Jan 18, 2024 - Technology

Meta begins training Llama 3, reshuffles AI responsibilities

headshot
Meta CEO Mark Zuckerberg, outlining the company's latest AI plans in a video posted to Instagram

Meta CEO Mark Zuckerberg, outlining the company's latest AI plans in a video posted to Instagram. Screenshot: Axios

Meta CEO Mark Zuckerberg said Thursday that the company has started training Llama 3, the next generation of its primary generative AI model. He also reaffirmed the company's commitment to releasing its AI models via open source — when possible — and said the company is once again shaking up its AI org chart.

Why it matters: Zuckerberg says the moves are designed to help the company build artificial general intelligence systems that will give its products more human-like capabilities in coming years.

What they're saying: "We've come to this view that, in order to build the products that we want to build, we need to build for general intelligence," Zuckerberg told the Verge. "I think that's important to convey because a lot of the best researchers want to work on the more ambitious problems."

Details: Zuckerberg posted a video to Instagram outlining his goals as well as the changes to the org chart.

  • Meta's Fundamental AI Research unit will move from the Reality Labs team run by CTO Andrew Bosworth to the company's Family of Apps unity run by Chris Cox. AI research heads Yann LeCun and Joelle Pineau will now both report to Cox.
  • The move, Meta says, will reunite the company's AI research and product work and help the company build toward what's known as artificial general intelligence.
  • Aiming to underscore the company's investment in AI, Zuckerberg said that the company expects to end the year with more than 340,000 Nvidia H100 GPUs. Including other chips, the company projects it will have the equivalent of nearly 600,000 H100s.
Go deeper