OpenAI’s solutions to around 300 tough problems put the math community on edge ... progress on 3 of 5 major challenges

[Financial News] The math community was shocked after AI solved hundreds of problems at a stroke, including tough problems that mathematicians had spent years grappling with.
The U.S. daily The Wall Street Journal (WSJ) reported on the 7th (local time) that OpenAI had made public new solutions to around 300 math problems spanning fields such as algebra and mathematical logic.
These included progress on three of the five Millennium Prize Problems, considered among the most challenging problems in mathematics, including a proof of the Quasi-Riemann Hypothesis.
Alex Kontorovich, a distinguished professor at Rutgers University, said of the proof of the Quasi-Riemann Hypothesis, “If a human had done this, they would unquestionably have received the Fields Medal immediately.”
The math community is unsettled by AI’s challenge. When rumors circulated early this week that OpenAI would soon pour out a large number of mathematical proofs, mathematicians rushed to publish their research findings ahead of it.
Until now, academia has expressed a negative view of AI’s attempts to tackle difficult mathematical problems.
Last month, after OpenAI disclosed that it had solved the Navier–Stokes problem, which carried a $1 million prize, 25 Fields Medal winners issued an open letter warning that AI-generated mathematical solutions risk undermining the fundamental purpose of mathematics as a discipline.
The Institute for Advanced Study in Princeton said, “Human understanding of mathematics is most important,” and pointed out that “AI is now producing mathematical arguments in situations that even the humans who entered the prompts cannot understand or verify.”
Another concern was that AI was tackling difficult problems by training on others’ prior research without authorization or using its own models that had not been made public to the broader math community.
In response, OpenAI established an independent advisory committee that includes prominent mathematicians and said it would seek ways to use AI to support mathematical understanding.
However, OpenAI rejected the math community’s request to stop testing advanced math problems with its own AI model, explaining that doing so was “very important in accelerating the development of tools to advance the field.”
[email protected] Lee Seok-woo, international affairs correspondent Reporter