OpenAI's math discoveries raise big questions from experts

OpenAI surprised the math world on Tuesday when it released more than 370 new mathematical discoveries. The company used some of its most powerful AI models to find results in areas like algebra, computer science, and mathematical logic. The release got a lot of attention — but it also made many math experts nervous.

Just last month, OpenAI's AI solved the Navier-Stokes equation, one of the hardest math problems in the world. Solving it comes with a $1 million prize for anyone who cracks it. That achievement made headlines around the world, but it also started a serious debate among mathematicians.

Many math leaders say there is a big problem with how OpenAI is working. They argue that the company is testing its powerful AI on the hardest math problems, but other mathematicians can't use the same AI tools. This makes it unfair for the wider math community. The Institute for Advanced Study, an independent group of math experts in Princeton, New Jersey, said it does not support this approach.

The Institute released a statement raising a key concern about AI and math. It said that AI can now produce math arguments that even the person who asked the question cannot fully understand or check. The group wrote, 'We believe that human understanding of mathematics remains of paramount importance.' They asked: how can we make sure humans still truly understand math in this new age of AI?

OpenAI responded by saying it would partner with the Institute for Advanced Study. The company said it wants to give mathematicians 'a voice in how we move forward.' However, OpenAI did not say it would stop using AI to tackle these tough problems. Many experts say they are still worried that OpenAI isn't doing enough to carefully check and verify its results.

One mathematician who spoke out is Tristan Buckmaster from New York University. He told the New York Times that people who ask AI models math questions might actually be giving the AI hints that help it reach the answer. 'There's likely to be a bunch of results where they take someone's work and then take it to completion,' Buckmaster said. This raises questions about how much credit the AI really deserves.

The Institute's advisory board also called on AI companies to give fair access to their tools. They want mathematicians all over the world to be able to use the same AI models that OpenAI uses internally. Without that access, they warn, AI labs could race ahead of the rest of the field. That could leave most of the world's mathematicians shut out from important discoveries in their own subject.