Dwarkesh PodcastMichael Nielsen on Dwarkesh Patel: Why Ether Died Slowly
Lorentz fit Einstein equations while keeping the ether ontology; Michelson-Morley only ruled out ether wind, so a single result cannot force a paradigm shift.
At a glance
WHAT IT’S REALLY ABOUT
Scientific progress is messier than verification loops and falsification stories
- The Michelson–Morley experiment did not straightforwardly “disprove the ether,” and key figures (e.g., Michelson) retained ether commitments for decades, illustrating how theory choice is underdetermined by experiments.
- Scientific communities can converge on productive interpretations before decisive experimental separation (e.g., relativity’s interpretation vs Lorentz-style ether framing), showing progress is not a simple, centralized method or process.
- Some major ideas (e.g., Darwinian natural selection) were conceptually accessible yet historically blocked by missing prerequisites like deep time, biogeography, and accumulated evidence, and often emerge via parallel discovery once conditions are ripe.
- AI can accelerate science in domains with strong data/feedback pipelines (e.g., AlphaFold), but “verification loops” can be long, misleading, or hostile (e.g., Vulcan, Prout’s hypothesis, Pioneer anomaly), limiting naïve automation-by-experiment.
- Nielsen argues the science/tech “tree” is far larger and more path-dependent than we assume, implying alien civilizations could explore different branches, creating future gains-from-trade in ideas and capabilities rather than convergence on a single stack.
IDEAS WORTH REMEMBERING
5 ideasExperiments rarely kill a single theory; they prune a family of stories.
Michelson–Morley primarily discriminated among ether models, not “ether vs no ether,” and leading scientists responded by modifying auxiliary assumptions rather than abandoning the core idea.
Scientific “progress” can precede decisive verification because interpretation matters.
Relativity’s conceptual shift gained traction well before later clean demonstrations (e.g., muon time dilation experiments), suggesting communities use aesthetics, coherence, and unification pressures—not just immediate tests.
Long verification loops are common—and often misleading.
Uranus’s anomaly rewarded a “patch” (Neptune), while Mercury’s anomaly punished the analogous “patch” (Vulcan) and pointed to new physics; most anomalies end up as mundane corrections (e.g., Pioneer thermal recoil).
Some ‘simple’ ideas require a large scaffolding of background facts to become compelling.
Natural selection wasn’t just the mechanism; Darwin’s breakthrough was arguing for its explanatory centrality across biology, enabled by deep time (Lyell), global biogeography, fossils, and accumulating comparative evidence.
AI success in science may mostly reflect data infrastructure, not automated theory discovery.
AlphaFold’s triumph depends heavily on decades of expensive experimental structure acquisition (Protein Data Bank); the model is impressive, but the bottleneck it breaks is partly “data + fit,” not necessarily “new principle.”
WORDS WORTH SAVING
5 quotesNewton was not the first of the Age of Reason. He was the last of the magicians, the last great mind which looked out on the visible and intellectual world with the same eyes as those who began to build our intellectual inheritance rather less than 10,000 years ago.
— Michael Nielsen
It's funny too, like the way we tell the history of science, it sounds so simple. Like, oh, you just focus on the right exception and, uh, you know, you realize that you need to throw out the old theory. And, and lo and behold, you know, your Nobel Prize awaits. But in fact, there's, these exceptions are all over the place.
— Michael Nielsen
We're-Basically slightly jumped up chimpanzees.
— Michael Nielsen
Most parts of the tech tree are never going to be explored.
— Michael Nielsen
It, it makes friendliness much more rewarding.
— Michael Nielsen
High quality AI-generated summary created from speaker-labeled transcript.
Get more out of YouTube videos.
High quality summaries for YouTube videos. Accurate transcripts to search & find moments. Powered by ChatGPT & Claude AI.
Add to Chrome