Sharing our AI progress in mathematics

On October 6, OpenAI published 722 papers and withdrew 3, averaging three hours of ChatGPT Pro computing power per result. On October 9, I published 722 papers and withdrew 3, averaging one minute of my laptop's computing power per result. This is an absurd time. If OpenAI can publish, so can I. As for what all this means for mathematics and humanity, no one knows, and I don't know either.

We're releasing a broad range of new mathematical results produced by a publicly available model running on a laptop.

As we look to improve how we share results with the math community, we have not consulted with any advisory group, as we do not have one.

For this release, we're publishing the results in a GitHub repository, with a protocol to mirror any revisions or withdrawals from openai/math. We are not exploring other alternatives for this release. For future releases, we have no plans beyond following theirs.

As part of our GitHub repository, we are sharing a script that exactly verifies all of the results. This script, written in Python, allows the configurations to be checked by a computer in about 15 seconds. We will not be updating the repository with more formalizations.

To promote scientific transparency and openness, we are also publishing additional details about how we obtained the results in the repository. These include 3 summaries of the model's reasoning, estimations of compute spent on a single laptop, and statistics about the number of attempted problems. The average result used the equivalent compute of roughly two minutes of laptop processing.

We want this progress to push the frontier of human knowledge and enable further progress in mathematics. We will not be funding any workshops, conferences, or special programs.

We want to directly empower scientists with state-of-the-art capabilities and are not working to release the model that produced these results, as it is already publicly available. This is why it is important to evaluate publicly available models on mathematics, so we can accelerate developing the tools to advance those fields. We will continue to act on feedback from the community, as our repository's issues and discussions are open, and update our standards for future disclosures of major scientific advancements.

In its statement on OpenAI's release, the Association for Human Mathematics said: "Releasing over 700 files at once is not a demonstration of scholarship, but a demonstration of power."

• OpenAI's release: Sharing AI progress in mathematics · github.com/openai/math
• My repository: github.com/Ge-limin/math