In a shocking interview that has stunned the tech world, former AI leader at Meta, Yann LeCun, revealed that the company engaged in "fraud" during the development of the large language model Llama 4. As a former head of the AI department, LeCun openly criticized the current Chief AI Officer, Alexandr Wang, for being young and inexperienced, claiming he did not understand the real needs of researchers. As Meta's strategy continued to change, LeCun eventually chose to leave the company and move toward more promising research on "world models."

LeCun traced his journey back to his successful early experiences at Bell Labs, reflecting on his connection with Facebook. In 2013, Zuckerberg even hosted a dinner at his home in California to attract him to join. However, upon joining, LeCun set several conditions, including maintaining his position at New York University and ensuring the openness of research results. In his view, Facebook was a blank slate, offering him unlimited research freedom.

However, as Zuckerberg's pursuit of the metaverse failed, Meta shifted its focus entirely to AI projects, especially the Llama series. LeCun revealed that Zuckerberg's pressure on AI development led to poor communication within the team, resulting in subpar performance of the Llama series. Worse still, after the release of Llama 4, Meta was accused of manipulating benchmark test results, leading to external doubts.

After this incident, Meta quickly entered the "post-Llama" era, with Wang's arrival marking a strategic shift in the company. LeCun stated that his work at Meta was severely restricted because most team members had been "brainwashed" into focusing on large language models. Although there was no direct conflict between him and Wang, he felt disappointed about the company's future direction.

Ultimately, LeCun left Meta and founded his new company, "Advanced Machine Intelligence Lab," focusing on what he believes is a more promising area of world model research. He hopes that through this model, not only will it understand language, but also the mechanisms of the physical world. He firmly believes that future intelligent development should break beyond the limitations of language.