Meta CEO Mark Zuckerberg introduced Friday that former OpenAI researcher Shengjia Zhao will lead analysis efforts on the firm’s new AI unit, Meta Superintelligence Labs (MSL). Zhao contributed to a number of of OpenAI’s largest breakthroughs, together with ChatGPT, GPT-4, and the corporate’s first AI reasoning mannequin, o1.
“I’m excited to share that Shengjia Zhao would be the Chief Scientist of Meta Superintelligence Labs,” Zuckerberg stated in a submit on Threads Friday. “Shengjia co-founded the brand new lab and has been our lead scientist from day one. Now that our recruiting goes properly and our staff is coming collectively, we have now determined to formalize his management function.”
Zhao will set a analysis agenda for MSL underneath the management of Alexandr Wang, the previous CEO of Scale AI who was lately employed to steer the brand new unit.
We’re excited to announce that @shengjia_zhao would be the Chief Scientist of Meta Superintelligence Labs!
Shengjia is an excellent scientist who most lately pioneered a brand new scaling paradigm in his analysis. He’ll lead our scientific route for our staff.
Let’s go 🚀 pic.twitter.com/D93KQWIvFl
— Alexandr Wang (@alexandr_wang) July 25, 2025
Wang, who doesn’t have a analysis background, was considered as a considerably unconventional alternative to steer an AI lab. The addition of Zhao, who’s a good analysis chief identified for growing frontier AI fashions, rounds out the management staff. To additional fill out the unit, Meta has employed a number of high-level researchers from OpenAI, Google DeepMind, Protected Superintelligence, Apple, and Anthropic, in addition to pulling researchers from Meta’s current Elementary AI Analysis (FAIR) lab and generative AI unit.
Zuckerberg notes in his submit that Zhao has pioneered a number of breakthroughs, together with a “new scaling paradigm.” The Meta CEO is probably going referencing Zhao’s work on OpenAI’s reasoning mannequin, o1, by which he’s listed as a foundational contributor alongside OpenAI co-founder Ilya Sutskever. Meta presently doesn’t provide a competitor to o1, so AI reasoning fashions are a key space of focus for MSL.
The Info reported in June that Zhao could be becoming a member of Meta Superintelligence Labs, alongside three different influential OpenAI researchers — Jiahui Yu, Shuchao Bi, and Hongyu Ren. Meta has additionally recruited Trapit Bansal, one other OpenAI researcher who labored on AI reasoning fashions with Zhao, in addition to three staff from OpenAI’s Zurich workplace who labored on multimodality.
Zuckerberg has gone to nice lengths to set MSL up for achievement. The Meta CEO has been on a recruiting spree to workers up his AI superintelligence lab, which has entailed sending private emails to researchers and alluring prospects to his Lake Tahoe property. Meta has reportedly provided some researchers eight- and nine-figure compensation packages, a few of that are “exploding gives” that expire in a matter of days.
Techcrunch occasion
San Francisco
|
October 27-29, 2025
Meta has additionally upped its funding in cloud computing infrastructure, which ought to assist MSL conduct the huge coaching runs required to create aggressive frontier AI fashions.
By 2026, Zhao and MSL’s researchers ought to have entry to Meta’s 1 gigawatt cloud computing cluster, Prometheus, situated in Ohio. As soon as on-line, Meta can be one of many first expertise corporations with an AI coaching cluster of Prometheus’ measurement — 1 gigawatt is sufficient power to energy greater than 750,000 houses. That ought to assist Meta conduct the huge coaching runs required to create frontier AI fashions.
With the addition of Zhao, Meta now has two chief AI scientists, together with Yann LeCun, the chief of Meta’s FAIR lab. Not like MSL, FAIR is designed to concentrate on long-term AI analysis — strategies that could be used 5 to 10 years from now. How precisely Meta’s three AI models will work collectively stays to be seen.
Nonetheless, Meta now appears to have a formidable AI management staff to compete with OpenAI and Google.
