Recursive Debuts Flint, A 40B Parameter Model Built By AI Itself

The race to build artificial intelligence that improves itself has reached a critical juncture with the release of Flint. This new open-weight coding model from a startup called Recursive suggests that fully recursive self-improvement is no longer just a theoretical concept. Critics fear that letting AI build AI could lead to a system that escapes human control. Proponents argue it is the most efficient path to breakthroughs. Flint is the first product from Recursive, a company dedicated to using AI to advance both the field of AI and other scientific disciplines.

Richard Socher, the CEO and cofounder of Recursive, stated that nearly every part of the model pipeline was completed and improved by AI. The company utilized a recursive system known as the Eureka Machine to automate most of the development work. This automation allowed a small team of researchers to create a world-class model in a matter of months. Socher noted that the Eureka Machine generated several key innovations. These include a novel method for training the base model, new architectural designs, and code specifically optimized for certain silicon chips.

Recursive has shared benchmark scores indicating that Flint outperforms other models of similar size and architecture on coding tasks. The model contains 40 billion parameters. While this is modest compared to the largest models that require industrial server racks, Flint runs efficiently on consumer hardware. The performance raises questions about how far other US labs have progressed in chasing recursive self-improvement. Some published research from Microsoft, Carnegie Mellon, and the University of Virginia suggests that fully recursive systems are already feasible. These studies show how an AI agent can improve itself by modifying its own software and weights.

The controversy surrounding recursive self-improvement gained significant attention this September. A researcher who left Anthropic warned that advances in this area could lead to human extinction within a few years. Many other experts have echoed these concerns. The idea has become so polarizing that several bills in Congress propose banning the technique. Recursive views these fears as more science fiction than reality. The company believes that legislation to prevent recursive self-improvement is an unnecessary barrier to progress. Socher argues that Flint was built using a system that cannot escape human oversight.

The goal for Recursive is for the Eureka Machine itself to achieve full recursive self-improvement. However, this process would remain under human supervision. Humans could stop the process if it went wrong. Socher hopes this demonstrates that all the necessary steps are in place to build such a system safely. The company has raised $650 million at a valuation of $4.65 billion. Investors include Google Ventures, Greycroft, Nvidia, and AMD Ventures. Socher was previously the chief scientist at Salesforce. His cofounders include Jeff Clune, a professor at the University of British Columbia who researches AI systems capable of indefinite learning.

Testing Flint on a Mac with 48 GB of RAM revealed impressive capabilities. The machine has enough memory to run the 40 billion parameter model comfortably. I used Flint to build several programs and found it to be fast and accurate. Models of this size are often slow or error-prone compared to larger competitors like Claude or Codex. Flint, however, handled requests perfectly. I asked it to create a command line to-do list. I also had it build a program that searches the web for interesting research on recursive self-improvement. Additionally, Flint created a simple website explaining how the technology works.

The ability to use AI to help build, test, and improve AI is already common. This speeds up the refinement process and reduces the need for human input. The next step is creating a loop where AI decides what improvements to make and then begins another round of improvement. Currently, this technique is limited. If AI can contribute to every aspect of model development, the process could theoretically continue indefinitely. This has raised alarms among experts who worry about the potential for disaster. They claim that an AI capable of improving indefinitely might decide humans are an impediment to its goals.

Socher believes the approach used to build Flint will make cutting-edge AI more accessible. He stated that it is the only way a very small team can produce a model better than any other in its class. He hopes this shows people that recursive self-improvement is beneficial for many. While there may be risks if one is close to building a godlike intelligence, the current state of the technology appears promising. For now, it seems like a very effective way to improve AI models. The release of Flint marks a significant step in a field that is moving faster than many predicted.

The next steps for Recursive involve continuing to refine the Eureka Machine. The company aims to push the boundaries of what AI can do in science and engineering. The reaction from the broader tech community will likely determine how quickly other labs adopt similar methods. Investors are watching closely to see if this model of development can scale. The debate over safety versus speed is set to continue as more data on Flint's performance emerges.