Meet Shengjia Zhao, Meta's chief scientist of Superintelligence AI group
What's the story
Meta Platforms has hired Shengjia Zhao, an ex-OpenAI researcher and co-author of the original ChatGPT paper, as the Chief Scientist of the new Superintelligence AI group. Zhao was instrumental in OpenAI's early breakthroughs, including ChatGPT and o1 (the company's first reasoning model). Mark Zuckerberg had announced Zhao's formal appointment in a post on Threads.
Leadership role
Zhao will set research agenda for MSL
Zuckerberg announced that Zhao has been "our lead scientist from day one." He added that they have decided to formalize his leadership role as their recruitment is going well and their team is coming together. Zhao will set a research agenda for Meta Superintelligence Labs (MSL) under the leadership of Alexandr Wang, ex-CEO of Scale AI who was recently hired by Meta.
Recruitment drive
Meta's aggressive recruitment strategy
Meta has been aggressively recruiting top-tier AI talent from OpenAI, Safe Superintelligence, Google DeepMind, Apple, and Anthropic. The company has also hired researchers from its existing Fundamental AI Research (FAIR) lab and GenAI unit. Zhao's appointment comes as part of this strategy to build a strong leadership team for MSL.
Infrastructure development
Zuckerberg personally recruiting researchers
Zuckerberg has been personally recruiting researchers and inviting some prospects to his Lake Tahoe estate. Meta has reportedly offered some researchers eight and nine-figure pay packages, some of which are "exploding offers" that will expire in a matter of days. The firm has also increased its investment in cloud computing infrastructure to support MSL's work on frontier AI models.
Training capabilities
Prometheus will be one of the largest AI training clusters
By 2026, Zhao and MSL's researchers will have access to Meta's 1GW cloud computing cluster, Prometheus, in Ohio. This will make the firm one of the first tech companies with an AI training cluster of such size. The company hopes this infrastructure will help conduct massive training runs needed to create competitive frontier AI models.