Yann LeCun, one of Meta’s most well-known artificial intelligence experts, will reportedly leave the company and begin preparations to establish his own startup. According to the Financial Times, based on sources familiar with the matter, Yann LeCun plans to leave his position at Meta in the next few months and has already started negotiations with investors. The focus of the initiative will be on “world model” studies in the field of artificial intelligence.
The world model is defined as an approach that enables an artificial intelligence system to internally understand its environment and predict future possibilities by simulating cause-effect relationships. This technology is also actively developed by leading research institutions and startups such as Google DeepMind and World Labs. What kind of innovation LeCun’s new initiative will bring in this field is already being followed with interest.
Yann LeCun may be leaving Meta due to chaos in AI structure
Meta has recently faced criticism that it has fallen behind in the competition in the field of artificial intelligence, and in response to this situation, it has made comprehensive changes in its organizational structure. The company established a new unit called Meta Superintelligence Labs (MSL) and transferred more than 50 engineers and researchers from competitors such as OpenAI, Google and Anthropic. In addition, a $14.3 billion investment was made in June in Scale AI, a company operating in the field of data labeling, and the company’s CEO, Alexandr Wang, was appointed as the head of MSL.
However, this rapid change in structure caused some unrest within Meta. According to information provided by TechCrunch in August, while new teams had difficulty moving within the large corporate structure, the authority of Meta’s previously productive artificial intelligence team was also narrowed. This changed the research balance within the company and resulted in the decline of the influence of the Fundamental AI Research (FAIR) unit, which focuses on long-term artificial intelligence developments.
The long-term research carried out by LeCun within FAIR also seems to be overshadowed by these new structures. After the company’s previous Llama 4 model family fell short of the competition, CEO Mark Zuckerberg’s restructuring decisions directly affected FAIR’s research direction. However, the aim of FAIR was to develop artificial intelligence techniques that could be used five to ten years later.
LeCun is one of the few top AI experts who think the potential of currently widely used large language models (LLMs) has been overrated. In his previous social media posts, he emphasized that today’s artificial intelligence systems have not yet surpassed the level of a house cat, and therefore he found discussions about controlling systems with superhuman intelligence unnecessary.
Meta did not comment on LeCun’s departure. It is not yet clear how the company will shape the future of the FAIR unit and its position in world model research following this development.