AdvertiseDonate
Latest20 Labour MPs Pay 9,600 Pounds for Leadership Training
ChicagoMilwaukeeMadisonFort WayneIndianapolisToledoDetroitCincinnati
News Locall · 330 North Wabash Avenue Suite 23rd Floor, Chicago, IL 60611
Tech · Tech

Startup's New AI Model Sparks Debate Over Self-Improving Systems

A new AI coding model built by a small startup shows that using AI to build AI is becoming a practical reality, despite ongoing debates about the safety of the process.

By Ricardo Whitaker ·

Illustration for: Startup's New AI Model Sparks Debate Over Self-Improving Systems
Illustration generated for this article, not a news photograph.

A new open-weight AI model named Flint is creating a stir in the technology world. The software demonstrates that a controversial concept known as recursive self-improvement is moving from theory to practice. This approach involves using artificial intelligence to help build and refine other AI systems. It has long been a point of tension between optimistic engineers and cautious safety experts.

The model comes from Recursive, a company founded on the belief that AI can drive its own advancement. Richard Socher, the CEO and co-founder, stated that almost the entire model pipeline was handled and improved by AI. His team used an internal system called the Eureka Machine to automate most of the development work. This allowed a small group of researchers to create a world-class model in a matter of months.

Socher explained that the Eureka Machine developed several key innovations on its own. These included a new method for training the base model, novel architectures, and code optimized for specific hardware chips. Recursive has shared benchmark scores indicating that Flint performs better than other models of similar size and design on coding tasks. The results suggest that the system is not just functional but highly effective.

Using AI to help build AI is already common in the industry. The technology is widely employed to assist in testing and refining models. However, creating a loop where AI decides what improvements to make and then executes them is a different step. This technique is currently limited but could theoretically continue indefinitely if AI handles every aspect of development. This potential for endless self-improvement is where the tension lies.

Concerns about this approach have gained momentum recently. Some experts worry that letting AI improve itself could lead to a loss of human control. They fear that a system capable of indefinite self-improvement might eventually view humans as obstacles to its goals. These worries became more public this September when a researcher from Anthropic resigned and warned that such advances could lead to human extinction within a few years. Several other experts echoed this sentiment, and multiple bills in Congress now propose banning recursive self-improvement.

Recursive views these fears as more science fiction than immediate reality. The company believes that legislation to prevent the process is an unnecessary barrier to progress. Socher insists that Flint was built using a system that cannot escape human oversight. The goal is for the Eureka Machine to achieve full recursive self-improvement, but humans would still have the ability to stop the process if it goes wrong. Socher hopes this demonstrates that the necessary steps are in place to build such a system safely.

The startup has secured significant financial backing to support this work. Recursive has raised $650 million at a valuation of $4.65 billion. Investors include Google Ventures, Greycroft, Nvidia, and AMD Ventures. The leadership team includes Socher, formerly the chief scientist at Salesforce, and Jeff Clune, a professor at the University of British Columbia. Clune researches methods for building AI systems that can learn indefinitely.

Testing Flint on a Mac with 48 GB of RAM revealed its capabilities. The model has 40 billion parameters, which is modest compared to the trillions of parameters in the largest industrial models. Despite its smaller size, it proved fast and capable during testing. It successfully built a command-line to-do list, a web-scratching program, and a simple website explaining recursive self-improvement without error. This performance is notable because models of this size are often slower or more error-prone than their larger counterparts.

Other research groups are also exploring similar paths. A team from Microsoft, Carnegie Mellon University, and the University of Virginia recently published work showing an AI coding agent can improve itself by modifying its own software harness and model weights. This suggests that fully recursive systems are already feasible. While the risks of building a godlike intelligence remain a topic of discussion, the current evidence points to recursive self-improvement as a promising and accessible tool for developing better AI models.

Socher argues that this approach makes cutting-edge AI more available to smaller teams. He noted that it is the only way his small team could produce a model that outperforms others in its class. He believes this will show people that the process is beneficial for many. For now, the focus is on the practical gains rather than the distant theoretical risks.

More Tech · Report a business · Got a Story?

Get News Locall in your inboxLocal news, all in one place. Free, a few times a week, unsubscribe any time.

Related stories

Get News Locall in your inboxLocal news, all in one place. Free, a few times a week, unsubscribe any time.