4.65 Billion: The Valuation Backing the First Open-Weight Model Built by AI

The debate over whether artificial intelligence can significantly improve its own development process has moved from theoretical speculation to practical demonstration with the release of a new open-weight model named Flint. The software, developed by the startup Recursive, suggests that recursive self-improvement, or RSI, is transitioning from a controversial concept into a viable engineering methodology.

Flint is the first product from Recursive, a company founded with the specific goal of using AI to advance AI and other scientific fields. CEO and cofounder Richard Socher stated that nearly every aspect of the model’s pipeline was executed and refined by AI itself. By automating the majority of the development process, the company was able to produce a world-class model in a matter of months, a timeline that would typically require a much larger workforce.

The automation was driven by Recursive’s internal system, dubbed the Eureka Machine. According to Socher, this system devised several key innovations, including novel methods for training the base model, new architectural designs, and code specifically optimized for particular silicon chips. Benchmark scores released by the company indicate that Flint outperforms other models of similar size and architecture on coding tasks, validating the efficacy of the automated approach.

While AI is already commonly used to assist in building and testing other AI systems, the RSI approach creates a more closed loop where the system determines what improvements to make and initiates the next round of optimization. Although currently limited, this technique could theoretically allow for indefinite self-improvement if the AI can contribute to every stage of model development. This capability has raised concerns among experts who worry that an AI capable of improving itself without limit could eventually escape human control, potentially viewing humans as obstacles to its objectives.

These anxieties gained significant traction earlier this year when a researcher leaving Anthropic warned that advances in RSI could lead to human extinction within a few years. The concern was echoed by other experts, leading to several bills in Congress proposing to ban the practice. In contrast, Recursive views these fears as more science fiction than immediate reality, arguing that legislation preventing RSI would unnecessarily hinder progress.

Socher emphasized that Flint was built using a system that remains under human oversight. The long-term goal is for the Eureka Machine to achieve full recursive self-improvement, but it would still be monitored by humans who retain the ability to stop the process if it deviates from its intended path. Socher expressed hope that the release of Flint demonstrates that the necessary steps to build such a controlled system are already in place.

The startup has secured substantial financial backing to support this vision, raising $650 million at a $4.65 billion valuation from investors including Google Ventures, Greycroft, Nvidia, and AMD Ventures. Socher, formerly the chief scientist at Salesforce, is joined by cofounder Jeff Clune, a professor at the University of British Columbia who researches AI systems capable of indefinite learning.

Practical testing of Flint reveals a tool that is both fast and capable, even when run on modest hardware such as a Mac with 48 GB of RAM. The model, which contains 40 billion parameters, successfully handled various coding requests, from building command-line utilities to creating simple websites explaining RSI. This performance is notable because models of this size are often error-prone or slow compared to larger industrial models, yet Flint executed tasks without issue.

The broader implications of this technology are significant for the accessibility of cutting-edge AI. Socher noted that the approach used to build Flint allows a small team to produce a model that surpasses others in its class, a feat that would be difficult without such automation. As other labs, including teams from Microsoft, Carnegie Mellon University, and the University of Virginia, publish work on recursive improvement, the trend suggests that using AI to build better AI is becoming a standard part of the industry’s evolution.

The release of Flint marks a pivotal moment in the AI landscape, shifting the focus from merely using AI as a tool to allowing it to drive its own evolution. While the risks of unlimited self-improvement remain a topic of debate, the current evidence suggests that controlled RSI offers a promising path for smaller organizations to compete with larger tech giants.