Yann LeCun to Exit Meta and Launch World Models AI Start-Up
Created on November 12|Last edited on November 12
Comment
Yann LeCun, Meta’s long-time chief AI scientist and one of the founding minds behind modern deep learning, is preparing to leave the company. His decision comes as Meta shifts away from open-ended research and leans heavily into commercial AI systems, particularly large language models. LeCun, who won the Turing Award in 2018, is now working on a new venture aimed at building “world models," which are systems designed to learn from real-world physical and visual inputs instead of just text.
Meta’s AI Strategy Shifts Toward Superintelligence
Under Mark Zuckerberg’s leadership, Meta has refocused its AI division toward large-scale commercial deployment. The FAIR lab, which LeCun founded in 2013 to explore foundational AI research, has been restructured. Zuckerberg is now channeling resources into Meta’s Superintelligence division, newly led by Alexandr Wang of Scale AI. Wang came onboard after Meta acquired a 49 percent stake in his company for 14.3 billion dollars. The new division also includes TBD Lab, a team designed to attract top-tier AI talent with massive compensation packages.
LeCun’s Disagreement Over LLM-First Approaches
LeCun has been openly skeptical of relying too heavily on LLMs, arguing that while they are powerful, they lack the capacity for planning and physical reasoning. His proposed “world models” aim to build systems that can understand and interact with their environments in a more human-like way. He sees this as a more viable path toward machines that can truly reason, rather than just respond.
Internal AI Leadership Turnover at Meta
LeCun’s departure is part of a broader shake-up within Meta’s AI ranks. Joelle Pineau, former VP of AI research, recently left for Cohere. At the same time, Meta laid off around 600 AI staffers. To reinforce its AI leadership, Meta brought in Shengjia Zhao, a former OpenAI researcher and one of the co-creators of ChatGPT, to lead scientific work inside the Superintelligence Lab.
What LeCun’s Exit Means for Meta and the AI Sector
LeCun’s start-up could emerge as a philosophical and technical rival to the LLM-focused arms race now dominating the industry, led by OpenAI, Google DeepMind, and Meta itself. Whether “world models” become a viable counterforce or remain a long-term research ambition will depend on whether LeCun can turn his theory into working systems.
Add a comment