Mathematician Terence Tao said OpenAI's release of AI-developed mathematics disrupted some junior researchers' projects after mathematicians working on overlapping problems received no advance notice of the release. He said the episode raised concerns about how researchers share questions they hope to pursue.
In an interview published Thursday, Oct. 15, Tao said he had reviewed about 10 papers. He found a few ideas inventive, but said much of the collection was difficult to understand.
Tao described moderately challenging problems as a way for students and postdoctoral researchers to develop expertise and demonstrate their work. He worries mathematicians may keep questions to themselves if they fear an AI system could solve them first.
The current OpenAI mathematics catalog counts 719 manuscripts in 372 result families, compared with 722 in its Oct. 6 edition, as EconoTimes previously reported.
Tao also discussed a claim that excludes zeros of the Riemann zeta function and Dirichlet L-functions whenever their complex input has a real part above 7/8. The Riemann Hypothesis says every nontrivial zero of the zeta function has a real part of 1/2. Tao called the result the biggest advance on the problem in five decades, but said it does not prove the hypothesis.
He criticized the competition among AI companies to release more capable systems, saying it comes at the expense of human understanding and safety. Tao cautioned that putting systems into real-world use without understanding how they work could cause problems. He expects mathematics to involve larger, more industrialized collaborations, while arguing that conventional methods should remain part of education, much as children learn arithmetic by hand before using calculators.
OpenAI pledged $5 million to support mathematicians reviewing AI-generated research. Researchers receiving foundation support would also get AI tools and computing resources from OpenAI and Anthropic.
