AI turns mathematics upside down

“We may actually need more mathematicians”

Mathematicians have been hit “amidships” by the rise of AI. Should they fear for the future of their field? Or will AI actually make mathematics more accessible? Several experts share their views, including TU/e mathematician Jim Portegies and Robbert Dijkgraaf: “We could see enormous breakthroughs.”

by
photo Robbert Dijkgraaf | Gabi Porter, Jim Portegies | Lieke Vermeulen

In October, OpenAI, the company behind ChatGPT, published solutions to hundreds of mathematical problems online. Armed with an enormous amount of computing power (which likely cost millions of dollars), the AI company has triggered an existential crisis in mathematics.

Researchers who have spent years working on a specific problem now have to watch as their expertise is taken over by a large language model from Silicon Valley. Mathematics has weathered all kinds of technological innovations in the past: calculators and computers became useful tools. But this?

“No future?”

“I think mathematicians should fear for their future,” says Piet Lammers of Sorbonne University in Paris. “Their raison d’être has always been their ability to prove new mathematical theorems, but that is disappearing.”

In October, Lammers wrote an alarming article about the issue for Dutch newspaper NRC. “Almost all the problems I was working on have now been solved,” the mathematician says.

His PhD candidates have also been affected, he explains in a follow-up email. They had been looking forward to spending years wrestling with mathematical puzzles, but now they have to focus on checking the solutions provided by OpenAI.

For the time being, Lammers is not taking on any new PhD candidates, even though he recently received a €1.4 million European research grant. “I don’t know whether they have a future in this field.”

The canary in the coal mine

Artificial intelligence is hitting mathematics “amidships,” says theoretical physicist Robbert Dijkgraaf. But scientists in other disciplines should not assume that mathematics is the only field affected, the former Minister of Education warns. They, too, will feel the impact.

People will remain important in this field. In the coming years, they will need to help shape the future of mathematics (...)

Jim Portegies
Mathematician at TU/e

“Already, we’re seeing a flood of AI-generated scientific papers, grant applications, and conference submissions,” Dijkgraaf says. “This affects the way we share scientific insights with one another. Reports are being flooded with AI-generated data. No one has the time to check it all.”

Optimism

Dijkgraaf wonders whether mathematicians will be able to retain control now that commercial AI models, in particular, are turning the field upside down. Scientists have little insight into how ChatGPT or Claude arrive at their answers. Does AI steal the ideas you discuss with it? “AI can be a very unreliable partner,” Dijkgraaf warns.

Like Lammers, Dijkgraaf also wonders what this means for young people who are about to become mathematicians. How do you train a new generation for a field that is undergoing such enormous changes at the same time?

But he also sees a positive side. “We can now solve some very difficult problems. AI is already being used to design nuclear fusion reactors. It’ also being used in weather and climate models. We could see enormous breakthroughs.”

 “More mathematicians”

Jeroen Zuiddam also sees many opportunities for mathematics. He is a mathematician at QuSoft, a collaboration between the University of Amsterdam (UvA) and the Centrum Wiskunde & Informatica (CWI), the Dutch national research institute for mathematics and computer science. “The field is becoming more accessible,” he says.

It’s not motivating to invest lots of time and effort in mathematical problems when you know they can be solved in a single day (...)

Piet Lammers
Mathematician at Sorbonne University

“Mathematics is such a broad discipline that most people only know a small part of it. Thanks to AI, they can also use other areas of mathematics. And we can focus more on the practical applications of mathematics.”

Mathematics can therefore flourish, he says. “We can broaden our ambitions and set bigger goals. If AI makes mathematics easier to apply, we may actually need many more mathematicians.”

Distrust

However, many mathematicians are digging in their heels, saying they want to avoid or ban AI altogether. Twenty-eight former winners of the prestigious Fields Medal described AI as a “threat to the field” in a statement. What damage could it cause? That will depend on “the people who are in control of this new technology.”

The Association for Human Mathematics (AHM), founded in September, already has more than 1,100 members. They pledge not to put their expertise at the service of commercial AI companies. One subgroup even wants to avoid using AI altogether.

Distrust of OpenAI is widespread within the AHM. Many accuse the company of not genuinely caring about mathematics itself. Instead, they say, it simply wants to demonstrate how impressive its AI tool is, at the expense of mathematical careers.

Ideological

There are all kinds of reasons to steer clear of AI, says TU/e mathematician Jim Portegies. The energy-hungry data centers required to run AI, for example, are far from sustainable. Portegies understands that some mathematicians will use AI, “but there must also be room in mathematics for people who don’t want to.” He believes there is quite a bit of evidence that people learn better without using AI.

AI is frightening, especially when you’re just starting out. But so much is changing that, in a way, we’re all back at the beginning of our careers

Robbert Dijkgraaf
Theoretical physicist and former Minister of Education

Portegies thinks aloud about the possibility of a mathematics institute where students could be educated without AI. “I think graduates from such an AI-free institution would be highly sought after in the wider mathematical community or in the business world.”

This is precisely the question that continues to occupy Dijkgraaf as well. In principle, students or early-career researchers could set a whole bunch of AI agents to work on a task. But how do you learn to understand exactly what those agents are doing? How do you learn to assess the answers AI provides?

You could argue that AI gives students a personal coach they can bounce ideas off while studying. But why would students want to learn at all if AI can simply give them the answers?

Changing

Jeroen Zuiddam of QuSoft is confident that students who want to pursue a career in science are genuinely curious. “Ambitious students are certainly motivated to work hard and learn.”

But will students still have that ambition in the future? Piet Lammers of Sorbonne University has his doubts: “It’s not motivating to invest a great deal of time and effort in mathematical problems when you know they can be solved in a single day as soon as OpenAI turns its attention to them.”

The work itself will also change significantly, Jim Portegies believes. “But people will remain important in this field. In the coming years, they will need to help shape the future of mathematics and the role AI will play in it.”

That is also Dijkgraaf’s message: get involved. He understands why young people are worried about the impact of AI on their careers. “AI is frightening, especially when you’re just starting out. But so much is changing that, in a way, we’re all back at the beginning of our careers.”

This article was translated using AI-assisted tools and reviewed by an editor.

Share this article