Deng Yu, the 37-year-old Chinese mathematician who won the 2026 Fields Medal, has a backup plan if artificial intelligence puts him out of a job.
On Tuesday, he wrote on social media that “if AI can solve all maths problems, I’ll go home and write my yuri novel”.
Deng is known to be a big fan of yuri, a genre of Japanese novels and comics that depicts intimate friendships and romantic relationships between women.
Do you have questions about the biggest topics and trends from around the world? Get the answers with SCMP Knowledge, our new platform of curated content with explainers, FAQs, analyses and infographics brought to you by our award-winning team.
In the post, which was later shared on another Chinese social platform, Deng wrote that AI has created “an atmosphere of restlessness and impatience” in mathematical circles.
“Once this wavefront passes in the next year or two, the situation will gradually become clear. Mathematics will naturally adapt slowly, and a new ecosystem is gradually taking shape.”
Deng told the South China Morning Post on Wednesday that he was “not currently writing a yuri novel. That statement was entirely in jest.”
“I don’t believe AI can develop to the point of solving all mathematical problems, at least not those I mentioned in my post,” he added.
Nevertheless, uncertianty is brewing in some parts of the maths world over AI’s impact on the field.
On September 8, OpenAI said it had used one of its models to solve the Navier-Stokes Millennium Prize Problem – one of the major questions at the frontier of mathematics.
Seven mathematical challenges were named Millennium Prize Problems by the US-based Clay Mathematics Institute on May 24, 2000.
Each problem comes with a US$1 million prize for its solution. In 2003, Russian mathematician Grigori Perelman solved one of the problems – the Poincare conjecture – but declined the prize and the Fields Medal he was subsequently awarded.
And now, a second Millennium Prize Problem may have been solved – but not by a human.
The Navier-Stokes equations are widely applied in fluid mechanics, yet our understanding of them remains limited. A successful proof would greatly advance humanity’s overall comprehension of mathematical physics.
According to OpenAI, the proof was generated using OpenAI’s unreleased next-generation model, which worked for 88 hours and consumed 130 billion output tokens worth millions of dollars.
However, a mathematician also working on the problem claimed that OpenAI began its efforts only after hearing of his progress towards a solution.
According to a report by Scientific American on September 8, the night before OpenAI announced its proof, New York University mathematics professor Tristan Buckmaster posted on social media that he and Anthropic researcher Levent Alpoge had proven the Euler equations.
The Euler equations are widely regarded in the field as a crucial step towards tackling the Navier-Stokes problem.
Buckmaster said that he and his collaborator had used several AI models to solve the equations, including those from OpenAI.
In a statement on Monday, Buckmaster claimed that OpenAI learned of the mathematicians’ progress the week before and adopted the approach they were using to solve the entire problem.
OpenAI later clarified on social media that it had independently proven the Euler equations, but did not completely deny it had drawn on ideas from other mathematicians.
“We did not see any of their work through any means until they released it publicly – in particular, no specific user data was accessed to solve this problem,” the company said.
“De-identified data from their use of our products may have helped improve our models.”
The American Mathematical Society said on Tuesday that while Alpoge and Buckmaster had made breakthroughs “assisted by new technologies”, the mathematicians at OpenAI had completed the final proof.
“The purpose of mathematics is human understanding, and this achievement, and the process that led to it, will bear fruit for a long time to come,” the society said.
This is not the first time that AI has been used to overtake human mathematicians.
For instance, OpenAI researcher and University of Pennsylvania Wharton School statistics and data science professor Su Weijie said on September 4 that he had used AI to make progress in one of the field’s biggest problems – the twin prime conjecture.
In 2013, Zhang Yitang proved that there are infinite pairs of consecutive prime numbers whose gap is bounded by a finite number. He proved that this gap was no more than 70 million.
This was one of this century’s most important mathematical breakthroughs.
The following year, the “Polymath8b” project, led by Terence Tao, proved that the gap was just 246.
Over the next 12 years, this value remained unchanged.
On August 31, Austrian mathematician Julia Stadlmann uploaded a 34-page preprint to arXiv. She proved that there exist infinitely many pairs of consecutive primes whose difference does not exceed 240. The paper does not mention the use of any AI tools.
According to an interview with China Science Daily on September 7, Su began working on the twin prime conjecture on August 15 and quickly advanced.
“I used an internal version of [OpenAI’s] GPT-6 Astra to first improve from 246 to 240, then quickly advanced it to 188, and after a period of exploration, finally reached 186,” Su said.
“In this proof process, my role was auxiliary. I only gave the AI simple prompts. The AI’s mathematical proof capabilities are far from reaching their limit,” Su told China Science Daily.
Tao, who won the Fields Medal in 2006, said in a September 6 social media post that “if AI had emerged 20 years earlier, Zhang would still be working as an adjunct lecturer, and Maynard would have switched out of analytic number theory to a different topic of study”.
James Maynard was the recipient of the 2022 Fields Medal.
More Articles from SCMP
Asian Games: China basketball stars packed in 3 to a container in Japan’s athletes’ village
China’s young investors embrace risk. Why are their portfolios so safe?
How Cantopop lyricist Wyman Wong pushed Hong Kong’s LGBTQ, style and emotional boundaries
Why AI experts are ‘freaked out’ over safety fears ahead of Xi-Trump summit
This article originally appeared on the South China Morning Post (www.scmp.com), the leading news media reporting on China and Asia.
