Yann Lecun, Meta’s Chief Ai Scientist, is reportedly leaving the tech giant to launch his own AI startup, according to the Financial Times, citing specific sources. His departure comes as the meta ramps up its efforts in developing advanced forms of AI and internal restructuring prioritizing the meta’s core mission of AI Chestration (Fair) in the background Surker.
Meta and Lecun did not respond to requests for comment from the Observer.
The French-American computer scientist, who moved to the US in the 1980s to work at AT&T Bell Laworatories, is known for his pioneering research on machine learning. Lecun joined meta in 2013, where he helped launch the company’s right team and became an ai-Ayi scientist in 2018. In the same year, he won a migration prize for his contribution to neural networks.
Lecun previously reported to Meta’s Chief Precip Officer Chris Cox, but after a recent restructuring, he now reports to Alexandr Wang, head of the Superintelligence Division. Wang was hired earlier this year after meta invested more than $14 billion in its data mining startup, AI AI.
The Meta Talent Talent Push coincides with the company’s restructuring of AI research under TBD Labs, a group focused on realizing AI that surpasses human capabilities. Three other divisions within Meta’s AI organization are dedicated to products, infrastructure and convenience.
Fair, who specializes in long-term, exploratory research, influenced the early versions of Meta’s Llama model. Lecun’s departure marks another setback for the group, which lost its leader, Joelle Pineau, in April. Pineau is now the head of research in Canada’s Ai Startup Core.
Global models are picking up steam
Lecun’s Open Venture will grow “global models,” according to financial times. It is said that in the discussion of learning money at the beginning of the startup they focus on training AI programs to understand the physical world rather than just generating language.
His move to global models is not surprising given his skepticism about the robustness of large-scale linguistic models (LLMS). Lecun once called the LLMS a “dead end” in achieving human-like AI, instead supporting systems that can see their surroundings and understand physical concepts like gravity.
He is not alone in pursuing this. Stanford’s Fei-Fei Li raised about $ 230 million for his $ 230 million for its first labs, which are equally sought to give “local intelligence.” Google Depmind has explored world models with its release of the genre, and nvidia is pushing virtual ai with products like its cosmos world model.
Lecun believes global models, not LLMSs, are the key to building AI that can take advantage, program complex objects, and make predictions. “We’re not going to get to Ani-Level Ai with just text training,” the researcher said during a Harvard Talk in September. “Except for what you can hear from some of the most optimistic CEOs of various AI companies in Silicon Valley, it’s just not going to happen.”
