Prominent Researchers Join $4 Billion Self-Improving AI Venture

As OpenAI and Anthropic roll out new AI systems strong in coding, expectations are growing that AI may soon be able to improve itself. Recursive Superintelligence, founded by Richard Socher together with 7 other researchers, aims for recursive self-improvement, and with offices in San Francisco and London has raised over $650 million, earning a valuation of over $4 billion. It has fewer than 30 employees.

Co-founders include Josh Tobin, Jeff Clune, and Tim Shi of OpenAI, and Yuandong Tian of Meta. They have worked on open-endedness research, pursuing goals over days, months, and even years, and the company has also brought on Peter Norvig.

However, current technology is not yet at the stage of fully excluding humans. New ideas still need to be created by people, and the goal is to gradually hand over more tasks and idea generation to machines.

OpenAI is also building an 'automated AI researcher,' aiming by this fall for a system that performs the work of a less-skilled researcher. Socher said it will take years for this technology to take hold, and that in the long term it could extend to drug discovery and life sciences research.