Berkeley Math Professor Stops Checking Homework Answers Because of AI

A math professor at the University of California, Berkeley has stopped checking whether his students' homework answers are correct. Richard Borcherds, a Fields Medal winner (a top prize in mathematics), now gives credit for simply handing the work in. He says AI tools make it hard to tell whether students are solving the problems themselves.

What changed

Borcherds noticed that students kept getting high scores. He saw little reason to keep paying teaching assistants to check their answers. So marks now depend on whether homework is submitted, not whether it is right.

He admits the method is not ideal but says he has not found a better one. In his view, AI is now so widely available that stopping students from using it on homework is increasingly difficult.

Why he is worried

His comments followed OpenAI's announcement that its AI systems had produced a claimed solution to the Navier-Stokes existence and smoothness problem, one of the hardest challenges in mathematics. The article describes it as a claim, and it raises questions about how AI could change math teaching and research.

Borcherds warned that if AI keeps improving at its current pace, machines could outperform human mathematicians across their work within a few years. Five years ago, he says, math looked like one of the jobs least likely to be disrupted. He even asked ChatGPT what advice to give students thinking about math careers. The chatbot could not give a clear answer.

What Berkeley is doing

UC Berkeley is reviewing how AI should be handled in classrooms. Anne Marie Richard, the university's assistant vice provost for undergraduate education and chief academic technology officer, says faculty are asked to explain their AI expectations in course outlines. Each professor can decide what is allowed, depending on the subject and course level.

The university tested AI detection tools, including Turnitin and Rumi, but found them not reliable enough to use as a standard system across campus. It also runs an online AI Hub with sample policies, training materials and guidance for teachers.

What happens next

Borcherds expects universities may take several years to respond. He says schools often keep old practices long after technology makes them less useful, and points to printed academic journals, which he calls unnecessary for around two decades. He compares AI to hand weaving giving way to machines and cars replacing horses, but says math has never faced a disruption like this.

This is one professor at one university, so it does not show what every school does. Still, it points to a wider problem: take-home work is getting harder to trust. What will replace it is not yet known.