Mirror Review
July 1, 2025
How do you build a team to create superhuman intelligence?
Mark Zuckerberg seems to have found an answer: hire the very people who were building it for your biggest rivals!
With the launch of Meta Superintelligence Labs (MSL), Meta is introducing a newly assembled group of elite minds poached from across their AI rivals to form Meta’s superintelligence team.
The AI Talent Wars
The competition for elite AI talent in Silicon Valley has become very fierce.
Zuckerberg’s company has taken a hands-on approach, personally reaching out to potential hires and even hosting them at his homes.
This aggressive strategy involves huge compensation packages, with some reports suggesting offers in the tens of millions of dollars.
This campaign has not gone unnoticed. In a memo to his staff, OpenAI’s Chief Research Officer Mark Chen expressed the impact of the departures, stating, “I feel a visceral feeling right now, as if someone has broken into our home and stolen something”.
He acknowledged that Meta has been “aggressively building out their new AI effort” and attempting to recruit top talent with “comp-focused packages”.
Meet the Recruits for Meta’s Superintelligence Team
Meta’s recent hiring spree has brought in at least 11 new researchers, with some standing out due to their previous work at competing labs. These individuals bring a wealth of experience from the front lines of AI development. They include:
From Google DeepMind:
1. Jack Rae
Education:
- PhD in Computer Science, UCL
- Master’s in Statistics, Carnegie Mellon University
- MSc in Mathematics & Computer Science, University of Bristol
Jack spent nearly eight years at DeepMind, rising to the position of Principal Research Scientist.
His work there focused on scaling memory-augmented neural networks, a key component in creating systems that can learn and remember over long periods.
Rae also had a stint at OpenAI, giving him a unique perspective on the inner workings of two of Meta’s biggest rivals.
2. Pei Sun
Education:
- PhD (Dropped out), Carnegie Mellon University
- Bachelor’s, Tsinghua University
Pei Sun was a Principal Researcher at Google’s DeepMind and a core contributor to the Gemini family of models, which are some of the most capable multimodal AI systems in existence.
Before his time at DeepMind, he spent nearly seven years at Waymo, Google’s self-driving car project, where he worked on machine learning research.
This background gives him a deep understanding of how to apply AI to real-world problems.
From OpenAI:
3. Jiahui Yu
Education:
- PhD, University of Illinois Urbana-Champaign
- BSc in Computer Science, University of Science and Technology of China (School of the Gifted Young)
Jiahui Yu joined OpenAI as the Head of Research for Perception, where he led projects like GPT-4.1 and the image generation capabilities of GPT-4o.
Before that, he was a key player at Google DeepMind, where he co-led the multimodal efforts for the Gemini model and was a core contributor to the PaLM-2 architecture.
His extensive experience also includes research internships at Nvidia, Baidu, Snap, and Microsoft.
4. Shuchao Bi
Education:
- PhD Candidate, Mathematics, UC Berkeley
- Master’s in Statistics, UC Berkeley
- Bachelor’s in Mathematics, Zhejiang University
Shuchao Bi brings a unique combination of research and product experience to Meta.
He led the multimodal post-training research organization at OpenAI, where his team worked on advancing reinforcement learning from human feedback (RLHF).
He also developed multimodal agents that can understand and interact with the world through images, video, and audio.
Before his time in AI research, he co-founded and was the Head of YouTube Shorts Discovery; therefore, he has a deep understanding of what it takes to build and scale a successful product.
5. Shengjia Zhao
Education:
- PhD in Computer Science, Stanford University
- Bachelor’s, Tsinghua University
- Exchange, Rice University
Shengjia Zhao is a Member of the Technical Staff at OpenAI.
He was a contributor to the technical report for GPT-4 and the system card for GPT-4o, both of which are foundational documents in the field of large language models.
His expertise lies in computational intelligence, machine learning, and deep learning, making him a valuable asset for the Meta ‘Superintelligence’ SuperTeam.
6. Hongyu Ren
Education:
- PhD in Computer Science, Stanford University
- Undergraduate, Peking University
Hongyu Ren is a rising star in the AI world. As a Research Scientist at OpenAI, he was the creator of the o3-mini and o1-mini models and a core contributor to GPT-4o.
He led a group focused on post-training, with the goal of making models “think faster, harder and sharper”.
His work has been recognized with prestigious fellowships from Apple, Baidu, and the Masason Foundation.
7. Trapit Bansal
Education:
- PhD in Computer Science, University of Massachusetts Amherst
- Integrated MSc in Mathematics & Statistics, IIT Kanpur
Trapit Bansal’s research at OpenAI focused on the cutting edge of reinforcement learning (RL) and reasoning.
He started the RL for reasoning effort with Ilya Sutskever, one of the co-founders of OpenAI.
Moreover, his background includes research internships at Microsoft, Google, and a previous stint at OpenAI.
From Anthropic:
8. Joel Pobar
Education:
- BIT (Hons), Computer Science, Queensland University of Technology (QUT)
Joel Pobar is a returning veteran to Meta, having previously worked at Facebook for over a decade.
He is a seasoned engineering leader who has built and managed teams working on some of the most critical infrastructure in the tech world, including compilers, programming languages, and AI/ML platforms like PyTorch.
He joins Meta from Anthropic, where he worked on inference systems, which are essential for running large-scale AI models efficiently.
What This Means for the Future of AI
Meta isn’t just accumulating talent; it is strategically acquiring deep expertise in areas critical to building next-generation AI.
This includes experience with large language models like Gemini and GPT-4, multimodal reasoning, reinforcement learning, and the essential infrastructure needed to support these massive systems.
Mark Zuckerberg believes this pursuit of advanced AI marks “the beginning of a new era for humanity” and has committed to Meta leading the way.
By assembling this dream team of researchers, Meta is making a bold statement. The race to build superintelligence is heating up, and with its new recruits, Meta’s ‘superintelligence’ AI team is now an intimidating contender.
