Mark Zuckerberg's former top AI scientist has revealed the reason behind his abrupt exit from Meta last year.
The tech world is no stranger to feuds, just look at the ongoing dispute between Tesla CEO Elon Musk and OpenAI's Sam Altman.
Shortly after receiving an award from King Charles for his work in artificial intelligence in November, Professor Yann LeCun announced he was leaving his position as Meta's chief AI scientist to launch a new company.
Now, in a revealing interview with Financial Times, one of the 'godfathers of AI' has explained his abrupt departure.
Advert
LeCun had been working at Meta for over a decade, where he could enjoy the freedom to conduct all kinds of experimental AI research without necessarily worrying about developing profitable products.
The computer scientist described the tech giant as a 'tabula rasa with a carte blanche,' adding that 'money was clearly not going to be a problem.'
Then, after AI chatbots like ChatGPT and Google's Gemini boomed in 2022, Zuckerberg ordered LeCun to develop Meta's own LLM, agreed on the condition that Llama would be open source and free.
According to LeCun, the Llama models 'changed the entire industry' and became a hit among AI researchers due to their power and open-source nature. That was until the launch of the latest Llama 4 model, which was criticised as highly outdated.
LeCun pointed the blame at Zuckerberg for pressuring his unit to accelerate AI development.
“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 explained. “When you do this, you fall behind.” In LeCun's view, LLMs represent a 'dead end' for achieving 'superintelligent" AI that exceeds human intelligence.
Instead, he believes 'world models' are key to the next leap in tech, able to understand the entire physical world rather than just language.
While Zuckerberg claimed to support LeCun's work on world models, his actions told a different story. The Facebook CEO created a separate Superintelligence Labs devoted entirely to LLMs, offering hundreds of millions in contracts to recruit top researchers.
Every person who joined was 'completely LLM-pilled,' LeCun argued.
Meanwhile, Zuckerberg's headline recruitment was Alexandr Wang, who had built and run Scale AI, a startup focused on data annotation. Zuckerberg paid $14 billion for nearly half of Scale AI, and Wang transitioned from there to Meta, taking charge of the new Superintelligence Labs. Here, LeCun found himself subordinate to Wang, whom he found to be 'young' and 'inexperienced.'
“The average age of a Facebook engineer at the time was 27,” LeCun said. “I was twice the age of the average engineer.”
LeCun made it clear in the interview that the change in authority didn't sit well with him.
“Alex isn’t telling me what to do either,” LeCun stated. “You don’t tell a researcher what to do. You certainly don’t tell a researcher like me what to do.”
LeCun's new 'world model' startup, called Advanced Machine Intelligence Labs, targets a $3 billion valuation.