OpenAI has published a collection of 372 mathematical results generated by an internal AI model, bypassing traditional academic journals in favor of a GitHub repository.
Key facts
- •The collection includes 372 mathematical results, including advances related to the Riemann hypothesis.
- •OpenAI used the Lean programming language to provide formalizations for many of the proofs to facilitate machine-checkable verification.
- •The company bypassed traditional peer-reviewed journals, citing the need for a faster process to handle the volume of AI-generated knowledge.
- •Fields Medal winners Timothy Gowers and Terence Tao have expressed concerns regarding the long-term impact of AI on mathematical literature and the training of young mathematicians.
- •OpenAI stated that it is working on a responsible release of the model to provide scientists with state-of-the-art capabilities.
OpenAI has released 372 new mathematical results produced by an internal frontier AI model. The proofs, which aim to solve or advance open mathematical problems, were published on GitHub rather than in peer-reviewed journals. The company intends for these results to undergo formal verification using the Lean programming language to assist the mathematical community in reviewing the large volume of data.
By the numbers
Methodology and Compute Costs
OpenAI reports that nearly every result was generated from a single prompt to a single AI agent, though some required multiple attempts. On average, each result consumed approximately three hours of compute time equivalent to ChatGPT Pro Thinking. This process differs significantly from a previously announced Navier-Stokes solution, which required a swarm of 10,000 agents and millions of dollars in compute resources.
Academic Engagement and Concerns
OpenAI consulted with the Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, which includes Fields Medal winner Timothy Gowers. While the company plans to fund workshops on AI-produced results, it maintained that the advisory group could not influence the speed or method of production. The release has drawn criticism from 25 Fields Medal winners, who signed an open letter warning that mass-producing mathematical statements may prioritize problem-solving over conceptual understanding and insight.
Advertisement
This article was independently rewritten by ManyPress editorial AI from reporting originally published by The Decoder.
