Terence Tao’s Surprising Turn: Why Math’s Star Now Backs AI as Research Partner

Terence Tao once viewed computers as tools for verification at best. Now he stands among the loudest advocates for artificial intelligence in mathematical discovery. The shift happened quickly. In October 2023 he posted on social media that he would learn the Lean4 proof assistant. He added a telling phrase. AI assistance would be necessary.

Three years later Tao has formalized parts of his own work. He has launched collaborative projects that mix human insight with machine generation. And he speaks openly about a future where mathematicians work alongside AI systems that handle tedious steps while humans steer toward insight. His evolution mirrors broader changes in the field. Proof assistants grow more capable. Large language models produce code and conjectures. Formalization turns sprawling arguments into verifiable code.

Quanta Magazine charted this transformation in detail. Tao knew about computer-verified proofs for years. Early successes impressed him yet formal math felt distant from daily research. Something changed. He began experimenting. The results surprised even him.

By 2024 Tao had become the most visible champion of machine-assisted mathematics. He didn’t arrive at this position through abstract theory alone. He sat down with Lean. He wrestled with its syntax. When stuck he turned to AI chatbots for help translating his ideas into formal statements. The process felt clumsy at first. Then it clicked.

Short sessions grew longer. Tao formalized small lemmas from his earlier papers. He found the exercise clarifying. Gaps in his own reasoning appeared under the harsh light of a proof checker. “It forces a level of precision that humans naturally gloss over,” he has said in interviews. The machine doesn’t forgive ambiguity.

But Tao also saw opportunity beyond verification. He envisioned breaking big problems into tiny chunks. Teams or even automated agents could tackle each piece. A central checker would reassemble them without fear of error. This modular approach could scale. One mathematician’s insight could combine with another’s calculation and an AI’s exhaustive case analysis.

And the field took notice. Other mathematicians began similar experiments. Some expressed skepticism. Others raced to adopt new tools. Tao’s endorsement carried weight. A Fields Medalist at UCLA, he reads more mathematics than almost anyone alive. When he says current AI models save more time than they waste, as he told an OpenAI forum in March 2026, listeners pay attention.

OpenAI Academy captured his updated stance. “Current models are now ready for primetime,” Tao declared. In math and theoretical physics they deliver net gains. He uses ChatGPT to generate images, code snippets and even drafts of arguments. The technology has become part of his daily workflow.

His equational theories project offered a concrete test. Launched in fall 2024, it examined simple algebraic structures called magmas. Tao formalized basic properties. He invited collaborators worldwide to contribute new results. AI tools helped generate conjectures and check proofs. The project demonstrated what a more experimental, collaborative style might look like. Progress came faster than in traditional solitary research.

Yet Tao never lost sight of limits. AI systems today act like jumping robots, he has explained. They can clear small obstacles with surprising agility. They sometimes scale taller walls through sheer persistence or clever shortcuts. But they lack the strategic vision to chart a path across an entire mountain range. Humans still provide the map.

This analogy appears repeatedly in his recent talks. It balances enthusiasm with realism. In a February 2026 Atlantic article, Tao predicted that 2026-level AI would function as a trustworthy co-author. That moment has arrived. Models now contribute at the level of a diligent junior researcher. They grind through cases. They suggest lemmas. They catch mistakes.

Tao and collaborators have tested these capabilities on fresh problems. Working with DeepMind researchers, he applied a system called AlphaEvolve. It generates long Python programs, then evolves them through genetic algorithms. The team attacked a new question every day or two for months. Some attempts failed. Others yielded unexpected insights. The process felt alive. Iterative. Almost playful.

Quanta Magazine reported on this wave of activity in April 2026. “2025 was the year when AI really started being useful for many different tasks,” Tao said then. Mathematics will soon look and feel different. The solitary genius in an attic may give way to hybrid teams where formalization serves as shared language.

His forthcoming lecture at the International Congress of Mathematicians carries the title “Mathematics in the Age of AI.” There he will stress the need to protect distinctly human aspects of the discipline. Not every goal reduces to efficient computation. Beauty, intuition and unexpected connections still matter. Tao co-wrote a philosophical paper with Tanya Klowden that explores these themes.

In “Mathematical methods and human thought in the age of AI,” uploaded to arXiv in March 2026, the pair examine how new tools reshape practice. They argue mathematicians must remain conscious of broader purposes. Verification is powerful. It should not eclipse discovery or understanding. The paper cites recent works by other authors who reached similar conclusions. Interest in these questions has surged.

Critics worry about over-reliance. If AI generates most routine steps, will young mathematicians lose the chance to develop deep intuition? Tao acknowledges the risk. He advocates deliberate use. Treat AI as an assistant, not a replacement. Learn the traditional methods first. Then augment them.

His own history supports this view. Tao entered mathematics as a child prodigy. He won gold medals at the International Mathematical Olympiad at ages 12 and 13. Decades of hard-won expertise now inform how he deploys new technology. He spots when an AI suggestion is elegant versus merely correct. That discernment comes from human experience.

So the conversation has matured. Early hype about AI solving open problems overnight has quieted. In its place sits a more measured assessment. AI accelerates exploration. It makes collaboration across distances easier. It turns proof into software that can be audited by anyone.

Formalization brings another benefit. It creates a permanent, machine-readable record of mathematical knowledge. Future systems could search this corpus, connect distant results and propose new avenues. Tao sees this as an extension of the mathematical tradition, not its disruption.

Recent discussions on X reflect growing acceptance. Mathematicians share Lean tactics suggested by AI. They post partial formalizations and ask for help. The community feels more open. Barriers to entry for certain tasks have dropped.

But challenges remain. Current models still hallucinate. They produce plausible but false proofs. Human oversight stays essential. Tao compares the situation to early days of computer algebra systems. At first they made mistakes. Users learned their quirks. Reliability improved. The same pattern may hold for large language models in math.

Tao’s willingness to experiment publicly has encouraged others. Where he leads, funding often follows. Conferences now feature tracks on AI and formalization. The 2026 ICM will devote significant attention to these topics. His public lecture is expected to draw a large audience.

Look closer at his recent output. Papers include AI-generated components. Blog posts describe interactions with chatbots that refined his thinking. He has even used generative tools to create diagrams that clarified complex dynamics.

None of this replaces the core of mathematical thought. Insight still arrives in flashes. Connections form across domains in ways that feel mysterious. AI can surface candidates for those connections. It cannot yet explain why they matter.

Tao returns to the mountain range image often. The range is vast. Some peaks yield to brute force computation. Others require creative leaps. The best mathematicians will learn to deploy both kinds of power. They will know when to climb carefully by hand and when to let the machine scout ahead.

His message is pragmatic. Adopt the tools. Master them. But hold onto the reasons mathematics has captivated humans for centuries. Rigor. Beauty. Surprise. These qualities persist even as the methods evolve.

The field stands at an inflection point. Tao’s transformation from cautious observer to active participant signals that the technology has crossed a threshold. What comes next depends on how the community chooses to integrate these new capabilities. If history offers any guide, the integration will be messy, fruitful and full of unexpected turns.

One thing seems clear. The mathematician who once needed no collaborators now actively builds systems that invite many more voices. Including silicon ones. The results so far suggest the partnership holds promise. How far it carries the discipline remains an open question. Tao, for one, intends to keep exploring it.