A beginning for mathematics

Mathematician Daniel Litt argues that as AI makes mathematical text easier to produce, a finished result and its author's expertise are becoming less reliable indicators of each other, so degrees may need stronger evidence that students understand the work they submit.

He proposes awarding a PhD primarily through a rigorous defense in which the student explains a deep topic to examiners' satisfaction, with regular opportunities to apply techniques to unfamiliar examples. Under that proposal, whether AI assisted would not determine eligibility; demonstrated understanding would. His institutional argument concerns how universities assess expertise, without requiring certainty about when AI might surpass mathematicians.