Former Chief AI Scientist calls Meta’s Alexandr Wang “young and inexperienced” — “You certainly don’t tell a researcher like me what to do.”

by

2025 could be a landmark year for Meta in the world of generative AI. The company behind Facebook made significant moves to push its AI capabilities forward, including reportedly recruiting leading AI experts from competitors like OpenAI. They also invested $14.3 billion in Scale AI and brought on its CEO, Alexandr Wang, to help lead a new research division focused on advanced AI.

It appears Meta is facing serious challenges, and its stability may be at risk. Yann LeCun, Meta’s head of AI research, recently explained in an interview with the Financial Times why he left the company in November to start his own business.

I learned that the executive, a long-time employee of over ten years, left because of a difficult relationship with Mark Zuckerberg at Meta. He also suggested that the company’s performance goals weren’t what they seemed, and there were other issues at play.

Before ChatGPT was released in November 2022, Yann LeCun characterized Meta as a fresh start with unlimited potential, allowing him to focus on research without the immediate pressure of creating a profitable product. He also noted that funding wasn’t a concern.

After ChatGPT became incredibly popular, Mark Zuckerberg asked Yann LeCun to create a similar AI for Meta. LeCun agreed, but only if the chatbot was released as open-source software and made available for free.

Zuckerberg asked LeCun to create Meta’s large language model, and LeCun agreed only if the model, called Llama, would be open source and available for free. He believes Llama has been a game-changer in the AI field, becoming popular with researchers because of its impressive capabilities.

The initial excitement didn’t last long, particularly after Meta’s Llama 4 model launched to disappointing reviews in April 2025. According to one scientist, the model’s failure was due to Mark Zuckerberg pushing his team to accelerate AI development too quickly.

Because Llama wasn’t performing as well as hoped, Zuckerberg grew frustrated. Even more troubling, LeCun admitted the AI team had altered some of the scores used to measure Llama 4’s performance.

Mark was deeply disappointed and lost trust in the team working on this project, so he effectively removed them from further involvement.

Despite Mark Zuckerberg’s apparent enthusiasm for Yann LeCun’s concept of ‘world models,’ he ultimately chose to direct Meta’s AI research efforts and funding towards a new lab called Meta Superintelligence Labs.

After Meta invested heavily in Scale AI and brought on its CEO, Alexandr Wang, to head a new AI division, the leadership structure changed. This meant Yann LeCun now reported to Wang, whom he considered to be young and lacking experience.

LeCun didn’t hold back any punches and outrightly expressed his feelings on Wang’s hiring:

Alex isn’t giving me instructions, and that’s fine. You shouldn’t tell a researcher what to do, especially not someone with my experience.

Ex-Meta Chief AI Scientist, Yann LeCun

I’ve been following Yann LeCun, and he’s now started a new company called Advanced Machine Intelligence Labs. It seems he’s using it to continue his work on what he calls ‘world models’ – basically, teaching computers to understand the world around them. From what I hear, the company is currently valued at around $3 billion.

As a researcher following the rapid advancements in AI, I’m curious to get your thoughts on whether Meta can effectively compete in this evolving landscape. I’d love to hear your perspective in the comments, and I’ve also created a poll to gather broader opinions on this topic.

Read More

2026-01-07 15:42