OpenAI has published hundreds of new mathematical findings generated by an unreleased AI model, posted to GitHub. Some of the papers prove long-running hypotheses and others disprove proposed explanations, drawing concern from experts who worry OpenAI is not doing due diligence to vet results and that its AI models are not accessible to the broader field of mathematics.
OpenAI published 372 AI-generated mathematical results on GitHub, including Lean formalizations for machine verification, with each result reportedly using about three hours of ChatGPT Pro compute on average.
Written by AI from The Decoder's published text. Read the original for full details.
More coverage of this story (6)
OpenAI publishes a large batch of AI-generated mathematical results
Primary source
Reporting from a publisher outside the organisation being covered.
Read original story ↗