Terence Tao – Kepler, Newton, and the true nature of mathematical discovery
Source
Dwarkesh Patel
Author
Dwarkesh Patel
Date
Terms in this piece · Glossary
eval — A repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.
Why it matters
A sharp counter-argument to the 'AI accelerates science via fast verification loops' narrative, using the Kepler/Newton history to show why frontier discovery resists RL-style optimization — useful for anyone building or reasoning about AI research agents and evalA repeatable test for AI quality — a set of tasks plus scoring — used the way software teams use test suites, because model output is too variable to judge by eyeballing.Full definition →.
Key quotes
“I think AI has driven the cost of idea generation down to almost zero, in a very similar way to how the internet drove the cost of communication down to almost zero.”
Terence Tao
“we’re going through a cognitive version of the Copernican revolution, where we used to think that human intelligence is the center of the universe”
Terence Tao
“whenever we do a systematic study, on any given problem an AI tool has a success rate of maybe 1% or 2%. It’s just that they can buy scale, and you just pick the winners.”
Terence Tao
“The progress is simultaneously amazing and disappointing. It is a very strange feeling to see these tools in action.”
Terence Tao
“2026-level AI would be stunning in 2021. A lot of it—face recognition, natural speech, doing college-level math problems—we just take for granted now.”
Terence Tao
Chapters
00:00:00Kepler was a high temperature LLM
00:11:44How would we know if there's a new unifying concept within heaps of AI slop?
00:26:10The deductive overhang
00:30:31Selection bias in reported AI discoveries
00:46:43AI makes papers richer and broader, but not deeper
00:53:00If AI solves a problem, can humans get understanding out of it?
00:59:20We need a semi-formal language for the way that scientists actually talk to each other
01:09:48How Terry uses his time
01:17:05Human-AI hybrids will dominate math for a lot longer
