Terence Tao on Math 2.0: who must understand the answer?

Terence Tao's "Math 2.0" paper argues that you would want at least one human cancer researcher to understand the mechanism before injecting a cure that an AI found. The thread split fast. One commenter said no, they'd take the cure. Another welcomed "sanity back in the conversation" but doubted that humans plus AI will make fundamentally more progress, partly because heavy AI use takes a mental toll.

A companion post, "Illustration Beyond Proof," frames the researcher's workflow bluntly: you feed the problem to an AI, and it either solves it or it doesn't. The interesting work becomes explaining and visualizing results. Readers floated using generative interface tools to make 3Blue1Brown-style explanations.

This is the same debate as the AI contribution bans, in a higher-stakes setting. The question is where human comprehension sits in the loop once machines can produce correct answers faster than people can check them.