Catenaa, Sunday, November 23, 2025- Yann LeCun, Meta’s chief AI scientist and one of the field’s “godfathers,” is reportedly leaving the company.
He plans to start a new venture focused on developing “world models,” an advanced approach designed to give AI systems a physical and causal understanding of the world.
LeCun, 65, criticized current text-trained large language models for their inability to plan, reason, or comprehend physical environments, comparing their capabilities unfavorably even to a house cat.
The new start-up will explore AI systems that integrate embodied data such as video, spatial information, and touch to form internal models of objects, actions, and cause-and-effect relationships.
Proponents argue this method could enable AI to predict outcomes and plan behaviors in ways current LLMs cannot.
LeCun has begun early fundraising discussions and is expected to work on technologies that may take a decade or more to fully mature.
His departure follows a restructuring at Meta that emphasized rapid deployment of LLM-based products under CEO Mark Zuckerberg, including the creation of Mega Superintelligence Labs and leadership changes in FAIR, LeCun’s research division.
The move highlights a growing divide in the AI industry between fast-to-market LLM commercialization and long-term research on human-level intelligence.
Stanford professor Fei-Fei Li, Google DeepMind, and Nvidia have also invested in long-term research on world models, emphasizing spatial and causal reasoning.
LeCun’s exit is seen as a high-profile signal that a new wave of AI innovation may focus less on text-based models and more on systems capable of understanding and interacting with the physical world.
