Science

Terence Tao, the mathematician asking what a proof nobody understands is worth

Penelope H. Fritz
Add us on Google
Terence Tao
Terence Tao
Photo: Institute for Pure & Applied Mathematics / CC BY 4.0, via Wikimedia Commons
BornJuly 17, 1975
Adelaide
OccupationMathematician
AwardsRoyal Medal · Salem Prize · Bôcher Memorial Prize

When Levent Alpöge released an enormous proof of the Generalized Riemann Hypothesis written entirely in Lean code and generated with heavy help from an AI model, Terence Tao did what he has done for most of his working life: he opened his blog and tried to explain to other mathematicians what had just happened. He did not call it a breakthrough. He did not call it a mistake. He ranked it last among that week’s results by how well anyone had digested it, and added, with some relief, that at least this time there was a human author to ask.

That restraint is the most revealing thing about the UCLA mathematician right now. Few people did more to make machine-checked proof respectable than Tao, who spent recent years persuading volunteers to translate theorems into Lean, the programming language in which a computer verifies every logical step. The machines are now producing verified arguments faster than anyone can read them, and the man who helped invite them in has become the field’s most persistent voice insisting that a proof nobody understands is not yet finished mathematics.

Tao was born on 17 July 1975 in Adelaide, the eldest of three sons of Billy Tao, a paediatrician from Shanghai, and Grace Leong, a Hong Kong-born graduate in mathematics and physics who taught in secondary schools. His parents had met at the University of Hong Kong and moved to Australia in 1972. The childhood record has been retold so often that it now reads like folklore: university classes at Flinders University by the age of nine, a bronze medal at the International Mathematical Olympiad at ten, and gold at 13 in 1988, which still makes him the youngest gold medallist in the competition’s history. He studied full time at Flinders from 14, finished a master’s degree at 17 and left for Princeton on a Fulbright scholarship, completing a doctorate in harmonic analysis under Elias Stein in 1996, at 20.

He has spent much of the time since trying to dismantle the legend those numbers built. In a 2007 essay titled “Does one have to be a genius to do maths?”, he answered no, arguing that sustained hard work matters far more than any imagined genius gene. Readers who know only the child prodigy often miss how much of his adult work is slow and collaborative, much of it done in public, in blog comment threads and shared files with dozens of co-authors.

UCLA hired him in 1996 and made him a full professor in 1999, at 24. The results that followed crossed borders most mathematicians never cross. With Ben Green he proved in 2004 that the prime numbers contain arithmetic progressions of every finite length, the result now known as the Green–Tao theorem. With Emmanuel Candès and Justin Romberg he helped lay the foundations of compressed sensing, the mathematics of rebuilding a signal from far fewer measurements than once seemed necessary, work that travelled from pure theory into medical imaging. The Fields Medal came in 2006, followed over the years by the Breakthrough Prize in Mathematics, the Crafoord Prize, the Royal Medal and, in 2020, the Princess of Asturias Award, shared with Candès.

He kept choosing problems with famous names. In 2015 he settled the Erdős discrepancy problem; in 2019 he proved that almost all orbits in the Collatz conjecture eventually reach almost bounded values, one of the most significant partial results on that problem in decades. Then his attention turned to how proofs are checked. In 2023, after he proved the polynomial Freiman–Ruzsa conjecture with Timothy Gowers, Ben Green and Freddie Manners, he invited volunteers to formalise it in Lean, and they finished in roughly three weeks. His Equational Theories Project, launched in 2024, combined human contributors and automated tools to settle all of the more than 22 million implications among 4,694 algebraic laws by April 2025.

Here the public image and the working mathematician pull apart. He has been the most prominent mathematician to use AI tools openly, and in 2023 he predicted that AI tools of the 2026 generation would be trustworthy co-authors; in February 2026 he said that forecast had arrived almost exactly on schedule. Yet in September he was among 25 Fields medallists who signed a declaration titled “A Severe Misalignment of AI in Mathematics”, and by October he said he no longer saw a way to engage constructively with OpenAI. He is open about the entanglements: after federal research grants to UCLA were suspended in 2025, including his own and the one that sustains the Institute for Pure and Applied Mathematics, where he oversees special projects, OpenAI money helped fund an institute workshop in March 2026. The tools he championed, formal verification above all, are exactly what make an unreadable proof possible to accept on a computer’s word. His argument now is that mathematicians should stay in the loop as interpreters of results, rather than just raw generators of results, as he put it to the Los Angeles Review of Books in November 2026.

He is building alternatives as well as arguing. Tao co-founded SAIR, a nonprofit that raises money for AI-for-mathematics work, and in October he and Yu Deng presented Open Math OS, a hosted platform that gives researchers access to open-source models without owning the hardware to run them. In November he published Six Math Essentials with Quanta Books, his first book for general readers, a tour of six ideas from numbers to dynamics, and launched it in conversation with the astrophysicist Janna Levin in Brooklyn. It arrives more than three decades after Solving Mathematical Problems, the problem-solving guide he wrote as a teenager. In June 2026 he was appointed a Companion of the Order of Australia.

He lives in Los Angeles with his wife, Laura Tao, an electrical engineer at NASA’s Jet Propulsion Laboratory, and their two children.

The verdict on Alpöge’s archive will come from the volunteers working through it on the Lean community’s chat, and from the paper Alpöge says he now intends to write. Tao has already said the digestion will take time. In the meantime he is doing something less visible: posting lecture notes for his graduate course on harmonic analysis, the subject he learned at Princeton, one week at a time, for students who will have to decide what understanding a proof means.

Tags: , , Terence Tao, , ,

Add us on Google

Discussion

There are 0 comments.