Notable Researchers Join $4 Billion Effort to Build Self-Improving A.I.

Recursive Superintelligence, a startup backed by former researchers from Google, Meta, and OpenAI, has secured over $650 million in funding from investors including Google Ventures, Nvidia, and AMD. The six-month-old company, valued at more than $4 billion, is led by Richard Socher, who previously headed AI research at Salesforce and co-founded You.com. Its seven co-founders include notable researchers like Josh Tobin, Jeff Clune, and Yuandong Tian, all of whom specialize in AI systems capable of long-term, open-ended tasks. The company’s goal is to develop AI that can recursively improve itself—meaning the system can generate new ideas and refine its own code without constant human intervention. This approach builds on recent advancements in AI coding, such as systems from Anthropic and OpenAI that excel at writing software. Recursive Superintelligence plans to hire Peter Norvig, a former Google research director and co-author of a foundational AI textbook, to support its efforts. While critics argue current AI is far from fully autonomous, OpenAI is already testing an 'automated AI researcher' to handle tasks previously done by junior researchers. Sam Altman, OpenAI’s CEO, stated the system could be operational by fall. Recursive Superintelligence acknowledges the challenge, noting years of development are needed before achieving its vision. The company’s focus on recursive self-improvement aligns with broader industry trends, as Silicon Valley firms increasingly explore AI systems that can evolve independently. Beyond AI development, Recursive Superintelligence plans to apply its technology to fields like drug discovery and biological research, though no timelines have been announced.