More than 700 preprints in various mathematical fields have been published simultaneously by OpenAI.Credit: Monty Rakusen/DigitalVision via Getty
The research community is in turmoil after OpenAI published on October 6 a trove of more than 700 mathematical preprints entirely generated by AI. The San Francisco, California-based creator of ChatGPT has posted the preprints to the Github software repository.
Although some mathematicians celebrated the solution of long-standing problems, others complained on social media about being shortchanged. Some were furious at what one physicist called a “slopocalypse,” even though the mathematical content could ultimately be proven formally correct.
The number of papers published at once is “staggering,” Alvaro Lozano-Robledo, a mathematician at the University of Connecticut in Storrs, wrote on the social media platform X. “Why is OpenAI trying to solve hundreds of math problems and publishing them at once? “, he added.
In an email to Naturean OpenAI spokesperson wrote that the company’s goal is to provide tools for mathematicians and scientists. “To achieve this, we evaluate and develop our models against difficult frontier problems across the sciences. »
Who gets the credit in the AI era? OpenAI math bomb sparks debate
Levent Alpöge, a mathematician at rival company Anthropic, also in San Francisco, commented on X that there would be “sad stories” for some mathematicians who might see their research projects preempted by the OpenAI version. “But we should put that aside for today. This is obviously the most important moment in the history of mathematics,” he wrote, because it shows how quickly technology has become useful to mathematicians.
Valuable issues
The large number of preprints contained no solutions to the remaining five “Millennium Prize Problems,” selected by the Clay Mathematics Institute, scientifically headquartered in Oxford, United Kingdom, at the turn of the 21st century. (OpenAI’s preprint on the Navier-Stokes equation, published in September, solved one of the original seven problems. Another was solved by mathematicians in the early 2000s.)
Yet some of the preprints focused on issues related to these five issues, including Hodge’s conjecture on complex algebraic varieties and the Riemann hypothesis on the statistical properties of prime numbers. The OpenAI spokesperson said: “We will leave it to the wider community to assess (their) importance. »
Gn bussni
