Question and Answer: Why developing mathematical skills still matters in the AI era
MIT mathematician John Urschel used AI tools in two arXiv preprints on random matrices and Gaussian elimination, showing how machine collaboration can accelerate mathematical research. Yet he argues that students should not outsource learning. The real value lies in understanding proofs, engaging in productive struggle, and building problem-solving skills through community. AI can broaden access to knowledge, but shortcuts may create a false feeling of understanding, so independent reasoning remains essential in mathematics education.
Key Takeaways
- MIT professor John Urschel used AI tools in two arXiv math proofs, but argues that mathematics education still matters because AI should support, not replace, human understanding and reasoning.
- He emphasizes productive struggle, independent problem-solving, and deep comprehension, urging students to internalize AI-generated proofs rather than treat them as shortcuts.
- AI can broaden access and speed up mathematical work, but it may also create a false sense of learning, so educators need to design challenging and meaningful learning experiences.
- Mathematics and quantitative reasoning remain important because they train independent thinking, a skill likely to be more valuable when machines can provide facts and methods.
- Community and in-person learning still matter, since students benefit from shared struggle, collaboration, and human connection in developing mathematical skill.
MIT Assistant Professor John Urschel Uses AI Tools in Two Math Proofs
Meta description: MIT assistant professor John Urschel used AI in two arXiv math proofs and urges students to prioritize understanding, productive struggle, and community.
Tags: AI in mathematics education, MIT, arXiv preprints, quantitative reasoning, higher education
Featured image: abstract friendly illustration of a human hand and a machine hand placing puzzle pieces, with no real people, logos, or text.
John Urschel, who earned his MIT doctorate in 2021 and is the Class of 1956 Career Development Assistant Professor in the MIT Department of Mathematics, recently published two preprints on arXiv, research papers posted before formal peer review. One addresses the growth factor of random matrices; the other examines worst-case behavior in Gaussian elimination, a common method for solving systems of linear equations. In both cases, Urschel combined his own mathematical insight with AI tools to help solve problems that had puzzled mathematicians for decades. For readers following AI in mathematics education, the story is less about machines replacing mathematicians than about what students should still learn: how to understand, struggle, and reason.
A football background and a classroom focus
Urschel is no stranger to physical training. Before earning his MIT doctorate, he built the strength needed as an offensive lineman for the Baltimore Ravens. In the classroom, he still asks students to work hard, but his focus is on critical thinking and quantitative reasoning. He tells students that his interest in math began with puzzles, from Sudoku and crossword puzzles to Martin Gardner math puzzles. The appeal, he says, is facing a problem that is not yet clear, finding a way through it, and reaching understanding.
When he teaches, Urschel wants students to strengthen their ability to solve problems and think quantitatively. He also wants them to absorb the big ideas, tools, techniques, and themes of the specific course. The goal is to turn subject knowledge into better problem-solving outside the classroom.
A new collaborator on arXiv
Urschel told MIT News that the two arXiv papers were the first two in which he had used AI tools. He described the experience as similar to working with a collaborator, except that the collaborator seemed to have access to unlimited resources and could do certain tasks much faster than he could. Just as computers perform calculations faster than people, he said, AI systems now appear able to do certain kinds of mathematics far more efficiently than human mathematicians.
Still, Urschel said the joy of problem solving is greatest when he works through a problem alone. He wants students to keep an independent sense of problem-solving ability and curiosity, even when machines can help. His main instruction for students who use AI is understanding: when a machine produces a proof, the value comes from a human recognizing how mathematical ideas are combined, internalizing that logic, and explaining it to others. For students who do not use AI, he focuses on traditional problem-solving skills that mathematics has built for centuries.
Struggle, shortcuts, and understanding
This point connects with a report from MIT's Ad Hoc Committee on AI Use in Teaching, Learning and Research Training, which stressed productive struggle and collaboration as essential parts of learning. Urschel sees AI as a major benefit for education because accessible technology can give more people broad access to knowledge. He recalled that neither of his parents is a professor and that he did not learn what a mathematician was until college. As a student, he used a laptop and Google to find advanced material and online papers by leading experts that were available to him at no cost. Today's tools go far beyond that, but he warned that the same convenience can create a false feeling of understanding.
A tired student can use AI to finish homework or produce a 10-page essay with very little effort, he noted, and mistake that speed for learning. People may dislike the word struggle, he said, but pushing through material on one's own is what connects a learner to the subject. He also said faculty and instructors have work to do: make recitations more useful, improve office hours, remind students that they have personal responsibility for their learning, and give them a learning experience that challenges them while remaining worth their time.
Why math remains a core skill
Asked why students should still study mathematics as AI advances, Urschel said the process of learning to think is becoming more important. In a world where many facts and methods are readily available through machines, he argued, the ability to think independently may be valued more highly. Training in mathematics and quantitative sciences, he said, helps develop that ability, especially if generations grow less practiced in solving problems for themselves.
The second reason is community. Learning with peers improves the overall experience, he said, and that is one reason he still values in-person, residential education. For him, campuses remain important because they let students encounter struggle, solve problems together, and connect with other people. As long as humans still have a role in thinking and solving problems, he said, building mathematical skill remains important.
Urschel's larger point is that institutions cannot prepare students the same way as 10 years ago. The world is changing quickly, and education needs to adapt. For readers watching AI enter classrooms and research, the message is practical: use AI to widen access and speed up work, but do not let it remove the hard thinking, the shared learning, and the human connection that make education lasting.
Comments (0)