Ankara:OpenAI has made public a large collection of mathematics research papers produced by an unreleased AI model. The papers, totaling 722 manuscripts, were released on the code-sharing platform GitHub and are grouped into 372 sets of related findings. OpenAI has cautioned that some of these findings may contain errors.

According to Anadolu Agency, The AI model, during its testing phase, tackled approximately 4,000 research problems. OpenAI noted that the published results are at various stages of verification. As part of this initiative, the company has also provided proofs written in Lean, a programming language that allows computers to verify the logical flow of mathematical arguments. More proofs in Lean are expected to be released as they become available.

OpenAI has also published 10 summaries that explain the model’s reasoning and estimates of the computing power utilized, which averaged about three hours of processing time equivalent to ChatGPT Pro for each result. The company sought advice from the independent Advisory Group on Mathematics and Artificial Intelligence at the Institute for Advanced Study on how to effectively present and share this research.

The advisory group emphasized that its role did not equate to an endorsement of the results. They described the release as a significant event for mathematics, with implications that reach beyond the individual findings to impact the broader mathematical community. They stressed that this public release is merely a starting point in integrating AI-produced work into mathematical knowledge, advocating for a future where human understanding remains central to mathematical research.