AI solves decades-old maths problem in 88 hours, sparking fears for mathematicians

An artificial intelligence system has reportedly solved a decades-old mathematics problem in 88 hours, a result that has left some of the world's leading mathematicians exhilarated and deeply worried. The development was described in a report featuring academics from Oxford, Imperial and Cambridge, who said the speed of progress had taken them by surprise.

According to the report, AI agents cracked a problem that had left humans stumped in just 88 hours. The same report noted that in September OpenAI announced a proposed proof of the Navier-Stokes equation, one of the Millennium Prize problems. Those seven problems were set in 2000 to inspire human mathematicians.

Mathematicians said the pace of change has been extraordinary. During this year, they said, AI systems have gone from mediocre PhD student level to very good PhD student level, then to experienced mathematician level, and on to top mathematician level. One described the progress as shockingly good. Another said the rate of progress was astonishing. The report said AI has reached a point where even the best mathematicians struggle to judge its capabilities.

The breakthrough has raised hard questions about the future of the profession. One mathematician said the first honest reaction was despair, while another said the field was still on the beach and could see the tsunami but it had not quite reached them. A big worry is the careers of future mathematicians. The situation for young people feels basically dire, one said. Many people are upset and traumatised by what is happening because it is unclear what the way forward for the profession is.

Some mathematicians argue that core parts of the work remain human. Mathematics is about exploration, understanding and figuring out which questions to ask, one said, adding that this still appears to be exclusively a human endeavour. The concern is that AI may solve problems without providing an understanding of how they were solved, or the ideas a human would have developed along the way. If that continues, researchers could lose sight of what to pursue next and fail to see the big ideas.

Others see opportunities as well as risks. One said the excitement is that science and mathematics could move forward faster than ever in human history, driven by human beings. The worry is that we are losing sight of the humanity. Another said the moment is a chance to push mathematics forward at a much faster rate than before. Some expect that in six months the systems will be much better and may be able to teach mathematicians things.

Yet the psychological and professional impact is already being felt. The report compared mathematicians to a profession on the front line of automation. It warned that automation can make people faster and more capable, but can also take away something that felt like it belonged to them: identity, purpose, expression and joy. Up until now, the advancement of mathematics and the advancement of humanity have been the same thing, one said. It feels like AI might prise those two things apart, bringing gains for humanity as a whole but losses on a personal level. That hard bargain may be one that more workers face as AI continues to progress.