OpenAI Claims Breakthroughs on Hundreds of Unsolved Math Problems
Translated & summarized from Ynet by baba
OpenAI has published solutions to 377 previously unsolved mathematical problems, including potential breakthroughs on three "Millennium Prize Problems." The company claims its advanced AI model solved these problems, with each solution averaging three hours of computation. The scientific community is divided, with some praising the achievement and others raising concerns about AI's creative capabilities and reliance on existing human research. Over 24 Fields Medalists have voiced opposition, warning of threats to the purpose of mathematics.
The story in 6 lines · by baba
- OpenAI published solutions to 377 unsolved mathematical problems, including potential breakthroughs on three "Millennium Prize Problems."
- Researchers are calling the achievement a "most significant moment in the history of mathematics."
- The AI model reportedly took an average of three hours of computation per solution.
- The scientific community is debating AI's creative capabilities and potential reliance on existing human research.
- Over 24 Fields Medalists have criticized AI's rapid pursuit of mathematical solutions, warning of threats to the field.
- Israeli startup doubleAI also announced a significant breakthrough in "extremal graph theory" using its AI agent.
OpenAI has announced a significant achievement, publishing over 700 papers detailing solutions to 377 previously unsolved mathematical problems across various fields including algebra, number theory, computer science, mathematical logic, and topology. Researchers cited in "The Wall Street Journal" and "The New York Times" are calling this a "most significant moment in the history of mathematics." Among the solved problems are potentially those related to three of the "seven Millennium Prize Problems," each with a $1 million reward for a solution. One of these is believed to be the "Riemann Hypothesis," concerning the distribution of prime numbers. OpenAI stated that each solution required an average of three hours of computation using an advanced, undisclosed AI model. This follows recent announcements from OpenAI regarding its AI's success in solving other complex mathematical challenges, including the "Navier-Stokes equations," another Millennium Prize Problem, just last month.
The use of AI to solve such profound mathematical problems has sparked debate within the scientific community. Some question whether these AI systems possess independent creative thought or are merely completing proofs by synthesizing existing human research. Concerns have been raised, with one academic claiming an AI model relied on his unpublished work. OpenAI has denied such allegations. Dr. Tristan Buckmaster, a mathematician at New York University, suggested that many solutions might be derived from existing work, and questioned the thoroughness of OpenAI's verification process given the sheer volume of published results.
Furthermore, over 24 Fields Medal recipients have expressed strong opposition, publishing an open letter titled "Serious Mismatch of AI and Mathematics," warning that the rapid pursuit of mathematical solutions by AI threatens the core purpose of mathematics. In response to scientific pressure, OpenAI announced it is collaborating with an independent advisory council of mathematicians, convened by the Institute for Advanced Study in Princeton, to establish guidelines for AI-generated mathematical findings. The council recently issued recommendations for how AI companies should publish their discoveries, based on feedback from hundreds of mathematicians.
OpenAI stated that it has taken the council's advice into consideration. In parallel, Israeli startup doubleAI, founded by Prof. Amnon Shashua, announced a breakthrough in "extremal graph theory" using its AI agent, solving a problem that had remained unsolved for 38 years with a budget under $2,000. DoubleAI focuses on developing "Expert AI" (AEI) for complex scientific and algorithmic problem-solving, aiming for expertise exceeding human specialists.