Meta has lately emerged as one of the most lavish spenders in the AI race, offering pay packages in the billion-dollar range to top AI talents. But at the same time, the company has also made some brutal cuts in its AI division, axing well-known researchers as it struggles to find mass acceptance for its MetaAI bundle and the Llama models. Internal organizational changes and talent upheaval might soon lead to one of the biggest departures from the company — Yann LeCun, a pioneer in AI for decades.

According to the Financial Times, LeCun has told his colleagues that he plans to leave Meta in the “coming months.” The AI trailblazer, who is a winner of the prestigious Turing Award, is reportedly in talks to raise funds for his own startup. LeCun’s exit could be a massive blow to Meta, where he joined in 2013 and is currently the Chief AI Scientist at Facebook AI Research (FAIR). So far, LeCun has not spoken publicly about his plans, nor has Meta (or any of its executives) made any comments. But it seems Meta has been planning to replace LeCun.

A report published by WIRED in July notes that Meta approached an OpenAI employee and offered them the role of chief scientist at the company. The poaching target, however, declined the offer. It was at the same time that Meta made internal changes and created the Meta SuperIntelligence Lab (MSL), which is now led by new acqui-hire Alexandr Wang as the Chief AI Officer. Gary Marcus, a prominent scientist and AI critic, also lambasted the decision and wrote on X, “LeCun just a wall decoration at this point?” The potential exit of Lecun would be a huge loss for Meta.

The potential departure of Lecun has sparked diverse online commentary. “An outcome that is inevitable after Zuck spent $15B to acquire Alexandr Wang and made Yann report to him,” wrote Yuchen Jin, technology chief at Hyeprbolic Labs. There are also theories floating around that LeCun attracted heat for criticizing Large Language Models (LLMs), the foundation tech behind generative AI products such as Meta AI, ChatGPT, Gemini, and more.

LeCun — who is widely hailed for his work in deep learning, computer vision, and machine learning — has been pretty vocal about what he thinks is the future of AI, and he doesn’t seem to be hedging all his bets on LLM. In a lecture at the prestigious Indian Institute of Technology last year, LeCun said LLMs aren’t the future of AI, adding that we need new architectures and that researchers should look beyond LLMs to move the field forward. In a separate interview with Newsweek earlier this year, he opined that LLMs will be obsolete within the next half-decade. He further added that world models will “ultimately supersede” the LLMs.

It seems LeCun wants to focus on building world models at his own startup. According to The Wall Street Journal, Lecun’s startup will reportedly focus on building AI world models. The report, citing interactions with anonymous insiders, notes that LeCun might decide to stay at Meta. The likes of Nvidia and Google are already making progress at developing world models, which could make Meta’s AI glasses even smarter. Fei-Fei Li, often referred to as the godmother of AI, launched a world model named Marble in November. LeCun, who is counted among the “AI godfathers,” apparently wants to move in a direction that keeps him at the vanguard of AI development.