AI pioneer Yann LeCun is not sold on Mark Zuckerberg's $14 billion bet on 28-year-old Scale AI co-founder Alexander Wang, who has been hired to lead Meta's superintelligence lab.
In a new interview with the Financial Times, Lekun, who was Meta's chief AI scientist until he announced his departure in November to start his own startup, said Wang was “inexperienced” and did not fully understand AI researchers.
“He's a quick learner and he knows what he doesn't know. … He has no experience with research, no way of doing research, no way of doing it, no sense of what's attractive to researchers and what's repulsive to researchers,” LeCun said.
Mr. Wang was the crown jewel of Mr. Zuckerberg's aggressive move in the AI talent war. Meta made a $14 billion investment in Scale AI, which included hiring Wang from the buzzy startup.
LeCun said Zuckerberg became frustrated with disappointing progress on Llama, the company's flagship open source AI model.
LeCun said in an interview that the AI team “fabricated” some of Llama 4's results. At the time, Meta was criticized for potentially gaming the results of benchmark tests. LeCun said the episode made Zuckerberg unhappy with Meta's existing AI team.
“Mark was really upset and had basically lost confidence in everyone involved in this,” he told the FT. “And basically sidelined the entire GenAI organization.”
As for his relationship with Wang, LeCun said the 28-year-old briefly became his boss after Zuckerberg's AI reorganization, but did not actually guide Wang.
“You can't tell researchers what to do,” LeCun told the magazine. “Certainly, we don't tell researchers like me what to do.”
Meta’s new AI team is “fully LLM-infused”
LeCun said that while Zuckerberg continues to support his views on the future of AI, the large talent that Meta CEO has hired is focused on developing language models at scale.
“I think there are a lot of people in the meta, probably including Alex, who would like us to stop telling the world that LLMs are basically dead-ends when it comes to superintelligence,” LeCun said. “But I'm not going to change my mind because some people think I'm wrong. I'm not wrong. My integrity as a scientist won't allow me to do that.”
LeCun repeatedly argues that LLM is too limited and that unlocking the true power of AI requires a different approach. This is why his startup is called Advanced Machine Intelligence, the very approach he claims is better suited than an LLM.
He will not be the CEO of the new company, but will be the executive chairman.
“I'm a scientist and a visionary. I can inspire people to do interesting things. I'm pretty good at guessing what kinds of technologies will work or not. But I can't be a CEO,” LeCun said. “I'm too disorganized and too old for this!”
