Will Top Talents Like Wang Hong and Wei Dongyi Be "Outperformed" by AI?

A professor who won one of the biggest awards in mathematics was moved to tears by AI.

"I cried on Tuesday night and woke up my daughter."

On October 8, French mathematician Hugo Duminil-Copin shared what had happened to him during an online conference. He won the Fields Medal in 2022 and has been one of the world's leading mathematics researchers for years. What made him cry was a huge collection of mathematics research results that OpenAI had just released.

Of those results, 16 problems were directly connected to his own research.

In a later radio interview, he compared AI to a bulldozer that was flattening all the problems he had been studying. People who had spent years working could understand his situation almost immediately. It was like this:

Imagine spending many years learning and building your skills. You finally discover a business idea that could make you rich. Then someone suddenly tells you:

AI already built that entire business model half a day ago, at no cost.

And Hugo Duminil-Copin is certainly not the only person affected.

OpenAI released 722 mathematics papers at once. They included hundreds of mathematical results across many areas, such as number theory, geometry, and theoretical computer science.

On arXiv, a well-known website where researchers share papers before they are formally published, one person can upload only two papers per month. To upload all 722 papers in one month, you would need more than 300 accounts.

Some of these very difficult problems have troubled scientists for 100 years.

For example, the Riemann Hypothesis was proposed in 1859. After 167 years, it still has not been solved. OpenAI said it had made important progress by finding a stronger result involving a "zero-free region."

Mathematicians have also spent more than 100 years trying to find out whether Catalan's constant is an irrational number, meaning it cannot be written as a fraction of two whole numbers. OpenAI said it had successfully proved that the number is irrational.

The Kakeya conjecture, which helped Wang Hong win the Fields Medal this year, was first proposed in 1917. A proof for two dimensions was not found until 1971. In 2025, Wang Hong provided a proof for three dimensions. Just one year later, AI produced a proof for four dimensions.

The newly announced results included many other major breakthroughs like these.

What makes this even more surprising is how quickly OpenAI reportedly solved mathematical problems that had troubled people for a century.

On average, the computing power used for each result was equal to about three hours of thinking by ChatGPT Pro.

That is an amazing level of power.

01

Three AI giants are racing to take over mathematicians' work

In a way, ordinary people have never felt closer to mathematicians than they do now.

We may still not understand their complicated formulas, but we understand what it feels like to work hard for many years and suddenly discover that we have fallen behind.

For the past decade or so, we have stayed up late studying for exams and working overtime. Mathematicians have stayed up late preparing for exams and working on proofs. We all hoped that our hard work would give us a secure future.

But almost overnight, even some of the world's smartest and most successful people are seeing AI take over their work.

When mathematicians saw this collection of AI papers, many were shocked and discouraged. It felt as if they had just watched an atomic bomb explode.

The article describes this as an event unlike anything seen before in the history of mathematics, because of the number of results, the many subjects covered, and the speed of the breakthroughs.

Generated by AI

Some people say AI can achieve these results only because it builds on the work of great mathematicians from the past. But human mathematicians also build on earlier discoveries.

It is also important to remember that other experts have not yet formally reviewed these AI research results. Their accuracy still needs to be checked. Even so, the huge number of results and the speed at which AI produced them are putting enormous pressure on people working in mathematics.

Of course, OpenAI is not stopping because mathematicians feel worried. It keeps announcing results that shake the mathematics world.

This competition to make AI better at mathematics has already been going on for some time.

Anthropic was the first company to make a major move.

On July 19, the night of the World Cup final, an Anthropic researcher posted on X that he had apparently disproved the Jacobian conjecture.

The Jacobian conjecture is a very important mathematics problem that has troubled researchers for more than 80 years.

The problem can be stated in one sentence: If the Jacobian determinant of a polynomial mapping is a non-zero constant, then that mapping must have a polynomial inverse function. In simpler words, it asks whether a certain kind of mathematical rule can always be reversed using another rule of the same kind.

The question sounds simple, but finding the answer is extremely difficult. Many mathematicians have spent years trying to solve it without success.

Over the decades, several mathematicians have announced that they had solved the problem. But other researchers later found mistakes in their proofs.

Many mathematicians, including Chinese mathematician Zhang Yitang, have struggled so much with problems like this that the article compares their experience to developing PTSD, a condition caused by severe stress or trauma. After decades of frustrating researchers, the conjecture was reportedly disproved in a single post on X.

OpenAI, Anthropic's competitor, did not want to fall behind. It joined this mathematical "arms race" too.

Over the following months, the two companies claimed to have solved many mathematics problems, including the ones shown below.

These may have been some of the most productive months in the history of mathematics.

But making such fast progress is not always a good thing. Many mathematicians have expressed serious concerns.

First, AI is moving ahead so quickly that mathematicians may have to give up research projects they have spent years working on, even before they finish them.

AI is not only making some mathematicians worry about losing their jobs. It is also making them question their own abilities.

For example, OpenAI recently produced a group of results in operator algebra, a branch of mathematics that studies certain kinds of mathematical operations. After reading them, one mathematician working in that field reached a troubling conclusion:

AI was right, and human researchers had spent 30 years going in the wrong direction.

Other mathematicians are facing more immediate problems involving their work and income.

Some have discovered that AI has already made major breakthroughs in the subjects they were researching. Now they do not know how to use the money they received for their projects. Others have not finished writing their papers and are unsure whether they will still be able to earn their degrees.

On Reddit, a person who said he was a PhD student studying operator algebra described his situation. He and his classmates had solved a special version of a problem. They had talked with their supervisor about turning a more general version into a doctoral thesis. But after OpenAI announced its results, they hurried to organize the work they had already completed, hoping to publish it as separate research. He still believes the subject is worth studying, but he is much less excited about it than before.

Big advances in technology can be exciting for society, but they can also have painful effects on individual people.

After OpenAI made these huge announcements, its CEO, Sam Altman, seemed ready to stop focusing on the shock and start thinking about what might come next.

Last year, AI could not even correctly compare 9.8 and 9.11 to tell which was bigger. This year, it can attempt problems as difficult as the Riemann Hypothesis.

It is hard to imagine what AI might be able to do next year.

When Wang Hong and Deng Yu won their awards, some people said these might be the last Fields Medals awarded for work done entirely by humans.

At the time, we did not believe that prediction. Now, it seems much more believable.

02

Will the first people replaced by AI be the very best students?

So why are AI companies going after mathematicians' work? Did mathematicians do something to upset them?

The answer is no. There is no conflict between the two groups. In fact, the biggest AI companies are actively hiring talented mathematicians.

That is because mathematics is one of the best ways for AI companies to test how powerful their systems are.

Mathematicians simply happen to work in a field that AI companies are using to compete for leadership around the world.

The article suggests that people who do not find ways to work with AI risk being pushed aside by it.

Jacob Zimmerman, who won the Fields Medal alongside Wang Hong and Deng Yu this year, has joined OpenAI to work on AI safety.

Alpöge, who recently challenged the Jacobian conjecture, is one of his students and now works at Anthropic.

Zimmerman once explained his reason for studying mathematics: He mainly does it because he enjoys it and is good at it.

But AI companies are not racing to solve mathematics problems simply because they love mathematics.

They are doing it because mathematics follows clear, exact rules of logic. That makes it a good way to show what AI can do. Solving famous mathematics problems also helps companies make big announcements that attract attention.

In this AI competition, mathematics has become an important battleground. The biggest AI companies seem very interested in mathematics, but their real goal is to become technology leaders. That leadership can help them attract more users, investment money, and business opportunities.

For mathematicians, research can be their life's work. For AI companies, the same research can be a way to advertise how powerful their technology is.

Competition between AI companies is now extremely intense. It is like running a 100-meter race at full speed without knowing where the finish line is.

If a company stops to rest, it could quickly lose its place in the race. Simply making progress is not enough. If it improves too slowly, competitors may move so far ahead that it is effectively falling behind.

Anthropic is moving quickly toward an IPO, when a company first offers its shares to the public.

From GPT-3 to GPT-6, the time between major model versions has become shorter: first three years, then two years, and then one year. Smaller updates now arrive almost every week.

AI companies must keep finding new areas to work in and showing impressive results if they want to remain among the leaders.

According to the article, Google's Gemini fell behind after one or two important model releases. Internet users began making fun of it by calling it the "American soybean bag." The article says Gemini has still not managed to return to the top group of AI models.

This helps explain why Sam Altman keeps announcing major breakthroughs that surprise people every few days. Anthropic CEO Dario also regularly makes announcements that create excitement and curiosity.

AI companies must keep finding new challenges as difficult as climbing Mount Everest. They need to prove that each new generation of AI is smarter than the one before it.

Mathematics is full of extremely difficult problems that AI can try to solve.

Most users cannot easily notice when an AI model's reasoning ability improves by a few percent or when it uses a few percent fewer tokens, the small pieces of text that AI systems process. But everyone can understand why solving a problem that has troubled mathematicians for decades is a big achievement.

If you do not believe it, visit some science forums for non-experts. Ninety percent