Meta has appointed Shengjia Zhao, a former OpenAI researcher and co‑creator of GPT‑4, as the Chief Scientist of its newly established Meta Superintelligence Labs (MSL).

The announcement was made on Friday by Mark Zuckerberg on Threads, highlighting that Zhao will lead the lab’s scientific agenda alongside him and Alexandr Wang, the former CEO of Scale AI, who Meta recently onboarded as Chief AI Officer.

“I am thrilled to assume the role of chief scientist for Meta Superintelligence Labs. I am eager to work on building ASI [artificial superintelligence] and aligning it to empower people with this incredible team. Let’s build!” Zhao expressed in his own Threads post.

“Artificial superintelligence” is a complex term in the AI industry used to describe systems more advanced and capable than any currently existing, surpassing even the smartest humans, thus making them challenging to control.

AI Scaling Hits Its Limits

Power caps, rising token costs, and inference delays are reshaping enterprise AI. Join our exclusive salon to discover how top teams are:

• Turning energy into a strategic advantage

• Architecting efficient inference for real throughput gains

• Unlocking competitive ROI with sustainable AI systems

Secure your spot to stay ahead: https://bit.ly/4mwGngO

Zhao’s robust commercial AI background

Zhao, who previously worked at OpenAI, was instrumental in developing foundational models like GPT-4 and GPT-4o, according to arXiv system cards and research papers listing him as a co-author. He is also recognized for his academic work on generative models and fair representations, with his papers being widely cited in venues such as NeurIPS, ICML, and ICLR.

Zhao joins Meta amid a high-stakes hiring spree across the AI industry. In recent months, Meta has attracted researchers from OpenAI, Apple, Google, and Anthropic as part of a multibillion-dollar investment in superintelligence, as CNN reported.

Meta recently invested $14.3 billion in Scale AI, acquiring a 49% stake and bringing on Wang to spearhead the superintelligence efforts. Former GitHub CEO Nat Friedman also joined the team.

Reports suggest that the company has offered compensation packages ranging from $100 million to $300 million over four years to attract top AI talent. A claim from a rival AI startup founder alleged Meta offered $1.25 billion over four years—around $312 million per year—to a single candidate who ultimately declined.

Other insiders indicate that Meta’s most senior AI scientists may be receiving $10 million+ per year, with first-year compensation for some new hires reportedly reaching $100 million.

Aspirations to lead the AI frontier

Zuckerberg has been transparent about his ambition to make Meta a leader in AI’s next frontier, repeatedly stating that the company plans to “invest hundreds of billions of dollars into compute to build superintelligence” using its own business-generated capital.

He mentioned that the Llama 4 rollout underscored the importance of elite talent: “You can have hundreds of thousands of GPUs, but without the right team developing the model, it’s inconsequential.”

Meta’s fundamental AI research group (FAIR), still led by renowned scientist Yann LeCun, will remain separate from the new lab.

The establishment of Meta Superintelligence Labs signals a more product- and mission-focused extension of Meta’s AI efforts, emphasizing building and aligning ASI with human interests.

Addressing the mixed reception of Llama 4

However, Meta’s venture into superintelligence follows a challenging rollout of its latest open-source foundation models.

The company released its Llama 4 model family in April 2025, presenting it as a significant advancement in multimodal reasoning and long-context understanding. Despite this, the release has struggled to gain traction amid the emergence of powerful Chinese open-source competitors like DeepSeek and Qwen.

Meta faced public criticism from researchers and developers who pointed out poor real-world performance, confusion around benchmark results, and inconsistent quality across deployments.

Some accused the company of “benchmark gamesmanship” and using unreleased optimized versions of Llama 4 to enhance public perception—a claim Meta has denied.

Internal sources attributed the issues to fast rollout timelines and bugs, yet the episode has cast a shadow over Meta’s generative AI credibility just as it embarks on its most ambitious endeavor yet.

Jim Fan, a former Stanford colleague of Zhao and now Nvidia’s Director of Robotics and Distinguished Scientist, expressed his support on X: “Shengjia is one of the brightest, most humble, and passionate scientists I know. Very optimistic about MSL!”

The move highlights Meta’s strategy of investing aggressively now to secure a leading position in what it views as the next foundational technology platform—one that could surpass the mobile internet. As Zuckerberg perceives it, ASI isn’t a moonshot—it’s the next frontier, and Meta aims to lead.