Michael Nielsen – How science actually progresses
TL;DR
Michael Nielsen dismantles the pop-science narrative of linear scientific progress through crisp experiments, revealing instead a messy, decentralized process where mathematical formalism often precedes conceptual understanding, expertise can blind researchers to truth, and communities adopt paradigm shifts long before experimental closure.
🧪 The Myth of Clean Falsification 3 insights
Michelson-Morley did not disprove the ether
The 1887 experiment only ruled out specific ether wind theories while leaving other ether variants intact, allowing Michelson to maintain belief in the ether until his death.
Einstein operated independently of the famous experiment
Einstein likely developed special relativity without direct influence from the Michelson-Morley result, stating he wasn't even sure if he knew of the paper when formulating his theory.
Experiments constrain rather than falsify
The null result eliminated some ether models but required decades of subsequent interpretation shifts before the scientific community abandoned the ether concept entirely.
🧮 When Mathematics Outruns Understanding 3 insights
Lorentz had the math but wrong interpretation
Lorentz derived the transformation equations underlying special relativity but interpreted them as dynamic effects of moving through the ether rather than as fundamental changes to space and time.
Poincaré understood principles but missed kinematics
Poincaré grasped the principle of relativity and constancy of light speed yet incorrectly explained length contraction as a dynamical compression of particles rather than pure kinematics.
Expertise can trap great minds
Both Lorentz and Poincaré possessed sufficient mathematical machinery for relativity but remained imprisoned by their deep expertise in classical frameworks, illustrating how knowledge can impede conceptual breakthroughs.
⏳ Progress Before Experimental Proof 3 insights
Heliocentrism won without stellar parallax
The scientific community adopted Copernican heliocentrism centuries before stellar parallax was finally measured in 1838, demonstrating that theories gain acceptance prior to experimental closure.
Accuracy and simplicity are not deciding factors
Copernicus's model was initially less accurate and more complex than Ptolemy's, yet its adoption suggests scientific progress values explanatory unity over raw predictive power or simplicity.
Newton unified the celestial and terrestrial
Newton's synthesis of planetary motion and terrestrial gravity under one law provided compelling explanatory coherence that transcended mere observational accuracy.
🧙 The Non-Algorithmic Nature of Discovery 3 insights
Consensus shifts without centralized authority
Scientific communities change interpretation through decentralized, non-standardized processes, adopting Einstein's framework before 1940s muon experiments could experimentally distinguish it from Lorentz's ether theory.
Great scientists can remain wrong for decades
Leading physicists like Michelson and Lorentz maintained commitment to disproven frameworks long after the broader community shifted, showing progress lacks standardized reconciliation procedures.
Newton was the last of the magicians
Newton applied the same rigorous methodological standards to alchemy and physics, functioning as a transitional figure who blended magical and modern thinking, revealing that scientific method alone doesn't distinguish valid from invalid domains.
Bottom Line
Scientific progress requires cultivating taste for explanatory frameworks that unify disparate phenomena while remaining skeptical that crisp experimental falsification alone drives paradigm shifts or that mathematical sophistication guarantees conceptual clarity.
More from Dwarkesh Patel
View all
Chip design from the bottom up – Reiner Pope
Reiner Pope explains how AI chips work from fundamental logic gates up, revealing that the physical cost of moving data between memory and compute units (via multiplexers) often exceeds the cost of the actual mathematical operations, and that circuit area scales quadratically with precision, making low-precision arithmetic exponentially more efficient than commonly assumed.
Building AlphaGo from scratch – Eric Jang
Eric Jang demonstrates how modern LLM coding tools and algorithmic improvements have democratized AI research, enabling a single researcher to rebuild AlphaGo for thousands of dollars rather than millions, while explaining how Monte Carlo Tree Search combined with neural networks solved a game previously considered computationally intractable.
David Reich – Why the Bronze Age was an inflection point in human evolution
Geneticist David Reich reveals that contrary to decades of evolutionary theory, natural selection has been rampant in human populations over the last 10,000 years, with the Bronze Age triggering an unprecedented acceleration in genetic adaptation to immune and metabolic challenges.
The math behind how LLMs are trained and served – Reiner Pope
Reiner Pope explains the mathematical mechanics behind LLM inference costs, demonstrating how 'Fast Mode' APIs charge premiums for smaller batch sizes that reduce latency, and why physical memory bandwidth constraints create hard limits on how fast or cheap inference can get regardless of budget.