What this covers
Matjaž Leonardis joins Dwarkesh Patel to challenge the notion that "science" and "the scientist" are useful categories for understanding how we learn about nature. The conversation centers on three linked arguments: that the identity of scientist is a recent invention with no bearing on the actual practice of investigation; that philosophers from Popper onward have cast doubt on whether a unified scientific method exists; and that becoming a polymath or solving difficult problems works better by pursuing genuine interest than by consciously applying a prescribed method. The discussion draws on Leonardis's work with David Deutsch on the Popper-Miller Theorem, a technical result about probabilistic reasoning that challenges whether evidence can confer inductive support in the way Bayesian updating assumes.
The episode ranges across intellectual history, philosophy of science, and practical advice for learners. Leonardis traces how the label "scientist" emerged only in the nineteenth century at Cambridge, yet people understood nature effectively for centuries before—suggesting the identity is not necessary to the activity. He argues that Feyerabend and even Popper himself, despite professing scientific method, doubted such a unified method exists. On learning and polymathy, he contends that self-conscious, systematic study creates artificial barriers; people are naturally interested in everything until trained out of it. The conversation also covers why textbooks lose their original problem-context, how people actually seek explanatory theories over merely probable ones, and why institutional support for intellectual work matters even if "science" is the wrong conceptual umbrella. Throughout, the emphasis falls on following problems where they lead rather than asking whether one is "really" doing science.
Leonardis argues that reifying 'science' and 'the scientist' as a special identity with a distinct method is counterproductive, and that understanding nature, learning, and polymathy all proceed better by following problems where they lead rather than self-consciously applying a prescribed method.
- The label 'scientist' is a 19th-century invention; people understood nature well before it existed
- Philosophers like Popper and Feyerabend doubted that a unified 'scientific method' exists
- Learning and becoming a polymath work better by pursuing interest than by systematic self-directed study
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.
The term 'scientist' is an early 19th-century invention traced to two professors at Trinity College, Cambridge, yet people were able to do science well before that, showing the identity is not necessary to the activity.
“the name scientist is actually um an early 19th century invention um it was traced back to i think two um two to professors at i think trinity college in cambridge um and it's uh and you know people were able to do science before that um quite well”
The most important thing for a young person seeking to contribute is to connect with other people or groups already doing something, where they will find something to contribute to and gain support; there is no such thing as an objectively 'unsolved problem' list, because value is only to a subsection of people who may not know in advance what they want.
“the most important thing is to connect um with other people or groups of people that are doing something because there you will... find something that you can contribute to”
The Popper-Miller result poses an interesting challenge to using Bayesian reasoning to build an AGI or for related purposes, because it questions whether probabilistic updating captures inductive support.
“i think they you know present kind of a very interesting challenge to this idea of of um you know sort of kind of trying to use bayesian reasoning to to either build an agi or or for various other purposes”
Karl Popper, despite being a professor of scientific method at the LSE, said the subject he professed does not exist, and Feyerabend wrote a book 'against method', so prominent philosophers of science dispute that a privileged scientific method exists.
“he was i think a professor of scientific method at the london school of economics... and he says in that lecture that that he is uh is a professor of of a subject that uh that he thinks does not exist um um his student firearm and uh i think uh actually wrote a book against method”
Explanatory theories enable progress in a way mere accumulation of experience cannot: by formulating universal theories, going out into the world, and finding where they fail, one replaces them with better ones, so knowledge increases and truths accumulate that are not present in experience alone.
“it's only if you kind of create this universal explanatory theories and you go out there um you know you will eventually find that something about it doesn't quite work so you will replace it... your kind of knowledge of the world is increasing um in a way it wouldn't if you just accumulated experiences”
The Popper-Miller theorem (Nature, 1983) argues that the probabilistic support a piece of evidence confers on a theory cannot be understood as the evidence conferring inductive support, because by splitting a theory relative to the evidence into a deductive part and an inductive part, the evidence always decreases the probability of the inductive part.
“they basically identified two propositions... this first proposition is a kind of a the deductive part of the theory relative to the evidence and this other part of the theory is uh the inductive part... and they showed that um the evidence always decreases the probability of this inductive part”
The constraints obeyed by a measure of a theory's logical content are exactly the same as the probability calculus, so a probability assignment can always be consistently reinterpreted as a statement of a theory's degree of logical weakness—saying one theory is more probable than another can mean it says less.
“the the constraints or the rules it has to obey are exactly the same as as the probability calculus... if i say that this theory is more likely than that theory i can say this theory says less than this theory”
What people actually want and believe in are not contentless-but-likely tautology-like theories, but highly informative, explanatory, and therefore less probable theories—Popper became skeptical of probabilistic approaches to knowledge for this reason.
“what people actually want what they believe in what they seek are not kind of these sort of tautology like contentless but likely theories as it were but kind of these incredibly informative... explanatory and therefore kind of less likely theories”
Becoming self-consciously aware of and classifying what one is doing can be counterproductive, paralleling the phenomenon of self-consciousness known to cause problems in many other domains.
“becoming you know becoming conscious of what one is doing um and... classifying it or telling yourself the story about it um is is one thing that is known you know in several other fields to sometimes lead to problems um you know hence the term self-consciousness”
The activity of understanding nature need not be conceived as unified: like entrepreneurs building different companies without anyone asking whether each is 'really entrepreneuring', researchers each pursuing distinct topics (liquid helium, star swarm evolution) can be institutionally supported without classifying them all as doing one thing called 'science'.
“people have this idea of an entrepreneur right um well... nobody nobody nobody is kind of asking the question well you know is any one of these people actually entrepreneuring”
The idea that there is a specific scientific method that one ought to use—rather than following where the logic of the problem itself leads—can stop people from doing sensible things that would help them pursue their problem.
“there is a specific method that that one ought to use to think about this particular question rather than just you know where were sort of the the question naturally leads you and i think that that um uh this attitude or this idea can often stop people from doing sensible things”
Becoming a polymath is not something you learn to be but something you unlearn to be: people are naturally interested in everything, and self-conscious, systematic attempts to learn—studying 'fundamentals, then intermediate, then advanced' levels—are misconceptions that lead in counterproductive directions.
“polymath is not something you learn to be it's something you unlearn to be”
Hume's problem of induction shows there is no way to proceed logically (by reason) from experience to the celebrated general theories of the Enlightenment such as Newton's, undercutting the Enlightenment idea that reason was the route to understanding nature.
“this is kind of the famous problem of induction you know there is no way to kind of proceed logically i.e by reason from experience to you know the the the celebrated you know general theories of the enlightenment such as newton's theory”
The conjunction fallacy—shown in Kahneman's Linda experiment where people judge 'banker and feminist' more likely than 'banker' alone—demonstrates that people seek explanatory theories that fit the given information rather than the most probable option.
“there is something known as the conjunction fallacy... the sort of case of linda... is it more likely that linda is a a banker or b a banker and a feminist and a lot of people uh pick option b even though um the probability calculus kind of tells you that... option a has to be more likely”
Everything in a textbook was originally created by someone trying to do or understand something in a specific context; once that context disappears only the textbook remains, leaving learners with seemingly irrelevant information—so reading history helps because it restores the original problem context.
“every single thing you see in a textbook wasn't written to be put in a textbook... it was created by somebody who was trying to do something or understand something... but what often happens is that then that context goes away and all that remains is a textbook”
Popper's psychological argument for the role of universal theories is that people have a need for regularity—placed in an environment without regularity they go mad or invent regularity out of thin air—though why this need exists is itself unclear.
“people seem to have this need for regularity um and... if you put them in an environment um without regularity uh people kind of go mad or... invent regularity out of thin air”
While certain patterns of thought seem more conducive to understanding nature than others, which patterns work and how to think about them remains more of a great mystery than an exact science.
“there are certain i guess patterns of thought that are more conducive to under that seem to at least be more conducive to understanding um the natural world than others um but i think that... all of these things are kind of um more of a great mystery than an exact science”
One can reach a better understanding of the natural world without ever becoming conscious that one is doing anything special, since people have always investigated nature without a special social role for it.
“it does seem to be the case that you can reach a better understanding of the natural world without it”
The narrative attributing the Enlightenment's success to a change in method may itself be a fiction; alternative explanations such as economic progress enabling more thought are difficult to rule out, and figuring out causality in such historical matters is very hard.
“there is perhaps yet another third explanation where... maybe it was just you know economic progress that that allowed for you know like uh all this thought to be produced at the time like it's very very difficult to sort of figure out exactly how causality works with with most of these things”
It is unclear whether one is ever using a 'method' at all, because thoughts constantly arise in the mind and it is a stretch to attribute all such thoughts to a method rather than to their simply arising.
“it is very unclear if one is ever is ever using a method um because... thoughts are constantly arising in the mind right and it's it's it often feels like a bit of a stretch to say that you know all these different thoughts one is having um are arising as a result of a method um as opposed to just just arising”
Excessive identification with 'science' and the question of whether one is 'really a scientist' has a counterproductive effect on what people actually do, because it makes them wonder what they should do in their capacity as scientists rather than just pursue the problem.
“people identify with science too much they they wonder whether what they are doing is science... and i think that often has a counterproductive effect on on on basically what they do”
The activity of understanding the natural world works better when institutionally and culturally supported—keeping relevant books and people in one place and organizing events—so universities are worth supporting even if the unified concept of 'science' is not the right way to understand the activity.
“there is an activity that is sort of worthy of of of being supported um but but whether that sort of particular way of looking at that activity is is the right one um is is less clear”
Connecting young people to mentors solving the right problems is a chicken-and-egg challenge—you can't contribute without knowing their world, yet can't join the conversation without something to contribute—but it is not impossible; the world is not hostile to it, there is tremendous goodwill toward young people, and the main barrier is simply lack of awareness that this is the key thing to do.
“it's definitely not the case that that the world is hostile to it i think people are very sympathetic to attempts to do that... there's tremendous goodwill for young people out there”
In German and Slovenian the word for probability translates roughly as 'believability', and the believability of explanatory theories is always higher than that of less likely theories—so believability isn't really probability.
“if you were translate the word for probability into english it would be something like believability... the believability of explanatory theories is always higher than the believability of of these sort of less likely theories and therefore believability isn't really probability”
The newest blog post is titled on whether prediction markets or policies should be legal, based on Robin Hanson's idea of a futarchy.
“The newest post i have out is titled if prediction markets or policies were legal based on robin hansen's idea of a futarki”