Meta is making a big push into advanced artificial intelligence with its Super Intelligence Lab. Now the Mark Zuckerberg-led tech giant has appointed Shengjia Zhao to head up its Superintelligence AI team. Zhao, one of the developers that helped create ChatGPT at OpenAI, will now be its chief scientist for the latest AI division of Meta.
Zhao joined Meta in June 2025. On Friday (July 25), Meta announced that Zhao will officially serve as its Chief Scientist. The announcement was made on Threads by CEO Mark Zuckerberg, who referred to Zhao as the team’s ‘lead scientist from day one.’
At OpenAI, Zhao was involved with some of the company’s most significant early efforts. He was a co-author of the first ChatGPT paper and helped to produce the initial reasoning model, titled “o1.” It’s the model that effectively ushered in the concept of “chain-of-thought” AI, in which the AI works its way through problems step by step, an idea they’ve gone on to license and which has been adopted by many other companies, including Google.
At Meta, Zhao will answer to Alexandr Wang, Meta’s new Chief AI Officer. Wang had joined Meta in June, after leading Scale AI. He now leads Meta’s quest to create synthetic general intelligence (AGI), or AI, that can think and reason like humans.
Meta’s Superintelligence Lab, which was launched in June 2025, is distinct from Meta’s older AI research group FAIR (Facebook AI Research). FAIR will retain its work under AI pioneer Yann LeCun, who now reports to Wang.
The hiring of Zhao is an indication of Meta’s aggressive talent acquisition approach. In recent months, Meta has hired more than a dozen AI researchers from top companies, including Apple, Google, OpenAI, and Anthropic. The list features industry veterans such as two of Apple’s most famous scientists, Tom Gunter and Mark Lee.
Some reports have said Meta is making huge pay packages to attract talent, including, in one case, up to $300 million. Meta has disputed those numbers but says it is also pouring significant resources into top AI talent.
Presently, Meta’s top open-source AI model, LLaMA 4, is still behind rivals such as GPT-4 and Gemini. But the company is due to put out a beefier model, codenamed Behemoth, later this year.
Zuckerberg said, “We are piloting a team that can build learnings with the tools and pass on the baton to the future of superintelligence AI.”
