OpenAI unleashes hundreds more math results upon a field in shock
3 hours ago
- OpenAI released 372 results from a new large language model, claiming breakthroughs in math and theoretical computer science, including progress on the Kakeya conjecture and Riemann hypothesis.
- Many results are verified in Lean, but mathematicians are skeptical due to OpenAI's lack of transparency and refusal to release the model or prompts.
- The model reportedly produced results from a single prompt to a single AI agent, a stark contrast to the costly swarm approach used previously for the Navier-Stokes problem.
- OpenAI formed an advisory group for responsible publishing but is not fully following its recommendations, choosing only to disclose average compute time.
- The math community is divided: some see releasing results as beneficial, while others, like Terence Tao, criticize the pace as 'insane'.
- OpenAI plans to continue, stating that solving math problems is an essential test for AI intelligence, despite internal confusion over the results.