Richard Borcherds, the UC Berkeley mathematics professor and Fields Medalist, has stopped grading his students' homework for correctness — a decision he attributes to the reality that students now routinely use AI tools to complete their problem sets.
In an interview with The Daily Californian published this week, Borcherds said that after seeing how many of his students used AI to consistently submit top-scoring work, he concluded there was "no point" in continuing to pay readers to grade homework for correctness, since AI use in the course has become unavoidable. For more context on this story, see our ongoing breaking AI news.
Instead, he now grades on completion: students who hand in their homework receive credit, regardless of whether the solutions were produced with machine assistance.
"It's not a very good system," Borcherds admitted, describing the arrangement as the least-bad option available at the moment.
A Fields Medalist Reassesses the Profession
Borcherds, who won the Fields Medal in 1998 for his work on vertex operator algebras and the monstrous moonshine conjecture, gave the interview shortly after OpenAI announced that its AI agents had produced an apparent solution to the Navier-Stokes existence and smoothness problem — one of the most celebrated unsolved problems in mathematics.
That claim remains contested by mathematicians, and the Clay Mathematics Institute has not awarded the associated $1 million Millennium Prize. But the announcement has already reshaped how some mathematicians talk about their field's future.
"If AI continues to improve at present rate, then, in a few years, there'll be nothing human mathematicians can do that AI can't do better," Borcherds told the student paper.
His comments place him among a growing group of prominent mathematicians publicly wrestling with the implications of frontier AI for their discipline. The Navier-Stokes announcement itself drew demands from researchers that OpenAI release verifiable proofs, and earlier this year a group of Fields medalists signed a declaration warning about AI's accelerating role in mathematics.
"I Even Asked ChatGPT"
Borcherds said he no longer knows whether to encourage students to pursue mathematics, a field that seemed — as recently as five years ago — to be among the safest careers from the AI boom.
"I have no idea what to advise. I even asked ChatGPT this exact question, and even ChatGPT said it didn't have any idea what was going to happen," he said.
He described substantial "inertia" in the academic system and predicted the math department could look much the same for a couple of years despite the pace of AI development. He compared the current moment to physical paper journals, which continue to publish despite being, in his words, "utterly pointless" for two decades.
History offers parallels, he noted — hand weaving was displaced by looms, and horses by motor cars — but he emphasized that nothing comparable has previously happened in mathematics.
"There have been plenty of technical breakthroughs in the past — hand weaving was put out of business by looms, and horses were put out of business by motor cars, and so on," Borcherds said. "This is the first time anything like this has happened in mathematics, though."
Campus Rules Struggle to Keep Up
Berkeley's institutional response illustrates how uneven AI policy remains across higher education. According to an email from Anne Marie Richard, assistant vice provost for undergraduate education and the campus's chief academic technology officer, the Berkeley Division of the Academic Senate asks faculty to include a statement about AI use expectations in course syllabi — but direct policies vary dramatically by course level and subject.
The campus has experimented with AI detection tools, including Turnitin and Rumi, but determined they are not reliable enough to standardize. AI policy therefore remains at the discretion of individual instructors, supported by a growing database of sample AI-use statements in the university's AI Hub.
Borcherds' completion-based grading is effectively a unilateral adaptation to that enforcement gap: rather than policing AI use he cannot reliably detect, he has redefined what homework credit means.
A Signal for Higher Education
The episode is a small but pointed data point in a larger debate about assessment in the AI era. If a Fields Medalist at one of the world's top mathematics departments has concluded that take-home problem sets can no longer measure what they were designed to measure, the implications reach well beyond Berkeley.
Educators across disciplines have been grappling with the same tension: detection tools lag behind generation tools, honor codes are difficult to enforce at scale, and assessment formats built for a pre-AI world increasingly fail to distinguish student understanding from model output.
For now, Borcherds' solution is pragmatic rather than optimistic — grade what can be graded, and acknowledge openly that the rest has become unverifiable. Whether mathematics education can preserve its traditional apprenticeship model in an age of increasingly capable AI mathematicians is a question even ChatGPT, by his account, declines to answer.
---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
