OpenAI dumps 372 AI-generated math proofs on GitHub, telling the academic world to keep up
decoder ·
OpenAI has published 372 mathematical results generated by an internal AI model on GitHub, aiming to solve or advance open problems such as those related to the Riemann hypothesis. These proofs were primarily produced using a single prompt and agent, with each result consuming approximately three hours of ChatGPT Pro compute. To address the overwhelming volume of output that could strain manual review processes, many proofs include formalizations in Lean for machine-checkable verification. The company chose GitHub over traditional peer-reviewed journals, signaling that conventional academic publishing is too slow for this pace of discovery. While OpenAI consulted with an advisory group including Fields Medalists, the mathematicians had no say in the production speed or quantity of the results. Reactions within the mathematical community are divided, ranging from excitement to frustration over the implications of mass-produced proofs. Critics, including 25 Fields Medal winners, warn that focusing on problem-solving outputs may undermine conceptual understanding and genuine insight. Consequently, there are concerns that the rapid expansion of AI-generated literature could eventually outpace human comprehension.
decoder