Meta’s head AI scientist is reportedly planning to leave the company and launch his own startup — the biggest exit yet in Mark Zuckerberg’s push to develop “superintelligence” at the social media giant.
Yann LeCun, 65, a Turing Award winner considered a pioneer in AI development, has told colleagues he will leave Meta in the coming months, people familiar with the conversations told the Financial Times.
The French-American scientist is in early talks to raise funds for his own venture, the people said.
Shares in Meta fell 1.2% Tuesday morning as investors grow antsy for CEO Zuckerberg to prove massive spending on the new tech will pay off.
Zuckerberg has been trying to inject fresh energy into the AI division as Meta struggles to compete with behemoth rivals like OpenAI and Google.
Zuckerberg has been pushing staffers to focus on more rapid product rollouts. That’s a pivot away from Meta’s Fundamental AI Research Lab (FAIR), which focuses on longer-term research — and which LeCun has run since 2013.
Meta — which owns Facebook, Instagram and WhatsApp — shelled out more than $14 billion over the summer to acquire a 49% stake in Scale AI and hire its founder, Alexandr Wang, to lead the new “superintelligence” division.
LeCun, who previously reported to chief product officer Chris Cox, has since reported to 28-year-old Wang, according to the report.
Zuckerberg has also been luring over staffers from rival firms with $100 million-plus pay packages, which has frustrated longtime Meta employees.
In July, Meta hired Shengjia Zhao, co-creator of OpenAI’s ChatGPT, as chief scientist of its new “superintelligence” division.
These new teams of staffers have been focused on the development of large language models to compete with market giants like ChatGPT after Meta’s Llama 4 model bombed. Its Meta AI chatbot has also failed to gain popularity.
LeCun and Zuckerberg seem to hold differing views on the future of AI.
The award-winning scientist has said the company’s large language models will be “useful,” but never able to completely replicate human reasoning.
Zuckerberg, meanwhile, has touted the potential of AI, arguing that it will be able to complete most of Meta’s coding within a year or so.
At FAIR, LeCun has been working on a new generation of AI systems that will be able to learn about the physical world through videos and spatial data instead of just language. But he has cautioned that it could take a full decade to fully develop these models.
His next project is reportedly focused on further developing these world models, two people familiar with the matter told the Financial Times.
Representatives for LeCun and Meta did not immediately respond to The Post’s requests for comment.
Zuckerberg previously said Meta’s “superintelligence” lab will cost hundreds of billions of dollars.
But the Meta founder — like other tech firms — is facing growing pressure to prove big spending will pay off. Investors led a massive tech stock sell-off last week as they panicked that AI potential has been vastly overvalued.
Shares in Meta tanked over 12% in late October — erasing nearly $240 billion from its valuation — after Zuckerberg said AI spending could exceed $100 billion next year.
The company has also lost valuable team members, like Joelle Pineau, vice president of research, who left Meta in May to join Cohere, a Canadian AI startup.
Last month, Meta slashed 600 roles in its AI research unit to cut costs and eliminate bureaucracy.
