one in new interview with The Financial TimesYann LeCun, one of the so-called godfathers of AI, finally decided to abruptly exit Meta in November.
Much of how he tells it is based on his increasingly strained relationship with CEO Mark Zuckerberg — and his new golden boy, Alexander Wang, who looms over LeCun despite being nearly four decades younger.
LeCun had been at Zuckerberg’s company for more than a decade, where, as chief AI scientist, he had the freedom to do all kinds of esoteric AI research without worrying about developing a profitable product. LeCun described Meta and then Facebook as “table juice with carte blanche”. “Money was obviously not going to be a problem,” he said. foot,
Then, in November 2022, ChatGPT came out and the whole world went crazy about AI chatbots. AI chatbots and their human-like capabilities for conversation are powered by the large language models LeCun helped pioneer in his foundational work on neural networks. When Zuckerberg ordered LeCun to develop Meta’s own Llama, he agreed on the condition that Llama would be open source and free.
The Llama model “changed the entire industry,” LeCun said, and was a hit among AI researchers because of its power and open source nature.
However, the success did not last; The latest Lama 4 model, released last April, was poorly received upon arrival and was immediately condemned as an outdated flop. LeCun blamed its failure on Zuckerberg putting pressure on LeCun’s unit to accelerate AI development.
“We had a lot of new ideas and really cool things that they should implement. But they were only going for things that were essentially safe and proven,” LeCun explained. foot“When you do that, you get left behind,”
However, the rift has deepened further. LeCun sees LLM as a “dead end” to creating even more powerful, “superintelligent” models that rival or surpass human capabilities. The next big leap in technology requires an entirely different architecture, called the “world model”, which seeks to understand not only language, but also the physical world.
According to Lacan, Zuckerberg really liked Lacan’s world model research, but he didn’t put his money where his mouth was. Instead, Zuckerberg last year launched a new LLM-focused Superintelligence Labs, separate from LeCun’s lab, and offered contracts worth several hundred million dollars to attract top talent. LeCun complained that whatever talent came in was “completely LLM-laden”.
Zuckerberg’s key new appointee was Alexander Wang, founder and former CEO of AI data annotation startup Scale AI, which provides an essential service for training AI models, but does not create or design them. Zuckerberg invested $14 billion in Scale AI to buy a 49 percent stake, and as part of that deal, Wang left and joined Meta to lead the new Superintelligence Labs. As a result, LeCun was forced to begin reporting to Wang.
The move raised questions from the start, including whether Wang, 29, had the experience and background in building large-scale AI models, which his company did not have. LeCun didn’t make us wonder where he stood on Wang’s appointment, calling him “young” and “inexperienced.”
Whatever the case, Lecun, who was considered the godfather of the entire region, was now taking orders from Wang. When the interviewer mentioned the new hierarchy, LeCun liked it at first. “The average age of a Facebook engineer at the time was 27,” LeCun said. foot“I was twice the age of the average engineer,”
But when the interviewer pointed out that the younger generation wasn’t telling them what to do until the 29-year-old Wang arrived, Laken revealed his true feelings.
“Alex isn’t even telling me what to do,” Laken said sarcastically. “You don’t tell a researcher what to do. You certainly don’t tell a researcher like me what to do.”
Later on he will be his own boss. LeCun has launched a new world-model-focused startup called Advanced Machine Intelligence Labswhich is aimed at $3 billion valuation. LeCun will serve as executive chairman, giving him the same kind of freedom to pursue research that he once had at Meta. foot,
More on AI: Police forced to explain why officer turned into frog as claimed in AI-generated police report
