OpenAI releases over 370 math results, raising concerns in academia

OpenAI on Tuesday released more than 370 mathematical results spanning algebra, theoretical computer science, and mathematical logic, showcasing the capabilities of its most advanced AI model. Earlier, the company used AI to solve the Navier-Stokes equations, raising concerns within the field that frontier labs should not test the hardest mathematical problems on proprietary models that the mathematical community cannot access. The Institute for Advanced Study in Princeton said it does not endorse this practice, noting that AI can produce mathematical arguments that the prompter cannot understand, verify, or take responsibility for, and called for preserving human understanding of mathematics in responsible scholarly output. OpenAI announced it will work with the Institute for Advanced Study to give mathematicians a say in how it proceeds, but did not indicate it would stop testing models on these difficult problems. Tristan Buckmaster, a mathematician at New York University, told The New York Times that the person prompting the AI to solve the equations may have provided information that helped the model reach its conclusion.