OpenAI has announced a broad range of new mathematical results produced by an internal frontier model. To improve how it shares results with the mathematics community, OpenAI said it has consulted the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study.
The company has drawn on the group’s advice and public recommendations to inform this release. It is also exploring other community-hosted alternatives that meet the committee’s guidelines.
For this release, OpenAI is publishing the results in a GitHub repository with protocols for paper revisions and citations. The repository also includes:
- Formalizations of many of the proofs in Lean
- Additional details about how the results were obtained
Lean is a programming language that allows mathematical proofs to be checked by a computer. OpenAI said it will update the repository with more formalizations as it obtains them. For future releases, the company plans to further improve the quality of its papers through citations, mathematical exposition and presentation of the results for better understanding.
As part of the additional information published in the repository, OpenAI is sharing details about the compute used and the problems attempted. These include:
- 10 summaries of the model’s reasoning
- Estimates of compute spent in terms of ChatGPT Pro usage
- Statistics on the number of attempted problems
- Compute equivalent to roughly three hours of ChatGPT Pro thinking for the average result
OpenAI will fund a series of workshops, conferences and special programs focused on understanding major results produced by AI. The company said it will share more information about these programs in the near future.
Meanwhile, OpenAI is working to responsibly release the model that produced these results. The company said it will continue evaluating its internal frontier models in mathematics and other sciences to accelerate the development of tools for those fields. It will also continue to act on feedback from the community and update its standards for future disclosures of major scientific advancements.