Physicist Sabine Hossenfelder calls AI the way forward and argues that human understanding should not be a prerequisite for scientific progress, while mathematician Hugo Duminil-Copin says human understanding of mathematics is ultimately the only thing that benefits society

Physicist Sabine Hossenfelder calls AI the way forward and argues that human understanding should not be a prerequisite for scientific progress, while mathematician Hugo Duminil-Copin says human understanding of mathematics is ultimately the only thing that benefits society

Duminil-Copin spoke on the Numberphile podcast on October 14; Hossenfelder posted on X on October 19. That same day, immunologist Derya Unutmaz addressed both positions, criticizing mathematicians and agreeing with Hossenfelder.

On October 6, OpenAI publicly released hundreds of mathematical manuscripts containing results that, according to the company, its internal model had obtained. Back in August, according to mathematician Bryna Kra, participants in a meeting with the company asked it to publish the results in papers that mathematicians could learn from and use. Hugo Duminil-Copin, who received the 2022 Fields Medal for solving longstanding problems in statistical physics, found results from his own field among the manuscripts. According to the podcast host, at least 16 papers concern his field, and Duminil-Copin estimates that they total at least 5–6 thousand pages.

In a post on October 19, Hossenfelder writes that scientific progress requires a willingness to accept that a correct description of nature will eventually be too complex for humans to understand. She concludes:

“I do not think it is reasonable to make human understanding a requirement. AI is the way forward.”

That same day, an X user shared a clip from the conversation with Duminil-Copin in which he says that people who do not want to use AI should be allowed to continue working as they have before. Unutmaz replied that mathematicians were acting as though they owned mathematics and asked how many people outside their profession understood it. He added that it was time to end the hostility toward AI. Later, he wrote that he fully agreed with Hossenfelder and considered AI the only way for science to progress.

In the full conversation, Duminil-Copin explains why understanding matters even after an answer has been found. In early September, he says, the most important conjecture in his field was proved. He asked the host, “Has your life changed?” The answer was “no.” Only the mathematicians who had worked on the conjecture, he explained, could say that the proof had changed anything at all. He then says:

“Human understanding of mathematics is the only thing that ultimately benefits society.”

He says that the mathematical concepts and techniques used in other sciences rarely come from the latest results. More often, they come from mathematics that people have absorbed and thought through, coming to understand where it can be applied. He reiterates that being able to calculate is pointless if you do not know when to do so, and that the same holds for humanity. If the purpose of mathematics is to produce as many proved and computer-verified results as possible, one can imagine it being fully automated: a black box produces an answer, and nobody knows why it is correct. Mathematics is also a human science, however, in which understanding must be developed so that it can be passed on to others.

He does not use AI in his own creative work. He may be slower and less original as a result, but he will at least contribute something unfamiliar to the algorithm. He worries that AI-generated ideas could become the norm, whereas mathematics needs people to explore diverse approaches, including mistaken ones. He plans to study OpenAI’s manuscripts and build on their ideas: the day after their release, his group in Geneva decided to work through them. He will not ask the model for ideas. He considers himself neither an advocate nor an opponent of AI and leaves his students free to choose.