OpenAI publishes AI proofs for hundreds of unsolved math problems

We’re releasing a broad range of new mathematical results produced by an internal frontier model. We’ve been consulting with the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study, and we have drawn on their advice and public recommendations to inform how we release these results.

OpenAI's Math post is up!

SITUATION DETECTED: OpenAI has released a GitHub repository with over 700 math problems solved by its models.

Completely unrelated to the rest of the ships today, but our models did solve some important math problems and the team documented it in the following blogpost. Remarkable new era of scientific progress.

mathematicians: frontier labs need to publish real proofs so we can check them, not blog posts OpenAI: fine. here’s 722 papers

This OpenAI math drop is nuts. This 7/8 quasi-Riemann hypothesis proof looks like one of the most important advances in analytical number theory in decades. And they also resolved one of the most famous decidability problems in number theory! Plus 3,998 more things! Oh my god.

"The average result used the equivalent compute of roughly three hours of ChatGPT Pro thinking."

We are entering a new era of discovery now:

My partner is a theoretical computer scientist (essentially a mathematician). She has mixed feelings about AI's impact on math research overall, but her feeling about today's release is one of overwhelming awe and excitement. Literal tears of joy. Quote: "so lucky to be alive rn"

towards acceleration of scientific discovery and improving quality of life for everyone

OpenAI’s internal frontier model looking at the world’s hardest open math problems rn:

.@theojaffee says OpenAI’s math results made him "less AGI-pilled." @cherylwoooo explains why they may actually make RSI more plausible: "There are capabilities that are narrow, where a self-sustaining acceleration could happen, but it doesn't generalize to a broader capability that could cause more economic consequences." "We've been seeing more discoveries over the last year on algorithmic progress. There were two pull requests on NanoGPT that basically made more than one year of improvements within one pull request. In the last one month, there were two of them." "That's evidence that there's gonna be more progress happening down the road, as well as math progressing, and more discoveries of vulnerabilities." @ElasticityInst

can't stop reading mathematicians' mixed reactions to OAI solving ~400 open problems yesterday, such as one below "as an early-career mathematician, these questions are quite personal: what will it mean to do research? Which contributions will our profession learn to value?"

Scott Aaronson on The Mathocalypse. 'The UGC proof invents a completely new bizarre code with a noise test. It's some crazy recursive construction. It's not the long code, not the short code - some alien craziness' I hear a lot of this today. Get used to alien craziness.

From Scott Aaronson on OpenAI's math drop - "Apparently they used about 3 hours of GPT-Pro level compute on average per problem solved", so maybe a couple hundred dollars at API prices, for problems that were untouched by humans for decades

Update from Scott Aaronson: Some AI companies are now using their latest internal models to take a run at breaking 'important cryptographic protocols and primitives.'

“Cryptography is a subfield that’s extremely conspicuous by its absence from OpenAI’s list of 376 papers! But my sources tell me that the AI cos have now started, gingerly and discreetly, investigating whether their latest internal models can break important cryptographic protocols and primitives.” - Scott Aaronson, CS chair at UT Austin