The Allen Institute for AI (Ai2) has made public Molmo, an innovative set of open-source multimodal models that contest the guiding influence of proprietary AI systems. With strengths in superior image recognition and actionable insights, Molmo is ready to assist developers, researchers, and startups by delivering an advanced yet easy-to-use AI application development tool. The launch brings attention to an important change in the landscape of AI, uniting open-source and proprietary models and improving everyone’s access to leading AI tech.

Molmo offers features that provide an exceptional degree of image understanding, permitting it to correctly read a wide variety of visual data—from mundane items to complex charts and menus. Instead of being like most AI models, Molmo surpasses perception by enabling users to interact with virtual and real environments through pointing and a range of spatial actions. This capability denotes a breakthrough, allowing for the introduction of complex AI agents, robotics, and many other applications that depend on a granular understanding of both visual and contextual data.

Efficiency and accessibility serve as major aspects of the Molmo development strategy. Molmo’s advanced skills come from a dataset of less than one million images, in stark contrast to the billions of images processed by other models such as GPT-4V and Google’s Gemini. The implemented approach has contributed to Molmo being not just highly efficient in using computational resources but has also created a model that is equally powerful as the most effective proprietary systems and features fewer hallucinations and quicker training rates.

Making Molmo fully open-source is part of Ai2’s larger strategic effort to democratize AI development. Ai2 enables a diverse array of users—from startups to academic laboratories—to innovate and advance in AI technology without the high costs of investment or vast computing power. It gives them access to Molmo’s language and vision training data, model weights, and source code.

Matt Deitke, Researcher at the Allen Institute for AI, told “Molmo is an incredible AI model with exceptional visual understanding, which pushes the frontier of AI development by introducing a paradigm for AI to interact with the world through pointing. The model’s performance is driven by a remarkably high quality curated dataset to teach AI to understand images through text. The training is so much faster, cheaper, and simpler than what’s done today, such that the open release of how it is built will empower the entire AI community, from startups to academic labs, to work at the frontier of AI development”.

According to internal evaluations, Molmo’s largest model, sporting 72 billion parameters, surpassed OpenAI’s GPT-4V and other leading competitors on several benchmarks. The tiniest Molmo model, including only one billion parameters, is big enough to function on a mobile device while outperforming models with ten times that number of parameters. Here you can see the models and try it for yourself.

At the recent Reuters’ Momentum AI conference in San Jose, California, Ramine Roane, AMD’s corporate vice president of data center, cloud, and AI, unveiled the company’s strategic pivot towards an open-source model to tackle the dominance of NVIDIA in the AI chip market, reports Business Insider. This shift comes as the industry grapples with a severe shortage of graphics processing units (GPUs), which are crucial for video gaming and AI development.

“The problem is vendor lock-in”

AMD and NVIDIA have been longstanding competitors in the production of GPUs. These chips are essential for handling complex computations required in AI and gaming. Despite AMD’s efforts, NVIDIA continues to hold a significant lead, controlling over 70% of the AI chips market, with major clients like Meta, Google, Amazon, and OpenAI, as per estimates by Mizuho Securities.

The surge in demand for generative AI has only intensified the need for more powerful chips, exacerbating the existing supply constraints. According to Roane, AMD is pushing its production capacities to the limit. “We are sending all the GPUs we can make right now,” Roane stated, highlighting the critical shortage within the industry.

A key issue Roane pointed out is the problem of ‘vendor lock-in,’ particularly with NVIDIA’s proprietary CUDA computing platform. Introduced in 2006, CUDA allows developers to create applications specifically for NVIDIA’s GPUs, which has not only cemented NVIDIA’s market dominance but also stifled competition due to its incompatibility with other GPUs.

AMD acquires Silo AI for $665 million

Roane criticized this closed ecosystem, noting that it restricts innovation and competition. He expressed AMD’s commitment to an open-source approach, aiming to provide a more flexible and developer-friendly alternative. This strategy could potentially enable easier access for developers across different hardware platforms, fostering greater innovation and reducing dependence on a single vendor.

As AMD champions this open-source initiative, it poses a direct challenge to NVIDIA’s entrenched position in the market. By advocating for open standards and interoperability, AMD hopes to attract a broader base of developers and industries, potentially reshaping the competitive landscape of the AI chip industry.

Featured image credit: Timothy Dykes/Unsplash

Mistral, a French AI startup, has made waves in the AI community with the release of Mixtral 8x7B, its latest open-source AI model. This model has garnered attention for potentially surpassing OpenAI’s GPT-3.5 and Meta’s Llama 2 in performance. The company adopted a unique approach by releasing its latest large language model unceremoniously via a torrent link on social media. This move contrasts the typical fanfare associated with AI releases, showcasing Mistral’s distinct, hacker-like attitude​​.

Mixtral 8x7B: A new AI powerhouse

Recently, Mistral raised an impressive $415 million in a Series A funding round, pushing its valuation to around $2 billion. This financial growth highlights the company’s success and potential in the AI sector​​. Mixtral 8x7B, employing a “mixture of experts” approach, integrates various models, each specializing in different tasks. This innovative technique has led to its impressive performance, equating or outperforming GPT-3.5 and Llama 2 in various benchmarks. Mistral released this model online, followed by an official blog post detailing its capabilities, and confirmed that it’s available for commercial use under an Apache 2.0 license​​.

Small footprint: It can run on a Mac

One of the notable features of Mixtral 8x7B is its ability to run on non-GPU devices, potentially democratizing access to advanced AI technology. The model achieves state-of-the-art results among open models, with strengths in language generation over long contexts and code generation​​​​.

For those who don’t follow AI closely:
1) An open source model (free, anyone can download or modify) beats GPT-3.5
2) It has no safety guardrails
There are good things about this release, but also regulators, IT security experts, etc. should note the genie is out of the bottle. https://t.co/nHvlNKaItw

— Ethan Mollick (@emollick) December 11, 2023

AI enthusiasts and professionals have quickly adopted Mixtral 8x7B, impressed by its performance and flexibility. The model’s small footprint allows it to run on machines without dedicated GPUs, including the latest Apple Mac computers. However, its lack of safety guardrails, as observed by Wharton School professor Ethan Mollick, has raised concerns about content deemed unsafe by other models​​.

6x faster than Llama 2 70B

Mixtral 8x7B stands out with its six times faster inference speed compared to Llama 2 70B, thanks to its sparse model architecture and eight different feedforward blocks in the Transformer. It supports multilingual capabilities, excellent code generation, and a 32k context window​​. Mistral’s valuation soared to over $2 billion in just six months, highlighting the growing importance of large Mixture of Experts models in the AI landscape​​.

Open-source with no limits

Mixtral 8x7B, an open-source model, is proving to be a game-changer. It not only outperforms some U.S. competitors like Meta’s Llama 2 family and OpenAI’s GPT-3.5 but also offers fast and efficient performance. The model’s open-source availability stands in contrast to OpenAI’s closed-source approach, aligning with Mistral’s commitment to an “open, responsible, and decentralized approach to technology”​​.

Mistral’s model is a high-quality sparse mixture of expert models (SMoE) with open weights, licensed under Apache 2.0. It has shown superior performance on most benchmarks compared to Llama 2 70B, achieving six times faster inference. This efficiency marks Mixtral 8x7B as the strongest open-weight model in terms of cost and performance​​.

There’s a controversy going on all over the internet regarding the Twitter algorithm. SpaceX and Tesla CEO Elon Musk has ambitions toward making it open-source. What would be the advantages and disadvantages from a user perspective?

Social media platforms are being heavily criticized for their content ranking algorithms. We discussed whether Elon Musk, who has controversially acquired the majority of Twitter recently, is playing with everyone, or whether the open-source algorithm can be the future of this business.

What would happen if the Twitter algorithm went open-source?

After buying Twitter for $44 billion, Elon Musk promised to make improvements to the popular social media platform, underlining free-speech and transparency he aims to make “the algorithms open source to increase trust.”

Musk told in a TED talk last month that the Twitter algorithm that controls how tweets are advertised and demoted might be uploaded to GitHub, allowing anybody to use it.

“People can look through it and say, ‘Oh, I see a problem here, I don’t agree with this. They can highlight issues and suggest changes, in the same way that you update Linux or Signal,” Musk said.

Musk’s desire to make the Twitter algorithm open-source is motivated by his long-standing concern about possible political censorship on the platform, but it is unlikely to have the impact he desires. Instead, experts caution that it may generate a slew of new issues.

Elon Musk has a deep distaste for authority, but his call for algorithmic transparency seems to correspond with the demands of national politicians. The idea has been a major pillar of several governments’ efforts to resist tech giants like Google and Facebook in recent years.

For example, Melanie Dawes, the CEO of Ofcom, which is the UK’s main authority on social media issues, has stated that social media platforms will be required to describe how their algorithm works.

The EU’s new Digital Services Act, which was approved on April 23, will also put pressure on platforms to be more transparent. Democratic lawmakers in the United States have proposed an Algorithmic Accountability Act since February 2022 to bring greater transparency and oversight of the algorithms that control our timelines and news feeds, as well as much more.

Seeing the code for an algorithm does not necessarily explain how it works, and it does not provide much information about the company’s organizational structures and processes that went into its development to the average person.

There are additional, more worrisome unintended consequences aside from this. One of the main concerns is that as people attempt to figure out how the algorithm works, they will get into fights. This might lead to yet another contentious and unproductive argument.

“I worry that it’ll be made into a mountain where it’s really just a molehill. There’s a lot of hype about the mysterious algorithm, but in reality it’s likely that bad behavior has social consequences that are reflected in the weightings of the tweets of those people,” told Catherine Flick, computing and social responsibility researcher at De Montfort University in the UK.

Pros of an open-source Twitter algorithm

From a user’s perspective, greater transparency makes sense on the surface. Users are largely unaware of the vast power and influence that social media platforms like Facebook, Instagram, and TikTok have. And just as the source code of a computer program allows you to look for flaws or backdoors in it, disclosing the code that drives Twitter may reveal whether the platform values one type of material over another.

It’s reasonable to assume that, if the Twitter algorithm would be made available for public examination, others can utilize the code once they are able to read it, either by proposing modifications to Twitter’s developers or developing their own applications based on it.

Cons of an open-source Twitter algorithm

Seeing how Twitter works internally would require more than simply uploading some code to GitHub. And proving the existence of biases that may be subtle in nature and rely on a slew of ever-changing variables might be far more difficult than Musk anticipates.

By making the algorithm flexible and open to public view, rivals may clone Twitter’s source code and rebrand it. But even though, seeing the code for an algorithm does not necessarily explain how it works, and it does not provide much information about the company’s organizational structures and processes that went into its development to the average person.

Taking action to address biases that are identified will no doubt be viewed through a political, rather than technical, point of view at a time when we’re already massively politicized. Some say taking the algorithm open-source won’t cure any issues with bias.

Revealing too much information about Twitter’s recommendation system may also lead to security concerns.

“That’s not necessarily because individuals would be able to understand the intricacies of how the code of the algorithm works. But they’d be able to discern roughly how Twitter recommends posts on users’ timelines,” says Eerke Boiten, a professor of cybersecurity at De Montfort University.

In some circumstances, an open-source Twitter algorithm might assist attackers in understanding the strategies behind the promotion of certain items over others.