Subvolts take Learn how slow, human judgment drove scientific discovery, challenging AI's fast-loop assumptions. Mar 20, 2026 - Developers
- Understand the historical context of scientific discovery and its potential parallels with AI research challenges.
- Business
- Consider the long-term nature of validation and the role of intuition in groundbreaking work.
This is a solid interview with a renowned mathematician discussing fundamental aspects of scientific discovery.
Terence Tao – Kepler, Newton, and the true nature of mathematical discovery
Audio streams from Dwarkesh Patel. Your place is remembered on this device.
From Subvolts
Summary and analysis by Subvolts editorial, AI-assisted.
Dwarkesh Patel interviews mathematician Terence Tao on the nature of mathematical discovery. They discuss Kepler's ingenious, yet slow, discovery of planetary motion laws. The conversation highlights how scientific progress can involve long verification periods, where better theories might initially make worse predictions, relying on poorly understood judgment and heuristics.
This episode explores how scientists like Kepler figured out how planets move. It shows that sometimes, the best ideas take a very long time to prove, and we don't fully understand why certain ideas eventually win out.
This episode offers a historical perspective on scientific discovery, contrasting it with modern AI's potential for rapid progress. It's for anyone interested in the philosophy of science, the history of mathematics, and the challenges of AI-driven discovery. The long verification loops discussed are relevant to current AI research, particularly in areas like automated theorem proving and scientific hypothesis generation, where robust evaluation remains a critical bottleneck.
- HostDwarkesh Patel
- GuestTerence Tao
- Guest: Terence Tao
- Title: Kepler, Newton, and the true nature of mathematical discovery
- Show: Dwarkesh Podcast
- Length: 84 minutes
- Key Example: Kepler's discovery of planetary motion laws
- Core Theme: Long verification loops in scientific discovery
We begin the episode with the absolutely ingenious and surprising way in which Kepler discovered the laws of planetary motion.
People sometimes say that AI will make especially fast progress at scientific discovery because of tight verification loops.
But the story of how we discovered the shape of our solar system shows how the verification loop for correct ideas can be decades (or even millennia) long.
During this time, what we know today as the better theory can actually make worse predictions.
And the reasons it survives this epistemic hell is some mixture of judgment and heuristics that we don’t even understand well enough to actually articulate, much less codify into an RL loop. Hope you enjoy!
Watch on YouTube; read the transcript.
Sponsors
- Jane Street loves challenging my audience with different creative puzzles. One of my listeners, Shawn, solved Jane Street’s ResNet challenge and posted a great walk-through on X. If you want to try one of these puzzles yourself, there’s one live now at janestreet.com/dwarkesh.
- Labelbox can get you rubric-based evals, no matter your domain. These rubrics allow you to give your model feedback on all the dimensions you care about, so you can train how it thinks, not just what it thinks. Whatever you’re focused on—math, physics, finance, psychology or something else—Labelbox can help. Learn more at labelbox.com/dwarkesh.
- Mercury just released a new feature called Insights. Insights summarizes your money in and out, showing you your biggest transactions and calling out anything worth paying attention to. It’s a super low-friction way to stay on top of your business. Learn more at mercury.com/insights.
Timestamps
(00:00:00) – Kepler was a high temperature LLM
(00:11:44) – How would we know if there’s a new unifying concept within heaps of AI slop?
(00:26:10) – The deductive overhang
(00:30:31) – Selection bias in reported AI discoveries
(00:46:43) – AI makes papers richer and broader, but not deeper
(00:53:00) – If AI solves a problem, can humans get understanding out of it?
(00:59:20) – We need a semi-formal language for the way that scientists actually talk to each other
(01:09:48) – How Terry uses his time
(01:17:05) – Human-AI hybrids will dominate math for a lot longer
This is a public episode. If you would like to discuss this with other subscribers or get access to bonus episodes, visit www.dwarkesh.com
- epistemic
- Dwarkesh Patel
- relating to knowledge or to the degree of its validation
- heuristics
- Dwarkesh Patel
- practical methods or shortcuts for problem solving
- RL loop
- Dwarkesh Patel
- a type of machine learning where an agent learns by trial and error
1 h 371 h 201 h 37AI researchers debate how close we are to recursive self-improvement
WithJohn Schulman, Beren Millidge, Charlie O’Neill
2 h 21Questions people ask
- What is the main topic of the podcast episode?
- The episode discusses the nature of mathematical discovery, using Kepler's work on planetary motion as a key example.
- How does the episode relate to AI progress?
- It contrasts the long verification loops in historical scientific discovery with the potential for faster progress in AI.
Page generated Oct 19, 2026. Summaries are Subvolts' own; the audio, artwork and notes belong to Dwarkesh Patel.