OpenAI announced that a new internal model solved over 100 long-standing mathematics problems across multiple fields, including the Hodge conjecture and the Navier-Stokes Millennium Problem, after approximately one month of training. The training run began on August 28, taking internal researchers by surprise due to the speed of development. The specific solutions, methodologies, and peer-reviewed implications have not yet been fully detailed or released to the broader scientific community.

The announcement coincides with OpenAI establishing the independent Advisory Group on Mathematics and Artificial Intelligence based at the Institute for Advanced Study. Featuring prominent mathematicians such as Fields Medalist Timothy Gowers, the unpaid group is designed to connect OpenAI with researchers, advise on public output, and help share benefits. However, OpenAI explicitly stated that the advisory board will have no control or oversight over the pace of its internal mathematical research.

This push comes amid growing criticism from academia, including an open letter from mathematicians warning that churning out automated solutions threatens true conceptual understanding. Observers note the aggressive public claims around math achievements may also serve to sustain investor interest amid stiff competition and shifting internal research priorities.

Why it matters

  • Rapid progress in AI math solving demonstrates increasing capability in self-verifiable domains, signaling upcoming shifts in automated scientific discovery.

  • Vendor governance boards may offer advisory input on distribution but rarely hold operational control over model development speed.

  • Startup founders should expect scrutiny when using automated problem-solving metrics without publishing verifiable technical methodologies.

Source: the-decoder.com