Meta hired a 29-year-old billionaire to lead AI, but its former chief scientist says he is inexperienced to lead real research

Meta

• Ex-Meta AI chief LeCun criticizes new leadership, warns of staff departures
• LeCun says Meta focuses on "safe" AI paths, risking innovation and talent loss
• LeCun claims LLMs like ChatGPT are a dead end for superintelligent AI

Did our AI summary help?

Meta, the company that owns Facebook, Instagram and WhatsApp, is once again making headlines, but this time it’s not because of a new feature or a social media trend. Former Meta Chief AI Scientist Yann LeCun has openly criticized the company’s AI leadership and warned that more employees could leave soon. And surprisingly, he says the reason is not Mark Zuckerberg, as per report by CNBC.

The controversy began when 29-year-old Alexander Wang joined Meta as Chief AI Officer in 2025. Wang is the billionaire co-founder of Scale AI, a startup that Meta invested in heavily. In 2025, Meta also bought a 49% stake in Scale AI, making Wang one of the youngest AI executives to lead a major tech company. Meta’s decision to bring him onboard happened during a fierce AI talent battle, where companies were throwing jaw-dropping signing bonuses to hire top minds. Reports even claimed Meta offered $100 million signing bonuses to attract talent from OpenAI, the company behind ChatGPT. It was clear Meta wanted to win the AI race at any cost.

But LeCun, who resigned from Meta in November at the age of 65, wasn’t too impressed. In an interview with the Financial Times, he described Wang as “young” and “inexperienced,” especially when it comes to AI research culture. He did acknowledge that Wang learns fast and knows his limitations, but also pointed out that he has no background in managing research teams or understanding what excites AI researchers. According to LeCun, that lack of experience could make Meta’s research environment less appealing for people who want to innovate instead of playing it safe.

LeCun also brought up the benchmark controversy surrounding Meta’s Llama 4 model. The company had been accused of tweaking benchmark tests to make the model look smarter than it actually was. That reportedly shook Zuckerberg’s confidence in the AI team. LeCun said Zuckerberg then sidelined the entire Gen AI division, which led to people quitting and could push even more to leave in the future. LeCun didn’t hold back, saying that Meta had exciting new ideas, but the company only focused on “safe” and “proven” paths. His point was simple. When you stop taking risks, you stop leading.

Another bold claim he made was that LLMs, or large language models like ChatGPT and Llama, are a dead end if the goal is superintelligent AI. He even joked that some people inside Meta might not like him saying this out loud. Instead, LeCun is now building a new startup called Advanced Machine Intelligence Labs. The company is working on world models, which learn from videos and real-world physical data, not just text. This kind of AI is meant to understand the world more like humans do, instead of predicting the next word in a sentence and hoping for the best.

Meanwhile, health tech startup Nabla, which recently partnered with LeCun’s new company, also shared a statement saying that LLMs still struggle with hallucinations, unpredictable reasoning, and handling real-time visual or multimodal information. In other words, they’re smart at talking, but not so great at making real-world decisions on their own.