opinion tech single source: Scientific American

AI in the Ivory Tower: Can Machines Make Math More Human?

Many mathematicians approach new ideas with caution. It’s a pattern we see across science; it took years for computers to start reshaping entire fields. Now, a similar shift feels imminent as artificial intelligence, especially large language models, starts entering mathematical practice.

Yang-Hui He, a mathematician and physicist from the London Institute for Mathematical Sciences, shared insights into this potential transformation. He noted that physicists were generally quicker to accept computational tools—CERN has been using machine learning since at least the 1990s. Mathematicians, however, remain much more skeptical of these innovations.

For years, He described his work like being a "traveling salesman," moving between fields looking for hidden structures within data. A concrete example came during Phase Four of a project in the summer of 2025: thirty mathematicians were isolated in a room tasked with creating difficult, unknown math problems. Astonishingly, within one month, the AI system managed to solve about ten percent of those challenges.

This isn't just small progress; it suggests something deeper is happening. The project has moved into Phase Five, tackling major open problems—raising the possibility that AI could soon achieve genuine mathematical breakthroughs.

The big question remains: will this powerful integration make mathematics more human or simply make it faster?