Need smarter insights in your inbox? Join our weekly newsletters to get solely what issues to enterprise AI, knowledge, and safety leaders. Subscribe Now
Meta has appointed Shengjia Zhao, a former OpenAI researcher and co‑creator of GPT‑4, because the Chief Scientist of its newly created Meta Superintelligence Labs (MSL).
The announcement was made Friday by Mark Zuckerberg on Threads, noting Zhao will lead the lab’s scientific agenda alongside him and Alexandr Wang, the previous CEO of Scale AI who Meta lately introduced onboard as Chief AI Officer.
“I’m very excited to take up the function of chief scientist for meta super-intelligence labs. Wanting ahead to constructing asi [artificial superintelligence] and aligning it to empower folks with the wonderful crew right here. Let’s construct!” Zhao wrote in his personal Threads put up.
“Synthetic superintelligence” is a nebulous time period used within the AI business to explain techniques extra highly effective and succesful than any at present, past even the neatest people, making them troublesome to manage.
The AI Impression Collection Returns to San Francisco – August 5
The subsequent part of AI is right here – are you prepared? Be part of leaders from Block, GSK, and SAP for an unique have a look at how autonomous brokers are reshaping enterprise workflows – from real-time decision-making to end-to-end automation.
Safe your spot now – area is proscribed: https://bit.ly/3GuuPLF
Zhao’s robust industrial AI background
Zhao, who beforehand labored at OpenAI, performed a key function within the improvement of foundational fashions like GPT-4 and GPT-4o, in response to arXiv system playing cards and analysis papers itemizing him as a co-author. He’s additionally identified for his tutorial work on generative fashions and truthful representations, with broadly cited papers in venues like NeurIPS, ICML, and ICLR.
Zhao joins Meta amid a high-stakes hiring blitz throughout the AI business. Over the previous few months, Meta has poached researchers from OpenAI, Apple, Google, and Anthropic as a part of a multibillion-dollar guess on superintelligence as CNN reported.
Meta lately invested $14.3 billion in Scale AI, buying a 49% stake and bringing on Wang to steer the superintelligence effort. Former GitHub CEO Nat Friedman additionally joined the crew.
The corporate has reportedly provided compensation packages price as a lot as $100 million to $300 million over 4 years to lure high AI expertise, in response to a number of studies. One declare from a rival AI startup founder alleged Meta provided $1.25 billion over 4 years—roughly $312 million per yr—to a single candidate who declined.
Different insiders say Meta’s most senior AI scientists could also be receiving $10 million+ per yr, whereas first-year comp for some new hires reportedly reached $100 million.
Aspirations of main the AI frontier
Zuckerberg has made no secret of his ambition to make Meta a pacesetter in AI’s subsequent frontier, repeatedly stating that the corporate plans to “make investments a whole lot of billions of {dollars} into compute to construct superintelligence” utilizing its personal business-generated capital.
He stated the Llama 4 rollout underscored the significance of elite expertise: “You possibly can have a whole lot of hundreds of GPUs, however should you don’t have the fitting crew creating the mannequin, it doesn’t matter.”
Meta’s basic AI analysis group (FAIR), nonetheless led by acclaimed scientist Yann LeCun, will stay separate from the brand new lab.
The creation of Meta Superintelligence Labs alerts a extra product- and mission-focused arm of Meta’s AI efforts, centered on constructing and aligning ASI with human pursuits.
Making up for the blended reception of Llama 4
Nevertheless, Meta’s push into superintelligence has come on the heels of a bumpy rollout of its newest open-source basis fashions.
The corporate launched its Llama 4 mannequin household in April 2025, positioning it as a leap ahead in multimodal reasoning and long-context understanding. However the launch has struggled to achieve traction amid the rise of highly effective Chinese language open-source rivals like DeepSeek and Qwen.
Meta confronted public criticism from researchers and builders who cited poor real-world efficiency, confusion round benchmark outcomes, and inconsistent high quality throughout deployments.
Some accused the corporate of “benchmark gamesmanship” and utilizing unreleased optimized variations of Llama 4 to spice up public notion—a declare Meta has denied.
Inner sources blamed quick rollout timelines and bugs for the problems, however the episode has solid a shadow over Meta’s generative AI credibility simply because it embarks on its most bold effort but.
Jim Fan, a former Stanford colleague of Zhao and now Nvidia’s Director of Robotics and Distinguished Scientist, provided his endorsement on X: “Shengjia is among the brightest, humblest, and most passionate scientists I do know. Very bullish on MSL!”
The transfer underscores Meta’s technique of spending aggressively now to safe a dominant place in what it views as the subsequent foundational know-how platform — one that might eclipse the cellular web. As Zuckerberg sees it, ASI isn’t a moonshot — it’s the subsequent frontier, and Meta intends to steer.
