ATRIUMsearch → argument graph
Audio · 2026-04-07 · 6 moments

Michael Nielsen – How science actually progresses

The true story of Einstein, Newton, and Darwin ✦ AI generated

Apple Podcasts ↗
GUEST Michael NielsenHOST Dwarkesh Patel

timeline · colored by role

01
Mechanism

Newton's theory became compelling not because it was simpler or more accurate than rivals, but because a single set of ideas explained three seemingly unrelated phenomena at once: planetary motion, terrestrial projectile motion, and the tides.

Nielsen suggests the real driver behind accepting the new cosmology was explanatory unification — one theory covering orbits, falling objects, and tides — rather than superior accuracy or simplicity, since Copernicus's model was actually neither.

transcript

Michael Nielsen: You have what seem like three very different disconnected phenomena all being explained by this one set of ideas. That starts to feel very compelling, at least to me. I think most people find that very satisfying once they eventually realize it.

02
Claim◆

The Michelson-Morley result did not falsify 'the ether' as a single entity — it only ruled out some of many competing ether theories, and how scientists responded to that ambiguous result was far messier than a clean falsification story.

Nielsen argues the textbook story that Michelson-Morley 'disproved the ether' and thereby triggered relativity is a misleading oversimplification — there were many rival ether theories, and Einstein himself said the experiment wasn't decisive for his thinking.

transcript

Michael Nielsen: Even just the word 'the' there is a misnomer. You actually had a ton of different theories and a couple of leading contenders. So yes, there's some version of falsification going on, but how you respond to this new experiment is very complicated.

03
Mechanism

Scientific bottlenecks form precisely where prior methods and heuristics stop applying, since researchers already exploit whatever has worked before — so genuine breakthroughs require many people pursuing different unproven approaches rather than a single reliable process.

Nielsen argues there's no universal 'crank to turn' for discovery — progress gets stuck exactly where old heuristics fail, and escaping that requires many parallel, unguaranteed attempts, with bigger bottlenecks yielding bigger eventual triumphs.

transcript

Michael Nielsen: If you're attempting to reduce science to a process, you're attempting to reduce it to something where there is just a method which you can apply, and you turn the crank and out pops insight. You can do a certain amount of that, but you're going to get bottlenecked at the places where your existing method doesn't apply.

extends · 2gives example · 1supports · 2

04
Claim

Opaque, high-parameter models like AlphaFold may not be explanations in the classical low-parameter sense, but they could be a genuinely new type of explanatory object — one from which we can extract embedded partial explanations (as arguably happened with AlphaZero and chess), and which we'll eventually gain new tools ('verbs') to merge, distill, and interrogate.

Nielsen lays out three ways to view AlphaFold-style models — not real explanations, mineable containers of small explanations, or an entirely new category of explanatory object — and leans toward the third, most radical view as the most interesting.

transcript

Michael Nielsen: The third and most interesting possibility is that they're a new type of object. They should be taken very seriously as explanations, but where in the past we haven't had the ability to really do anything with them, now we have interesting new actions we can do. We can merge them, we can distill them.

05
Example

Prout's hypothesis that all atomic weights are whole numbers was defended with ad hoc patches for 85 years — because chlorine's fractional weight couldn't be chemically resolved — until isotopes, which are chemically indistinguishable, were discovered; there was no way to know in advance that this stubborn anomaly pointed to something real.

Dwarkesh recounts Lakatos's example of Prout's hypothesis: an 85-year 'hostile' verification loop where the anomaly (chlorine's 35.5 weight) couldn't be resolved until isotopes were discovered, showing there's no ex ante rule for knowing which anomalies matter.

transcript

Dwarkesh Patel: So you have 85 years before we realize what an isotope is, where the verification loop is actively hostile against the correct theory. You just need this remnant to be defending… There's no ex ante reason it's the preferred theory.

06
Prediction◆

The science-and-technology tree is vastly larger and less explored than commonly assumed, so different civilizations — or even different research choices within one civilization — could end up discovering entirely different fundamental ideas, meaning alien civilizations would likely have a very different tech stack than humans.

Nielsen argues we're still near the bottom of an enormous, sparsely-explored tech tree (citing computer science's constant discovery of deep new ideas since the 1930s), so path-dependence means alien civilizations likely diverge sharply from human science and technology.

transcript

Michael Nielsen: The tech tree or the science and tech tree is probably much larger than we realize... My expectation is that there will be different ways of exploring this tech tree, and we're still relatively low down. We're still at the point where we're just understanding these basic fundamental theories, and we haven't yet explored them.

Highlight slides
"The Ether" Was Never Just One Theory✦ from: The Michelson-Morley result did not falsify 'the ether' as a single entity — it only ruled out some of many competing ether theories, and how scientists responded to that ambiguous result was far messier than a clean falsification story.Falsification Was Messy, Not Clean✦ from: The Michelson-Morley result did not falsify 'the ether' as a single entity — it only ruled out some of many competing ether theories, and how scientists responded to that ambiguous result was far messier than a clean falsification story.Einstein Himself Downplayed the Experiment✦ from: The Michelson-Morley result did not falsify 'the ether' as a single entity — it only ruled out some of many competing ether theories, and how scientists responded to that ambiguous result was far messier than a clean falsification story.The Tech Tree Is Still Mostly Unexplored✦ from: The science-and-technology tree is vastly larger and less explored than commonly assumed, so different civilizations — or even different research choices within one civilization — could end up discovering entirely different fundamental ideas, meaning alien civilizations would likely have a very different tech stack than humans.Evidence: Computer Science Keeps Yielding Deep Ideas✦ from: The science-and-technology tree is vastly larger and less explored than commonly assumed, so different civilizations — or even different research choices within one civilization — could end up discovering entirely different fundamental ideas, meaning alien civilizations would likely have a very different tech stack than humans.Implication: Aliens Would Diverge From Human Tech✦ from: The science-and-technology tree is vastly larger and less explored than commonly assumed, so different civilizations — or even different research choices within one civilization — could end up discovering entirely different fundamental ideas, meaning alien civilizations would likely have a very different tech stack than humans.
Related episodes
Video · Machine Learning Street Talk

Intelligence is collective, not artificial — Prof. Michael I. Jordan (UC Berkeley / Inria)

→
Video · Machine Learning Street Talk

He won a Nobel here for AlphaFold. Then he left. - John Jumper

→
Audio · Lex Fridman

#460 – Narendra Modi: Prime Minister of India – Power, Democracy, War & Peace

→