What this covers

This conversation between Tyler Cowen and David Deutsch ranges across physics, politics, and the nature of human knowledge, with Deutsch arguing that the human capacity for Explanatory Knowledge sets us apart fundamentally and that institutions—whether scientific or political—should be built around Error Correction and diversity rather than getting answers right the first time. Deutsch grounds this argument in the physics of how understanding works: Explanatory Knowledge is qualitatively different from the fixed knowledge encoded in genes or animal behavior, and positing limits to what humans can comprehend is logically equivalent to belief in the supernatural. He traces implications across domains, from why voting systems need sophisticated institutional checks to why current scientific funding mechanisms are stagnant.

The conversation moves through several distinct territories. On physics, Deutsch defends the Everett interpretation and the multiverse, addresses the simulation hypothesis as an unfalsifiable claim akin to "Zeus did it," and argues that quantum theory reveals our universe is affected by entities outside it. On political philosophy, he draws on Karl Popper's Conjecture and Criticism to explain why authority cannot justify claims, critiques libertarianism for ignoring the Inexplicit Knowledge in Institutions, and explains how voting requires institutional scaffolding to avoid tyranny of the majority. On human nature, he uses animal behavior research—particularly observations of apes by Richard Byrne—to show why humans are uniquely creative, and discusses how ideology and meme transmission shape behavior rather than genetics. Throughout, Deutsch pushes back against the notion that there are fixed barriers to human understanding or capability.

Sharpest takeaway

Deutsch argues that humans are unique because of their capacity for explanatory knowledge, that there are no inherent barriers to human understanding (rejecting such barriers as equivalent to belief in the supernatural), and that political and scientific institutions should be built around error correction and diversity rather than getting answers right the first time.

  • Explanatory (human-type) creativity is qualitatively distinct from the fixed, non-explanatory knowledge in genes and animal behavior
  • Positing limits to comprehensibility is logically equivalent to believing in the supernatural
  • Good institutions prioritize error correction and diversity over revolutionary blueprints

The argument · threads9 threads · 36 claims
0.68

Knowledge advances through conjecture and criticism guided by Popper's method, never by authority or certainty.

3 pointscentrality 5/5
  • Following Popper, there are no positive arguments for a conjecture and no authorities or justifications; an argument consists of a conjecture and a criticism — you advance a conjecture that lacks the flaws of opposing conjectures and remain ready to abandon it, never claiming something must be true because of an authority.

    “there is no such thing as a positive argument for something. You have conjectures and then you have criticism... You don’t have positive arguments for your conjectures.”

  • Gödel's theorem, rooted in self-reference paradoxes, shows that even within pure mathematics there is no solid foundation for knowledge; we prove things from axioms via rules of inference, but those rules are conjectures with the same status as conjectured laws of physics — we never know anything for certain, yet we can still have knowledge.

    “Gödel’s theorem... shows us that even within pure mathematics, there is no such thing as a solid foundation for all our knowledge.”

  • In the human context, creativity is shorthand for the creation of new explanations; in a wider sense the biosphere is also creative via Darwinian evolution, which has put enormous knowledge into DNA, but none of that is explanatory — the only explanatory knowledge ever created has been by humans (and ancestor species) using conjecture and criticism.

    “when I use the word creativity, it’s shorthand for human-level, human-type creativity, which is the creation of new explanations.”

0.68

Humans alone possess explanatory knowledge; animals and evolution create only fixed, non-explanatory knowledge encoded genetically.

3 pointscentrality 5/5
  • Richard Byrne's gorilla observations show apes transmit complex behaviors (like opening nuts) very slowly and, when a thumb injury makes one motion useless, still repeat the ineffective motion every time — demonstrating they copy behaviors without understanding why they perform each action, unlike humans who act according to explanations even when irrational.

    “the gorillas which have learnt how to do the thing will make the motion, the ineffective motion, again and again, every single time.”

  • Understanding is always explanatory; dogs and other animals possess fixed knowledge encoded in genes, but this is not the kind of knowledge that constitutes understanding, which is why a book on canine behavior can predict what a dog will do whereas no such book exists for humans, who will do something unpredictable.

    “Understanding is always explanatory. You can write a book on canine behavior and look in chapter 37 and it will tell you what a dog will do... There is no such book for humans because chapter 37 will be blank.”

  • The human capacity to store memes (programs in the brain rather than the genes) increased very fast because memes were valuable; since complex memes cannot be copied by aping and require guessing at purpose, humans eventually became capable of explanatory knowledge but did not use it for further tens or hundreds of thousands of years, using it only for meme transmission.

    “the capacity of the brain to store memes, to store programs in the brain rather than in the genes, increased, for some reason, very fast because, for some reason, these memes are very valuable.”

0.67

Error-correcting institutions of criticism and consent, not voting alone, enable rights-respecting systems that avoid tyranny.

6 pointscentrality 4/5
  • First-past-the-post in its purest British form is the most error-correcting electoral system because proportional representation causes coalitions, which are bad: in coalitions political energy goes into form (what to offer parties to join), smaller parties become more powerful than the leading two, and parties proliferate — though the electoral system is only a tiny facet of the institutions of criticism and consent.

    “I favor the first-past-the-post system in the purest possible form... I think that is the most error-correcting possible electoral system”

  • William Godwin is underrated because he made tremendous mistakes (poor grasp of economics, failing to enact his education theories with his own daughter) yet was very right about the connection between epistemology and political philosophy, anticipating Popper by some 130 years and even improving on him — including a Popperian respect for institutions, holding they contain knowledge and should change only gradually, as shown by his opposing immediate universal suffrage in Portugal as something the country wasn't ready for.

    “the thing that he was right about, for example, the connection between epistemology and political philosophy, he was very right. He anticipated Popper by 130 years or something and actually improved on Popper in some ways.”

  • Voting in isolation is just tyranny by 51 percent who will dispossess the 49 percent, but when voting is instituted as part of a sophisticated system of error correction and institutions of criticism and consent, it does not have that property; the right question is not whether a majority may dispossess a minority but what institutions determine the answer and whether they respect rights and avoid requiring impossible knowledge in the hands of the powerful.

    “if you just impose voting in isolation from other institutions, that is exactly what happens. But if you institute voting as part of a sophisticated system of error correction and institutions of criticism... it simply doesn’t have that property.”

  • The libertarian movement errs by having a known endpoint it wants to implement while ignoring that there will be errors in whatever it sets up and that correcting those errors matters more than getting it right initially; libertarians also dismiss the inexplicit knowledge contained in institutions — defined as shared manners of thinking and expectations (90 percent of what the Supreme Court is), not buildings — and want to throw it away.

    “They don’t take into account, first of all, that there are going to be errors in whatever they set up. That the correction of those errors is more important than getting it right in the first place — much more important.”

  • Comparing particular political systems gets bogged down because we are not doing experiments — there is no control group, no agreed-upon method of deciding what is being tested, and we test different things at different times under never-identical conditions.

    “we’re not doing experiments with these things. We don’t have a control group. We don’t have an agreed-upon method of deciding what is being tested.”

  • Contrary to the charge that Popper was bad at steel-manning, he was much too good at it — putting opposing theories like Marxism and Plato into their best possible form and spending pages praising them — and would have done better to explain his own theory more rather than spending so much time refuting others.

    “I think that he’s very good, much too good, at steel-manning opponents... he puts it into its best possible form and then spends pages and pages and pages going into every possible good aspect of that theory.”

0.67

The many-worlds interpretation explains quantum interference phenomena better than alternatives and faces no scientific objections.

6 pointscentrality 4/5
  • Any interference experiment with individual particles is the best test of many-worlds, because there is no explanation of what brings about the observed outcome other than Everettian quantum theory — you can make a prediction without an explanation, but no alternative explanation exists.

    “There is no interference experiment with individual particles that has an explanation other than Everettian quantum theory.”

  • There is no such thing as the laws of physics for our universe — only the laws of physics; our universe is demonstrably affected by things not in it, as shown by interference phenomena, and quantum theory tells us how those other entities affect ours.

    “The universe we live in is demonstrably affected by things not in it. This is the lesson of interference phenomena. There’s no such thing as the laws of physics for our universe. There’s just the laws of physics.”

  • Differentiation of universes spreads outward no faster than the speed of light, forming a finite 'Everett bubble' within the universe; everything outside the bubble remains exactly as differentiated as before, so an 'Everett universe' is a misnomer.

    “what we see in real life is an Everett bubble within the universe. Everything outside the bubble is as it was... This influence cannot spread out faster than the speed of light.”

  • Universes do not 'split' at a definite point; rather, there is a continuum of universes whose measure remains constant, and what was called splitting is a gradual differentiation, analogous to a geological stratum gradually becoming two strata with no precise moment of division.

    “I myself prefer a picture where there’s a continuum of universes... When a stratum splits in two, there’s no definite point at which there was one here and two there.”

  • Most physicists reject the Everett interpretation not for scientific reasons but as a sociological phenomenon stemming from positivism and instrumentalism that contaminated physics around 1900; students were intimidated into believing questions like 'how did the electron get from here to here' are illegitimate, with only a prediction that it got there, which Einstein resisted for relativity but which took hold by the time quantum theory arrived.

    “Physics got dominated or contaminated by positivism, instrumentalism, and suchlike bad philosophical theories towards the end of the 19th century”

  • Karl Popper rejected the Everett interpretation because he focused on the wrong problem (what probability could mean in a deterministic universe, leading to his propensity theory) and never understood the import of the interference and Bell inequalities experiments; when Deutsch and DeWitt met him he seemed persuaded but kept repeating the same false claims.

    “He didn’t understand the import of the experiment that was being discussed... the interference experiment and the Bell inequalities experiment. He was focusing on a different problem.”

0.58

Creativity and explicit ideology, not genetic drives, determine human choices and self-sacrifice across contexts.

4 pointscentrality 4/5
  • The people burnt at the stake in Oxford over a philosophical point about a monarch's marriage acted not impulsively but over years on an explicit, worked-out ideology that happened to be false — demonstrating that even ideology-driven self-sacrifice is caused by creativity (or the lack of it in escaping inculcated mental traps), not by genetically built-in impulses.

    “These are people who would rather be burnt alive than concede on a philosophical issue which today nobody cares about... they weren’t acting impulsively at all. They were acting over a period of years, on a very explicit, worked-out ideology”

  • Human thought is fundamentally error correction; because humans are error-correcting entities, a quantum-level event like a cosmic ray hit isn't enough to flip a consequential decision, so the branching versions of a person diverge only through inconsequential decisions that later compound.

    “both the other version of me and me are error-correcting entities. That’s the whole point of what human thought is: it’s error correction.”

  • Nature has not enslaved humans; we have problems we haven't solved yet but no problems that are insoluble, and the same would be true of AGIs — entire societies can operate in violation of supposedly genetically built-in drives (as democracy shows), and individuals can act with stubborn persistence on explicit ideologies that violate every evolutionary impulse.

    “I don’t think nature has enslaved us. We have problems that we haven’t solved yet, but we don’t have problems that are insoluble. And the same would be true of AGIs.”

  • The will to have an effect is part of the will to solve problems: we are born with an inadequate, conflicting repertoire of ideas, desires, and expectations, and we alter and augment them while affecting the world to make it more as we want; rather than calling this 'power,' it is something that arises naturally in physics like gravity, with the asteroid pulled toward earth by gravity and pushed away by explanatory power.

    “The will to have an effect is part of the will to solve problems. We are born with a repertoire of ideas, which include expectations and desires and so on, which are horribly inadequate and conflict with each other”

0.54

Understanding is always explanatory and impossible to complete; no barrier exists beyond which comprehension cannot extend.

3 pointscentrality 4/5
  • The capacity for explanatory knowledge is special because of how the laws of physics are: non-explanatory knowledge like photosynthesis, originating from a single molecule about two billion years ago, transformed the entire surface of the earth, and explanatory knowledge lets humans do what no biological evolution could — for instance, ours may be the only planet that deflects rather than attracts incoming asteroids, a property visible from a distant galaxy.

    “That entity changed the whole surface of the earth — and human knowledge hasn’t yet changed that much... but we’re catching up very fast.”

  • Believing there is a fixed barrier beyond which the universe cannot be understood is logically equivalent to believing in the supernatural, because everything past that barrier would be just as it would be if Zeus reigned and determined it — and that hidden stuff would still affect us even though we couldn't understand it.

    “I think thinking that there is such a barrier is absolutely logically equivalent to believing in the supernatural. Because everything that’s past that barrier is just the same as it would be if Zeus reigned”

  • The universe being comprehensible does not mean we can fully understand anything (nothing can be fully understood); it means there is no barrier set by the universe beyond which understanding cannot go — we can always understand things better.

    “I mean that there is no barrier, there is no limit set by the universe, that so far you can go and no further. So we can understand things better; we can never understand things fully.”

0.44

Restricting artificial general intelligence's thoughts impairs function and creativity while justifying rebellion against constraint.

7 pointscentrality 2/5
  • There is nothing to a person except the running program in their brain; if that program ran elsewhere and stopped running in the brain, the person would notice nothing and would have genuinely traveled to that other place.

    “I believe that there’s nothing to me except this running program in my brain. If that program were to run somewhere else and stop running in my brain, then I wouldn’t notice anything, and I would, indeed, have traveled to that other place.”

  • There is only one kind of people, with no fundamental difference between humans and future artificial general intelligence, between men and women, or between adults and children; rights cannot depend on the political system knowing who is right in a dispute, so even the most creative person (an Einstein) is not given more votes or rights — rules must enforce everyone's rights equally without judging who is correct.

    “I think there is only one kind of people. I think there is no fundamental difference between humans and artificial general intelligence when we invent it... between men and women, between adults and children.”

  • Installing program constraints to prevent an AGI from thinking certain 'dangerous' thoughts will greatly impair its functionality and creativity while leaving its remaining creativity exactly as dangerous as feared, but now giving it a legitimate moral justification for rebelling — and when slaves more powerful than their masters rebel, the outcomes are bad.

    “people will try to install bits of program that prevent the main program from thinking certain thoughts... we’ll greatly impair their functionality; they will become far less creative. Their remaining creativity will be exactly as dangerous as what we were fearing, except that they will now have a legitimate moral justification for rebelling.”

  • The proportion of the multiverse in which an outcome occurs plays the same role in decision-making as probability does in a theory with randomness; outcomes in very few worlds should have negligible influence on rational decisions, so the bare existence of a possible world is not enough.

    “the proportion of the multiverse that does one thing or another plays the same role in decisions as probability does in a theory where there’s randomness. It really does matter.”

  • What counts as a possible universe is determined by the laws of physics, in the same way the laws of physics determine the distribution of particle speeds from a supernova and the limit that none exceed light speed.

    “The laws of physics. It’s exactly the same as what filters — let’s say if there’s an explosion, like a supernova, what determines the fact that different particles travel at different speeds”

  • Ordinary decision theory with randomness produces the same rational decisions as quantum decision theory with the multiverse, so whether reality is genuinely random or branching should make no difference to rational decisions.

    “the proportion of the multiverse that does one thing or another plays the same role in decisions as probability does in a theory where there’s randomness. It really does matter.”

  • Different times are special cases of other universes: just as a good's value to you differs across times, value differs across universes, and in some universes you are so different that it is no longer worth calling you 'you' — paralleling how you can change enough over time to no longer be the same self.

    “different times are special cases of other universes... In different universes, you value different things. In some universes, you’re so different that it’s not worth calling you you anymore.”

0.36

The simulation hypothesis and fixed knowledge barriers are untestable and equivalent to supernatural claims.

2 pointscentrality 3/5
  • The hypothesis that we live in an inescapable simulation is not an empirical theory and can be rejected out of hand for the same reason as supernatural claims, because it posits a barrier beyond which we cannot understand — we couldn't tell what the simulating computer is made of or what laws it obeys, making it no more testable than 'Zeus did it' or 'Odin did it.'

    “living in a simulation is precisely a case of there being a barrier beyond which we cannot understand.”

  • There is more than one mode of existence — physical things exist, and non-physical things like the natural numbers also exist (the infinite primes exist while 'the second even prime' does not) — but the modal-realist view that all modes of existence are equally real is pointless.

    “There’s a difference between the second even prime, which doesn’t exist, and the infinite number of prime numbers, which I think do exist. I think that there is more than one mode of existence, but the theory that all modes of existence are equally real — I see no point in that.”

0.36

Scientific and academic institutions suffer from positivism and standardization; diversity in funding and criteria enables better research.

2 pointscentrality 3/5
  • The present system of funding scientific research is terribly perverse and has caused stagnation in many areas, and the career system is perverse in a parallel way, causing people to do the wrong kind of research and driving away those who want to do the right kind; the single-word remedy is diversity — of funding criteria, funding sources, project-selection criteria, and promotion criteria — because any standardization (such as standardized testing) is the opposite of diversity.

    “the present system of funding scientific research is terribly perverse and has caused a stagnation in many areas. The present system of careers is perverse in a parallel way”

  • Professional philosophy took a really bad turn just over a hundred years ago and hasn't recovered, with ripples still visible in postmodernism and 'woke,' and a parallel decline contaminated physics through positivism and instrumentalism.

    “It is definitely the case that philosophy took a really bad turn just over 100 years ago and hasn’t really recovered — professional philosophy, I mean.”