What this covers

Michael Nielsen and Dwarkesh Patel discuss how scientific discovery actually unfolds, departing sharply from the textbook image of hypothesis-testing and falsification. Nielsen argues that the space of possible theories vastly exceeds what we realize—a tech tree so large and contingent that different civilizations would likely discover entirely different scientific stacks. The conversation traces this through major historical cases: Darwin and Wallace arriving at natural selection simultaneously, Copernicus being preferred over Ptolemy before precision tests could settle it, and Einstein adopting special relativity ahead of decisive experimental confirmation. The core claim is that verification loops are long and often hostile, bottlenecks perpetually shift to wherever existing heuristics fail, and which deep ideas get discovered depends heavily on historical accident—what building blocks happen to exist, which individuals are alive and curious at the right moment, which instruments mature when.

The discussion ranges across physics, biology, and the structure of modern science. Nielsen examines how Lorentz and Poincaré possessed the pieces for relativity but couldn't escape their attachment to the ether, while a teenage Einstein had the freedom to let go. He unpacks why the Michelson-Morley experiment didn't actually drive Einstein's thinking the way popular history claims, and why apparent anomalies—like the Pioneer spacecraft's slight trajectory error—turn out 99.9% of the time to be mundane effects, creating a survivorship bias in how we remember science. The conversation also addresses AI's actual contribution to discovery: AlphaFold succeeded not because of AI but because the Protein Data Bank represented decades of costly experimental work, with the algorithm a small final step. When verification requires infinite theories to fit any experiment, human judgment about what to pursue becomes irreducible, and no amount of compute easily resolves the choice. Throughout, Nielsen emphasizes the need for research program diversity—supporting many parallel bets even when most will fail—because you cannot know in advance which bet will pay off.

Sharpest takeaway

Nielsen argues that scientific progress is not a centralized, falsification-driven 'process' but an unpredictable exploration of an enormously larger-than-realized tech tree, where verification loops are long and hostile, bottlenecks shift to wherever existing heuristics fail, and which deep ideas get discovered is highly path-dependent.

  • Famous discoveries (special relativity, natural selection, heliocentrism) were adopted long before experimental verification and via no standard method
  • The science/tech tree is far larger and more contingent than the linear-tech-tree assumption, so different civilizations would discover different stacks
  • AI helps with data-acquisition bottlenecks (AlphaFold) but cannot easily resolve theory selection where infinite theories fit any experiment

The claims · ranked44 claims · weighted by value

This asset isn't compiled yet

You're seeing its claims, ranked. Compile it to build the argument threads, weight them, and check each claim against your library — the full view.

0.60

The same idea — positing an unseen perturbing planet — succeeded spectacularly for Uranus (predicting Neptune, found by Le Verrier in 1846) but failed completely for Mercury (the planet Vulcan didn't exist; the real cause was general relativity); you cannot tell a priori which case you're in, so you must support diverse parallel research programs even though it's hard to do.

causalestablishednovelty 3/4durability 4/4· Michael Nielsen

“You’ve pursued very similar ideas, and it’s been very successful in one case, and it’s been completely and utterly unsuccessful in the other case. A priori, you can’t tell which of these is the thing to do, and you actually need to do both.”

0.60

The near-simultaneous independent discovery of natural selection by Wallace and Darwin, like the broad history of parallel innovation in science, is evidence that prerequisite building blocks (deep time, paleontology, biogeography from colonization-era voyages) had to be in place — when an idea undiscovered for ages is suddenly hit upon by many people at once, it shows the building blocks were necessary.

causalestablishednovelty 3/4durability 4/4· Dwarkesh Patel

“if it’s not discovered for a long time and then spontaneously many different people are coming up with it, that shows you that the building blocks were in some sense necessary.”

0.60

Newton was not the first modern scientist but a transitional hybrid figure — 'the last of the magicians' — who looked at the world with the same superstitious eyes as ancient thinkers while also being modern, applying the same careful method and sobriety to his esoteric, theological, and alchemical work as to the Principia.

factualestablishednovelty 3/4durability 4/4· Michael Nielsen

“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 ten thousand years ago.”

0.60

Darwin's genius wasn't having the idea of selection — animal breeders knew artificial selection — but understanding how central it was to all of biology and doing the hard work of marshalling enormous evidence to show it was intertwined with everything, which is conceptually different from a single overwhelming experiment like Newton's tides.

causalestablishednovelty 3/4durability 4/4· Michael Nielsen

“Darwin’s genius wasn’t in having that idea, it was understanding just how central it was to biology.”

0.60

Apparent anomalies that look like they could overturn a fundamental theory — like the Pioneer spacecraft being slightly off course in the 1990s — are 99.9% of the time mundane effects (here, asymmetric thermal radiation causing tiny acceleration toward the sun), not new physics; the history of science suffers heavy selection bias because we only tell the rare stories where the exception was real.

factualestablishednovelty 3/4durability 4/4· Michael Nielsen

“Most of the time when there’s these apparent exceptions, it’s just something like that going on.”

0.60

Nature has gifted us an immense library of biological machines — hundreds of millions of proteins, kinesin walking, the ribosome as a miraculous factory — that we understand only on a painstaking one-by-one basis (tens of thousands of papers on hemoglobin and we still don't understand it); treating these as a gift from an alien civilization shows how much deep knowledge remains latent and how slow extraction is, mirroring how slowly we'd dig through alien code.

factualestablishednovelty 3/4durability 4/4· Michael Nielsen

“We’ve been gifted by biology this immense library of machines, no doubt containing an enormous number of very interesting ideas, and we’re just at the very, very beginning of understanding it.”

0.60

The hope that AI will make disproportionate scientific progress because science has tight verification loops (like running experiments) fails to appreciate that there are infinitely many theories compatible with any given experiment; why we latch onto the retrospectively-correct one is hard to articulate, as shown by hostile multi-decade verification loops like Prout's hypothesis where measuring chlorine at 35.46 looked like evidence against whole-number weights until isotopes were understood ~85 years later.

causalcontestednovelty 4/4durability 4/4· Dwarkesh Patel

“there’s an infinite number of theories that are compatible with any given experiment. Over time, why we latch onto the one we think is more correct in retrospect is, as we’re discussing, hard to articulate.”

0.58

Modern collective science like the LHC is so complex that no single person can understand everything in depth — finding the Higgs boson requires deep expertise in accelerator physics, inverse problems for particle-track simulation, quantum field theory, detector physics, vacuum physics, and data processing, each a lifetime of work — which is why thousand-author papers exist: people communicate at a high level but don't understand each other's specialties in depth.

factualestablishednovelty 2/4durability 4/4· Michael Nielsen

“I don’t think it’s possible for one person to understand everything in depth. Lots of people broadly understand a lot of these ideas, but they don’t understand everything in the depth that is actually utilized.”

0.58

Science always gets bottlenecked precisely where previous processes and heuristics don't apply, almost by definition; because smart people study and apply what worked before, they don't get stuck in the same places, so bottlenecks keep moving to new places where there's no crank to turn and you need many people trying different ideas — the harder the idea, the greater both the bottleneck and the triumph.

causalcontestednovelty 3/4durability 4/4· Michael Nielsen

“where we always get bottlenecked is where the previous processes and heuristics don’t apply. That’s almost definitionally what causes the bottlenecks.”

0.55

The standard story that the Michelson-Morley experiment created a crisis Einstein solved with special relativity is misleading; Einstein wasn't even sure he was aware of the paper, it wasn't dispositive for his thinking, and Michelson and Morley were actually testing competing theories of the ether against one another, not proving the ether didn't exist.

factualestablishednovelty 3/4durability 3/4· Michael Nielsen

“he stated later in his life he wasn’t even sure whether he was aware of the paper at the time. There’s a lot of evidence that he probably was aware of the paper at the time, but it actually wasn’t dispositive for his thinking at all.”

0.55

Lucretius's first-century idea, though analogous to natural selection, was extremely different from real natural selection: he posited a one-time generative period producing all species followed by a single filter, not an ongoing gradual process or a tree of life connecting all organisms, which is why the analogy went nowhere for nineteen centuries.

factualestablishednovelty 3/4durability 3/4· Dwarkesh Patel

“He thought there was this generative period in the past where all the species came about, and then there was this one-time filter which resulted in the species that are around today”

0.55

The scientific community adopted what we now consider the more correct (Einsteinian) interpretation over Lorentz's ether-based interpretation before it was experimentally shown to be preferred — the muon decay experiments confirming time dilation came around 1940, decades later — showing human science can distinguish theories ahead of experimental validation.

factualestablishednovelty 3/4durability 3/4· Dwarkesh Patel

“the scientific community adopted what we in retrospect consider the more correct interpretation before it was actually experimentally shown to be preferred. So there’s clearly some process that human science does which can distinguish different theories.”

0.55

Aristarchus proposed heliocentrism in the 3rd/2nd century BC and Athenians dismissed it because they should see stellar parallax, which wasn't measured until 1838 — yet Copernican heliocentrism was preferred long before that, despite the Ptolemaic model being more accurate and Copernicus actually needing more epicycles, so one couldn't have known ex ante that Copernicus was correct from accuracy or simplicity.

factualestablishednovelty 3/4durability 3/4· Dwarkesh Patel

“it’s only in 1838 that stellar parallax was actually measured. And so, we didn’t need to wait until 1838 to have heliocentrism.”

0.55

Quantum computing emerged in the 1980s (Feynman 1982, Deutsch 1985) rather than the 1950s — when von Neumann could have invented it — because two historically contingent things matured around 1980: computation became far more salient as personal computers (Apple II, Commodore 64) appeared, and the Paul trap gave the first ability to manipulate single quantum states; the spark required someone like Feynman who was both excellent at quantum mechanics and excited by the new machines.

causalestablishednovelty 3/4durability 3/4· Michael Nielsen

“You got these two separate things that for historically contingent reasons had both matured around 1980 or so.”

0.54

Poincaré understood the principle of relativity and constant speed of light but clung to a wrong dynamical interpretation of length contraction because he knew too much; older experts like him and Lorentz may have been 'prisoners of their own expertise,' while teenage Einstein, less attached to the ether, could let go and see space and time were simply different.

causalcontestednovelty 3/4durability 4/4· Michael Nielsen

“It’s almost like he knew too much. He had almost too grand a vision in mind. Einstein subtracts from that and says, “No. Space and time are just different than what we thought, and here’s the correct picture.””

0.54

Genuine deep learning requires a creative artifact and a forcing function that clamps understanding — implementing the transformer to understand it, or doing physics exercises — because without it you only map inputs and outputs of a black box, which fades incredibly fast; spending time stuck is perhaps the most important part of the process, which is why essays written in three months are remembered 15 years later while those written in a couple of days taught the author little.

causalcontestednovelty 3/4durability 4/4· Michael Nielsen

“if you don’t really understand the deeper mechanism, you’re just mapping inputs and outputs of a black box. That just fades incredibly fast”

0.54

The real history of science shows you can't apply falsification as easily as the scientific-method picture implies: it's unclear what is being falsified (one version of the ether? all of them?), and you certainly cannot induce special relativity from the disconfirmation of one ether theory.

factualestablishednovelty 2/4durability 4/4· Dwarkesh Patel

“you really can’t apply falsification as easily as you might think. It’s not clear what is being falsified.”

0.54

Lyell and other geologists establishing deep time (tens to hundreds of millions of years) in the early 1800s was a crucial precondition for evolution, because on a 6,000-10,000-year timescale you'd need to observe evolution at a massive rate within human lifetimes, which isn't seen — so insufficient time was a genuine blocker to the idea.

causalestablishednovelty 2/4durability 4/4· Michael Nielsen

“If you don’t have at least tens or hundreds of millions of years, evolution starts to look like a non-starter.”

0.53

The comparative-advantage argument that humans will stay employed post-AGI breaks down for two key reasons: transaction costs (like the cost of maintaining roads compatible with both horses and cars, analogous to AIs thinking 1000x faster finding human supply-chain interaction too costly), and that comparative advantage doesn't guarantee trade terms above subsistence for a producer — just as a horse's mathematical comparative advantage doesn't make it worth the $100,000/year to sustain in San Francisco.

causalcontestednovelty 3/4durability 3/4· Dwarkesh Patel

“comparative advantage does not tell you that the terms on which the trade happens are above subsistence for any given producer.”

0.50

What made Newton's theory compelling was that one set of ideas — his theory of gravitation — explained three seemingly disconnected phenomena: planetary motion via Kepler's laws, terrestrial parabolic motion, and the tides via the moon and sun's gravitational pull on water; this unification feels deeply satisfying even before decisive experiments.

causalestablishednovelty 2/4durability 4/4· 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.”

0.50

By Dean Keith Simonton's equal-odds rule, the probability that any given output (paper, book) will be extremely important is roughly constant across a person's life, so the era in which someone is most productive is determined by how much they're publishing — meaning many of the most successful scientists and creatives, like Shakespeare, simply produce a lot, though there are counterexamples like Gödel who published almost nothing.

factualcontestednovelty 3/4durability 3/4· Dwarkesh Patel

“the probability that any given thing you release... will be extremely important for a given person through their lifetime is not that different. What really determines in what era they are the most productive is how much they’re publishing.”

0.50

AlphaFold's success isn't really about AI — a massive fraction is the Protein Data Bank built over decades through X-ray diffraction, NMR, and cryo-EM costing billions to obtain ~180,000 protein structures; the AI was a tiny fraction of total investment, fitting a model at the end of a giant data-acquisition story.

causalcontestednovelty 3/4durability 3/4· Michael Nielsen

“A massive fraction of the success there is the Protein Data Bank. It’s X-ray diffraction, NMR, cryo-EM, and the several billion dollars that were spent obtaining those 180,000-odd protein structures.”

0.50

Raw gradient descent could keep adding epicycles to a Ptolemaic model and improve local accuracy, but it wouldn't make the paradigm-level switch to Copernicus that locally doesn't improve accuracy yet is globally more progressive; such switches require a forcing function (an internal contradiction, like Newtonian gravity allowing faster-than-light signaling in special relativity) plus many independent research programs kept alive simultaneously over long verification loops.

causalcontestednovelty 3/4durability 3/4· Dwarkesh Patel

“you could just keep adding epicycles upon epicycles, but it really took one mind to integrate it all in and say, “Here’s what makes more sense overall.””

0.50

The popular idea that high performers should be in flow all the time is false; like athletes who are in flow while playing but stuck or doing things badly during training, high performers are in flow only some of the time, which implies learning requires deliberately structuring work output to include the un-flowy, stuck, demanding parts.

factualcontestednovelty 3/4durability 3/4· Michael Nielsen

“There’s this terrible idea that lots of people have that they should be in flow all of the time. And as far as I can tell, high performers just don’t believe this at all.”

0.49

There is no standard procedure or centralized authority by which the scientific community reconciles competing theories; great scientists like Lorentz, Poincaré, and Michelson can remain wrong long after the community has changed its opinion, so 'process' is the wrong frame.

factualcontestednovelty 2/4durability 4/4· Michael Nielsen

“It’s not as though there’s some standard procedure that we’re all using to reconcile these things. Great scientists can remain wrong for a very long time after the scientific community has broadly changed its opinion. But there’s no centralized authority or centralized method.”

0.47

Scientific progress was slow and irregular until ~1700, partly because good ideas were lacking but also because you must build institutions — training, capital allocation, even basic security for researchers from things like the Inquisition — around ideas before progress bursts; so observed diminishing returns may be a symptom of stagnant external circumstances rather than anything intrinsic to the structure of knowledge, and changing those circumstances (or improving the minds doing the work) can restart progress.

causalcontestednovelty 3/4durability 4/4· Michael Nielsen

“If there’s some kind of stagnation, if you’re not changing those external circumstances, yes, you may start to get diminishing returns again. But that doesn’t mean there’s anything intrinsic about the situation.”

0.47

The low-hanging-fruit argument for diminishing returns holds in a static snapshot — at a dessert buffet of 30 desserts the best go first because people have similar preferences — but in science someone keeps restocking the table with new fields (computer science arose as a side effect of esoteric philosophy-of-mathematics questions), so the diminishing-returns argument repeatedly fails as new fields open up and let twenty-one-year-olds make major breakthroughs.

causalcontestednovelty 3/4durability 4/4· Michael Nielsen

“if somebody is standing behind the dessert table and is replenishing and restocking the desserts and keeps adding new ones in, it may turn out that a little bit later, much better desserts appear”

0.47

The credit/attribution economy of science is something we construct by agreement, not a natural fact: it took over a century to move from Galileo and Kepler publishing discoveries as scrambled anagrams to the modern paper-plus-attribution reputation economy; today we can share code, data, and in-progress ideas but with no agreed credit, and any change to that economy yields a different system for constructing knowledge — illustrated by physicists and biologists giving opposite competitiveness rationales for preprint culture.

factualcontestednovelty 3/4durability 4/4· Michael Nielsen

“this kind of attribution economy is just something we construct. It’s something we do by agreement. Any attempt to change that economy results in a different system by which we construct knowledge.”

0.45

There may be an interesting transition with quantum computers because they are probably capable of a strictly larger class of potentially interesting computations than classical computers, so artificial general intelligence built on quantum machines (AQGI) might be qualitatively different — though after only ~40 years of thinking about it without the devices, it's unknown whether quantum advantage is narrow (like Shor's algorithm breaking encryption) or very broad.

forecastcontestednovelty 3/4durability 2/4· Michael Nielsen

“But actually, there’ll be an interesting transition with quantum computers as well. They’re probably capable of a strictly larger class of potentially interesting computations.”

0.45

As AI removes the coding-time bottleneck for programmers — prototypes that took three weeks now take three hours — the bottleneck is shifting to having interesting design ideas, for which there's no verification loop; programmers formerly weren't bottlenecked on ideas because writing code consumed their time and let ideas accumulate, but now they implement faster and lack as many good design ideas afterward.

causalcontestednovelty 3/4durability 2/4· Michael Nielsen

“one thing that a lot of them seem to be bottlenecked on now is having interesting ideas, and in particular, having interesting design ideas. There’s not really a verification loop for knowing that a design idea is very interesting.”

0.43

Large opaque models like AlphaFold can be viewed three ways as scientific explanations: (1) conservatively, not real explanations because they lack few-parameter simplicity; (2) as containers of many little extractable explanations via interpretability/archeology, as strategies were extracted from AlphaZero and adopted by Magnus Carlsen; (3) most interestingly, as a genuinely new type of object that should be taken seriously as explanation but on which new operations like merging and distilling can be performed.

definitionspeaker onlynovelty 4/4durability 3/4· 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.”

0.43

Most of the tech tree will never be explored because there are too many interesting combinations and deep ideas, so choices about how to explore matter a great deal; this is why technological-determinist arguments are wrong higher up the tree — we are already shaping exploration through institutions that preemptively ban paths (DDT, chlorofluorocarbons, nuclear weapons restrictions, the Nuclear Non-Proliferation Treaty).

normativecontestednovelty 3/4durability 3/4· Michael Nielsen

“Most parts of the tech tree are never going to be explored. There are just too many interesting ways of combining things.”

0.42

The science and tech tree is probably much larger than we realize and we are still near the bottom; a 'theory of everything' wouldn't finish physics any more than Turing and Church's 1930s theory of computation finished computer science — we've spent ninety years discovering deep hidden principles like public-key cryptography that lay latent in that theory, and phases of matter keep multiplying (superconductors, superfluids, Bose-Einstein condensates, fractional quantum Hall systems) that we'll increasingly design.

factualspeaker onlynovelty 3/4durability 4/4· Michael Nielsen

“the tech tree or the science and tech tree is probably much larger than we realize. We’re in this funny situation.”

0.42

We keep finding deep new fundamental primitives: the Church-Turing idea of universal programmable devices contained within it public-key cryptography, which in turn contained the ideas of cryptocurrency and collectively-maintained agreed-upon ledgers — a recurring pattern suggesting there may be an effectively unbounded supply of extremely deep principles still latent in our foundational theories.

factualspeaker onlynovelty 3/4durability 4/4· Michael Nielsen

“you understand later, this also contains within it the ideas of public-key cryptography. Then you understand later, that also contains within it the ideas people refer to as cryptocurrency.”

0.42

There is a seductive failure mode where you learn a complex system (Alan Kay called Linux 'a great big ball of mud') and confuse mastering its idiosyncratic bits with understanding fundamentals; most of what you learn is non-transferable, so whether it's a good use of time depends entirely on your objective — fine if you want to be a sysadmin, but poor if you want to understand the fundamentals of computing.

normativespeaker onlynovelty 3/4durability 4/4· Michael Nielsen

““It doesn’t have anything to do with computer science. It’s just a great big ball of mud. There are a few interesting ideas in there which are worth understanding, but mostly all you’re learning is stuff about Linux. You’re not actually learning anything which is transferable.””

0.41

Deep learning's apparent uniqueness as a field where twenty-one-year-olds make progress yet that now requires tens of billions of dollars is largely a consequence of the architecture of attention — there is always a 'biggest thing,' and if deep learning weren't it we might be celebrating CRISPR, or framing protein structure prediction as a triumph of curve fitting rather than AI; it's substantially fashion-driven, not intrinsic difficulty.

causalspeaker onlynovelty 3/4durability 3/4· Michael Nielsen

“I think that’s a consequence of the architecture of attention. At any given time, there’s always a most successful thing. If deep learning wasn’t a thing, maybe you’d be talking about CRISPR.”

0.39

Creative projects involve two kinds of work — routine work, where you should avoid procrastination, get good at it, or outsource it and finish as fast as possible; and high-variance work, where you must be willing to spend lots of time, wander, and talk to people even though most of it won't be an input — and because preferring one over the other is almost a personality trait, most people over-invest in one and under-invest in the other, so deliberate balancing is essential.

normativespeaker onlynovelty 3/4durability 3/4· Michael Nielsen

“There’s routine stuff, and there you just want to avoid procrastination... Then there’s high-variance stuff where you actually need to be willing to take a lot of time.”

0.39

Objects formerly too complicated to use — like a 100-page equation in 1920, where you'd simply give up — are now workable with tools like Mathematica, and people sometimes get simple answers out of them; large ML models may follow the same trajectory, becoming objects on which interesting operations can be performed even if they aren't classic explanations.

factualspeaker onlynovelty 3/4durability 3/4· Michael Nielsen

“today, with tools like Mathematica, you can just keep going. That’s an object now, a thing that you can work with.”

0.39

People consistently under-operate: when students fail to solve a week's problem, asking whether they'd have put in the same effort if a million dollars were at stake reveals the answer is invariably no — they tried but didn't really try — so the key to deeper learning is engineering a sufficiently demanding context or 'taskmaster' that raises the stakes and makes preparation far more intense.

causalspeaker onlynovelty 3/4durability 3/4· Michael Nielsen

“If a million dollars had been at stake, would you have put the same effort in? And the answer is no, invariably. They’ve tried, but they haven’t really tried.”

0.38

If the tech tree is so wide, contingent, and path-dependent that different civilizations stumble onto entirely different technology stacks, then there will be enormous gains from trade far into the future via comparative advantage, making friendliness more rewarding and shaping how civilizations coordinate rather than a 'go forth and exploit' dynamic.

forecastspeaker onlynovelty 4/4durability 3/4· Dwarkesh Patel

“There’s a very interesting implication that there will be gains from trade into the far, far future, which might actually be one of the most important facts about the far future in terms of how civilizations are set up”

0.38

Because the tech tree is enormous and exploration is path-dependent and shaped by contingent biases (humans are very visual, other creatures aurally based; more exotic civilizations would perceive and manipulate the world differently), different civilizations would explore it differently and end up in very different parts of the tree — contradicting the assumption that all civilizations converge on the same science.

forecastspeaker onlynovelty 4/4durability 3/4· Michael Nielsen

“It’s potentially the case that different civilizations or different choices mean we end up in different parts of that tree.”

0.38

AI lets you move faster but may not improve learning, because the most demanding thing is so aversive that you take any excuse to escape it; a back-and-forth LLM conversation where you gloss over is entertaining and not useless, but it substitutes for the hard intermediate thinking you should be doing — there's always a next question to ask the chatbot instead of working through the difficulty yourself.

causalspeaker onlynovelty 3/4durability 2/4· Dwarkesh Patel

“The hardest thing, the thing that is most demanding, is so aversive that you try to take any excuse you can to get out of it. Just having a back-and-forth conversation with an LLM where you gloss over…”

0.37

Every single life form on Earth has a common ancestor, which is not incredibly weird if you think the origin of life was very hard — a bottleneck — making a single origin unsurprising.

factualcontestednovelty 2/4durability 3/4· Michael Nielsen

“It’s not incredibly weird. If you think that the origin of life must have been very hard, that there’s a bottleneck there, then it’s not so surprising.”

0.34

Nielsen chose quantum computing in 1992 by applying a 'low-hanging fruit algorithm' — relative to his own interests and abilities, finding where to pick up a shovel and dig — after a mentor (Gerard Milburn) gave him a stack of foundational Deutsch and Feynman papers; the papers asked fundamental questions on which one could tractably make progress at a time when almost nobody worked on it, so he benefited from another person's taste.

factualspeaker onlynovelty 2/4durability 3/4· Michael Nielsen

“you are applying this low-hanging fruit algorithm. You’re like, “relative to my particular set of interests and abilities, where should I pick up my shovel and start digging?””