Yann LeCun, one of the most prominent figures in artificial intelligence, is said to be considering a move away from Meta and the establishment of his own startup. The Financial Times was the first to report this development on Wednesday, referring to anonymous sources that have knowledge of the issue.

LeCun is no ordinary researcher. Besides being the chief AI scientist at Meta, he is a lecturer at New York University, and a holder of the A.M. Turing Award (which is essentially the Nobel Prize of computing) among others. If he actually quits, it would be a notable setback for Meta.

The article mentions that LeCun is ready to go in the next few months and is already having discussions about funding. His new company would deal with world models, which has been the subject of his research for a long time.

What are world models anyway?

A world model represents an artificial intelligence system that creates an inner picture of its surroundings. The process of predicting outcomes through the simulation of cause and effect scenarios is no longer merely the response to the inputs. You may consider it as an AI which does not merely react, but actually grasps the principles involved.

Leading research institutions like Google DeepMind and new companies like World Labs are all working on the world models. It is one of the most hopeful paths for developing AI that is intuitively intelligent rather than a mere massive pattern matcher.

The timing couldn’t be worse for Meta

LeCun's possible move to another company or department within Meta is indeed unfortunate timing. Meta has made a big change to its AI development strategy lately, indicating a fear of being left behind by competitors such as OpenAI, Google, and Anthropic.

The reports indicate that Meta has started a new AI unit called Meta Superintelligence Labs (MSL) by gathering more than 50 engineers and researchers from the competition. Moreover, in the month of June, a whopping $14.3 million was invested by Meta in Scale AI, coming with the CEO Alexandr Wang for the new division.

The reports have been suggesting chaos in Meta's AI organization by sources that spoke to media in August. The new hires are tired of dealing with the corporate red tape of a big company. On the other hand, the area of generative AI, which was previously handled by Meta, has now become more limited.

LeCun's research is part of the Meta's Fundamental AI Research (FAIR) department that emphasizes techniques for long-term research which could be five to ten years later MSC is working on the opposite mandate.

However, it seems that CEO Mark Zuckerberg's decision to reshuffle when Meta's Llama 4 models could not measure up to competition's as if they were already obsolete, has put the light on FAIR's research. Long-term research is often left aside when short-term competitive pressure dominates.

What was the response from Meta

Meta did not respond right away to the inquiries for a statement. It's still unclear if LeCun's leaving is true or just a negotiating tactic for better conditions at Meta.

However, the broader trend is clear. The company is prioritizing short term competitiveness for long term research. And those researchers who are dealing with the most fundamental questions about how to create really intelligent systems are getting less support and attention.

If LeCun departs, it will be a validation of the worries that the recent reorganization at Meta has set up an atmosphere where serious AI research is going to be a slow process compared to the fast pace of ChatGPT.

It is a personal fact for LeCun that starting a company dealing with world models would be logical. He could conduct research in the way he believes without corporate pressure for immediate results from the commercial side. And his good name would ensure that capital-raising should not be a problem.

The question is still open as to whether Meta will be able to hold its fort in AI research without one of its most honored scientists. In this area, talent counts, and LeCun means a whole lot of expertise from years that simply cannot be passed on to others easily.