AI in Mathematics: Genius or Tool? Terence Tao Stops the Nighttime Idolatry

When AI solves an open mathematical problem that has remained unsolved for 50 years, headlines explode. But there’s a problem: we are idolizing what is really just an improved search engine.

Recently, Terence Tao — one of the leading proponents of AI research in mathematics — decided to pause the collective euphoria. He didn’t do it at a conference but published late at night on GitHub, almost as if he wanted to put out the fire before it engulfs the mathematical community. His message is direct: stop taking things out of context.

The fairy tale no one questioned

The narrative is seductive: “AI conquers problems that mathematicians have been unable to solve for decades!” For those dreaming of AGI, it’s music. For those defending human capability, it sounds like an existential threat.

The curious thing is that Terence Tao doesn’t say that AI is useless. He says something much more uncomfortable: that we are comparing apples to oranges.

When you look at the Erdős “solved” problems by AI, you need to ask yourself: Were they really unsolved? Or simply poorly documented? Many problems labeled as “Open” on the web are just “low-hanging fruit” with little review. Some already had solutions in the literature that no one had updated. Others have original statements that are ambiguous, which the AI “solved” through literal interpretation.

The true work of AI: engineering, not intuition

Here’s where it gets interesting: Terence Tao isn’t denying the achievements. He’s redefining what “achievement” means.

AI is excellent at:

• Systematic search: scanning vast problem spaces
• Formal verification: turning proofs into checkable Lean code
• Rewriting: polishing existing arguments
• Routine tasks: following known patterns

What AI (still) does NOT do is what defines real mathematics: asking the right questions, creating new concepts, connecting a result with the entire network of existing knowledge.

A technically correct proof generated by AI often lacks context, motivation, comparison with existing literature. It’s like a map without a compass: it works if you know where to go, but doesn’t help you discover new territories.

The mathematician of the future: strategist, not solitary thinker

This is where Tao finds balance. It’s not “AI vs Humanity.” It’s “Humanity + AI.”

The future of the mathematician isn’t less exciting, just different. We will shift from being solitary thinkers to commanding silent armies of machines. We provide the vision; AI opens the way. We identify which questions matter; AI helps build the answers.

What Terence Tao is trying to stop with his late-night post is simple: the idolization of isolated cases. Not because AI is weak, but because overestimating it distracts us from real work.

The call: stop idolizing, start collaborating

Mathematics has always been collaborative — among humans, across generations, between disciplines. Making AI part of that conversation is logical. But a revolutionary tool remains just that: a tool.

What matters now is honesty: What problems has AI truly solved? How difficult were they? What value do they add to the discipline?

Because the true revolution in mathematics won’t be when AI solves more problems than humans. It will be when humans and machines discover truths that neither could have found alone.