Enterprise & Industry

Ex-Meta researcher's Recursive Superintelligence raises $650M for self-improving AI

A $4.6B startup aims to build AI that autonomously refines its own code.

Deep Dive

Tian Yuandong, a star Chinese researcher laid off by Meta Platforms, has co-founded Recursive Superintelligence with seven other co-founders. The startup announced on Wednesday it raised over $650 million at a $4.65 billion valuation, led by venture capital firms GV and Greycroft, with major participation from Nvidia and Advanced Micro Devices. Yuandong previously served as research scientist director at Meta's FAIR (Fundamental AI Research) team.

Recursive Superintelligence is focused on what’s often called “recursive self-improvement” – AI systems that can automate the AI research and development process, enabling them to autonomously refine their own code and reasoning. This concept has long been considered the holy grail of AI research. The company believes this is the fastest path to superintelligence, as a system that can improve itself exponentially would rapidly outpace competitors. The investment from chip giants like Nvidia and AMD signals strong industry belief in the approach, which could reshape the competitive landscape of AI development.

Key Points
  • Raised $650M at a $4.65B valuation from GV, Greycroft, Nvidia, and AMD
  • Founded by Tian Yuandong, former research director at Meta's FAIR lab
  • Aims to achieve recursive self-improvement for AI to autonomously enhance its own capabilities

Why It Matters

Self-improving AI could disrupt the entire AI industry, making current models obsolete quickly.