For most of the last century, the practice of mathematics changed remarkably little. A mathematician notices a pattern, writes a conjecture, spends months or years constructing a proof, and other mathematicians verify it. Now, according to an in-depth report by Benjamin Skuse in IEEE Spectrum, artificial intelligence is starting to bypass that slow, deliberative process and is forcing the field to confront what it actually means to be a mathematician.
The piece, published on June 25, 2026, traces how quickly large language models have evolved from what critics once dismissed as stochastic parrots into genuine mathematical reasoning engines and what that shift means for the humans who have always done this work. For more context on this story, see our ongoing breaking AI news.
AI's rapidly growing footprint in mathematics
Computation has accelerated mathematics for decades. The turning point came roughly 50 years ago, when a computer was used to prove the four-color theorem, which asks whether any map can be colored with no more than four colors so that no adjacent regions share a color. The answer was yes, but the proof was controversial because it relied on checking 1,936 cases that no human could realistically verify by hand.
Even through that computational era, the human mathematician remained central, proposing conjectures, devising strategies, and verifying results. AI is now challenging that arrangement directly. Last summer, systems from Google DeepMind and OpenAI reached a level equivalent to the world's most mathematically gifted high school students, achieving gold-medal status at the International Mathematical Olympiad (IMO), the annual competition where contestants must solve six notoriously difficult problems.
The milestones have kept accelerating. Earlier in 2026, Google DeepMind's experimental system Aletheia autonomously produced publishable PhD-level research results. Although the work itself is mathematically obscure, calculating structure constants in arithmetic geometry, the significance lies in the complex reasoning required to tackle a previously unsolved problem. More recently, a new general-purpose AI system from OpenAI disproved an important conjecture in combinatorial geometry, a result that top mathematicians hailed as a milestone for the field and that would have been worthy of publication in a major journal.
Proof assistants and the formalization bottleneck
A second shift has come from pairing large language models with proof assistants. Systems such as Isabelle, Lean, and Rocq are specialized programming languages that check mathematical proofs step by step, verifying their logical correctness. For more than a decade, the bottleneck has been formalization, the laborious process of translating theorems and proofs by hand into machine-readable code.
LLMs are beginning to remove that bottleneck by automating the translation of informal proofs into formal code that proof assistants can verify. In February, the AI company Math, Inc. used its reasoning agent, named Gauss, to formalize the work that earned EPFL's Maryna Viazovska a Fields Medal in 2022. Gauss first helped human mathematicians complete the formalization of Viazovska's solution to the 8-dimensional sphere-packing problem in a matter of days, then autonomously formalized the more complicated 24-dimensional case in just two weeks.
Such achievements suggest that AI is already handling tasks long considered uniquely human, and that much of the day-to-day work of a mathematician may soon be fair game for automation.
Mathematicians debate whether they become 'priests to oracles'
Reporting from the 12th Heidelberg Laureate Forum in September 2025, Skuse describes palpable unease in the audience as speakers described a future in which superhuman AI systems form conjectures, search solution spaces, prove theorems, verify proofs, and generalize results without human involvement.
Yang-Hui He of the London Institute for Mathematical Sciences memorably warned that human mathematicians could become "priests to oracles." Jessica Randall, a mathematician for Google Developer Groups, said she sensed collective existential dread rising among young researchers. "We certainly started realizing AI has the potential to replace us," she told IEEE Spectrum.
Not everyone is alarmed. Some established mathematicians simply want answers to the biggest open questions, such as the six remaining Millennium Prize Problems, regardless of who or what produces them. "A lot of mathematicians are pragmatic and just want to understand," Carnegie Mellon University's Jeremy Avigad observed. "They would sell their soul for the solution to a problem."
Three competing visions for the future
The IEEE Spectrum report identifies three broad visions for how mathematics and AI might coexist.
The first is a human-centric view that prioritizes human understanding and treats AI as a tool, much like a calculator. Fields Medalist and Princeton mathematician Akshay Venkatesh argues that mathematics is, at heart, a way of reaching agreement. "Sometimes I think when we use numbers, it's not so much that we are describing phenomena that are intrinsically numerical, but that we can all agree exactly what the numbers mean," he said.
The second emphasizes the irreducibly human experience of struggle. Maia Fraser of the University of Ottawa argues that an AI proof of a stubborn conjecture is useful only if it is comprehensible to people. "That the statement can be proved by AI is already useful information," she conceded. "But then it's still an open problem to come up with an elegant, beautiful human proof."
The third, and perhaps the most ambitious, comes from Terence Tao of UCLA, who first competed in the math Olympiad at age 10 and is now one of the most decorated mathematicians alive. Tao sees AI as the catalyst for what he calls "big mathematics," a future of large-scale, decentralized collaborations between humans and machines in which complex problems are divided into pieces, with humans claiming the creative work and AI handling the technical grunt work.
A discipline asking why it exists
What makes the moment unusual is that the debate is no longer hypothetical. AI systems are already earning IMO gold medals, producing publishable research, disproving conjectures, and formalizing Fields Medal-winning proofs in days rather than years.
For students like Trill White of Australia's Deakin University, who attended the Heidelberg forum, the realization landed hard. "That's devastating," White recalled thinking. "What will people have to contribute to mathematics? Will it become something that no one understands?"
The answer is not yet settled. But as Skuse's reporting makes clear, mathematics, one of humanity's oldest intellectual disciplines, is now confronting the same question that is reshaping nearly every field AI touches: when a machine can do the work, what is the work actually for?
Source: Benjamin Skuse, "What It Means to Be a Mathematician When AI Does the Math," IEEE Spectrum, June 25, 2026.---
Stay Ahead of AIGet the latest AI news, analysis, and breakthroughs — all in one place.
Read more AI news →
