A Meta spokesperson confirmed to Bloomberg on Wednesday that AI giant Yann LeCun is exiting Zookland and striking out on his own. According to a memo from LeCun, which Bloomberg claims to have read, LeCun’s new effort aims to “bring about the next big revolution in AI: systems that understand the physical world, have persistent memory, can reason, and plan complex task sequences.”
Sources apparently told Bloomberg that LeCun “clashed with others internally.” Meta recently created an entirely separate AI research department focusing on generative AI, and in its latest story, Bloomberg now claims that Meta had begun to overshadow Laken in favor of high-profile recent hires. Recent hires include Shengjia Zhao, co-creator of ChatGPIT.
As previously discussed here at Gizmodo, Laken is fascinated by an area of AI called “world models.” He has spent over a year saying that he thinks LLM research, the backbone of systems like ChatGPT, is no longer a worthy area of exploration – at least as far as hypothetical advanced AI functions with terms like “AGI” and “superintelligence” are concerned.
LeCun, who was born and raised in France, is among a handful of researchers often referred to as the “Godfather of AI” or more specifically the Godfather of deep learning, and he shared the Turing Award in 2019 with fellow Godfathers Geoffrey Hinton and Yoshua Bengio. Influential cognitive scientist and AI researcher Gary Marcus is a longtime critic of Lacan, and his public disagreements go back years.
LeCun joined Meta in 2013, when it was still called Facebook, as the head of a research operation with a location in New York where LeCun could walk from his office at NYU, where he works as a professor. At the time, it was not entirely clear what a company like Facebook wanted from a scientist working with deep neural networks. Another prominent AI researcher, Andrew Ng, explained Facebook’s hiring decision to Wired in terms that now seem quaint and social media-centric:
“Machine learning is already used in hundreds of places across Facebook, from photo tagging to ranking articles in your News Feed. Better machine learning will be able to help improve all of these features, as well as help Facebook create new applications that none of us have even dreamed of yet.”
As AI began to dominate all priorities in the tech world following the release of ChatGPIT in 2022, LeCun became notable for his skepticism about the need for AI security. He told the Wall Street Journal last year that the idea that AI is a threat to humanity is “complete BS.”
But LLM is not LeCun’s cup of tea anyway. He clarified last month that he had almost no involvement with Meta’s llama model, and that such generic AI-related work did not occur in any other department of Meta. He explained that LeCun worked in META’s Fundamental AI Research (FAIR) department and was trying to go “beyond the LLM.”
Lacan believes there is a need for AI models that can comprehensively understand the physical world through sensory inputs such as vision, and how to navigate their way through interactions and changes in that world. He believes that the current generation of AI systems can’t even come close to this, and that they are actually dumber than cats.
You can already see the beginning of LeCun’s world model research under the auspices of META in V-JEPA-2. That model is trained not on text, but on videos of the physical world, and like Sora is not designed to simply replicate all that video, but to model the cause and effects of actions in the world as things move and interact. That’s the principle anyway.
Bloomberg writes that Meta “plans to partner with LeCun on its startup, though details are still being finalized.” In Lecan’s memo, he wrote that his former company “will be a partner of the new company and will have access to its innovations.”
It’s not yet clear exactly how the partnership between LeCun’s new company and Meta will be structured, but the tech companies are famous for being almost inseparable from each other as far as AI is concerned. Microsoft owns about 27% of OpenAI, and has exclusive rights to use its technology. Similarly, Google has 14% stake in Anthropic. The way interdependent investment leads to higher valuations in the AI world has been compared to “circular dealmaking”.
LeCun’s memo stated that his new technology would have “far-reaching applications in many sectors of the economy, some of which overlap with Meta’s business interests, but many of which do not.”
LeCun famously favors the term Advanced Machine Intelligence (AMI) in place of something like AGI (nota bene: “ami” is French for “friend”). In his memo, he reportedly wrote that “pursuing AMI’s goal in an independent entity is one way to maximize its broader impact.” It’s an appropriately obscure turn of phrase. Presumably the “independent entity” is the new, non-meta company, not an intelligent entity. Although it could also mean this.