired design, meaning its movements are modeled after the human body. This allows for more natural, flexible motions compared to the rigid movements of many older robots.

1X NEO Beta weighs around 25 kg (66 lbs), which is lighter than many similar robots designed for industrial or home use. This lighter weight reduces the risk of damage or injury in case of a fall or collision, making it more suitable for household use. By comparison, other humanoid robots like Tesla’s Optimus GEN2 and Figure 02 are heavier, weighing 57 kg (152 lbs) and 70 kg (187 lbs), respectively.

The robot is designed to learn from its environment, adapting its behavior over time. This is achieved through continuous interaction with its surroundings and input from human operators who guide and train it on various tasks. 1X employs operators who work with the robot to refine its abilities, enabling it to perform a variety of household tasks more effectively.

That's def a dude in a suit

— 1X (@1x_tech) August 31, 2024

Is the 1X NEO Beta robot safe?

Safety is a primary design focus for NEO Beta, especially since it will operate in home environments where it may encounter people, children, and pets. It includes sensors and software to detect and avoid obstacles and prevent collisions.

The exterior of NEO Beta is soft and cushioned in specific areas, unlike other robots that use hard plastic or metal casings. This makes any accidental contact less likely to cause injury.

However, humanoid robots like NEO Beta face several challenges, including developing comprehensive safety standards and gaining consumer trust. Organizations such as IEEE are working to establish guidelines to ensure robots operate safely around people. As these guidelines develop, robots like NEO Beta will need to meet strict safety and performance criteria to be widely adopted for home use.

Is the 1X NEO Beta robot just a concept?

NEO Beta is being prepared for mass production at 1X’s factory in Moss, Norway. The goal is to produce these robots on a large scale, making them accessible for use in more homes.

This production readiness shows that the robot is not just a concept but is intended to be a real product used in everyday situations.

A small number of NEO Beta units will be deployed in selected homes to gather data and improve the robot’s capabilities. This testing phase will help understand how the robot performs in actual living spaces and what adjustments are needed. The feedback collected from these test environments will be crucial in refining the robot’s design and functionality.

Comparison with other humanoid robots

NEO Beta is part of a new wave of humanoid robots being developed for different environments. Compared to others, like Tesla’s Optimus GEN2 and Unitree’s G1, NEO Beta is lighter and designed specifically with home use in mind. Its softer exterior and bio-inspired design set it apart, especially in terms of safety and adaptability in a home environment.

Get to know 1x Robotics

1X Robotics is a company specializing in humanoid robotics, focusing on developing robots that can operate safely in human environments. The company is based in Norway, with additional offices in Sunnyvale, California; Irving, Texas; and Oslo, Norway. 1X Robotics has been working on humanoid robots for over a decade, starting with its earlier model, EVE, introduced in 2017. The company has a factory in Moss, Norway, where it is preparing for the mass production of its latest humanoid prototype, NEO Beta. This robot is designed for home use and aims to gather real-world data to improve its performance in various household tasks. 1X Robotics employs a team of operators who train these robots to enhance their capabilities in different settings, ensuring they are adaptable and capable of handling diverse environments.

Featured image credit: 1x

Artificial Intelligence is becoming a game-changer for businesses across various sectors. Many companies, acutely aware of the competitive edge AI can provide, are choosing to keep their technological advancements close to the chest. This strategic silence extends beyond mere product development. It encompasses AI-driven strategies and innovations that could reshape industries.

While many companies keep their AI advancements secret, today, we’re revealing five well-known businesses that use AI to boost profits, work smarter, and create innovative products.

Whether you’re a business leader seeking inspiration or simply curious about how AI is shaping our world, these examples will provide a fascinating glimpse into the future of business – a future that, in many ways, is already here.

Almost all these companies are leveraging OpenAI‘s models. But remember, this is just the tip of the iceberg. There’s a whole world of AI solutions, from open-source projects to commercial offerings, each with unique strengths.

Moderna

Moderna, famous for its COVID-19 vaccine, pioneered AI use in the pharmaceutical industry. In 2023, they partnered with OpenAI to implement ChatGPT Enterprise across their entire company, aiming for 100% AI adoption by all employees in just six months.

Moderna is using AI for different departments:

Clinical research department

At the heart of Moderna’s research efforts is the Dose ID product, a custom-built AI assistant that transforms clinical data analysis.

It helps with the following processes:

• Data processing: Dose ID sifts through vast amounts of clinical trial data, uncovering insights that might otherwise remain hidden.
• Visualization: The AI tool generates clear, informative visual representations of complex data sets, enabling researchers to identify patterns and trends easily.
• Dosage optimization: Perhaps most critically, Dose ID aids in determining optimal vaccine dosages, a crucial factor in developing safe and effective treatments.

Legal department

Moderna’s legal team has embraced AI to optimize document handling and improve internal communications:

• Contract analysis: AI algorithms swiftly summarize lengthy contracts, distilling complex legal language into concise, actionable briefs.
• Policy assistance: An AI-powered system provides rapid responses to employee inquiries about internal policies, ensuring consistent interpretation of guidelines across the organization.

Brand department

The brand team at Moderna utilizes AI to enhance its messaging and expand its audience:

• Presentation preparation: AI assists in creating comprehensive, visually appealing slides for quarterly earnings calls, effectively communicating complex financial data.
• Language simplification: Advanced natural language processing helps translate intricate biotech terminology into more accessible language for diverse stakeholders.
• Audience targeting: AI tools aid in tailoring Moderna’s narrative for different audiences, from investors to the general public, ensuring effective communication across various channels.

The company has seen remarkable success in adopting custom GPT models. Employees across the organization have enthusiastically embraced this technology, creating an astounding 750 custom GPTs tailored to various business needs. The widespread adoption is evident, with 40% of weekly active users taking the initiative to develop their own GPTs. Furthermore, the engagement levels are noteworthy, as each user averages an impressive 120 AI-driven conversations per week, highlighting the significant integration of AI into daily workflows.

Moderna’s CEO claims that thanks to AI, they’re now doing the work of a 100,000-person company with just a few thousand people. That’s not just efficiency—it’s a complete paradigm shift in medicine development.

Moderna’s AI implementation opens up exciting opportunities for the entire pharmaceutical industry. By dramatically reducing the time and resources needed for drug development, AI could lead to faster discoveries of life-saving treatments. Other pharma companies could adopt similar AI-driven approaches to streamline their research processes, improve clinical trial designs, and enhance stakeholder communication. Beyond pharmaceuticals, any research-intensive industry could benefit from AI-powered data analysis and knowledge management systems similar to Moderna’s, potentially accelerating innovation across fields like materials science, renewable energy, and biotechnology.

Oscar Health

Oscar Health, an American health insurance company, uses AI to reduce costs and improve patient care.

They’re using AI in three main areas:

• Claims processing: In claims processing, they’ve developed an AI assistant that efficiently navigates claim traces, resulting in a 50% reduction in escalation resolution time.
• Medical tecord analysis: The company also utilizes AI for medical record analysis, enabling faster retrieval of relevant patient information for insurance claims, summarizing records to prepare providers for patient encounters, and identifying suitable patients for specific treatments by analyzing multiple records.
• Clinical documentation: Oscar Health сut the time spent documenting medical care conversations by nearly 40%.

Oscar isn’t just using AI—it’s building its company around it. It’s created a centralized AI Pod to guide other teams and is leading a coalition of 37 major healthcare players to develop ethical AI use principles.

Oscar Health’s AI initiatives demonstrate the potential for technology to make healthcare more efficient and accessible. By automating time-consuming tasks like claims processing and medical record analysis, AI frees up healthcare professionals to focus on patient care. Other insurance companies could adopt this approach to reduce costs and improve service quality. Moreover, AI-assisted medical record analysis could be valuable for hospitals, clinics, and research institutions, potentially leading to better patient outcomes through more informed decision-making. The healthcare industry could benefit from similar AI applications to tackle challenges like personalized medicine, early disease detection, and resource allocation.

Klarna

Klarna, a fintech company, uses AI to revolutionize shopping and customer service. They’ve developed an AI-powered shopping assistant to recommend products based on user preferences.

Their AI assistant has made significant impacts on various departments:

• Managed 2.3 million conversations in its first month.
• Equivalent to the work of 700 full-time agents.
• Matches human agents in customer satisfaction.
• More accurate in resolving issues, leading to 25% fewer repeat inquiries.
• Available in 23 markets, communicating in over 35 languages.

Klarna estimates a $40 million profit improvement in 2024 due to AI implementation. They’re encouraging all employees to explore and test AI applications, aiming for a new level of employee empowerment and enhanced customer experience.

The company’s success with AI in customer service and personalized shopping experiences opens up new possibilities for retail and financial services sectors. Companies can significantly improve customer satisfaction while reducing operational costs by providing instant, accurate customer support at scale. The concept of AI-powered shopping assistants could revolutionize e-commerce, offering personalized recommendations and seamless customer journeys. Other fintech companies, banks, and retailers could adopt similar AI strategies to enhance customer interactions, streamline operations, and gain valuable insights from customer data. This approach also paves the way for hyper-personalization in marketing and product development across various industries.

Verizon

Verizon, one of the world’s largest telecommunications companies, uses AI to handle an astounding 170 million customer calls annually.

Their AI strategy focuses on:

1. Speeding up internal processes.
2. Enhancing products with AI technology.
3. Providing robust network capabilities for customers running AI programs.

Here’s how AI is transforming Verizon’s customer service:

• Quick information retrieval: Pulling up customer history instantly.
• Predictive problem solving: Anticipating call reasons based on past interactions.
• Recommendations: Suggesting solutions to agents.
• Call summaries: Automatically summarizing conversations.
• Follow-up tracking: Ensuring necessary actions are taken.

The results are impressive: agents can now anticipate the reason for a customer’s call 80% of the time and recommend resolutions with about 90% accuracy.

Verizon’s AI implementation in customer service showcases how large-scale operations can be optimized using intelligent technologies. AI improves customer satisfaction and reduces operational costs by predicting customer needs and streamlining problem-solving. This approach could be valuable for any company with a large customer base, from utilities and airlines to e-commerce platforms. Using AI for hyper-personalization at scale opens up exciting possibilities for marketing, product development, and customer relationship management across industries.

Duolingo

Duolingo, the popular language learning app, uses AI to revolutionize how over 50 million monthly active users learn languages.

They’ve introduced features like:

• Role Play: It’s an AI conversation partner that can discuss a wide range of topics.
• Explain My Answer: It’s a tool that breaks down grammar mistakes using GPT-4.

These AI-powered features are deeply integrated into the app, leveraging Duolingo’s gamified approach, which users love.

Duolingo’s innovative use of AI in language learning demonstrates the potential for personalized, interactive education at scale. This approach could revolutionize not just language learning but education as a whole. Other edtech companies could adopt similar AI-powered conversational partners and personalized feedback systems for subjects ranging from mathematics to music. Beyond education, AI-driven personalized practice and feedback could be valuable in professional training, skill development, and even therapy applications.

Conclusion

AI isn’t just a future possibility – it’s here now and changing the game across industries. AI delivers tangible business value from medicine to finance, telecommunications to education.

The question isn’t whether AI will impact your industry – how soon and how dramatically. As these companies show, embracing AI can increase efficiency, provide better customer experiences, and create innovative products and services.

This is just the beginning. If companies are achieving this with AI now, imagine what’s next. The future looks exciting, and it’s clear that AI will play a crucial role in shaping it.

Featured image credit: Freepik

Can machines truly understand human emotions? This question propels the intriguing field of emotion AI, a frontier in artificial intelligence that seeks not just to interpret, but to empathize with human feelings.

Can AI feel emotions?

Absolutely not—AI cannot feel emotions. It’s a common misconception, fueled by the often human-like interactions people have with AI systems. What AI does is simulate responses that might be expected from an emotional being based on patterns it has learned from vast amounts of data. These simulations can be remarkably convincing, but at their core, they are the result of complex algorithms processing information and producing programmed responses.

AI systems, including those designed to converse or interact in seemingly emotional ways, lack consciousness or self-awareness. They operate based on coded instructions and learned data, without personal experiences or emotional states. Essentially, while AI can mimic emotional intelligence and react in ways that suggest understanding or empathy, these responses are not backed by genuine feelings but by calculated outputs designed to fulfill specific functions or tasks.

What is emotional intelligence?

Emotional intelligence, often abbreviated as EI or EQ (Emotional Quotient), is the ability to recognize, understand, and manage one’s own emotions, as well as to recognize, understand, and influence the emotions of others. This concept goes beyond mere emotional awareness; it involves the application of this awareness in a way that enhances personal, professional, and social interactions.

Key skills of emotional intelligence

• Self-awareness – This is the foundational skill of emotional intelligence. It involves being aware of your own emotional state and how your emotions can affect your thoughts and behavior. Self-awareness enables individuals to recognize their strengths and weaknesses and to operate with a clear understanding of their emotions.
• Self-regulation – This refers to the ability to control or adjust one’s emotions to adapt to changing circumstances or to achieve one’s goals. It involves managing impulsive feelings and behaviors, managing stress effectively, and adapting to evolving situations flexibly.
• Motivation – Emotionally intelligent people are generally motivated by things beyond external rewards like fame, money, or recognition. They are driven to achieve for the sake of achievement and personal satisfaction.
• Empathy – This is the ability to understand the emotions of others and to treat them according to their emotional reactions. This skill is crucial for leaders, managers, and professionals who need to deal with various people and situations, as it helps in navigating social complexities and managing relationships.
• Social skills – This encompasses a wide range of skills, such as effective communication, managing relationships, inspiring and influencing others, and working well in a team. People with strong social skills are typically adept at managing social networks, fostering relationships, and understanding social norms.

Developing emotional intelligence can greatly improve one’s interactions and relationships at work and in personal life.

Is AI creative: Answering the unanswerable

What is emotion AI?

Emotion AI is experiencing a renewed wave of interest, as highlighted in the recent Enterprise SaaS Emerging Tech Research report from PitchBook. This technology, viewed as an advanced version of the earlier sentiment analysis techniques, promises to enhance the interpretation of human emotions beyond text by integrating multimodal inputs such as visual, auditory, and other sensor data. Employing a blend of machine learning and psychological principles, Emotion AI aims to discern human emotions during interactions, marking a significant evolution in how AI understands human sentiments.

Despite not being a fresh concept—having been offered previously as a cloud service—the increasing incorporation of bots in the workplace has catapulted emotion AI into a more prominent position within the business sector than ever before. This resurgence brings with it not just opportunities but also renewed scrutiny and challenges.

Around 2019, when the focus of the AI and machine learning community was predominantly on computer vision and before the shift to generative language and art technologies, emotion AI was subjected to critical academic review. Researchers conducted a meta-review of studies and concluded that human emotions cannot be reliably deduced from facial expressions alone. This finding casts doubt on the foundational premise of emotion AI—that AI can effectively interpret human emotions through the same cues humans use, such as facial movements, body language, and vocal tones.

Regulatory frameworks, particularly in regions like the European Union, are poised to have a significant impact on the future of emotion AI. The EU’s AI Act, for instance, includes provisions that prohibit the use of computer-vision-based emotion detection systems in specific applications, such as in educational settings. Such regulations could severely limit the application scope of Emotion AI, potentially stifling its development and integration in affected domains.

Will AGI have emotional intelligence?

The question of whether AGI will possess emotional intelligence has philosophical, technical, and practical aspects. AGI, which refers to a type of AI that can understand, learn, and apply intelligence across a broad range of tasks at human levels or beyond, presents unique challenges and possibilities in this regard.

AGI could be designed to simulate emotional intelligence, much like current narrow AI systems that mimic empathetic responses. This would involve sophisticated algorithms capable of processing and responding to human emotions in a way that appears understanding and sensitive. Such simulation would likely be based on vast datasets detailing human emotional interactions, allowing the AGI to perform convincingly in scenarios requiring emotional sensitivity.

Beyond mere simulation, AGI might develop a form of functional emotional intelligence. This would not just be about reacting in emotionally intelligent ways, but using these reactions in decision-making processes, learning from emotional data, and adapting its behavior based on an understanding of human emotions. This could enhance an AGI’s ability to perform tasks involving complex social interactions and negotiations.

A significant distinction must be made between understanding human emotions and actually experiencing them. While AGI could be developed to understand and predict emotional responses accurately, the subjective experience of emotions—a core aspect of genuine emotional intelligence—is likely beyond the capabilities of AGI as we conceive it today. Without consciousness, an AGI’s emotional understanding would remain computational, devoid of true empathy.

Programming AGI to behave in emotionally intelligent ways introduces complex control issues. Developers would need to create and enforce ethical guidelines on how AGI interprets and acts on emotional data. There’s also the risk of manipulation if an AGI learns to use emotional intelligence to achieve ends that may not align with human ethics.

The deployment of AGI with emotional intelligence capabilities raises ethical questions.

How do we ensure that such AGI respects human emotional boundaries? What regulations are needed to prevent exploitation of emotional manipulation by AGI systems?

For AGI to be safely integrated into society, it may need to align with human values and ethics, which includes appropriate responses to emotional cues. This alignment is crucial to ensure that as AGI systems become more integrated into everyday life, they augment human interactions rather than disrupt them.

Developing AGI with emotional intelligence capabilities will require advancements in computer science and insights from psychology, neuroscience, and ethics. This interdisciplinary approach can help ensure that AGI’s emotional intelligence is both effective and ethical.

Image credits: Kerem Gülen/Midjourney

Now that Brazil has imposed a comprehensive ban on the social media platform X, there’s an escalating discussion over free speech and government censorship. This decision comes after X’s refusal to comply with Brazilian court orders demanding the removal of certain anti-government commentators, a standoff that questions the balance between regulation and freedom on the internet.

Why did Brazil ban X (Twitter)?

X, the prominent social media platform, now finds itself barred from operating in Brazil. The ban stems from its refusal to comply with a Brazilian Supreme Court order demanding the removal of seven anti-government commentators accused of fueling unrest in the region. X has contested the order, claiming a lack of sufficient evidence to justify such actions and is sticking firmly to its stance, advocating for “free speech” over compliance.

The story began in April when the Brazilian government identified specific accounts allegedly disseminating misinformation and hate speech. Despite the government’s directive, X resisted, citing the demand as legally unfounded under Brazilian law and inconsistent with its newly adopted moderation policies promoting free expression.

As tensions escalated, the threat of arrest loomed over local X employees, compelling the company to close its Brazilian office and evacuate staff two weeks ago. Subsequently, Brazilian authorities insisted on X appointing a legal representative within the country. The failure to meet this demand—due to the risk of potential arrest for the appointee—led to the platform’s ban from operating in Brazil just last week.

The people of Brazil are not happy with the current regime https://t.co/04n2aSsMA8

— Elon Musk (@elonmusk) September 1, 2024

Enforced penalties and platform restrictions

Brazil’s response to X’s non-compliance has been stringent. The country has not only banned the platform outright but has also mandated that the App Store and Google Play Store withdraw the app from availability. Moreover, individuals attempting to bypass this ban through VPNs, disguising their location to access X, face hefty fines—up to $8,900 for each violation.

The controversial posts and their aftermath

The heart of the controversy revolves around baseless assertions that the 2022 Brazilian election was manipulated. Following the election results, former President Jair Bolsonaro, unsuccessful in his re-election bid, claimed the electronic voting machines were compromised, leading to a skewed outcome against him. Echoing the aftermath of the U.S. Capitol riots on January 6th, Bolsonaro’s assertions spurred his supporters to launch protests, which escalated to attacks on key governmental buildings, including Brazil’s Congress, Presidential Palace, and Supreme Court.

The repercussions were significant: over 300 arrests were made, and Bolsonaro, having fled Brazil post-election, now faces an eight-year ban from political candidacy for his role in destabilizing Brazilian democracy. The accounts targeted by the Brazilian Supreme Court are linked to Bolsonaro’s administration and persistently deny the legitimacy of the 2022 election results, potentially fueling further discord.

Notably, Elon Musk, a vocal supporter of Bolsonaro during his presidency, had engaged in multiple business initiatives, including plans to expand internet services in Brazil through Starlink. However, these agreements have been discontinued by the succeeding administration.

Brazil might fine VPN users

The New York Times has shed light on the escalating tensions surrounding X’s ban in Brazil, highlighting the substantial fines—nearly $9,000 per day—imposed on anyone in Brazil attempting to access the platform via VPNs. This extraordinary measure by Brazil’s Supreme Court underscores the seriousness with which the country is addressing the dissemination of disinformation.

The narrative surrounding the ban is not flattering for X or its CEO, Elon Musk. Initially prompted by the government’s directive to halt the spread of false information, Musk labeled the action as “censorship.” This standoff has painted Musk into a corner, as any retreat now could appear as capitulating to Brazilian authorities.

Musk’s firm stance, if genuinely driven by legal concerns and principles, aligns with his vision for X. However, this conflict has already begun to impact the platform financially and in terms of user engagement in Brazil, with many users migrating to alternative apps.

Amidst this controversy, Bluesky, a platform launched earlier this year as a counter to the monopolization by platforms like Twitter, has seen a resurgence in popularity, particularly in Brazil. Its interface, mirroring X’s blue and white design, has appealed to former users of X. Although still significantly smaller than giants like X and Threads, Bluesky’s user base reportedly doubled last month, reaching an estimated 7.79 million globally.

A spokesperson for Bluesky expressed excitement about hosting users seeking alternatives in the wake of Brazil’s ban, affirming the platform’s commitment to transparency, openness, and user-centric values. Whether Bluesky will maintain its momentum or if users will revert to larger platforms like X remains uncertain. However, for now, Bluesky appears to be benefiting from X’s exclusion from the Brazilian market.

Celebrity critiques on platform freedom

Not only has Brazil taken a firm stand against platform X, but prominent figures like Mark Cuban are also voicing skepticism about the freedom Elon Musk purports to uphold on the platform. Cuban expressed his concerns directly on X, contradicting the notion that it’s a paragon of free speech.

“I know everyone thinks this platform is a bastion of free speech. I see the opposite,” Cuban stated in a post. Musk has publicly committed to enhancing the platform by introducing new features, making algorithms open source to foster trust, combating spam bots, and ensuring all users are verified humans. However, Cuban observes a discrepancy between these promises and the current state of affairs on X.

“It’s only as free as Elon wants it to be. He can intimidate. He can alter the algorithm. He can suspend anyone he wants at any time,” Cuban articulated, underscoring Musk’s control over the platform. “He is the ultimate arbiter of everything on here. And the biggest power play is him making everyone think their speech is free speech until he decides it isn’t, and there is nothing any of us can do about it.”

I don’t think it’s tragic. I like to engage with different people on any number of topics. Kills time when I’m doing cardio or in between emails.

I know everyone thinks this platform is a bastion of free speech. I see the opposite.

It’s only as free as Elon wants it to be.…

— Mark Cuban (@mcuban) September 1, 2024

Despite his criticisms, Cuban continues to engage on X, finding some enjoyment in the interactions. “Which is why whenever he calls me out or calls me names, I’ll throw something back at him. It’s fun,” Cuban revealed in his exchange with Voulgaris. “And you know I always enjoy and respect our back and forth.”

Cuban also mentioned the personal value he finds in these interactions, “I don’t think it’s tragic. I like to engage with different people on any number of topics. Kills time when I’m doing cardio or in between emails.”

This isn’t the first instance of Cuban accusing Musk of personalizing X’s operations to his preferences. In a recent appearance on “The Daily Show,” Cuban suggested that Musk might have tailored X’s algorithm to align with his personal biases.

Where do we draw the line between free speech and unlawful activity?

While platforms like X offer unprecedented opportunities for expression, they also face challenges in navigating the complex interplay of global legal standards and users’ diverse expectations.

Free speech involves the ability to express opinions without unwarranted censorship. However, when such expressions incite violence, spread misinformation, or disrupt public order, they veer into territories that most legal systems define as criminal. The role of tech companies in moderating such content is pivotal and often contentious, particularly when their policies intersect with national sovereignty and security concerns.

Justice in this digital age requires a delicate balance, ensuring that while the freedom to speak one’s mind is preserved, it does not become a shield for harmful activities. As platforms and policymakers grapple with these issues, the dialogue between freedom and regulation remains ever crucial. This balance will ultimately shape digital communication, ensuring that free speech does not come at the cost of public safety and order.

Featured image credit: Kerem Gülen/Midjourney

Imagine having a smart assistant that you can customize to help with just about anything, from managing projects to learning new skills. That’s exactly what Google Gemini Gems offers—personalized, AI-driven tools designed to meet your specific needs. Similar to ChatGPT’s GPTs, Google Gemini Gems allows you to create your own “mini-experts” that provide guidance, suggestions, and support on a wide range of topics.

Whether you want a Gem to act as a coding tutor, a fitness coach, or a writing editor, you can shape it to be exactly what you need, and here is how.

What are Google’s Gemini Gems?

Google Gemini Gems is a feature within the Gemini AI platform that allows users to create customized, AI-driven tools tailored to their specific needs and preferences. Think of Gems as personalized assistants or mini-experts that provide guidance, suggestions, and support on a variety of topics. Whether you need help brainstorming ideas, learning new skills, managing projects, or refining your writing, Gems can be configured to deliver just the kind of assistance you’re looking for.

At their core, Gems are built on the Gemini AI model, which is capable of understanding and processing natural language. However, what makes Gems unique is their ability to be customized based on user instructions. Users can create their own Gems by providing specific instructions and preferences, which the AI then uses to generate tailored responses. This means you can shape a Gem to act like a coding mentor, a career advisor, a fitness coach, or anything else you might need.

Here are all the things you need to understand before you start using Google Gemini Gems:

• Customization: The defining feature of Gems is their high level of customization. Users have the flexibility to define detailed instructions that tailor the Gem’s responses to their exact requirements. This allows for a highly personalized user experience.
• Memory and reusability: Gems remember the instructions and preferences set by the user, enabling them to provide consistent and relevant responses each time. This is particularly useful for repetitive tasks or ongoing projects where the same type of assistance is needed regularly.
• Ease of use: Setting up a Gem is straightforward and does not require any technical expertise. The interface is user-friendly, allowing anyone to create, edit, or delete Gems with ease.
• Cross-platform compatibility: Gems can be accessed and used across both desktop and mobile devices, making them highly accessible whether you’re at home, in the office, or on the go.
• Diverse use cases: Gems are versatile and can be applied to a wide range of scenarios, from professional development and education to personal hobbies and fitness. This flexibility makes them valuable tools for users with varying needs.
• Premade Gems: To help users get started quickly, Google provides a range of premade Gems tailored to common tasks and activities. These can be used as they are or customized to better fit specific requirements.

How to use Google Gemini Gems

Using Google Gemini Gems is quite easy if you have Gemini Advanced; follow these steps:

• Access Gemini Apps
  • Web App: Go to gemini.google.com on your computer.
  • Mobile App: Open the Gemini app on your phone or tablet.
• Choose a new Gem: 
  • On the web app, find and click on the “Gem Manager” icon.

• • On mobile, look for the “More” option next to “Chats & Gems.”
• Start a new Gem: Click “New Gem” or choose from the ones premade by Google. You can also customize premade Gems.

How to create a new Google Gemini Gem?

To create effective custom Google Gemini Gems, be clear and specific with your instructions, providing detailed guidance on what you want the Gem to do. Include preferences, examples, or scenarios to help the AI better understand your needs. To create a Gem, users follow a straightforward process in the Gemini web app:

1. Access the Gemini Platform: Start by logging into the Gemini web app or mobile app.
2. Open the Gem Manager: Navigate to the “Gem Manager” within the app.
3. Create a New Gem: Click on “New Gem” and provide a name for your Gem.
4. Provide instructions: Write specific instructions detailing what you want your Gem to do. These instructions guide the AI in responding to your queries.
5. Save and use: Once your Gem is created, you can begin interacting with it, asking questions or seeking advice based on the predefined instructions.

That’s all!  Now, whether you create your own Google Gemini Gem or choose from Google’s, it’s time to type your question or prompt into the text box and hit “Submit.” For instance, if you’re using a “Writing Editor” Gem, you might type, “Please review this paragraph for grammar.”

Limitations

While Gems are powerful tools, there are some limitations to keep in mind:

• Feature availability: Not all features are available in every language or region. The availability of Gems may vary depending on your location and the language settings of your Gemini app.
• Terms of service: Users must adhere to Google’s Terms of Service and Prohibited Use Policy when creating and interacting with Gems. This is especially important for users with work or school accounts.
• No image generation: Gems do not currently support image generation and cannot be used with Gemini Live.
• Gradual rollout: Gems’ availability is gradually rolling out, which means some users might not have access to them immediately.

Google has once again upped the ante for artificial intelligence with its recent announcement of improved versions of its Gemini AI models.

As the tech giant accelerates toward the release of Gemini 2.0, the company is making waves with the introduction of the Gemini 1.5 Flash-8B, an enhanced variant of the existing Gemini 1.5 Flash, and a more robust version of the Gemini 1.5 Pro.

These updates, according to Google, represent significant strides in performance, particularly in areas like coding, complex problem-solving, and the ability to handle extensive data inputs.

Gemini’s evolution

The latest iterations of the Gemini models are not just incremental updates but reflect Google’s strategy to lead the next wave of AI innovation. The Gemini 1.5 family, first introduced earlier this year, was designed with the capacity to manage long contexts and process multimodal inputs, such as documents, video, and audio, over large token sequences. This capability alone set a new standard for how AI can be applied in various domains, from research and development to practical applications in coding and content generation.

With the introduction of the Gemini 1.5 Flash-8B, Google has provided a more compact yet powerful variant that retains the core strengths of its predecessor. This model is tailored for efficiency without sacrificing the ability to process and reason over fine-grained information. It’s a move that aligns with the growing demand for AI models that can be deployed across a range of devices and platforms without the heavy computational costs traditionally associated with large language models (LLMs).

Today, we are rolling out three experimental models:

– A new smaller variant, Gemini 1.5 Flash-8B
– A stronger Gemini 1.5 Pro model (better on coding & complex prompts)
– A significantly improved Gemini 1.5 Flash model

Try them on https://t.co/fBrh6UGKz7, details in 🧵

— Logan Kilpatrick (@OfficialLoganK) August 27, 2024

Gemini 1.5 Flash and Pro

Google’s latest updates are particularly noteworthy for the performance enhancements in the Gemini 1.5 Flash and Pro models. The Gemini 1.5 Flash, which has been described by Google AI Studio’s product lead Logan Kilpatrick as “the best in the world for developers,” shows massive gains across internal benchmarks. This model has been optimized for developers who require fast, reliable processing power for complex tasks. Whether it’s generating code, analyzing large datasets, or engaging in intricate problem-solving, Gemini 1.5 Flash is now better equipped to handle these challenges with improved speed and accuracy.

On the other hand, the Gemini 1.5 Pro model, which has always been geared toward more specialized applications, has seen a marked improvement in its ability to tackle math-related tasks and complex prompts. This is a crucial development for industries that rely heavily on precise calculations and the generation of complex code structures. The enhanced Pro model is also touted as a “drop-in replacement” for the previous iteration released in August, making it easier for developers to transition to this new version without the need for significant adjustments to their workflows.

Google’s strategic approach to AI innovation

The rapid rollout of these Gemini updates reflects Google’s broader approach to AI innovation, which is characterized by frequent iterations and the incorporation of user feedback. According to Kilpatrick, these experimental models serve as a critical testing ground that allows Google to refine and perfect its offerings before releasing them on a wider scale. By making these models available for free testing through platforms like Google AI Studio and the Gemini API, Google ensures that developers have the opportunity to engage with the latest technology and provide feedback that can shape future versions.

Imagen 3 is now available for free via Google AI Test Kitchen

This strategy is particularly important as Google races toward the release of Gemini 2.0, which is expected to bring even more advanced features and capabilities to the table. The iterative process not only helps Google stay ahead of its competitors, but it also fosters a sense of community and collaboration within the developer ecosystem. This approach contrasts with the more traditional, slower-paced development cycles seen in other tech companies, where major updates are few and far between.

Community reactions are mixed but engaged

As with any major release, the updated Gemini models have sparked a range of reactions from the AI community. On platforms like X (formerly Twitter), feedback has ranged from enthusiastic praise to pointed criticism. Some users have lauded the speed and efficiency of the new models, particularly in image analysis and processing tasks. Others have expressed frustration with the frequency of updates, arguing that they would prefer a more substantial leap forward with the release of Gemini 2.0 rather than a series of incremental improvements.

Critics have also pointed out some lingering issues, such as the models’ occasional tendency to repeat phrases or generate less coherent outputs when tasked with producing longer texts. These concerns echo similar critiques leveled at other LLMs, suggesting that while Google’s Gemini models have made significant strides, there is still room for improvement, particularly in the realm of natural language processing and generation.

The path to Gemini 2.0

Despite the mixed reviews, it’s clear that Google is committed to pushing the boundaries of what’s possible with AI. The rapid development and release of the Gemini 1.5 variants underscore the company’s dedication to staying at the forefront of AI innovation. As we look ahead to the anticipated release of Gemini 2.0, there’s no doubt that Google will continue to refine its models, taking into account the feedback from its community of developers and AI enthusiasts.

In the meantime, the Gemini 1.5 Flash and Pro models represent significant advancements in the capabilities of large language models, offering developers powerful new tools to tackle increasingly complex tasks. Whether these models will fully meet the high expectations set by the community remains to be seen, but one thing is certain: Google is not slowing down in its quest to dominate the AI landscape.

As the AI arms race continues, the introduction of stronger and more capable models like Gemini 1.5 Flash-8B and the enhanced Pro variant show that Google is not just keeping pace with its competitors—it’s setting the standard for what the future of AI will look like.

Featured image credit: Google

The battle between tech giants has heated up with the Google-Yelp lawsuit. Yelp, the popular review platform, is accusing Google of using its power as the top search engine to push its own services and shut out competitors. This isn’t just a fight between two companies—it’s a challenge to Google’s dominance that could change how we search for and find information online. Here’s what Yelp is claiming and why this case could have a big impact on the future of the internet.

What we know about the Google-Yelp lawsuit so far

Yelp has filed an antitrust lawsuit against Google. The company accuses Google of using its dominant position in the search engine market to promote its services and unfairly hurt competitors like Yelp. The Google-Yelp lawsuit is part of a larger battle over how much power big tech companies should have in controlling what users see online. Here are Yelp’s main allegations against Google:

• Manipulating search results: Yelp claims that Google unfairly promotes its own services, such as Google Maps and Google Reviews, when users search for local businesses. For example, if you search for a restaurant on Google, the top results might show Google’s own reviews and information rather than results from Yelp or other competing platforms. Yelp argues that this manipulation keeps users from visiting its site, reducing its visibility and traffic.
• Monopoly power: The Google-Yelp lawsuit states that Google’s strong position as the most widely used search engine allows it to dominate the local search market as well. This dominance means that Google can control what users see, keeping them within its own ecosystem and preventing them from going to rival sites like Yelp, Expedia, or finance sites.

• Exclusive contracts with mobile companies: Yelp also points out that Google has made exclusive deals with companies like Apple to be the default search engine on mobile devices. This strategy ensures that most mobile searches are automatically directed to Google, giving it an unfair advantage over competitors who don’t have the same level of access.
• Lower quality of Google’s local services: Yelp argues that Google’s local search services are not as good as Yelp’s. It cites a report from the Federal Trade Commission (FTC) showing that 32% of reviews on Google have no text, while Yelp requires detailed text reviews, resulting in higher-quality content. Yelp claims that Google’s dominance is due to its market power, not because it offers better services.
• Harm to Yelp’s business: Yelp claims that Google’s practices have directly hurt its business by lowering the traffic to Yelp’s website, which in turn reduces its advertising revenue and increases costs. Yelp also says that Google’s dominance forces local businesses to rely more on Google for advertising, allowing Google to charge higher fees and making it harder for Yelp to compete.

For more than ten years, Yelp has complained about Google’s practices, arguing that the search giant manipulates its search results to prioritize its own services over those of competitors. This isn’t the first time Yelp has raised these concerns, but it is the first time the company has filed a formal lawsuit to address these issues.

Why is Yelp suing Google now?

Yelp is suing Google because it believes Google has been using its dominant position in the search engine market to promote its own services, hurting competitors like Yelp unfairly. The Google-Yelp lawsuit comes after a recent ruling by a federal judge that found “Google is an illegal monopoly” by abusing its market power to stifle competition. Yelp has long argued that Google manipulates search results to favor its own platforms, like Google Maps and Google Reviews, over those of competitors, making it harder for users to find content from other sites like Yelp.

Yelp’s lawsuit aims to address these concerns by seeking monetary damages and asking the court to order Google to stop these practices. The timing of the Google-Yelp lawsuit is significant because the recent antitrust ruling against Google has created a more favorable environment for companies like Yelp to challenge Google’s practices. By filing the lawsuit now, Yelp hopes to take advantage of this momentum and push for a more competitive and fair online marketplace. If successful, this case could lead to significant outcome, including:

• Potential for more lawsuits: If Yelp wins its case, it could encourage other companies that feel disadvantaged by Google’s practices to file their own lawsuits. Other specialized search providers like Expedia, Glassdoor, and Zillow might see this as an opportunity to challenge Google’s dominance and seek compensation for similar grievances.
• Possible changes to Google’s business practices: A ruling in favor of Yelp could force Google to change how it displays search results and interacts with competitors. This might mean showing more results from outside sources like Yelp, rather than prioritizing its own services, which could provide a more level playing field for other companies.
• Increased regulatory scrutiny: Google-Yelp lawsuit is part of a growing wave of regulatory scrutiny on big tech companies. Governments and regulators worldwide are looking closely at the power of companies like Google, Facebook, Amazon, and Apple, and are considering new rules to ensure fair competition. Yelp’s lawsuit could add to the momentum for stronger antitrust enforcement and regulation in the tech sector.
• Impact on consumer choice: At the heart of the Google-Yelp lawsuit is a concern about consumer choice. Yelp argues that Google’s practices limit users’ access to a wide range of information by keeping them within Google’s ecosystem. If Yelp’s claims are proven true, it could lead to calls for more transparency and fairness in how search engines operate and display information.

Google’s response to the lawsuit

Google has denied Yelp’s allegations, calling them not new. A spokesperson for Google said similar claims have been made before and were dismissed by the FTC and other regulatory bodies. Google argues that users prefer its services because they are of higher quality and provide the best search experience.

Google also plans to appeal Judge Mehta’s recent ruling, stating that the judge’s decision recognized Google as the best search engine, which is why users prefer it over competitors like Yelp.

DeepMind’s internal revolt: Employees demand Google cut ties with Israeli and other military contracts

A timeline for Google, Yelp & other antitrust cases

Yelp has created a detailed page about the Google lawsuit. According to this page, here’s a quick recap of the events that led up to it:

2007:

• Yelp ends a licensing deal that allowed Google to access its ratings and reviews.

2009:

• Google’s attempt to acquire Yelp is unsuccessful.

2010:

• Google starts incorporating content scraped from competitors, including Yelp, without permission or attribution.

2011:

• Jeremy Stoppelman testifies before the U.S. Senate about Google’s anticompetitive practices.

2013:

• The FTC closes its probe into Google’s anticompetitive search behavior; later leaked documents reveal staff had recommended tougher action.

2014:

• The EU reopens its antitrust investigation into Google following complaints, including from Yelp.
• The EU proposes an antitrust settlement requiring Google to display more competitor offerings in search.

2015:

• EU competition chief Margrethe Vestager files formal antitrust charges against Google.

2017:

• Yelp notifies the FTC of Google’s continued scraping of third-party content.
• The EU fines Google €2.42 billion for illegally self-preferencing its Google Shopping vertical in general search.

2019:

• Jeremy Stoppelman participates in the U.S. House antitrust probe into Google.

2020:

• The House Judiciary Committee Report finds Google leverages dominance by misusing data and self-preferencing.
• Yelp testifies before the U.S. Senate Judiciary Subcommittee on Antitrust regarding self-preferencing.

2021:

• The Turkish Competition Board fines Google for self-preferencing hotel and local content.

2023:

• In a San Francisco antitrust lawsuit, Epic wins a jury trial against Google for illegally monopolizing Android app distribution and in-app billing.
• The U.S. DOJ sues Google in federal court in Virginia for illegally monopolizing digital advertising technologies.

2024:

• March: The European Union (EU) opens an investigation into Google for self-preferencing in violation of the Digital Markets Act.
• May: The U.K. enacts the Digital Markets, Competition, and Consumers Act, which Yelp had urged the House of Commons to pass in 2023.
• A case brought by the U.S. DOJ and state attorneys general results in Judge Mehta ruling that Google used illegal means to maintain its monopolies in the general search and general search text advertising markets.
• August: Yelp files its antitrust lawsuit against Google in federal court in San Francisco.

The outcome of the Google-Yelp lawsuit could set a precedent for how tech giants are regulated in the future.

This is a developing story.

Featured image credit: Eray Eliaçık/Bing

Google’s Gemini AI is making strides in enhancing user experience with the introduction of two highly anticipated features: “Ask about this screen” and the ability to summarize YouTube videos.

These new functionalities are poised to seamlessly integrate into the Google ecosystem, offering users a more efficient and informative way to interact with their devices.

No more confusions thanks to Gemini

Ask about this screen is a feature that allows users to query Gemini about any information displayed on their screen. Whether it’s a document, webpage, or image, Gemini can provide summaries, answer questions, or offer insights based on the content. This feature is particularly useful for quickly extracting key information from complex or lengthy documents.

Yet, some people like umang on X did not really like it.

I opened a presentation in Google drive.

I didn’t realize I had Gemini open.

It gave a great summary of the content in bullet points.

This was my first experience with Gemini.

I then asked it to create images, but wasn’t impressed.

How has your experience been with…

— umang 🚢 (@umangv_) August 25, 2024

Are you tired of noisy YouTubers?

Another exciting addition is Gemini’s capability to summarize YouTube videos. By analyzing the video’s content, Gemini can generate concise summaries that capture the main points and themes. This is a significant time-saver for users who want to quickly understand the content of a video without having to watch it in its entirety.

Moreover, Gemini’s ability to take notes on Meet video calls further strengthens its position as a productivity tool. Users can now have Gemini automatically generate summaries of their meetings, capturing important discussions and action items. This feature is particularly valuable for professionals who need to keep track of multiple meetings and tasks.

How does it work? See it from Saadh Jawwadh‘s post on x below!

📝 AI note taker for Google Meet:

Google Meet is getting a Gemini powered note taking feature called “Take notes for me”

This will automatically take notes and there will be a quick summary available at a glance and also after the meeting the notes document will be available. pic.twitter.com/5YHfbaCitR

— Saadh Jawwadh (@SaadhJawwadh) August 27, 2024

The Google ecosystem advantage

These new features seamlessly integrate into the Google ecosystem. Users can access them through Google Assistant, Google Search, or other Google products. This integration makes it easy for users to leverage Gemini’s capabilities without having to learn a new interface or platform.

Google Gemini’s latest features demonstrate its potential to revolutionize how we interact with technology. By providing users with the ability to ask questions about on-screen content, summarize YouTube videos, and take notes on video calls, Gemini is becoming an indispensable tool for productivity and information retrieval. As Gemini continues to evolve, we can expect even more innovative features that will further enhance our digital experiences.

Featured image credit: Google

A major energy struggle is unfolding in the United States, as big tech companies and cryptocurrency miners clash over power supplies. As artificial intelligence (AI) and cloud computing data centers grow rapidly, they are competing fiercely with bitcoin mining for electricity. This competition is changing the way energy is used and who gets access to it.

The new power struggle

According to Reuters, U.S. tech giants are snapping up energy resources from bitcoin miners to fuel their expanding AI and cloud computing centers. These data centers are experiencing a huge rise in electricity demand, which is expected to reach up to 9% of all U.S. electricity by the end of the decade. This is more than double their current usage and is outpacing the growth of power grids. As a result, tech companies like Amazon and Microsoft are scrambling for electricity wherever they can find it. Just recently, Donald Trump pointed out this issue in an interview, too.

Donald Trump says AI will be "the oil of the future" and to meet the electricity requirements it needs will require investing in nuclear power and building lots of small plants at the sites of data centers pic.twitter.com/A6gBAgY5DT

— Tsarathustra (@tsarnick) August 26, 2024

This scramble for power is impacting the bitcoin mining industry. Some miners are making significant profits by leasing or selling their energy infrastructure to tech firms, while others are struggling to maintain their operations due to reduced access to electricity.

Is AI green, how sustainable is it?

Bitcoin miners’ new challenges

Bitcoin miners are facing a tough situation. Some are able to make good deals by renting out or selling their power resources, but many are losing access to the energy they need. Greg Beard, CEO of Stronghold Digital Mining, emphasizes the intensity of this competition: “The AI battle for dominance is a battle being had by the biggest and best capitalized companies in the world, and they care like their lives depend on it that they win.” This fierce competition is reshaping the energy market.

Shifting to AI

Bitcoin miners are starting to pivot towards AI and cloud computing, but this transition comes with significant challenges. Analysts predict that by 2027, up to 20% of bitcoin miners’ power capacity might shift to AI. However, turning a bitcoin mining facility into an AI data center is not straightforward. It requires expensive upgrades, such as advanced cooling systems and new infrastructure.

The timeline for setting up new AI data centers is also much longer compared to bitcoin mines. While bitcoin mines can be set up in six to twelve months, a high-tech data center may take up to three years. This difference in setup time is crucial for tech companies that need to move quickly.

Financial disparities

The financial resources of tech giants make a big difference in this energy competition. Companies like Amazon have large capital reserves and can afford to invest heavily in acquiring and developing energy resources. In contrast, many bitcoin miners are struggling financially and cannot compete with the financial power of tech giants. For example, Marathon Digital Holdings, the largest publicly traded bitcoin miner, was interested in a nuclear-powered data center but lost out to Amazon in the deal.

Looking ahead

The battle between AI-driven tech companies and bitcoin miners over energy resources is transforming the U.S. energy landscape. As data centers and cryptocurrency mining vie for power, the energy market is evolving rapidly. Technology companies are investing heavily in securing energy assets, while bitcoin miners face the challenge of adapting to this new competitive environment.

As this energy race continues, the way these two sectors interact will shape the future of energy use and availability. The outcome of this competition will have lasting effects on both industries and the overall energy market.

Featured image credit: Eray Eliaçık/Bing

Imagine waking up in the morning, grabbing a cup of coffee, and then replaying your dreams from the night before like a movie. It sounds like something straight out of a sci-fi thriller, but according to some experts, the technology to make this possible may be closer than we think. Dr. Peter Bentley, writing for BBC’s Science Focus, suggests that the idea of dream-recording AI isn’t a question of “if,” but “when.” And he might be onto something, as recent developments in brain imaging and AI seem to indicate that we are inching toward this once far-fetched concept.

The science behind dream recording

Before we dive into the future possibilities, it’s important to understand the current science that might one day make dream recording a reality. A 2023 Japanese research study has already laid some of the groundwork. The researchers used functional magnetic resonance imaging (fMRI) scanners to record brain activity in sleeping participants. But there’s a catch: the study didn’t actually focus on dreams. Instead, it concentrated on sleep onset—the first couple of stages of sleep where people experience visual imagery, something akin to mild hallucinations. By using machine learning, the scientists were able to classify various perceived objects, like a key or a person, based on the brain activity observed during this phase.

However, these were not actual dreams, just the initial flickers of visual imagery that come as we drift off to sleep. The researchers chose this approach for a practical reason—participants could wake up immediately and describe what they saw, making it easier to match brain activity with specific images. This, while groundbreaking, is still a far cry from recording and replaying full-blown dreams.

To move from sleep onset imagery to full dream recording, researchers would need an enormous amount of detailed fMRI data from dreaming subjects. This is where things get tricky. For one, the volunteers in these studies would need to be exceptionally good at remembering their dreams in vivid detail, something not everyone can do. Accurate dream recollection is crucial because it would serve as the basis for verifying the AI’s predictions about what was dreamed.

Even more challenging is the question of how to gather this data consistently and reliably. Dr. Bentley acknowledges that while we have vast datasets of fMRI brain activity recorded while participants are awake—watching videos, listening to spoken words, or reading text—capturing similar data during REM sleep, when dreams are most vivid, is far more difficult.

Is GenAI the missing link?

Let’s assume, for a moment, that researchers manage to collect the necessary fMRI data from dreaming participants. The next step would involve using this data to train an AI capable of translating brain activity into a visual or textual representation of the dream. This is where generative AI could come into play. Generative models, such as OpenAI’s Sora and Google DeepMind’s Lumiere, are already adept at creating video sequences that mimic the chaotic and surreal nature of dreams.

In theory, you could feed a textual description generated by a dream-recording AI into one of these generative models to produce a video sequence that reflects the content of the dream. But there’s a significant caveat. These AIs aren’t actually “reading” minds. They’re matching patterns of brain activity to images they’ve encountered before and then stringing those images together into a coherent, if somewhat surreal, narrative.

As one of the sources explains, the final video might look “spookily dream-like” and contain many elements from the original dream, but it won’t be an exact replica. The generative AI wouldn’t know whether its output accurately reflects your dream; it would just be piecing together what it thinks you might have seen, based on the brain activity it was trained to interpret.

How artificial intelligence can enhance binary options trading

The idea of recording and replaying dreams raises a host of ethical questions. For starters, how would this technology impact privacy? Dreams are intensely personal experiences, often reflecting our deepest fears, desires, and memories. The ability to record them, especially if the technology becomes widely available, could lead to situations where individuals feel pressured to share or even manipulate their dreams.

Moreover, the accuracy of such technology would be under constant scrutiny. As noted earlier, these AIs are not perfect mind-readers; they’re pattern-matchers. While they might produce dream sequences that feel eerily familiar, they are unlikely to be perfectly accurate representations. This could lead to potential misunderstandings or misinterpretations, especially if dream recordings were to be used in legal or psychological contexts.

There’s also the question of technological feasibility. While fMRI machines have become more advanced, they are still bulky and noisy, making it difficult to imagine sleeping through a full night’s scan. Portable, silent fMRI machines that could record brain activity without disturbing sleep would be a significant hurdle to overcome before dream recording could become a reality.

The road ahead

So, where does this leave us? Dr. Bentley suggests that given the rapid pace of AI research and development, it wouldn’t be surprising to learn that someone, somewhere, is already working on dream-recording AI. Yet, he also acknowledges that this technology is likely still years, if not decades, away from practical application.

For now, the idea of replaying our dreams remains in the realm of speculation. But as research into brain activity and AI continues to evolve, what seems like science fiction today could very well be tomorrow’s reality. Until then, our dreams remain our own—fleeting, mysterious, and unrecorded.

Image credits: Kerem Gülen/Midjourney