OpenAI has released hundreds of new mathematical findings, publishing over 370 results across various topics including algebra, theoretical computer science, and mathematical logic on Tuesday. This follows the company's announcement last month that its advanced artificial intelligence models had solved the Navier-Stokes equation.
The achievement has led to concerns from leaders in the field. The Institute for Advanced Study in Princeton, New Jersey, an independent group of mathematical experts, stated it does not endorse the practice of testing advanced mathematical problems on proprietary AI models that are not accessible to the broader mathematical community.
The Institute for Advanced Study highlighted that AI can now produce mathematical arguments that humans may not be able to understand, verify, or take responsibility for. They believe that human understanding of mathematics remains paramount.
OpenAI has announced it will collaborate with the Institute for Advanced Study to give mathematicians a voice in future developments. However, the company did not indicate it would cease testing its AI models with these advanced problems. Experts are concerned OpenAI may not be conducting the necessary due diligence to vet these results.
Tristan Buckmaster, a New York University mathematician, suggested that mathematicians prompting AI models might be providing information that helps the model achieve its results. The advisory board has also requested that AI labs grant equitable access to their AI models for the global mathematics community, warning that the use of proprietary models risks creating a two-tier system.