As AI systems start proving theorems, mathematicians are asking an anxious question: what is left for us? A guest essay on Terence Tao's blog, by Karthik C. S., Amit Sahai and Jie Xue, offers an unusual answer. If mathematical truths are discovered rather than invented, they argue, a result loses none of its worth when a machine finds it, and the real target of the anxiety is an older habit of measuring mathematicians by their output.
The argument rests on a Platonic view of mathematics. On this account there were infinitely many primes before anyone proved it; mathematical truth does not wait for human recognition or draw its authority from us. We meet that independence, the authors say, whenever a construction surprises us, a definition reveals an unexpected connection, or a conjecture we trusted turns out to have a counterexample. We cannot make an argument valid by wishing it so. From this it follows that the significance of a proof lies in its relation to that independent order, not in who produced it, so a discovery made by a machine is no less a genuine discovery.
Their sharper move is to separate correct derivation from understanding. An axiomatic system allows endlessly many formally correct deductions, but producing them does not by itself say which are worth pursuing. What makes one definition illuminating and another merely valid is its relation to the structures it brings into view, which is a matter of mathematical taste understood as more than private preference. On this reading the real worry predates AI: mathematicians had already come to see themselves as collections of productive abilities measured by publications, citations and awards, and AI only exposes the consequences of that self-image. The authors disclose that they used GPT models to help prepare the piece.
It is a thoughtful counter to a genuine fear, and the move from status to understanding is a healthy one for a field that has tied worth to prizes. But the essay is only as strong as its starting point. The whole argument leans on mathematical Platonism, which is a contested philosophical position rather than a settled fact, and a reader who does not share it loses the comfort the essay offers.
The essay also says little about the part of the anxiety that is not philosophical. Jobs, funding, who controls the tools and who gets credit are where most working mathematicians' worries actually sit, and a meditation on egoless truth-seeking does not touch them. Read it for the argument about what mathematics is for, not as a guide to how the profession should respond.