From Inc. Magazine, January 6:
Yann LeCun didn’t hold back on his [former] employer’s AI shortcomings.
After more than a decade as Meta’s chief AI strategist, AI pioneer Yann LeCun abruptly left the company in November, sparking questions about the circumstances of his resignation. Now, in a recent interview with the Financial Times, he’s offering his own account of what led him to exit Meta—and it’s not generous.
From the outset of his time at Meta, then Facebook, LeCun was granted unusual freedom to pursue AI research, free from the immediate pressure to develop a profitable product. He told the FT that “money was clearly not going to be a problem,” a reality that allowed him to follow his intellectual curiosities and conduct extensive research without market pressure.
That latitude, however, began to narrow as Meta’s AI ambitions sharpened. LeCun’s largely unconstrained research role came into clearer focus when Mark Zuckerberg tapped him to lead the development of Meta’s own large language models, later dubbed Llama. LeCun agreed on one condition: the models had to be open source and free. LeCun said Llama models “changed the entire industry,” quickly becoming a favorite among AI researchers thanks to their power and open nature.
Llama’s success, however, soon faltered. When Meta released Llama 4 last April, it was already trailing competitors. LeCun has since blamed Zuckerberg for pushing the product out too quickly, arguing that speed was prioritized over innovation.
“We had a lot of new ideas and really cool stuff that they should implement. But they were just going for things that were essentially safe and proved,” LeCun said. “When you do this, you fall behind.” He even claimed that the AI team “fudged” some of the Llama 4 benchmark results.
LeCun isn’t sold on LLMs as the future of AI anyway. Instead, he argues that in order for AI to rival human intelligence, engineers need to develop “world models” that aim to understand the physical world, not just language.
Zuckerberg supposedly liked LeCun’s world model research but still decided to pursue an LLM-focused Meta Superintelligence Labs, offering lofty compensation packages, some reportedly in the hundred of millions, to attract top talent.
Zuckerberg’s most notable hire was 29-year-old Alexandr Wang, co-founder and former CEO of Scale AI, with whom Zuckerberg struck a $14 billion deal for 49 percent stake of the company and for Wang’s move to Meta. The move had immediate consequences for LeCun as he was then forced to report to Wang....
....MORE
Mr. LeCun's new company, Advanced Machine Intelligence (AMI) Labs, will target a $3.5 billion valuation because why not?
That said, Zuckerberg still creeps me out, even after all these years.
Previously on Monsieur LeCun:
February 2024 - "Meta’s A.I. Chief Yann LeCun Explains Why a House Cat Is Smarter Than The Best A.I."
November 2024 - Chief AI Scientist at Meta, Yann LeCun: "I don't wanna say "I told you so", but I told you so."
November 2025 - "He’s Been Right About AI for 40 Years. Now He Thinks Everyone Is Wrong."
And many more.
