Episode 85: Critical Rationalism and Douglas Hofstadter (Part 1)
- Links to this episode: Spotify / Apple Podcasts
- This transcript was generated with AI using PodcastTranscriptor.
- Unofficial AI-generated transcripts. These may contain mistakes. Please check against the actual podcast.
- Speakers are denoted as color names.
Transcript
[00:00:07] Blue: Hello out there! This is the first of our two -part series where Bruce delves into the work of Douglas Hofstadter, specifically the book, surfaces, and essences. We consider, what is the relationship, if there is any, between critical rationalism and Hofstadter’s idea that analogy is a core mechanism of human cognition? Is it fair to criticize Hofstadter’s ideas as being inductivism in disguise? Could something like what Hofstadter suggests be central to human consciousness and the creation of AGI? I hope you enjoy this.
[00:00:44] Red: Welcome to the Theory of Anything podcast! Hey, Peter! Hello, Bruce! How are you doing today? Good! We’re going to talk about a subject that I’ve mentioned several times I wanted to cover and never got around to it, and it’s Douglas Hofstadter and his theories and some of his books. So I’ve mentioned it in passing, and so I thought maybe it was time to start to dig into it a little and talk about some of his ideas. So you’ve read at least one of his books, right?
[00:01:14] Blue: Yes. My background with him is that I read I Am a Strange Loop, and then in preparation for this podcast, well, there was no way I was going to read that, is it Surfaces and Essences book? But I did listen to several lectures where he summarized his ideas from the book, so I feel somewhat prepared to at least listen to you and ask a couple questions.
[00:01:39] Red: So Surfaces and Essences is not including all the appendices. It is 530 pages long. I noticed the audiobook was 33 hours.
[00:01:52] Blue: It is. I’m guessing that’s 500 fairly dense pages too.
[00:01:57] Red: Yes. And the first time,
[00:01:59] Unknown: so
[00:01:59] Red: let me just admit, I am a Douglas Hofstadter fan. He’s not the most well thought up amongst a lot of crit rats, and so I’m kind of an exception in this case. We’re actually going to talk about that today. However, I’m actually somewhat sympathetic to their negative viewpoint, not sympathetic to the point where I’m actually negative myself, but like I can kind of see where they’re coming from.
[00:02:25] Blue: Well, let me say I have not had the opportunity to discuss his work too much with critical rationalists. I kind of thought that a lot of people read his books through Deutsch because he mentions them a couple of times in the beginning of infinity.
[00:02:46] Unknown: And
[00:02:46] Blue: then if you look at the back, everyone should read these lists. And then below that, there’s a longer list of further reading and both of Hofstadter’s books, Godel, Escher, Bach, and I Am a Strange Loop appear on that list. But looking back into how Hofstadter is referenced in beginning of infinity, it’s not clear to me that he’s actually that much of an influence on David Deutsch. He kind of just takes like a couple of sort of random examples of that Hofstadter uses in his book and seems to sort of riff on that and basically talks about how he disagrees with some of his conclusions. So I’m not really seeing Hofstadter as too much of an influence on David Deutsch. I will admit, like I said, this could be just my personal thing. I found I Am a Strange Loop very boring. It seemed to have a central idea that was quite interesting. And then he just rift on it for like hundreds and hundreds of pages until I was like bored to death of it. And maybe I should cut that. I might be being… No, no, no, no, that’s fine. I’m not qualified to…
[00:04:06] Red: Whatever is your feelings, they’re real,
[00:04:08] Blue: right? I’m just being honest here. So I think that probably maybe the trajectory of this podcast is that you can convince me that Hofstadter is an incredibly interesting thinker that I should be more interested in. So I hope you’ll do that. Okay, I will try.
[00:04:30] Red: Okay. So I’m pretty sure that it was after reading Fabric of Reality, because I don’t think Beginning of Infinity was even out yet. I had seen that Hofstadter was listed as further reading in the back of Fabric of Reality. And I think that’s how I picked it up. Check me on that, because I don’t think he’s even mentioned in Fabric of Reality but his book is listed as one of the further reading books.
[00:04:55] Blue: Oh, yeah, I should have checked that too. Okay.
[00:04:59] Red: Well, guess what? We were both wrong. Douglas Hofstadter is at the back of Fabric of Reality under Everyone Should Read These.
[00:05:07] Blue: Oh, okay. But I was looking at the beginning of Infinity, though. He’s not under the Everyone Should Read These. He’s on the…
[00:05:15] Red: Yeah, he got devoted to the reading.
[00:05:20] Blue: All right. Maybe he is more of an influence then, if it’s on the Everyone Should Read Me list.
[00:05:27] Red: So Everyone Should Read These, Douglas Hofstadter, Godel Escher Bach in Eternal Golden Braid.
[00:05:33] Blue: Okay.
[00:05:33] Red: And in fact, that was the book that I read. I remember reading after I read Fabric of Reality. And I highly recommend that book. That is an excellent, excellent, excellent book. And it’s a published Pulitzer Prize winning book, right? And it’s the book that established his career because it was so famous and popular. You’re
[00:05:51] Blue: talking about Godel Escher Bach.
[00:05:53] Red: Yes. Yeah.
[00:05:54] Blue: Okay.
[00:05:54] Red: Now, here’s the thing, though. So I devoured that book. I went through it. It’s a fun Alice in Wonderland, Lewis Carroll sort of romp that it first seems to be going nowhere in particular, which is one of the problems that you mentioned with Hofstadter is that he at times seems unfocused.
[00:06:11] Blue: So
[00:06:12] Red: the book starts teaching you these fun little mathematical games. And you have no idea why he’s teaching you these mathematical games, right? Because he does not tell you, right? And so you have to kind of accept it for what it is. You have to kind of think each game individually as kind of a fun little thing to learn that may have no real connection to anything else, right? And you go through and then suddenly you reach the climax of the book. It’s a nonfiction book that has a climax. And you suddenly realize that the little games that you’ve been taught throughout that because of these games, you now understand Godel’s proof as well as the church -turing thesis. And it was really reading Godel Escherbach that taught me in particular the church -turing thesis and how powerful it was. And I was able to understand it at a very detailed level because of the fact that I played the little games in the book as I was reading it and then got to the end and went, oh, I get it. And I understood it at a much deeper level than Deutsch explains it in his books. Because he’s not trying to show you the proofs and Hofstadter is actually trying to show you the proofs, trying to teach you what you need to know so that you can make sense of the proofs.
[00:07:37] Blue: I think the reason I read I am a Strange Loop instead of Godel Escherbach is because it was described to me as really an updated version of Godel Escherbach where he felt like people weren’t really understanding what the book was actually about. So he wanted to make it more clear in I am a Strange Loop.
[00:07:56] Red: That sounds right. All the math is gone, all the little games. So he does not teach it to you at anywhere near as deep a level in I am a Strange Loop. Instead, he’s trying to write to a lay audience to try to explain it.
[00:08:10] Blue: So
[00:08:11] Red: he’s got two theories that he is advancing throughout Godel Escherbach. And obviously, one is the church -turing thesis, which Deutsch would completely agree with. And maybe that’s why he ended up in Everyone Should Read These is because a great deal of the book is actually teaching you the church -turing thesis and its significance to AGI.
[00:08:33] Blue: Does he mention Deutsch in the book?
[00:08:35] Red: No, this book is written far before Deutsch. Well,
[00:08:38] Blue: he’d already written papers at least about the church -turing -Dutch thesis, hadn’t he? I don’t know, but I think it’s true. That’s right. That’s really old. Yeah. Deutsch
[00:08:51] Red: was probably a kid in the
[00:08:54] Blue: 80s. Okay.
[00:08:58] Unknown: And
[00:08:58] Red: then the second thing is this idea of the Strange Loop,
[00:09:01] Unknown: which
[00:09:01] Red: Deutsch does not agree with. I’ve seen him tweet that he thinks it’s wrong. It’s this idea of the mind as a recursive mechanism. By the way, I think Hofstetter’s correct. So I think there is some significance in the fact that the mind self -references. Yeah, it seems to make some sense.
[00:09:20] Unknown: It
[00:09:20] Red: does seem to make some sense. Now, Godel’s proof, does that prove the mind self -references? No. It’s more like he’s trying to show you what a recursive proof is like, and then he’s imagining the brain working in a similar sort of way. But the connection is kind of tenuous at best. So it’s really an analogy, which is what we’re going to talk about today.
[00:09:45] Blue: And that’s what the whole services and essences is about. It’s a role of analogy, which is also an intriguing idea, I think.
[00:09:54] Red: It is. And so well, it took me a while to realize that it was an intriguing idea. Okay. So if you don’t understand church -turing thesis, I think Godel Ashabak is one of the best places you can go to learn it,
[00:10:07] Blue: by the way. Okay. Okay.
[00:10:09] Red: All right. So I then read, I’m a strange loop, and my experience with that was kind of similar to yours. Like I felt like it was kind of a warmed over version of Godel Ashabak, and I didn’t find it as interesting. I also read the mind’s eye, which he’s the editor. He wrote some of the articles in it, but it’s several different famous philosophers. And the eye here is not EYE. It’s the letter I, right? The mind’s eye.
[00:10:37] Blue: Okay. Get it? Yeah. Yeah.
[00:10:40] Red: So, and that’s a super interesting book because he’s collected together several different philosophers about the concept of self -consciousness, awareness, having a sense of self, things like that. And that’s where I first read, you know, what is it like to be a bat by Nagel? Like there’s a whole bunch of really famous articles in there that I’m really glad I read that one. So I actually highly recommend the mind’s eye quite a bit. And I think that one’s super entertaining too. I then tried to read fluid concepts and creative analogies, and I bounced off that one. I got about halfway through it before I stopped. I intend to eventually finish reading it, but like I first started reading it decades ago, and I still haven’t finished. So that kind of shows you.
[00:11:28] Blue: I think the issue there is it’s a far more technical book.
[00:11:31] Red: And so I came away with a much stronger understanding of what his theory is, his theory of AGI is, and actually was impressed by some aspects of the theory. But there’s a great deal of, you know, trying to explain his coding projects to implement his ideas. And they feel like a waste of time to me, like just they probably were very cutting edge at the time, but they feel like a complete waste. Like I don’t see how you would ever take these projects that he’s done and turn them into something that would turn into AGI. I think they were kind of a really, really super basic early look at what are analogies and how would you implement them. And it only works for analogies between letters, right? I don’t see any way you could generalize it the way he’s set it up. Obviously that’s because he doesn’t know how to either, right? And this is the whole problem of like if he knew how maybe he’d have solved AGI, right? So maybe it’s not that surprising that he didn’t know how to do this. But the coding was just too simple to be useful here. So then I came across services and essences and man, that’s a long book. So I decided to listen to it and I hated it. Like I thought it was boring, annoying. I forced myself to listen to the whole thing because I’m a big fan of Douglas Hofstadter in general.
[00:12:49] Blue: Yeah.
[00:12:49] Red: And I just thought it was like one of the worst books ever, right? Okay. And I felt like it was just, oh my gosh, I can’t believe he’s still dwelling on the same subject. And I’ve already, he’s already covered this 12 other times, you know? Okay. Okay. Now here’s the thing though. It aged well because as I, once I had listened to it, I spent years and it kept coming back to my mind and I would go, oh, that’s an interesting point. And then I would, you know, years would go by and I’d think, oh, that’s an interesting point. I hadn’t noticed that before. I like, this is just like reviewing it in my mind, right? And I started to realize this is a really good book. And for all that, the repetition in it, like he’s making some really interesting points that really need some further thought. Okay. And so I actually went and bought the physical book, re -listened to it and then highlighted everything that I had found interesting the first time, now that I had had a chance to think about it. And eventually ended up buying the audio book so I could re -listen to it again. So I guess I’d have to say I’m a big fan of the book now, even though I hated it initially. There’s one idea that I did really like from the book the first time I listened to it. And it probably deserves its own podcast and we’ll touch on it briefly, but like maybe we’ll dwell, maybe we’ll hit it more in a future podcast. But it was how words don’t have a set number of definitions that they’re actually fuzzy collections of interconnected categories. Okay.
[00:14:19] Red: We’re going to talk a little bit about that today, but like specifically how that relates to words and therefore relates to essentialism. Okay. The error of essentialism sort of deserves its own podcast. I will touch on it in this podcast, but like, I’m only really in this podcast, and this is going to be an epic life podcast, but I’m really only covering chapter one and part of chapter five in this podcast to, for the kind of the introduction. And then we’ll come back to it later and look at some of the other chapters and what he had to say. Hopefully I can shrink them down and make it simpler than this first podcast. So I mentioned in passing to a couple of crit rats that Hofstadter had some interesting ideas about AGI that needed some more exploration. And to my surprise, the two crit rats I brought this up with immediately responded, that’s a waste of time. It’s inductive. So Popper disproved it already. And they were not open to any further discussion on the subject. We are not discussing him. He’s an inductivist. They were so sure of this, so infallibly sure that they weren’t interested in the topic at all in terms of further discussion. Now I understand induction and its relationship to critical rationalism. And one of the things I understand is that strictly speaking, there is no such thing as induction. So I would agree with them on that. Now it’s true though that there are some things such as statistical inference that we refer to as inductive inference.
[00:15:40] Red: And I think this is one of the things that actually strangely enough, this is really what Hofstadter is talking about as you’ll see that what I’m saying right now fits right into his theory. So you’ve got this problem where induction, strictly speaking is refuted wrong and incorrect. And then you’ve got this reality that we refer to things that actually work like machine learning as induction. And if they actually work, they can’t be refuted. I mean, I don’t know how to get around that. If you’ve got something and it’s based on a false philosophy, then it doesn’t work. That’s how you know it’s a false philosophy, right? And machine learning actually works. So I don’t see these as equivalent like machine learning and statistical inference. I don’t see that as any way equivalent to inductive logic because there’s no such thing. In fact, these inductive examples entirely rely on a simple regular inductive logic, in fact, okay, just like Popper explains. So I couldn’t make sense of their argument that was ignore that it’s inductive. Like that argument did not make sense to me, okay? If there’s no such thing as something being inductive, so if by inductive you mean, if by inductive you mean uses inductive logic. So it’s impossible for something to be inductive in the sense in that sense. So you clearly can’t use this sense of the word inductive to ignore an idea that’s actually implementable. And Hofstadter’s ideas are actually implementable. That was the book Fluid Concepts and Creative Analogies. He was implementing the basic ideas that he was getting at on a computer, okay? So that argument doesn’t make sense to me.
[00:17:28] Red: You can’t take something that works like Hofstadter’s ideas and you can’t dismiss it as, oh, that’s inductive in the sense of inductive logic because inductive logic doesn’t exist. Therefore, his ideas can’t work. So there’s something wrong if that wasn’t what was intended. Now, perhaps these crit rats were referring to the idea that it was inductive in the looser sense of being statistical. My guess is that they were thinking too deeply about this. But let’s assume they were. And so they’re thinking more in the statistical sense. But Hofstadter’s ideas are not statistical. And statistical inference is not at odds with Popper’s epistemology anyhow. And Popper’s very clear about this fact in logic of scientific discovery. So if they were rejecting Hofstadter’s ideas on these grounds, they were factually wrong on two counts. Because of this, I couldn’t make sense of their objection. It’s an objection that seems self -defeating to me. It doesn’t even make sense to dismiss something as I can ignore that idea because it’s inductive if it’s an idea that actually works. Try to talk to them about this. There was this wall I couldn’t break through. To both of them, it was just obvious that Hofstadter’s ideas were inductive and thus disproven and that there was no need to look into them any further. I had assumed back then that my understanding of Popper’s epistemology was the same as other crit -rats. This was my first hint that I was understanding critical rationalism quite differently than other crit -rats, which is obviously one of our past episodes, what was about that, where I tried to summarize my understanding of Popper’s epistemology. I later noticed that even Deutsch was making comments to this effect. Like, I saw him make a tweet.
[00:19:04] Red: I didn’t find the tweet, so I don’t have the exact tweet. But he’d say something to the effect of this new AI technique is inductive so it’s bad epistemology. I do not see any coherent way to make sense of such a statement. Okay, so if we’re talking about a technique that actually works and accomplishes impressive things, then it can’t possibly be based on bad epistemology. On the contrary, it must be using correct epistemology or it wouldn’t work. Now, it’s tempting here to say, okay, Bruce, you’re being pedantic. Deutsch just means that it’s statistical, so it’s not explanatory, and thus it is not a path to AGI. In fact, I have seen Deutsch make other tweets that say something closer to that. And of course, famously, he said something to that in his article that was on Aeon about why can’t we do AGI, that we covered on this podcast. So I really wanted to agree with this sentiment. Okay, but if it’s true, then why is he calling statistics bad epistemology? That doesn’t make sense, right? He’s actually, I’ve seen him use the term bad epistemology. Statistics is integral to how popper’s epistemology works. There are few scientific discoveries that aren’t statistical, at least in part. Popper was well aware of this. Popper realized that his epistemology can’t that if his epistemology can’t accommodate probability theory, then it is his epistemology that is refuted, not probability theory. Here’s the actual quote from Popper, by the way, since some people may not believe me. Ideas involving the theory of probability play a decisive part in modern physics. In investigating this problem, we shall discover what we shall discover that will at first seem an almost insuperable objection to my methodological views.
[00:20:46] Red: For although probability statements play such a vital, important role in empirical science, they turn out to be in principle impervious to strict falsification. This is from page 133 of Logic of Scientific Discovery. He then spends an entire chapter, just painfully, it’s the worst chapter to read, at least, trying to work out a new theory of probability because he thoroughly understands that if he cannot subsume probability theory into his falsificationist theory, that it is his epistemology that is wrong and he’s the one that’s refuted. It’s really important to him that he comes up with this new theory of probability that allows him to work it into this falsificationist epistemology of his. This is really where the propensity theory of probability comes from, which, by the way, I’m not sure I buy the propensity theory of probability, but I understand why he was so interested in coming up with it is because he needed to make sure that statistical theories fit comfortably within his falsificationist epistemology, and it needed to be a realist theory. It couldn’t, it could not be an instrumentals theory. It needed to actually have a real explanation behind it. So Popper was well aware that you can’t use his epistemology refute the science that science, science’s use of probability theory, but you can potentially use science’s use of probability theory to refute his epistemology. So you understood the real weight of the need to show that his epistemology could accommodate science’s use of probability theory. Now this stands in contrast to what many crit rats believe today that I’ve seen.
[00:22:33] Red: I’ve been told at various times that from crit rats that probability theory is unempirical and unscientific, that it’s one of the great myths of our times, that Popper’s epistemology was not about probability theory, but it was about falsification and you can’t falsify probabilistic theories so they are not part of Popper’s epistemology. He had something more narrow in mind. I’ve been, I’ve heard numerous things to this effect and they’re all wrong. Those would actually refute Popper’s epistemology if those were true and Popper knew this. So what is Hofstadter’s theory that crit rats reject so quickly as inductive? Well, let’s get into that now and let’s actually kind of describe, I’m going to read quite a bit from the first chapter to try to explain at a high level what his theory was. Okay. And let’s do our best to steal Manit. There is obviously a lot that’s very speculative here, to be honest. Okay. So here is Hofstadter explaining it in his own words. In 1977, I began my new career as a professor of computer science aiming to specialize in the field of artificial intelligence. My goals were first to uncover the secrets of creativity and second to uncover the secrets of consciousness by modeling both phenomena on a computer. This is from Fluid Analogies, page 13. By the way, a shout out to Melanie Mitchell, even though Douglas Hofstadter kind of gets the most credit for this book, she is co -author of several of the papers inside of Fluid Concepts and Creative Analogies. So and she’s obviously older. I don’t hear as much from him anymore, but I hear things from her all the time.
[00:24:14] Red: We’ve covered her our episode on Microsoft’s LLM study with AGI’s and if they’re early sparks of artificial general intelligence, we read her negative review of that paper as the counterpoint. So I just kind of want to bring her out because she’s kind of a very important figure in this space also. And in some sense, she’s the current torchbearer, not Douglas Hofstadter, because she’s still highly active because she’s obviously she was his student and she’s younger. So Hofstadter then goes on to talk about how pattern, pattern fighting is rooted in analogy making. I don’t like this term analogy making, by the way, I feel like it’s deeply misleading as to when you read the book, you start to get an idea of what he’s really talking about and it doesn’t feel very much like the word analogy is maybe the best word for it. However, that is his word, so I’m going to stick with it. That is to say, we try to find a pattern in one situation that we see as analogous to another situation. So this can be quite mundane. You might understand the pattern of getting a cup of coffee at a coffee shop. No matter which coffee shop you go to, you understand that you get in line, you come up, you look up, make an order. There’s certain things you might add into it, and then you have to wait for it to be brewed. There’s each of those situations, even though they may be happening in different states, in different countries, that you have this analogy between them.
[00:25:39] Red: This is what this is Hofstadter’s term that allows you to kind of understand the essence of the situation and how to behave, even though you’ve never been in this particular coffee shop before. So it might be something less mundane as well, such as realizing that the heart is a kind of pump. Now, prior to that point, the word pump might have been purely mechanical devices and a heart clearly is not mechanical. So it is not analogous to a pump in its original sense as this mechanical device. And yet that insight that the heart is a kind of pump, just not a mechanical pump, a metal pump or whatever, that’s a really interesting, powerful insight and might be a deeper sort of analogy that is getting made that’s actually very useful. So this fits well with Popper’s idea of theory driven observations. We observe according to some sort of analogy that sorry, let me try to restate that. We observe according to some sort of following or deviation from some expected pattern and that pattern is an analogy of some sort to some other situation you’ve experienced in the past. So because this is something that is analogous to something that’s in the past, you know which observations to look for and either things continue to follow your expectations or they don’t and then there’s some sort of deviation and that forces you to reassess the situation. So far, that actually fits Popper’s idea of theory driven observation quite well. In fact, according to Popper, we try to assert patterns on the world by perceiving patterns that don’t exist. This is Popper in conjecture refutations page 60.
[00:27:30] Red: Without waiting passively for repetitions to impress or impose regularities on us, we actively try to impose regularities upon the world. Okay, so again, this still seems like it fits with Hofstadter’s ideas so far. So there is little controversial here to a crit wrap so far. Now here is where things get interesting and controversial. Hofstadter then goes on to say, this is again in fluid concepts and creative
[00:27:58] Blue: analogies page 63.
[00:28:00] Red: Once all of this has been laid out so explicitly, it seems a trivial thing to insert that analogy making lies at the heart of pattern perception and extrapolation. What could be more obvious? And when this banality is put together with my earlier claim that pattern making is the core of intelligence, the implication is clear analogy making lies at the heart of intelligence. This is the part of the theory that the crit rats were strongly reacting to.
[00:28:30] Blue: Okay, I think I’m starting to get my mind around this more. So their criticism is that an analogy is a bit like another word for induction.
[00:28:42] Red: Yes. Now, why would crit rats so strongly reject this? It’s for exactly what you said. Now, an issue here is the fact that it seems or maybe even is very inductive. Okay. Now, maybe this is already obvious. It seems to be obvious to you already, Peter. But if not, I’m going to spend the next hour laying out in more detail, Hofstadter’s arguments, and we will get around to why this has a relationship to inductivism. So page one, in this book without thinking, analogies and concepts about thinking, sorry, in this book about thinking, analogies, concepts and concepts will play the starring role for without concepts, there can be no thought and without analogies, there can be no concepts. This is the thesis that we will develop and support throughout the book. If you’re a crit rat, you probably already had nails in the chalkboard over the word support. By the way, you shouldn’t have nails in the chalkboard over the word support, but I know many people very react to that word. So that was page one of his book. It is easy. This is again from page one. It is easy, in fact, almost universal to underestimate the subtlety and complexity of concepts, all of the more because of the tendency to think of concepts in overly simple terms to is reinforced by dictionaries, which lay out the various different meanings of a given word by dividing the main entry into a number of sub -entries. So he uses the example of a band and he says, you know, a dictionary might look it up and a dictionary might say that it’s like a piece of cloth wrapped around something or it’s a colored stripe or it might be a small group of musicians.
[00:30:18] Red: So you’ve got kind of these three different, maybe initially seeming unrelated uses of the word band. So he says on page two, the dictionary will clearly set out these various concepts, all fairly distinct from each other and all covered by the same word band. And then it will stop as if each of these narrow meanings has been made perfectly clear and were cleanly separable from the others. But in reality, there’s an enormous number of things that we might refer to as a band, he claims. So this might be easier to use his example of the letter A, which is the front of the book, by the way, is like a whole bunch of different letter A’s written in different stylistic fonts to kind of make his point. So let’s say that I asked for you to explain to me what the essence of an A is. So it’s tempting here to say that it’s an upside down V with a line between it. But what Hofstadter points out is that it’s actually easy to find actual A’s used in real life that are stylized that are not upside down V’s with a line between them. So if you’re an essentialist, you might argue that’s not a real A. But anyone who reads the word with that letter A in it, even though it stylized knows plain well that it is in fact an A, and it’s immediately recognizable as an A, even though it isn’t an upside down V with a line between it. This is going back to Hofstadter now, page four.
[00:31:41] Red: One would need many pages of text in complex quasi -legal language if one were trying to pin down just what it is that we recognize in common among the countless thousands of shapes that we effortlessly perceive as a member of the category of A. Something that goes way beyond the simple notion that most people have of the concept A, namely that it consists of two opposite leaning diagonal strokes connected by a horizontal crossbar. Then he goes on to say, any a priori notion that one might have dreamt up of Anus will be contradicted by one or more of these letters, and yet each of them is perfectly recognizable, if not effortlessly so, that when displayed all by itself, then certainly in the context of a word or sentence. Page four. This is the real reason why the philosophy of essentialism is bad philosophy, by the way. This is something that I’m not sure people quite get as to what exactly is wrong with essentialism. So the early Greeks were trying very hard to work out what is the category known as beauty or what is the category known as knowledge or what is the category known as creativity. And no matter how hard they tried to define these categories, what they found was that there was always an exception case refuting their criteria. Now, does this sound familiar? It should. It is not an accident that I spent six episodes playing exactly this game with the Dwight’s constructor theory of knowledge and the two sources hypothesis. What the Greeks got wrong was that they wanted their categories to be watertight and that is not how actual conceptual categories work, at least not every day categories.
[00:33:30] Red: So now an important aside here is that, especially since the advent of science, we have learned to make categories that are far more precise. I could, for example, define prime number in such a way that it would be watertight and precise. It would not be fuzzy. So the claim being made here is not that it’s impossible to make watertight categories. We do that all the time, but that it isn’t the natural way of speaking or thinking or speaking or how humans use terms. What humans actually do might be thought of as something like fuzzy categories, where the boundaries are not watertight. The mistake the essentialists was really then that they wanted regular everyday concepts to be hard categories with distinct sets of attributes that define that category precisely when really they were fuzzy categories. Or in other words, this is just not how human beings actually think for the most part on a day -to -day basis. We do not use hard categories. We use fuzzy categories. There used to be this debate online, probably still is, that is a good example of this. Let’s actually just talk about it quickly. I love video games and I love role -playing games. And so there’s always an ongoing debate somewhere on the internet over whether such and such a game is really a role -playing game or not. And people will debate this endlessly and they’ll get angry with each other and they’ll they’ll have flamores over this. And the simple truth is there is no essence of role -playing game. There are certain attributes that the more of these attributes that you have, the more likely you’re going to call it a role -playing game.
[00:35:14] Red: And the fewer the attributes you have, the less likely you’re going to call it a role -playing game. And everybody has a different breaking point. And they may even weigh these attributes differently. So do you have statistics that show your strength and your dexterity or whatever? Do you interact with people and have conversations? Is it mostly about combat? And certain of these things, the more of these different attributes, the most about combat would actually be a negative one. People would say that makes it less a role -playing game. The more that you have these positive attributes that kind of in people’s mind make something a role -playing game, the more likely you’re going to say, yes, that’s a role -playing game. And the fewer of those and the more you have attributes that pull you away from that, the less likely you’re going to call it a role -playing game. And there’s literally no essence of role -playing game. So the entire conversation is silly. Okay, these flamores are silly because there just is absolutely no essential set of attributes that define a role -playing game. It’s really a fuzzy category. The more you have of it, the more likely that you’re going to say it. You can have some games that have so many of the traditional attributes of a role -playing game that they sit at the center of the essence of role -playing game. And then you’ve got others that sit near the edge. So like is Zelda a role -playing game, right? Most people would probably say no, but it has many, many of the attributes of a role -playing game. And I could give you other examples, but I’m not going to go any further into it. Maybe we’ll,
[00:36:46] Red: I didn’t actually write this one into my notes. I just thought it was an interesting example. These arguments over what is the essence of something, unless the category was intentionally designed to be precise like we do sometimes in science, there’s almost never an essence of the category in everyday human usage. So why do humans by default use fuzzy categories rather than precise categories? Isn’t that a disadvantage? Like we kind of have this idea that there should be an essence of role -playing game. There should be something that is the essence of what is creativity, that there should be something that is the essence of what we call knowledge and that we can define it. We can say knowledge is, you know, a adaptive information that causes itself to remain instantiated. That, that is the essence of knowledge. And I promise you, there is no essence of knowledge.
[00:37:43] Blue: That’s kind of an intriguing way to look at it. That was kind of an aha moment for me right there when you think, you know, I mean, I’ve obviously familiar with the arguments against word essentialism and, you know, what a waste of time that that is. But I never, never really thought of it in terms of it could be a strength of humans that we have that we don’t insist, at least on our own minds, of precise definitions of these words.
[00:38:15] Red: And this is what Hofstadter is going to argue, is that this is a strength. And in fact, if you don’t have this capacity, you can’t be intelligent. Okay. And he even gives an explanation for why it had to be this way. And guess what? It’s a Popperian explanation. I don’t think he knew about Popper. He doesn’t show any influence from Popper. But the explanation he gives is an evolutionary one, which fits very well into Popper’s epistemology. Okay. So fuzzy categories turn out to be a really useful thing to do when we’re allowed to make categories blurry or slip like this. This is the point Hofstadter is going to end up making, that intelligence needs the ability for conceptual slippage to allow us to constantly step from one category or analogy to the next open -endedly. Here is Hofstadter explaining this in his own words from page five. The everyday concept, band, chair, teapot, mess, and letter A are very different from specialized notions such as prime number or DNA. The later have unimaginably many members, just like band, chair, etc. But what is shared by all of their members is expressible precisely and unambiguously. By contrast, in the mental structure underpinning a word like band, chair, mess, or teapot, there lurks a boundless, blurry richness that is completely passed over by dictionaries. Ormney words don’t have just two or three but unlimited number of meanings. In fact, Hofstadter is going to argue that human everyday categories constantly evolve over time through analogy making. And that’s why they must always forever be blurry to be able to work properly. So Hofstadter explores this idea using the linguistic notion of a Zugma.
[00:40:08] Red: So there are sentences that make you smile by intentionally exploiting the conceptual slippage inherent in human concepts. Kurt was and spoke German. The bartender gave me a wink and a drink. In each of these sentences, Kurt was and spoke German, you’re clearly slipping between two uses of the concept or the word German. Okay. And yet you understand exactly what I’m saying, but it makes you kind of smile. This is what a Zugma is. Makes you smile because you can tell that we’re exploiting the fact that this is really two uses of the word German. Okay. And what word is this?
[00:40:44] Blue: Zugma?
[00:40:45] Red: Z -E -U -G -M -A.
[00:40:48] Blue: Okay.
[00:40:50] Red: My favorite.
[00:40:51] Blue: Never, never heard that. Okay.
[00:40:53] Red: My favorite was this one. The book was cloth bound, but unfortunately out of print. That’s almost like a normal sentence, right? This one really makes you stop and think about what you mean by the word book. Here you’re simultaneously using it to mean a literal object called a book. And at the same time, an abstract idea of a book as a group of identical copies of that book. Okay. So you are slipping between two entirely. One is physical and one is abstract concept, and you’re using them both inside the same sentence. The book was cloth bound, but unfortunately out of print. Now let’s say somebody says, I’m translating this book into English. This doesn’t fit either of the two previous concepts. Okay. It’s no longer a physical book that we’re talking about, but it’s not even the abstract idea of a bunch of physical copies of that book. It’s now an entirely different book that’s conceptually the same, but written in a different language that doesn’t even exist yet because that’s because you’re the one translating it. Okay. And all of these fit under the word and also the category book, even though some of them are really hard nouns and some of them are abstract concepts. And some of them are even more abstract concepts. Okay. And yet they all are somehow fit under the concept of book. So he points out that there’s something like a distance in conceptual slippage. I found this really interesting.
[00:42:23] Unknown: Okay.
[00:42:24] Red: Edmund plays basketball and soccer. Is that one concept or two? You could probably argue over this that you could say, well, to play basketball is analogous to playing soccer, but it’s not quite the same thing. So this is really two concepts. It doesn’t feel like it though. It really feels like it’s a single concept to most people. Okay. They would say it’s just one concept. It’s the concept of playing sports. So it doesn’t even seem like a Zugma to us, even though you could argue that it is because they’re two very different kinds of sports. But now consider this sentence. Sylvia plays tennis monopoly and the violin. This one seems very much like a Zugma to us precisely because there is a much larger conceptual distance between playing tennis, tennis monopoly and the violin, even though we immediately recognize them as all in the category of playing something. Okay. Notice that you can literally feel the conceptual distance. You probably even have an opinion of which of these has the most conceptual distance. So I’m going to just ask your opinion. You’ve never heard me explain this before. So which of the two are more alike? Playing tennis and playing monopoly or playing tennis and playing a violin?
[00:43:44] Blue: Playing tennis and playing monopoly or playing monopoly and violin?
[00:43:50] Unknown: No, tennis. Oh,
[00:43:51] Blue: tennis and violin.
[00:43:52] Red: Violin. Yeah. Which of those two is more analogous? Maybe tennis
[00:43:58] Blue: and violin because it’s a little bit more physical, involves your arms moving.
[00:44:03] Unknown: It
[00:44:03] Blue: would
[00:44:03] Red: have said tennis and monopoly. And I suspect most people probably would. But both games.
[00:44:10] Blue: Yes.
[00:44:10] Red: Yes. Okay. Okay. Okay.
[00:44:12] Blue: Yeah. I can see a case for that.
[00:44:14] Red: There’s a case either way. It’s not like this is actually a correct answer.
[00:44:18] Blue: Sure.
[00:44:19] Red: And it probably depends on how you happen to store these concepts in your mind. Right. It’s most people would probably sense a greater distance between tennis and violin than tennis and monopoly because that’s how they happen to have those concepts are stored in their minds. Okay. Well,
[00:44:35] Blue: you know, my wife is a violinist and she just got a repetitive stretch or repetitive like a stress injury through playing violin. So maybe I probably had that in my mind that it’s almost like a sport that you can be injured in. Yeah. Okay. So I
[00:44:51] Red: don’t know maybe where
[00:44:52] Blue: my mind was going.
[00:44:53] Red: Okay. And that is how this works, right? Conceptual distance is this kind of unconscious. What are the analogous points between them? Okay. And as the number of, as the number of attributes becomes more distant, the concept, you sense a greater conceptual distance. Okay. So most people would probably say, I’m guessing, I’ve never done a survey on this, but I would probably say most people would say playing the violin is a further conceptual distance. Now,
[00:45:20] Blue: yeah,
[00:45:20] Red: given that it shouldn’t surprise us that in Italy, there is no word for playing the violin and there is an entirely different word and category for that. And Italians would find it strange that to English speakers, those all fit into one category because they would say, what, what is this playing the violin? What is making music on a violin? How is that in any way analogous to playing tennis? Right. And they would be find it very strange that we would actually try to put those into the same category at all. Okay. This, the point that Hofstadter’s making here is that what items go into the category is in fact, culturally and language dependent. And it depends a lot on how we currently conceive things. So here’s another one that I thought was interesting. The asparagus tips and the potato dumplings were delicious. Is this a Zugma? Do you think? Does it feel like a Zugma?
[00:46:22] Blue: Oh, because they’re both delicious?
[00:46:24] Red: Yeah. It doesn’t really feel like it.
[00:46:26] Blue: Not really.
[00:46:27] Red: Okay. How about this one? The asparagus tips and the after dinner witticisms were delicious.
[00:46:35] Blue: That seems a little weird.
[00:46:37] Red: Okay. That’s a Zugma. Okay. You immediately deeply feel the conceptual distance that you sense in a Zugma, exactly like what you just said. Okay. And then this is, how about this one? The asparagus tips and the expressions of surprise on Anna’s face were delicious. Now you sense it. At least I do sense an even greater conceptual distance. This is all from page 19 of his book.
[00:47:02] Blue: Did you have a comment? Oh, I was just going to say, so okay, so this is all interesting. But I think what, you know, what Carl Popper, one of the things I get from Carl Popper is that these kinds of controversies about words are just sort of a, it’s more of like a banal fact of life that words just mean different things. And there’s these kinds of idiosyncrasies. But it seems like you’re going, Hofstadter is making a case that there’s something deeper here.
[00:47:29] Red: That is, that is the case Hofstadter is making, that there is something deeper here.
[00:47:33] Blue: So he’s kind of departing from something that what Popper emphasizes there. Yes. Okay.
[00:47:39] Red: Okay. But he’s also similar to Popper in some ways. Okay. So an example, the spontaneous categorization, this is page 13, the spontaneous categorizations that are continually made by and in our brains, probably should have said minds, and that are deeply influenced not just by the language we are speaking, but by our era, our culture and our current frame of mind are quite different from the standard image according to which categorization is placing of various entities surrounding us into preexisting and sharply defined mental categories as one sorts items of clothing into different drawers of a chest of drawers. Okay. Now notice the similarity here with the idea of observations being theory impregnated. He’s openly admitting that there’s this giant connection to things like culture, era, current frame of mind. Notice you just talked about how maybe the violin seemed closer to you because your current frame of mind included this idea that the violin could have injuries like a sport. Okay. So we’re under new situations by, when we’re under new situations by observing patterns that we’re already familiar with, we use a theory to determine what to observe. Okay. This is really what Hofstadter saying. He’s saying we’re in some sort of new situation and then we categorize that situation as being similar to some other situation we’re familiar with or set of situations that we’re familiar with, a category of situations that we’re familiar with. And then that then serves like a theory that we use to then observe and say, okay, this is following this, this is different from this and allows us to adjust ourselves to that situation, even though really technically every situation is brand new and completely different than every other situation. Okay.
[00:49:30] Red: So you can see how that would have survival advantages. If this was true, if you’re able to kind of grasp and say, oh, I’m in the following similar situation, or this is similar to some other thing, that would be immediate survival advantages.
[00:49:44] Blue: Okay. So I’m kind of just thinking off the cuff here, but so the severe war of hypothesis is this idea that language really shapes the way people view the world.
[00:50:00] Red: Yes.
[00:50:01] Blue: Is Hofstadter aligning himself with that? Would you say?
[00:50:05] Red: That’s a good question. I would, I don’t think he ever comes out and says that he is. And I can’t think of anything in particular that would seem like he definitively does agree with that. But I do think we are talking about something somewhat similar.
[00:50:22] Blue: But then Chomsky, here’s, here’s just, I’m just going to give you my layman’s understanding of the controversy. So Chomsky might more say that there are, would criticize that as linguistic determinism, but his thing was more like that there were innate, there’s like an innate universal grammar in our brains that goes beyond particular language that we’re using. Whereas I think Deutsch would probably go a completely different direction and see our minds as something like software that’s more malleable.
[00:50:59] Red: Right.
[00:50:59] Blue: And that the language is just kind of like something that’s more on the periphery maybe.
[00:51:06] Red: Yes.
[00:51:06] Blue: Is that on right?
[00:51:08] Red: Yes, it does. And in fact, I have seen a great deal of hostility amongst crit rats, not just to Hofstadter’s theories, but to the severe war hypothesis. Like, like it makes them see red for some reason.
[00:51:19] Blue: Okay. Well, I can kind of see that. I mean.
[00:51:23] Red: I should note that the severe war hypothesis is a really good example of a theory that nobody actually believes is strictly true, but that there’s a great deal of corroborating evidence for. So my guess is, is that it’s a high verisimilitude theory, but is ultimately false.
[00:51:41] Blue: Something like we have, if you have different words, like, you know, different languages have different words for colors. So you’ll see colors in different ways. Like if you only have four words for colors, you’re only going to see four colors or something like that.
[00:51:55] Red: I think the problem with the severe war hypothesis is that obviously you’d never get off the ground if you only thought in terms of words, right? There must be some underlying layer that humans have that allows them to pick up words. But I do think once you pick up language that does allow you to think of concepts that natively you’d be unable to. And there’s even some really strong evidence of this in animal intelligence, like if they teach language to an ape, that ape seems to then gain access to concepts that a non -language speaking ape that testably doesn’t have access to, right? So I feel like it’s almost like there must be some layer that humans naturally have, even if they never learn language. But once they do learn language built on whatever that first layer is, which I’m not even sure mental ease sometimes is referred to, that’s like such a strange term. And I don’t even know if it’s a may even be a misleading term. But once you start picking up language, I know I think in terms of language a lot, right? I don’t have to. Like I can think in other ways, I can picture stuff, I can have an image in my mind that communicates without words. But I know that I do think very heavily in terms of words. So I’m fairly certain that words drastically affect how we think, right? Yeah.
[00:53:23] Blue: Yeah. Well, you know, I had a similar conversation with a friend who was kind of more, I think he was more into more of a postmodernism kind of thing, but he, where he thought that humans were just locked up in like maybe like a prism prison, I should say, of our own language. And even Papa said that. Yeah. Okay. Well, I mean, my pushback was, well, you know, there are things in this world that are going to save your, if your child is sick, I mean, there’s a surgery and they can be saved by a surgery, you know, that you don’t need to express how to do that. And the right, I mean, there’s a thousand different ways. You could even use hand signals maybe to explain how to do that in some ways. Right. You know, it’s not like, there must be things that go beyond language and things that could save the life of your own child. So
[00:54:20] Red: I have a book about this, and it’s very similar to Hofstadter’s theory. In fact, I’m fairly certain Hofstadter may have gotten it from this book. So we’re going to cover it in a future podcast. What they think is the connection between these.
[00:54:33] Blue: Okay. I’m taking this down some tangents. We should get back to Hofstadter. Sorry.
[00:54:37] Red: So it’s actually very relevant because like, how would you ever get analogies off the ground to begin with, right? If they’re all done in terms of words. Hofstadter never claims they’re done in terms of words, by the way, that would be a misunderstanding of Hofstadter’s theory. And yet there’s clearly some sort of connection. Now, here’s the thing that’s interesting though. And this is where I think things, I try to explain this and I feel like I never am understood. There’s always way more concepts than there are words. So it’s sort of makes sense that we would need to use a single word with lots of meanings. If we made a separate word for every concept, first of all, that would be a problem as we’re going to see. It would make it so that we were unintelligent, but it would also just be impossible. There are so many different concepts you would need. If you needed a new word for every single one, it would just be unworkable. Okay. So a single word may actually contain many categories. It is not obvious to me that a musical band and a rubber band are somehow part of a single category of any sort, no matter the conceptual distance. And not all categories have a word for them. So this is where Hofstadter tries to show that words and categories aren’t the same thing. Hofstadter gives the example of an argument over the size of the military budget.
[00:55:57] Red: So if two people are arguing over the size of America’s military budget, you’re going to immediately see one side trot out the need to protect the nation and the need to develop newer technology and the other sides going to immediately trot out the great inefficiencies of the military and other sectors are more important like feeding the hungry first. Okay. Like we have almost stock phrases and grammatical patterns that we immediately slip into depending on which side of the debate we’re on. Now these are Hofstadter argues a very high level category that might be thought of as something like need for a bigger military budget versus need for a smaller military budget. There’s sense them and we use them freely and we slip into them in relevant and the relevant ways of speaking that that category kind of contains within it when the situation calls for it. Okay. He also gives examples of where people find analogies between two things and the analogies are completely useless. Like I don’t have this and I’ll have to probably do it in a future podcast, but like somebody has like two identical cars and then they see two identical dogs and they think of the cars. Like the mind is almost like it’s trying to find analogies and a great number of them are just dead ends. They’re so completely worthless that they leap into the mind and you sense some sort of connection and one thing reminds you of another and then it’s like it’s a completely useless analogy and you let it die, right? And there’s never a name or a word for this category and so it’s completely analogies can be completely outside of words in this sense. Okay.
[00:57:45] Red: So on page 14, what is a category?
[00:57:48] Unknown: A category is a mental structure that is created over time and that evolves sometimes slowly and sometimes quickly and that contains information in an organized form allowing access to it under suitable conditions.
[00:58:01] Red: The act of categorization is in the tentative and graduated gray shaded linking of an entity or a situation to a prior category in one’s mind, not unlike a tentative theory about a situation. Okay. Again, you can see some connection with Popper’s epistemology. Again, on page 14, the tentative and non -black and white nature of categorization is inevitable and yet the act of categorization often feels perfectly definite and absolute to the categorizer. Since many of our most familiar categories seem at first glance to have precise and sharp boundaries and this naive impression is encouraged by the fact that people’s everyday run -of -the -mill use of words is seldom questioned. In fact, every culture constantly almost tacitly reinforces the impression that words are simply automatic labels that come naturally to mind and that belong intrinsically to the thing and entity. Page 14. Continuing on page 14 into page 15, the resulting illusory sense of the near -perfect certainty and clarity of categories gives rise to much confusion about categories and the mental processes that underlie categorization. The idea that category membership always comes in shades of gray rather than just black and white runs strongly against ancient cultural conventions, i.e. the essentialists like Plato and Aristotle, and is therefore disorienting and even disturbing. Accordingly, it gets swept under the rug most of the time. A category pulls together many phenomena in a manner that benefits the creature in whose mind it resides. It allows invisible aspects of objects, actions, and situations to be seen. Aha, this is the whole idea of theory -impregnated observations.
[00:59:48] Red: Categorization gives one the feeling of understanding a situation one is in by providing a clear perspective on it, allowing hidden items and qualities to be detected by virtue of belonging to the category person and entity is known to have a stomach and a sense of humor. Future events allow future events to be anticipated. The glass that my dog’s tail just knocked off the table is going to break, and the consequences of the actions to be foreseen. If I press the G button, the elevator will go to the ground floor. Categorization thus helps one to draw conclusions and to guess about how a situation is likely to evolve. He goes on to claim that nonstop categorization is every bit as dispensable to our survival in the world as is the nonstop beating of our hearts, page 15. So he then says, page 19, there is no clear cut distinction between analogy -making and categorization, since each of them simply makes a connection between two mental entities in order to interpret new situations that we run into by giving us potentially useful points of view on them. Now this is something that Hofstadter spends enormous amounts of time on, trying to show that there’s no clear cut distinction between analogy -making and categorization. To me this seems obvious, so I didn’t have to spend the chapters and chapters he spends trying to convince you of this fact, because I was already well aware that categories were fuzzy, long before I came across Hofstadter. Let me say this, I’m going to use the term analogy, category, and schema interchangeably. I kind of prefer schema. I think the main problem with the word schema is that it doesn’t mean much to most people where analogy category does.
[01:01:34] Red: And I kind of have some issues with the word analogy is because people understand it in a specific way that isn’t really what I think Hofstadter has in mind. And I think so maybe categorization is the best word, it’s not a great word either maybe, but it at least means something immediately to most people. I think the biggest problem with the word category is that people think of categories in terms of something like category theory or set theory where you’re automatically guaranteed that you’re going to make the categories very precise. And so it’s almost more like I have to call it fuzzy categories to specify, I mean something different than the way you would mathematically use the term category. And yet it is the way you probably use the term in real life. Most times when you categorize things, you do not do so in a precise mathematical way. This is like the example of role playing games. You can immediately tell this is a role playing game, have an opinion whether Zelda is a role playing game or not. And yet that category role playing game has no precise boundaries and will never have precise boundaries. And the reason why is very straightforward. It’s because we’re constantly inventing new games that force us to keep reimagining what counts as a role playing game. And so in fact, if you were to go back and look at the history of role playing games, the first pen and paper role playing game that isn’t dragons was actually just a tactical war game and had very few of the elements that today we would think of as a role playing game. And yet it bore the term role playing game.
[01:03:03] Red: And that was where the term came from over time. How you played that is a dragon turned into this thing that involved a lot more quote role playing, even though there was still always this tactical war games element where you would often use miniatures and engage in combat. And it was a very huge part of the original game because it was really a tactical war game. And over time that became more important to the concept of role playing game. And today if the game is only a tactical war game, even though that’s what a role playing game originally was, you probably won’t call it a role playing game today that has slipped out of the umbrella of the category of role playing game. Okay. And you would call that a war game instead. And we’re constantly inventing new concepts that go into this idea of a role playing game when Skyrim comes out or World of Warcraft or something like that. It forces us to have new attributes that we associate with the term role playing game. And that category evolves. It expands. It contracts as games are forgotten. It gets new attributes as new attributes are invented. Okay. And this is why the categories that humans use must be fuzzy categories, must forever be fuzzy categories to be useful. Okay. Because we live in a world that evolves. So our categories must evolve. So page 20. Why do we categories? Thanks to categorization through analogy making, we have the ability to spot similarities and to exploit the similarities in order to deal with the new and the strange.
[01:04:43] Red: By connecting a fresh encountered situation to others long ago encountered encoded and stored in our memory, we are able to make use of our prior experiences to orient ourselves in the present. Consider the idea of a desktop. This comes from later in the book. At one point, the word desktop and the category desktop was a wooden surface of a physical desk or maybe didn’t have to be wood, but the surface of a physical desk. Then Apple and Microsoft created a software desktop. That analogy to a physical desktop instantly gives you access to useful, but sometimes misleading information about how to use that user interface without having to be instructed specifically. And the process, and in the process, the category of desktop expanded to include the user interface of a computer that you interact with, a mouse. And now if we were to talk about the category desktop, that is an integral part of the category desktop today where it didn’t used to be and the category had to expand to include it. And the reason why they used it as an analogy was because it was so useful. Think about the days back in the days of DOS. Maybe this is you’re too young for this, but we did everything through this user interface and we would type little commands. That’s the way we interact with computers. And then windows came out and Apple Macintosh came out and you had this mouse and you could interact with a desktop. And it used to be that you had to read up on which commands to use and memorize them. And it was hard. It was a lot of work.
[01:06:19] Red: Suddenly you could throw your grab on front of a computer and tell her, this is a desktop. And she would immediately start to use the computer and would start to shuffle things around. There’s also an analogy with folders and things like that, right? Which is actually a competing analogy for a desktop. And your grandma suddenly can use a computer when she couldn’t even begin to use DOS because she was given instant repackageable information from her own personal use of a physical desktop that she could now reapply to a brand new situation. That’s why they called it a desktop, right? As they wanted you to be able to grasp the essence of a physical desktop and reapply it to their software so that you didn’t have to be trained in their software. So you utilize your existing knowledge of a physical desktop and you jump right in and you infer what you need to do based on this analogy that they’re offering you, okay? And again, I want to emphasize the analogy is often misleading. And I think we’ve all bumped into this. We have this idea that it’s a desktop and yet we can’t just we can’t get it to do what we want to do because we’re trying to use two hands and you can only use the one mouse or like there’s things that are definitely different about it, right? There was some author that joked that Windows desktop is only a desktop if it was a desktop that was on a plane and it had your other hand tied behind your back, you know. So the analogy’s imperfect, okay? So now moving to page 21, a term that will be useful to us in this context is inference.
[01:07:54] Red: Okay, this is where things get I think get very interesting. We will use the term much more broadly than it is used in the field of artificial intelligence where it is synonymous with formal logical deduction. What we mean by making an inference is simply the introduction of some new mental element into a situation that one is facing. Basically, this means that some facet of a currently active concept is lifted out of dormancy and brought into one’s attention. Okay, so again, notice the similarities to Popper’s idea of observations always being theory impregnated. I keep emphasizing that, but there’s like a huge tie between Hofstetter’s theories and that aspect of Popper’s epistemology. Whether this new element is right or wrong is not the point. Notice how this matches the idea of tentative theory making, conjectures. Nor does it matter whether it follows logically from prior elements. Notice how this is basically the same as saying that it’s a conjecture. For us, inference will simply mean the fact that some new element has been activated in our mind. So he uses examples. He says observation, a crying child, inference, the child is in distress, observation, the table is set, inference, a meal will soon be served, observation. You see a dog, so you infer that it can bark, it bites, it has a heart and two lungs, etc. And then he says it is not merely for idle fun that one calls a cat -like thing that one encounters cat, thereby assigning it to a pre -existing category in one’s mind. It is principally because doing so gives one access to a great deal of extra information, such as the likely fact that it will show pleasure by purring,
[01:09:34] Blue: page 21.
[01:09:36] Red: Okay, now I can imagine a crit -rat pushing back here, maybe rightly so, that say okay, but are these really categories? Or are they explanations? I have an explanation that a crying child, that a child cries when it’s in distress. So of course I make that inference. I have an explanation that a person who wants to eat has to set the table. So of course I infer that from the table being set. And I have an explanation as to what dogs are, what their lineal genes are in the space of animals. And so of course by that explanation, I know that likely a dog is going to bark and bite and have a heart in two lungs. So I could foresee a crit -rat pushing back here and saying, look, you’re not really talking about categories. You’re really talking about explanations. And so this is where Hofstadter is wrong. He’s not really categories we’re talking about, it’s explanations. Now in a past podcast, I gave Hun Rosling’s story of seeing a baby on its back and flipping it to its front to save it from the poor ignorant mother that didn’t realize that if you don’t do this, the baby will choke on its vomit just like an unconscious soldier does. In fact, this inference is entirely incorrect. Babies are more likely to die if you put them on their stomach. Despite this being a false inference, the medical establishment refused to change their mind about putting babies on their stomachs until well after there was overwhelming empirical evidence that they were wrong. This is all part of what he talks about in his book, Factfulness. As it turns out, babies aren’t analogous to unconscious soldiers.
[01:11:24] Red: A baby’s reflexes work even when asleep while unconscious soldiers do not. The reason the medical establishment struggled to accept that they were wrong about this inference was that it didn’t feel like a conjecture. It felt like a logical inference. Unconscious soldiers are the category of unconscious person, and so is a sleeping baby. But while this is a valid category, it is the wrong category for the inference you want to make. For the purposes of the inference you want to make, these need to be thought of as two separate categories, unconscious people with no reflexes, unconscious people with reflexes. Note how this is all just a fancy way of saying that we assign the situation to categories, but that it is only a conjecture. Also note that it’s a little difficult to take this example and fit it into the idea that we’re talking merely about explanations. It starts to blur the lines between the concept of a category, as Hofstadter uses it, and the Apparian -Deutschian concept of an explanation. Literally, you could say that you have two explanations here, and they’re really just categories. In one case, you’re trying to use the explanation that unconscious people choke on their vomit, which is a wrong explanation and a wrong category. And in the other case, you have to realize that there’s an actual difference between an unconscious soldier and an unconscious baby, and that they really kind of belong in two different categories. Here, things are getting so blurred between explanations and categories, I’m not sure I can tell the difference apart anymore. Now, at this point, Peter, let me kind of just stop and get some feedback from you on what we’ve talked about so far.
[01:13:13] Red: So far, I have been intentionally steel -manning Hofstadter’s theories to sound as paparian as I can possibly make them sound. Let me admit that that’s what I’m doing. So why did my crit -rat friends reject Hofstadter’s theory? Well, it’s because it feels very inductive. That word feels maybe is a bias, but let me actually say that they’re totally right. It is inductive. So I’ve tried to point out several similarities to Popper’s epistemology to make it as appealing to paparians as possible. I actually think there’s a lot there that paparians should find very interesting. But the crit -rats are right. This is totally inductive. Page 28, he in fact starts to really hint at this. The same happens when a new situation reminds you of another situation or a family of situations that you previously encountered, and this on the surface, and this while on the surface totally different, but that share an abstract essence with the new one. This is now starting to sound quite inductive, isn’t it, that we generalize from specific circumstances to generalities. Hofstadter gives two interesting examples of this that I want to dwell on a bit. So the first is a male professor that I’m going to probably get some of the details a little bit wrong, so my apologies to Hofstadter, but I’ll get it close enough for that it won’t make a difference the stuff I get wrong. There’s a male professor that meets with a female reporter for publication for his university. The professor ends up being very attracted to this reporter. Nothing comes of it. Turns out she’s actually married, but they become good friends. They get along very well.
[01:15:01] Red: The next time an opportunity comes along to meet with a female writer for publication for his university, he jumps at the chance to do it. Now, even this professor that’s involved in this story could tell you that there’s something a bit silly about this. There is just no reason to believe that just because the situation was somewhat analogous with the previous situation, that somehow that means he’s going to be attracted to this new female writer that the professor married to. I didn’t he doesn’t say he’s married. It doesn’t say one way or the other. I think he was single and she was married. Here’s the thing though. This is a very human thing that he’s doing, and we’ve all experienced this in real life. Okay.
[01:15:53] Blue: Sure.
[01:15:54] Red: It is as if the analogy forces itself upon us. This is what Hofstadter is getting at. Okay. Furthermore, it’s an analogy of one. Right? I mean, like he’s only had one similar experience in the past. It’s not like he’s had a whole bunch of repetitions of meeting with female reporters and finding them attractive. It has happened once, and yet the analogy is already forcing itself upon him. Okay. And he’s anxious to meet with this new female writer. I don’t think it was the second one was even a reporter. It was just it was enough of an analogy that he could not escape it. There’s no way you could call this an explanation. Okay. Like there is no explanation here at all. All right. In fact, if anything, you would say it’s a completely bad explanation just because you were once attracted to a woman that was a writer that wrote for your university. The fact this other woman who’s a writer that wrote for university, you’re going to be attracted to her too. It doesn’t even make sense. Right. And yet we’ve all experienced the force of analogies like this in our lives. Okay. Now the second analogy is even more interesting. So you have an author and his book has been translated into another language. One that he has some familiarity with. So he reads parts of the translation. I don’t think he’s actually fluent in the language, but he’s able to read some paragraphs out of the translation and he finds that he’s very disappointed in the translations. These translations are in European languages, by the way. Okay.
[01:17:29] Red: And so he’s very disappointed with the translation and he has a very bad feeling about the fact that his book was translated into the other languages in such a poor way and it’s going to reflect badly on him. So when later the book has a chance to be translated into Korean, a language that he does not know, he can’t bring himself to do it despite overt assurances from the publisher that this time it will be different. We’ve got top notch translators. He has no explanation for why he can’t just bring himself to agree. And in the end, this costs him dearly when he refuses to do the translation. Okay. There is simply no way for him to know if he has done the right thing or not. And again, we’ve all experienced situations like this. This is so much the human experience. It is like the analogy, and this is a quote now, exerts nearly irresistible pressure on us. Page 305, 306. Unlike our previous examples, there’s just no explanation possible here. So page 306, Hofstadter says, what would the readers do in this man’s shoes? On the one hand, there was strong pressure coming from the magazine’s upper echelons to publish the book in Korean, which would have been good for his reputation. On the other hand, there was an analogy whose relevance was not in the least clear. What possible link could the disappointing performance of a few European translators have with the performance of unknown translators in a country that was so distant geographically, linguistically, and culturally? Hofstadter points out that it would be unthinkable that the editor would say, your concerns are merited because they were not merited.
[01:19:07] Red: There really was no merit to his concern in terms of any actual explanation that we can offer. This is page 307 now. It’s as if you were telling me you would never drive a Korean car after having test -driven only cars built in Eastern Europe. This is what he imagines the editor saying to this person who’s trying to object. There is simply no good explanation for why the analogy was such a strong overriding part of how he and we reason. What then, this is a quote from Hofstadter, what then was the bedrock set of beliefs on which this author thought it was reasonable to ground his analogy? What if there had been four different bad translations? I actually think there was. I think there was like four bit different bad translations that were in Europe, but they’re all in Europe. Okay, so it’s unanalogous in that it’s a totally different language. What if it had been a thousand? What if it had included at least one bad translation in Korea? At what point does this author get the right to claim that his concerns are warranted because now they are a correct extrapolation?
[01:20:22] Blue: Okay.
[01:20:24] Red: These antidotes, this is Hofstadter again, these antidotes are cases of induction, a form of thought in which one extrapolates to a new situation, certain observations that one made in one or more prior situations. So Hofstadter himself calls it induction. Induction has no claim on logical validity in the sense that no rigorous rule of reasoning allows a person to arrive at an inductive conclusion with absolute certainty, pages 307 to 308. On to page 308, of course, the fact that a given conclusion is not derivable through watertight logical reasoning doesn’t mean that the conclusion has to be false. It’s perfectly possible that the Korean translation will be miserable or that the professor will be smitten beyond belief when he meets the new writer. In short, the likelihood of a conclusion should not be confused with the logical validity. And indeed, what usually matters in everyday life is how likely something is rather than how logically deducible it is. What matters overwhelmingly to these men is simply the question of likelihood. If this isn’t like totally making the crit rats roll over in their graves right now, I don’t know what would. And then Hofstadter goes on and argues that this is correct. He says, if on some grade A grade A, you suddenly decide to reject every inductive conclusion you reach simply because it wasn’t the result of an ironclad form of reasoning, then your thinking would have to grind to a halt because every thought that everyone has, no matter how tiny it is, no matter how spontaneous or mindless it may seem, it is an outcome of this kind of mental activity that has no logical validity.
[01:22:08] Blue: I was going to tell a story that, see if we can fit this into your framework, Hofstadter’s framework that the professor made me think about. And I’ll cut it out if I or my wife wants me to. So I’m going back 20 years now when I met my wife and I was about 23 or so, just gotten out of college. And this was in the very early days of internet dating, right? And I knew, I really, I logged on to these sites and including Friendster, which you might recall, were you ever on Friendster, Bruce?
[01:22:53] Unknown: No.
[01:22:54] Blue: This was before MySpace. So we’re talking about the original social media thing. And I realized fairly quickly that on these dating sites, they’re primitive by our current dating sites that I could get actually sometimes two dates a day. You meet for coffee and then a drink later or someone else later at night. I mean, I went on about 50 dates. Many of them were quite bad. And I could have concluded in, I mean, actually, a few of them were bad, really, but they were, they were, most of them were fun. And we had nice, nice chats, but not like there wasn’t a lot of romantic interest in a lot of them. But I knew I had an explanation that said, if I keep at this, I will meet the right person, which is exactly what happened. And we’ve been married over 20 years, happily married, not over 20 years, almost 20 years, we’ve been happily married. So, you know, this could be a situation where inductively, this was just the wrong way to think of it. And I have plenty of friends who are single and they’ve sworn off internet dating because they think, oh, I’ve been on so many bad dates, it’s gone horribly. And, you know, maybe I need to get into some epistemology with them. And no, you just have to think about the explanation. There’s someone out there. And if you go on enough dates, you’ll find them. You
[01:24:31] Red: know what? I think that’s a perfect, perfect for what I was about to say. So, Hofstadter is admitting this is induction, so we can dismiss it, right? Except that we just pointed out that this really is how human beings think.
[01:24:51] Blue: Okay.
[01:24:52] Red: And you just admitted this, right, that there’s many people who go on these dates and they’re bad and then they inductively decide internet dating is awful and they stop. And there’s somebody else who’s maybe a little smarter who says, you know what, I have an explanation that tells me I should ignore creating an inductive category like this.
[01:25:13] Blue: And
[01:25:14] Red: so I’m going to override my tendency to want to stop and want to say, oh, they’re all just bad or whatever, right? Okay. And you can see that human beings really do think inductively exactly like this.
[01:25:28] Blue: Yeah.
[01:25:29] Red: Okay. And we’ve all felt it and we’ve all experienced it, which brings us to the center of the problem that I’m trying to raise. If human beings really do induce things, like Hofstadter is saying, and we’re basically agreeing that they do, then it is Popper’s epistemology that is refuted, not the theory of induction. Okay. Or at least that would be the case. I’m going to argue against this now. I mean, I’m trying to set this up.
[01:25:57] Blue: Okay.
[01:25:58] Red: But let me just point out that this is why Popper has taken so much criticism over the idea that his theory leaves a whiff of induction. Okay.
[01:26:09] Blue: Criticism from critical rationalists who take things a little further is that what you’re saying?
[01:26:15] Red: So obviously he’s taken a lot of criticism from inductivists. So maybe we don’t count them.
[01:26:20] Blue: Okay.
[01:26:20] Red: He’s taken criticism from his own students over this. And I’m going to give you some examples of this. Okay.
[01:26:25] Blue: Okay.
[01:26:26] Red: So these are all from the Popperian podcast, which is excellent podcast where he often interviews students of Popper. So Anthony Oh here in episode 14 of the Popperian podcast, he claims Popper was trying to reconstruct science without using any kind of induction. And then he goes on to say, and this seems to me to be just plain wrong. He goes on to say that we do reason via induction, but that he agrees with Popper that it offers no certainty or justification. And then this is a quote, Popper tries to explain it away. I think he says he tries to explain it away, but anyhow, it’s more or less a quote. So Anthony then claims Popper tries to get rid of induction and that it bumps that every time he tries to get rid of induction, it just bumps up somewhere else like it just did for us in this example. Okay. Now Anthony wasn’t one of the students of Popper, but even some of Popper’s own students felt this way. I’ve actually heard more than one student of Popper on the Popperian podcast mentioned the fact that Popper failed, there’s still a whiff of induction poppers failed to get rid of it. I couldn’t find all of them, but I did remember one of them was Joseph Agassi. So I looked that one up and that’s episode 18. And here is some actual quotes. The context of the discussion is that Agassi is saying that Popper should have defended inductivism before attacking it, like to steelman it. Okay. So you know how Popper like often steelmans the case before he attacks it? He doesn’t do this to induction.
[01:28:00] Red: And Agassi felt that that was a mistake on Popper’s fault, Popper’s part, that he should have thoroughly steelman induction and said lots of positive things about it before attacking it and showing that it had problems. Okay. And the interviewer at this point says when Popper talked about corroboration, he’s dipping into inductivism, whiffs of inductivism. And Agassi responds, he said that it is a whiff of inductivism. Yes. So Agassi is saying here not merely that he personally feels it’s a whiff of inductivism, but that Popper felt it was a whiff of inductivism still in his epistemology and that he didn’t get rid of it all the way because of corroboration. Okay. This is around nine minutes at episode 18. By the way, this is obviously Agassi’s interpretation of Popper. I’m not sure I could point to a specific quote. Okay. But you can see how Agassi has is bringing out, even though he’s a student of Popper, that there is this understanding that there’s still a whiff of inductivism that hasn’t been entirely removed through Popper’s epistemology. Now, Agassi does not stop there. He says, if there is a whiff of inductivism, then it isn’t just conjectures and refutations. It is conjectures, corroborations and refutations. And he, Popper, said, corroboration is necessary to encourage research. Now, I found this in total discord with the whole of what Popper did, which is not, I cannot pronounce this term, psychologistic. What Agassi is getting at here is that Popper was very strong on there’s no point in having a theory that is purely psychological when you’re talking about epistemology and yet Popper would ask why we need corroborations, at least according to Agassi. He tried to use a psychological explanation. Well, they’re necessary to encourage research.
[01:30:02] Red: Now, I’ve pointed out in our epistemology episode that I really feel like Popper goes much, much further than that, explaining why we need corroboration, that it exists not just for psychological reasons. However, I do agree with Agassi here that Popper does at times dip into psychological explanations because there was always a little discomfort over how does corroboration fit in. Or could it be that human beings do not induce things and that that is itself the mistake? Okay. And I’ve had many crit rats and David Deutsch talks about this, I think quite a bit. There is in fact, I think this is actually Popper’s actual view. He believed that all induction was really conjecture refutation, at least for science or variation selection for ideas outside of science. You might recognize this as being a form of Campbell’s evolutionary epistemology of blind variation and selective retention. But then I have to ask, did these men in our examples make a series of conjectures and refute them? Or did the analogy force itself on them? Did they try out several variations first and then reason about them? For the professor that was attracted to the journalist, what are the conjectures and refutations that caused him to induce his theory that the next woman writer might be attractive to him too? What about the writer that was worried about the new translator? What are the conjectures and refutations process that he followed to arrive at that inductive idea? It is not readily apparent how to explain this kind of induction that Hofstadter is referring to in terms of conjectures and refutations. This is why even some of Popper’s own pupils like Agassi felt that Popper was unable to banish induction entirely.
[01:31:54] Red: Now, on the Popperian podcast, episode 15, Jogdish Hattigadian, totally butchered that name, I’m sure. He knew Popper personally and even says he took me under his wing. I don’t know if he was officially a student of Popper or not, but he is clearly someone who Popper was a mentor for him. And the interviewer says Kuhn and Popper didn’t give induction any validity whatsoever. And Jogdish response, Karl Popper and Thomas Kuhn did not think that induction in any form is a valid method to follow. The former argued energetically against induction of all kinds, and he did so with great effect. That’s Popper he’s referring to. He goes on to say that Popper was defensive. Oh, skip that part. We’re not going to do that in this episode. He, Jogdish goes on to claim that Popper misunderstood bacon and that Baconian induction was a more meaningful theory than Popper gave it credit for. Because he had this, what he felt was a common misunderstanding of bacon that was going around academia at the time, but that if you actually go read Bacon, you find something different there. Now, you all know how much I love interesting problems. I’ve talked about this problem or a form of it as far back as episode 26, where I mentioned the problem that you can’t cast all inductive machine learning algorithms easily into variation and selection like Donald Campbell thought either. This is another form of the same problem, at least on the surface. Does this mean Popper did not refute induction? I don’t think so. On the contrary, I think Popper is correct that induction, strictly speaking, does not exist. Though there are real things we call induction that do exist.
[01:33:48] Red: And of course, this is the case of it depends on what you mean. And that’s almost always true. In fact, it’s almost always true for the reasons Hofstadter is explaining that the category induction is itself a fuzzy category and can mean different things in different circumstances. There are kinds of things that we call induction like machine learning that are real and actually work. And there’s kinds of things that we call induction like the famous Baconian induction and logical induction that are not real and do not work and do not exist. And they all get kind of mixed up into one kind of category called induction in our minds that’s actually a fuzzy category, just like Hofstadter is saying. And a great deal of the confusion around induction and Popper and the relationship between those two comes from the fact that we intermix these terms, these concepts that are kind of all part of this big fuzzy category, when really we need to think about them a little more precisely to make sense of them. Okay. So then how would you cast Hofstadter’s examples in Popperian epistemology? If it is the case that there is really no such thing as induction, strictly speaking, and if it is the case that you could cast it in terms of Popperian epistemology, how would you do that? That’s the question. Interestingly, Popperians do not seem to know how to do this in any real detail from what I’ve been able to tell talking to them. There seems to be a couple of opinions on how to do this. So one of them is that there is no such thing as induction and I hate induction and induction is all wrong.
[01:35:30] Red: Anything that looks like induction is misleading point of view. And if this is true, then why reject Hofstadter’s theory then as clearly it cannot be induction because it’s something that human beings actually do. The other point of view seems to be the induction’s a real thing, but it has nothing to do with epistemology point of view. This is turned them out to saying induction is really a second epistemology.
[01:35:57] Blue: So these are two divisions within critical rationalism, you would say that I’ve come across. So
[01:36:03] Red: let me explain why the inductivism is not epistemology doesn’t work. So when it comes to epistemology, we must follow the Highlander principle, which is there can be only one. This is a logical necessity. Suppose there are in fact two epistemologies. So I’m going to assume contrary to the Highlander principle that there are two true epistemologies. That is to say there are two ways to learn about the world and discover true patterns or generalizations. Suppose Popper’s epistemology was one of these ways and induction was the other that you by wish you could find true patterns about the world. Then Popper’s epistemology would be wrong in that it ignores induction. It would be partially it would be right to a degree, but it would be partially wrong. Induction would be wrong for the same reason because it ignores the importance of conjecture and refutation and explanations. The correct epistemology would start like this. There are two ways in which we learn about the world, induction and conjecture and refutation. And here is how they relate and interact. So in other words, if we had induction as a separate category from Popper’s epistemology, and that was really the case, like I’ve had some crit rats argue to me. If that were really the case, then the correct epistemology would be the combination of the two. They would not be separated anymore. So there can’t ever be two true epistemologies. If there were, then we’d be able to form a new epistemology that combines the two and that would be the correct epistemology. So therefore I’m going to argue that the correct answer to this is that Popper’s epistemology, his evolutionary epistemology, if not his scientific epistemology, those are a little bit different, must entirely subsume Hofstadter’s theory.
[01:37:54] Red: For that matter, it must entirely subsume the entire concept of induction. There must be a Popperian explanation for how human beings manage to induce from specifics to generalizations. What is it? Interestingly, the answer I think already exists or almost already exists, a big part of it already exists, but I did not find it in philosophy books. I actually found it in Tom Mitchell’s machine learning textbook, which is an older machine learning textbook that I had to study as part of my master’s degree in machine learning. And it’s a book that is different than the way they write them today because he really is trying to figure out the epistemology of machine learning. I don’t think he even knew what the term epistemology meant, so he never uses that term. But he’s very concentrated on pack learning, guarantees, how do you guarantee that this can learn, things like that. Probably pack learning remembers, probably approximately correct from Wesley Valiant. And he’s actually trying to work out what I would say is the epistemology of machine learning. And he comes up with all sorts of really interesting answers. And it’s way outside the bounds of this podcast episode to get into that. But let me just say that it’s going to have to be a discussion for another time. In fact, it would probably not be a single podcast. It would probably be like a class on machine learning where I would have to take you through and show you the connections between machine learning and poppers epistemology that exist and how those actually enlighten these problems in philosophy and give potential answers to these problems in philosophy. For example, how it shows,
[01:39:44] Red: Tom Mitchell’s book actually goes to some lengths to explain that induction is really just a deductive process that is based around conjecture refutation, although he doesn’t use those terms. And does so in examples that I would never have guessed were really secretly conjectures and refutations.
[01:40:06] Blue: That does sound intriguing, Bruce. I mean, it seems like that’s kind of your contribution to the philosophy or whatever idea. The world of ideas is you’ve got one foot in machine learning and one foot in philosophy. And it seems to be a really, really intriguing place to be. You know, I think that if you’re interested in poppering epistemology, you should learn machine learning probably from Tom Mitchell’s book.
[01:40:35] Red: I don’t think that guarantees you, but I have seen so much that I’ve learned about poppers epistemology due to my knowledge of machine learning.
[01:40:44] Blue: And I have
[01:40:45] Red: found it super enlightening, right? Where I’ll suddenly go, oh, no, wait, I remember. And my knowledge of epistemology when I read Tom Mitchell’s book, I can understand things that he doesn’t understand. He doesn’t realize he’s answering important philosophical questions. He’s got no clue he’s doing that. But because I’m familiar with epistemology, I’ll go, oh, that was an answer to a really interesting philosophical question. And he doesn’t even know it’s there, right? And I feel like that the two worlds need to talk more, that they have a much deeper relationship than people realize.
[01:41:20] Blue: Well, what was that you said? The Aristotle is the first AGI researcher?
[01:41:25] Unknown: Yes.
[01:41:25] Blue: Yeah. I thought that was quite interesting idea.
[01:41:29] Red: And there is this interesting intertwining between these. So let me just, we’re actually to the end of the first part of talking about Hofstadter’s theories. Let me say this. Yes, Hofstadter’s theories are inductive, but we shouldn’t be afraid of the word inductive. We shouldn’t even be afraid of the concept of induction. Human beings really do reason from specifics to generalities. We do do that, right? And it’s fine to call that induction. It’s just a word. The question is, how do we do that? Popper’s epistemology is incredibly enlightening on how human beings do that, but he never fully answered the question. Obviously, part of the reason why, and you’ve kind of heard me mention this before, is that we apparently do something like Hofstadter. Whether he’s got this exactly right or not, I don’t know. I don’t want to overly offer up Hofstadter’s theories. I feel like they’re still very vague, but I feel like that he’s correct that we somehow use analogies to make conjectures. And those conjectures feel like logical inferences to us. And we do infer things from them. Right or wrong, we do. Okay. And that much is all true. And it’s difficult to say that the problem is, is that this is somewhat of an answer, albeit a very vague answer and maybe even incorrect answer, partially correct, partially incorrect answer, of how human beings make conjectures. This is, I’ve mentioned, we need a universal conjecture engine. This is the only book I’ve ever read that makes a sincere attempt to explain what the universal conjecture engine is. And yes, he casts it in terms of induction. And it is in fact, in some sense, inductive. And so I can understand why crit rats kind of try to avoid it.
[01:43:34] Red: And yet this is the issue in a nutshell. If we do make conjectures, the conjectures we come up with, we then try to refute in science. But how do we come up with the conjectures in the first place? Does that itself come out of some sort of conjecture refutation process? And if it does, is that then an infinite regress? And that’s the problem that needs to be solved. Right. How do we actually come up with these conjectures in a way that doesn’t just create an infinite regress of conjectures and refutations? And the fact that Popper never solved that problem, and he did not solve that problem, right? He kind of left conjectures as just a mystery is why his theory has a whip of induction still and why it has not been removed yet. I am of the opinion that it can be removed and it will be removed. And that that is one of the advances in epistemology that we still need to make. And that machine learning already has a lot of the answers available to us, that we just need to re take a look at and then use and add into epistemology and vice versa, that you can share and learn from each other. And you’ll end up with a theory that is an actual improvement on poppers that removes the whiff of induction that is currently there. That’s my point of view. Or to put it a different way, I fall into a third category. If there’s one category that says everything if an induction is wrong and you should always avoid it because it’s worthless. And if the other one is no, induction is a real thing, but it has nothing to do with epistemology.
[01:45:19] Red: My answer is induction and poppers epistemology are the same thing. And poppers epistemology subsumes induction and induction is no longer needed because poppers epistemology is better than induction. Or at least it has the potential to be that way, even if we haven’t solved all the problems yet. That’s my opinion. And that’s a third category.
[01:45:43] Blue: Extremely intriguing take that I’ve not heard anywhere else. And I’m very interested to explore that more. Well, I have, as usual, your take on Hofstadter is also intriguing. You may have failed in that I don’t know if I’m ready to run out and read surfaces and essences. But I will say, I might read actually Gotel Escher Bach. I like how you said that it explained the church -turing thesis in a very, very coherent and understandable way. And I think I might benefit from that.
[01:46:27] Red: I’ve recommended on this podcast in the past, if you really want to understand Deutch’s theories, you have to read Penrose.
[01:46:33] Blue: Penrose, yeah.
[01:46:35] Red: Reading Penrose is a tough go because he throws the math in there.
[01:46:39] Blue: That’s why I’m afraid.
[01:46:40] Red: Which is why he’s such a great read though. You come up with a much stronger understanding.
[01:46:45] Blue: I was reading Gotel Escher Bach and Penrose around the same time.
[01:46:50] Red: And I don’t think I would have had a chance against Penrose, but for the fact that Penrose is using Gotel’s theorem just like Hofstadter is, but for a totally different purpose. And they’re at odds with each other. One’s using it to show you can make an AGI and one’s using it to show you can’t make an AGI. Having an understanding of like, because Hofstadter actually explains Gotel’s theorem and you don’t actually seriously get it. When I read Penrose, I was ready for him. I understood Gotel’s theorem to a deep degree, at least at the time. I don’t know if I could repeat it right like today, but at the time I actually understood it. And I was able to detect the mistakes that Penrose was making because of that. And also, Penrose uses a lot with the Church Turing thesis. He actually named the Turing principle. He named it the Turing thesis, but Deutsch gets it from Penrose. And again, having read Hofstadter, I was able to detect the mistakes in Penrose’s reasoning because I already knew the math and I already knew the proof and things like that. So I actually do think that these theories, these books all do kind of belong in the same category. It probably makes sense to read Deutsch, then to read Gotel Asher Bach, then to read Penrose. And that’s probably the right order so that you’re in a position to kind of understand what Penrose is getting at, but to also detect his mistakes.
[01:48:26] Blue: So thank you, Bruce.
[01:48:28] Red: You’re welcome.
[01:48:28] Blue: I appreciate everything you’ve said here. Thank you.
[01:48:32] Red: All right. Thanks. The theory of anything podcast could use your help. We have a small but loyal audience and we’d like to get the word out about the podcast to others so others can enjoy it as well. To the best of our knowledge, we’re the only podcast that covers all four strands of David Deutsch’s philosophy as well as other interesting subjects. If you’re enjoying this podcast, please give us a five star rating on Apple podcasts. This can usually be done right inside your podcast player, or you can google the theory of anything podcast Apple or something like that. Some players have their own rating system and giving us a five star rating on any rating system would be helpful. If you enjoy a particular episode, please consider tweeting about us or linking to us on Facebook or other social media to help get the word out. If you are interested in financially supporting the podcast, we have two ways to do that. The first is via our podcast host site, Anchor. Just go to anchor.fm slash four dash strands f o u r dash s t r a n d s. There’s a support button available that allows you to do reoccurring donations. If you want to make a one time donation, go to our blog, which is four strands.org. There is a donation button there that uses PayPal. Thank you.
Links to this episode: Spotify / Apple Podcasts
Generated with AI using PodcastTranscriptor. Unofficial AI-generated transcripts. These may contain mistakes; please verify against the actual podcast.