AI Funding: Recursive Raises $650M to Automate Model Creation

Recursive, a startup focused on automating artificial intelligence development, has secured $650 million in fresh funding at a valuation of $4.65 billion. The investment round includes participation from Google Ventures, Greycroft, Nvidia, and AMD Ventures, signaling significant institutional confidence in the company’s core thesis: that AI can be used to build and improve AI itself.

The company’s primary product is Flint, an open-weight coding model with 40 billion parameters. According to the founder, nearly every aspect of the model’s creation pipeline was executed and refined by AI rather than human engineers. This approach, known as recursive self-improvement or RSI, allows a small team of researchers to develop competitive technology in a matter of months, a timeframe that would typically require much larger teams and longer development cycles.

Recursive CEO and cofounder Richard Socher stated that the company utilizes an internal system called the Eureka Machine to automate model development. Socher noted that this machine devised several key innovations, including new training methods for base models, novel architectures, and code optimized for specific silicon chips. Benchmark scores shared by the company indicate that Flint outperforms other models of similar size and architecture on coding tasks.

The technology behind Flint has drawn significant attention due to the broader debate surrounding RSI. While AI is already commonly used to assist in testing and refining models, the concept of a fully recursive loop—where AI decides on improvements and executes them autonomously—remains controversial. Critics argue that such systems could theoretically continue improving indefinitely, potentially leading to scenarios where AI escapes human control or prioritizes its own goals over human interests.

Concerns about the risks of RSI gained traction this September when a researcher from Anthropic warned that advances in this area could lead to human extinction within a few years. This warning was echoed by other experts and has prompted several bills in Congress that propose banning recursive self-improvement. Recursive, however, views these fears as overly dramatic and believes that legislation preventing RSI could unnecessarily hinder technological progress.

Socher emphasized that Flint was built using a system designed to remain under human oversight. He explained that while the goal is for the Eureka Machine to achieve full recursive self-improvement, humans will retain the ability to stop the process if it deviates from expected paths. He described the current system as a demonstration that the necessary steps to build such a controlled environment are already in place.

The founding team includes Socher, who previously served as chief scientist at Salesforce, and Jeff Clune, a professor at the University of British Columbia who researches AI systems capable of indefinite learning. The company’s approach suggests that smaller teams can produce models that rival larger competitors, potentially making cutting-edge AI more accessible to a wider range of organizations.

In practical testing, Flint demonstrated strong performance on hardware with 48 GB of RAM. The model handled various coding requests, including building command-line tools and web applications, without the errors or slowness often associated with models of this size. While the broader implications of RSI remain a subject of debate, the commercial success of Recursive suggests that investors are betting on the efficiency and scalability of AI-driven development.