Yann LeCun, Meta's Chief AI Scientist and a Turing Award laureate, is reportedly leaving Meta to establish his own startup. This significant departure, reported by the Financial Times on November 11 (local time), is viewed as a major blow to Meta due to LeCun's prominent status as one of the "four AI kings." His new venture is expected to concentrate on advancing "World Models," a next-generation AI system that aims to understand the physical world by learning from language, video, and spatial information, a project LeCun estimates will take approximately 10 years to fully realize.

LeCun's exit is set against the backdrop of a major strategic shift within Meta's AI division, driven by CEO Mark Zuckerberg. This reorientation prioritizes accelerating the market release of AI products, particularly Large Language Models (LLMs), to enhance competitiveness against rivals such as OpenAI and Google. This contrasts with the long-term, fundamental research focus of the 'Fundamental AI Research (FAIR)' lab, which LeCun has led since 2013. Zuckerberg has directly initiated the 'TBD Lab,' an exclusive internal team dedicated to developing cutting-edge LLMs, leading to a change in LeCun's reporting structure from the Chief Product Officer to Alexander Wang. A contributing factor to LeCun's departure appears to be his philosophical divergence regarding LLMs; he has publicly asserted that while LLMs are useful, they lack the capacity for human-like reasoning and planning, a capability his 'World Model' research, distinct from current LLMs, seeks to achieve. This marks another high-profile departure from Meta's AI division, following Joel Pino, VP of AI Research, who joined Canadian AI startup Cohere in May.