OpenAI publishes 722 AI-generated math discoveries in major scientific milestone

OpenAI's team has unveiled 722 math papers created using an undisclosed artificial intelligence model. The papers, posted to GitHub, address a range of mathematical disciplines, including a significant contribution to the Riemann hypothesis, an unresolved problem concerning prime numbers that has perplexed mathematicians for over a century.

While the full answer remained elusive, the AI model proved the quasi-Riemann hypothesis, a crucial piece in the puzzle. The research also explored theoretical computer science, with the model generating over 80 papers on matrix multiplications, a critical operation in AI model computations, leading to a clearer definition of computational limits.

Additionally, the AI model developed a novel algorithm for integer multiplications, a fundamental aspect of many programs. The breakthroughs extend to physics, with proofs of De Giorgi’s conjecture, a hypothesis related to equations in metal alloy studies, and clarifications concerning the Navier–Stokes equations, pivotal for understanding fluid dynamics.

OpenAI has taken steps to enhance the discoverability and verifiability of these proofs, integrating Lean files into the papers and committing to publish Lean proofs for additional findings. Looking ahead, OpenAI aims to host research events and programs centered around evaluating the impact and reliability of AI-generated mathematical discoveries.