Amit Sahai, a mathematician who has spent years working with AI systems, has a simple warning for his colleagues: the machines are about to outpace you, and you should not only accept that but train more people to keep up. His piece, titled “We’re gonna need a lot more mathematicians,” argues that as AI produces more and more new ideas, the human research community needs to expand, not shrink.
Sahai recalls his own undergraduate days, when he watched classmates struggle to keep up with the fastest math students. Those students eventually gave up on research mathematics. Now, Sahai says, AI is doing to the whole field what those top students did to their peers.
The Machines Are Already Outthinking Us
Sahai’s argument starts with his direct experience. The AI systems he has worked with are already producing new ideas, not just doing impressive calculations or confirming arguments that human researchers already understand. He believes we cannot even imagine what future systems will produce.
The result, he says, is that mathematicians will soon feel what his undergraduate friends felt: unable to keep up. The temptation, he warns, will be to leave the field.
Why Giving Up Would Be Wrong
Sahai is blunt about the stakes. He calls it “a profound abdication of our responsibility to humanity” if mathematicians give up understanding the machines’ output. The responsibility is collective, not individual: humans must be able to understand and contribute to the discoveries that will shape our world.
Struggle is essential to understanding, he says, and that struggle can be shared. He envisions research groups spending a term or a year trying to understand an extraordinary set of ideas produced by an AI system, with the help of AI systems themselves.
The Terawatt Nuclear Plant Test Case
To make the stakes concrete, Sahai uses a hypothetical example. Imagine an AI system proposes a design for a one terawatt nuclear fusion power plant, a machine that could produce an insane amount of electrical power. The design is novel and untested.
Before approving construction, Sahai says, he would want communities of humans to understand why the design works and what justifies confidence in its safety. A theorem exists only within a model, he argues, and understanding that model — its experimental evidence and uncertainties — is demanding work that mathematically sophisticated people must be available to do.
Why the Humans Matter
Sahai acknowledges that AI systems should handle routine, reliable tasks. He sees no reason to insist that humans manually repeat work an AI system might perform more reliably, even including proving mathematical guarantees. But there is a limit: a theorem can only exist within a model, and understanding that model is demanding work that mathematically sophisticated people must be available to do.
One might respond that AI systems should handle those questions too, and ultimately decide whether the plant should be built. That is a serious position, he says, but it asks us to accept a future in which decisions of enormous consequence rest on reasons that no human community understands.
The Case for More Mathematicians
Sahai argues that society has a strong immediate interest in making human understanding possible. A civilization focused on depth of human understanding, he says, is one worth living in. But there is also a practical reason: decisions about novel technology rest on reasons that humans must understand.
He calls for a “deployable intellectual reserve” — communities of mathematically sophisticated people available to help understand consequential AI-enabled breakthroughs. Our ability to understand difficult and unfamiliar ideas, he argues, may become one of the most important contributions we can offer to society.
A Vision for the Future
Sahai’s vision is optimistic, even utopian. He wants a world where mathematicians and physicists are plentiful, where society invests in the capacity to understand, and where humans retain the ability to meaningfully consider alternatives.
The stakes are high. A future in which decisions of enormous consequence rest on reasons that no human community understands is a future we should reject, he says.
The Numbers Behind the Warning
| Element | Detail |
|---|---|
| Source title | “We’re gonna need a lot more mathematicians” |
| Author | Amit Sahai |
| Key figure | One terawatt nuclear fusion power plant design |
| Core claim | AI systems are already outthinking human mathematicians and will accelerate |
“Before approving construction, I would want communities of humans to understand why the design works and what justifies confidence in its safety. I would hope that we all would.”
Our Take on Sahai’s Argument
Sahai’s piece is a genuine call to optimism, not a warning. He is asking mathematicians to embrace AI as a partner rather than an adversary, and to think about how humanity can understand and trust the machines’ output.
The key judgment is that the struggle to understand is essential, and that it can be shared. Sahai’s own career, spent working with AI systems that produce new ideas, is proof that humans can keep up — and that the effort is worth it.
The machines are coming faster. The question is whether humans will choose to keep up, or to step aside.
Source material: “We're gonna need a lot more mathematicians,” terrytao.wordpress.com.
Get the Notebook.
The day's best stories and every fresh verdict, in plain English, in your inbox by seven. One email a day, no more.