Episode 112: Words vs Concepts: Does ‘Randomness’ Exist?

• Links to this episode: Spotify / Apple Podcasts

Unofficial, reviewed transcript. It may still contain mistakes; check the podcast when wording matters.

Speakers are labeled by first name where identified. Unidentified-speaker labels may refer to different people. Bracketed question marks indicate uncertain wording.

Transcript

[00:00:00]  Peter: Hello out there. This week on The Theory of Anything Podcast, Bruce examines the assertions about probability and randomness in the critical rationalist community. Is there any such thing as true randomness? What does the many worlds interpretation suggest? And is this really a conflict between words and concepts or something more fundamental? Bruce's thoughts on this helped me to understand a little more, the discussion on probability Bruce had with David Deutsch on episode 100 of our podcast. And I hope someone out there gets as much out of this as I did.

[00:00:45]  Bruce: Welcome to The Theory of Anything Podcast. Hey, Peter.

[00:00:48]  Peter: Hello, Bruce.

[00:00:49]  Bruce: We've got some video today because you don't need it, but we've got some optional visuals for this episode that I think will make things a little more clear.

[00:00:59]  Unidentified speaker: Okay.

[00:01:00]  Bruce: They don't show anything that we aren't gonna exactly talk about, but for some reason, I think it just makes it more clear to actually see a graphical version of it. I whipped up some visuals, so you're gonna see us right at the beginning, you're gonna see our faces, and then don't worry, we get rid of ourselves and we go to the visuals after that, and we'll turn off our video. Here's my little presentation today, Concepts versus Words, also known as, Why is Deutsch so Insistent that Randomness Doesn't Exist? We're going to be talking mostly about this idea of Concepts versus Words, but we're gonna also talk a bit about probability. It's something, like, I'm asking the question, why is Deutsch so Insistent that Randomness doesn't exist? And I had a chance to ask him about that when we had him on the show, and I still don't feel like I understand the answer to that question. And I'm going to use this idea of Concepts versus Words to help explain why it seems so confusing to me, what he keeps saying about probability. But also, I'm going to be throwing in there, what I see is a lot of confusion between Concepts and Words and just my own thoughts on it. Not very necessarily well thought out thoughts, but just something that's been bouncing around in my head. I've been wanting to say for a while, I talk about this to people and they don't seem to understand what I'm talking about. So I wanted to draw a picture so I can point to people, to a podcast episode and say, listen, I'm trying to say this.

[00:02:32]  Unidentified speaker: Okay.

[00:02:33]  Peter: Well, this will be a good one for me because I have a, I'm not sure I completely follow the debate either. I have a lot of questions and it's even your conversation with Deutsch on our podcast as many times as I edited that, it seemed to lose me at times, I will say.

[00:02:55]  Bruce: But I had a specific question I was asking him, and I kept re-asking it, and I kept feeling like he wasn't answering the question. And I'm sure he felt like he was answering the question. So I'm going to have to explain in this episode in detail what my actual question was. Mind you, I've had a long more chance to refine my thoughts on this. So my thoughts weren't nearly so refined back when I was doing the episode with David Deutsch. So, I mean, it's a little bit unfair. I mean, like, I'm utilizing what I've been able to figure out since then to make myself more clear in a way that just wasn't available to me back then. So I'll acknowledge that up front.

[00:03:39]  Peter: Well, it seemed like what I got is that he he kept coming back to probability doesn't exist in the real world because of the multiverse. Yes, and that, you know, he wants to favor explanations above beliefs. Yes, and beliefs are kind of like probability. Whereas what I hear you saying is more, well, probability is still a useful thing in the world.

[00:04:13]  Bruce: So, there's two levels to this. And I'm only going to cover the most basic level today. There's the first question of why is he saying randomness does not exist? He actually was insistent he meant randomness did not exist or stochasticity does not exist, not necessarily probability does not exist, which a completely fair question is, what's the difference? And I'm a little unclear on what the difference is. Like, what does he mean? Okay, then there's the question of whatever he means, do I agree with him? I'm at the level of I don't know what he means. So when you say he thinks this and I think that, you might be right. Like, you might have grasped his point of view and my point of view, and you may have actually grasped, and this isn't unthinkable, that you have figured out what the difference between his and my views are. I'm so unclear as to what he's getting at. I'm not even sure if I have a well-developed view. This episode is going to be mostly about concepts versus words and why this has led to why I feel so confused by what he's saying. I want to harken back to our episodes 85 and 86, which was the Douglas Hofstadter episodes, where Hofstadter, we were talking about his book, Surfaces and Essences, and he was trying to explain this idea that I called fuzzy categories, okay, and how words correspond to a fuzzy category, not to a specific category, at least in most cases, not always. So this is the Hofstadter example. He had this idea of mother. So we've got this fuzzy category called mother.

[00:06:00]  Bruce: And what he claims is that there's like a halo that exists around a concept like mother that extends out by analogies. So you've got the most basic idea of mother. Timmy or Tommy or whatever his name was from the book, he would refer to mommy, and mommy is a specific person. And for Tommy at the right age, mommy literally just means my mommy, and that's it. There is no other concept of mommy, which today, as adults, we might say he's got the concept of biological mother, but only for himself, okay? Then Tommy discovers that other people have mommies too, and then he has to extend his concept of mother, and the halo grows, that it's any biological mother, not just my biological mother. And then from there, it just keeps growing. You've got this idea of adopted mothers. It's not even a biological mother, yet somehow it's like a mother. And so that becomes part of the mother, but notice how it's a little further away from the center of the halo now, okay? Then you've got like the stepmother, okay, or a surrogate mother. That's also even now further away. It's still part of the concept of mother, but it's not the center of the halo anymore, okay? Then you've got this idea of a dog having a mother, okay? Well, a dog is not a person. A dog is an animal. Wait, animals have mothers too? Well, obviously, what we mean here is a female parent, okay? So we can see how a human biological mother has some sort of analogy to the biological parent of an animal. We don't always refer to animals as having mothers. Because sometimes we reserve that term more centrally for humans.

[00:07:50]  Bruce: But then if we say this is the dog's mother, we know exactly what we mean, okay? Because there's this analogy that we can immediately look to. But notice that I'm putting it a little bit further out on the halo, okay? And now this idea of a female parent comes with, and I've got these words, they're grayed out, gives birth and protection. So no, we're not talking about different definitions of mother anymore. We're talking about a stereotype of mother, that mothers give birth, but not always, because they might be an adopted mother. And mothers, they protect us, okay? These are kind of idealized stereotypes of mothers. That stereotype is part of the halo, okay? And it becomes a part of how we understand this fuzzy concept of mother. Now, by using those, we might have this idea of a point of origin, which is now not even a person we're talking about, motherland, okay? It's the place that gave my birth, and it's, you know, a place of protection or whatever, right? So you can see why we might use the word motherland, and that still is a type of mother, even though now we're not even talking about parents, biological or otherwise. It relies on this stereotype of mother for us to take the meaning, okay? And then from there, we might have part of the stereotype of mother being nurture. So we have this source of nurture, and mother earth is a type of mother, okay? Still a part of the concept of mother, but way further out on the halo. We're no longer near the center of the halo, okay? But, and yet people know exactly what we're talking about.

[00:09:33]  Bruce: I say mother earth and I don't even have to explain it to people because they immediately understand the connections. Okay, likewise, this idea of a point of origin based on gives birth and protection and things like that might lead to things like Marie Curie as the mother of radioactivity, meaning a female inventor. You don't often hear of a male inventor called the mother. You might sometimes, but you've got this idea of a female inventor is the mother of something, right, whatever they invented. And then you've got this saying, the American Revolution is the mother of the French Revolution. People know exactly what that means, even though we're like way at the edge of the halo now. Okay, now that is Hofstadter's example. Now, one of the things that I have argued, this is from now Episode 91, is that this idea that a word points to a fuzzy concept that actually is splitable into multiple sub-concepts causes no end of confusion to people. So let's take induction as an example. Again, listen to Episode 91 if you wanna know what I'm talking about here. Okay, because I went into detail on this. So we've got this concept of induction. And at the center of the halo is to generalize using specific observations in some way, okay? Now when people first tried to understand this concept, they, philosophers in particular, had this idea that you generalize only using specific observations, okay? So somehow we're taking specific observations and we are generalizing into a general theory from that, okay?

[00:11:18]  Bruce: And then based on that, you've got this idea that we then use it to justify knowledge. This becomes part of the halo of, quote, induction, okay? And then we've got this, so based on that, we got this idea that it creates knowledge. Induction creates knowledge in some way. It's the source of scientific knowledge, okay? And then we've got this idea that it's a new form of inductive logic. You've got deductive logic where you can, with surety, move from if A, then B. Given A, therefore, I know B. And now you're doing this kind of inverse inductive logic. I know B, therefore, with some level of probability, I can justify A, or something along those lines. The various things that people have tried to experiment with, with this idea of inductive logic. Okay, and then we've got this idea of learning [from?] repetition. So part of the theory of induction was that you see patterns over and over, and then it repeats, and then you go, oh, I'm going to use that pattern to generalize. Okay, and so this then becomes part of the halo of induction. And then you got this idea, based on this idea of justifying knowledge, of how does it justify knowledge? Well, it does it through the senses. And so this idea of empiricism and induction became, in people's minds, kind of one thing, part of the same halo.

[00:12:38]  Unidentified speaker: Okay.

[00:12:40]  Bruce: And then based on this idea of inductive logic, we've got this thing called probability theory. And well, isn't that what we mean by inductive logic?

[00:12:48]  Unidentified speaker: People say.

[00:12:49]  Bruce: And so probability theory is a form of inductive logic or somehow associated with inductive logic. And then probability theory gives birth to machine learning. So then it's like, well, machine learning, that is inductive, it's induction.

[00:13:03]  Unidentified speaker: Okay.

[00:13:04]  Bruce: So you've got this built up over time, this idea of induction that actually has this giant halo.

[00:13:11]  Unidentified speaker: Many of these are sub-concepts.

[00:13:12]  Bruce: They're not necessarily the same thing, right? But they've somehow in people's minds become associated together.

[00:13:19]  Unidentified speaker: Okay.

[00:13:19]  Unidentified speaker: Now here's the problem.

[00:13:21]  Bruce: This one right here generalize only using specific observations.

[00:13:27]  Unidentified speaker: That one's just wrong.

[00:13:28]  Bruce: Like it's literally impossible to generalize using only specific observations. And it's not that hard to make a logical proof of that. In fact, in episode 91, we went over Popper's logical proof of that, although he's not the only one who's proven that.

[00:13:42]  Unidentified speaker: Okay.

[00:13:42]  Bruce: I showed Tom Mitchell proving the same thing.

[00:13:44]  Unidentified speaker: Okay.

[00:13:45]  Unidentified speaker: A machine learning expert.

[00:13:47]  Bruce: So once we realize that that one is false, and Popper proves it false, let's say.

[00:13:53]  Unidentified speaker: Okay.

[00:13:53]  Bruce: He refutes induction by proving it false. Then suddenly justifies knowledge and justification through senses and a new form of inductive logic. Those are all now false.

[00:14:05]  Unidentified speaker: Okay.

[00:14:05]  Bruce: Because they were based on this concept that's actually false. That's part of the halo of induction, but it's not the whole of the halo of induction.

[00:14:14]  Unidentified speaker: Okay.

[00:14:15]  Bruce: So then based on that, we suddenly realize that learning repetition and creates knowledge. Well, those are real things, of course, like you could learn from repetition.

[00:14:26]  Unidentified speaker: Right.

[00:14:26]  Bruce: But it's not really working the way the philosophers thought it did. And it's not really that strongly at all associated with what we were calling induction.

[00:14:35]  Unidentified speaker: Okay.

[00:14:36]  Bruce: And in the meantime, we've got this probability theory and machine learning, which are supposedly induction because they're based on inductive logic.

[00:14:45]  Unidentified speaker: And well, those are totally real.

[00:14:47]  Bruce: Like those are absolutely real forms of quote induction, but they've got nothing to do with inductive logic. So they're in a totally different part of the halo than we thought they were.

[00:14:58]  Unidentified speaker: Okay.

[00:14:59]  Bruce: Now, this is actually the state of induction today, okay, as a fuzzy concept. So, it's not too surprising that you're going to hear Deutsch, and in Episode 91, I actually quote his tweets where he says this. He says, machine learning is not induction.

[00:15:14]  Unidentified speaker: Okay.

[00:15:15]  Bruce: And presumably what he means, like I'm not a mind reader, but presumably what he means is that it is not a form of inductive logic. And that's absolutely true.

[00:15:23]  Unidentified speaker: Okay.

[00:15:24]  Bruce: And in the meantime, you hear me saying, machine learning is induction. Of course, what I really mean is machine learning is a kind of generalizing from using specific observations.

[00:15:35]  Unidentified speaker: Okay.

[00:15:36]  Bruce: So we've got this conflict, or apparent conflict rather, between me saying as a critical rationalist saying, yeah, machine learning is a form of induction, and Deutsch saying machine learning is not a form of induction. So what people really want to know is, is machine learning really a form of induction?

[00:15:57]  Unidentified speaker: And they really want to know.

[00:15:59]  Bruce: This is the question people want to answer to, okay? And a word holy war erupts over this. And millions die over word holy wars. And yes, I mean literally millions die over word holy wars. For example, there was massive holy wars over whether the Trinity should be described as homoousios or homoiousios, I can't even pronounce these Greek terms, right? But they're literally led to wars over this, trying to argue over these words and what they meant and how they should be applied to the Trinity doctrine.

[00:16:37]  Peter: Yeah, that's actually a good point that, you know, I thought you were joking at first, but I think you could probably look at history and find millions of people for or through not much more than arguing about words, for sure. Yeah, that's why philosophy to me is so actually so consequential if you look into history. I mean, it really has real world ramifications.

[00:17:02]  Unidentified speaker: Yes.

[00:17:03]  Unidentified speaker: Okay.

[00:17:04]  Unidentified speaker: And the crit rat, war on words has begun.

[00:17:07]  Bruce: Don't you dare ever say evidence for or even insufficient argument, because they're going to jump down your throat and then human sacrifice, dogs and cats living together, mass hysteria because you used the phrase evidence for and Popper proved there was no such thing as evidence for.

[00:17:25]  Peter: Well, luckily just fought with mean words on Twitter and not actual death.

[00:17:31]  Bruce: But so, you know, and I wish I was kidding here, but really, I'm not even exaggerating, right? Like people get really hung up on what words you use. OK, here's the thing. The question, so is machine learning really induction is literally a meaningless question. It is a totally completely meaningless question because it presupposes we should care what label we slap on a concept.

[00:17:58]  Unidentified speaker: So let's go to Popper here.

[00:18:00]  Bruce: Popper says, this is in Open Society and its Enemies Vol. 2, page 14, while we may say that the essentialist interpretation reads a definition normally, that's in scare quotes, that is to say from left to right, we can say that a definition, as it is normally used in modern science, must be read back to front or from right to left, for it starts with a defining formula and asks for a short label for it. I want to say that I think Popper is completely spot on here. I would even go further than what he's saying here. This isn't merely true of modern science. This is literally what definitions always are, even just outside of science. All definitions are just labels slapped on to concepts as a short label, right?

[00:18:49]  Unidentified speaker: That's it.

[00:18:50]  Bruce: That's all they ever are. They never are anything but that.

[00:18:53]  Unidentified speaker: Okay.

[00:18:55]  Bruce: So now getting back to our induction example, there are clearly, clearly, clearly multiple concepts of induction going on here. At a minimum, there's more than these two, but at a minimum, there's this idea of generalizing with use of observations in some way, which is a very real concept, actually exists in real life. And then there's this idea of some sort of inductive logic that justifies theories through use of observations, which does not exist. It's literally just false.

[00:19:26]  Unidentified speaker: Okay.

[00:19:27]  Bruce: There are several more nuanced concepts possible here, but let's just stick with these two for the moment to keep it simple.

[00:19:33]  Unidentified speaker: Okay.

[00:19:34]  Bruce: Some of these concepts are valid and correct. Some of them are complete nonsense. It makes sense that over time, the concepts all got confused and commingled.

[00:19:42]  Unidentified speaker: Okay.

[00:19:43]  Bruce: And so it's understandable that people are really kind of struggling to tease the concepts back out because in their minds, induction somehow is the single concept. And instead of seeing it as a sort of catch fuzzy term for a fuzzy category, they're trying to make it mean a single essential concept that's not a fuzzy category but is a non-fuzzy category.

[00:20:05]  Unidentified speaker: Okay.

[00:20:06]  Bruce: And that just isn't how human language and words and concepts work.

[00:20:11]  Unidentified speaker: Okay.

[00:20:13]  Bruce: So what you should really be interested in is the concepts, not the words, not the definitions. What's the underlying concept?

[00:20:21]  Unidentified speaker: Okay.

[00:20:22]  Bruce: So let me give an example here. So first of all, let me just say that Deutsch is totally correct that machine learning is not induction. And I am also totally correct that machine learning is induction. But just under different definitions of induction, okay, which I don't understand why this is a problem, right? Of course, words have multiple definitions, and a person can be right under one definition and wrong under a different definition. So let's take this word induction. There's always more concepts out there than there are words, okay? Words are far more parsimonious than concepts. Hofstadter points out that for every word, you look up at a dictionary, it's got like five definitions. There's probably actually thousands of slightly different definitions, and they're just trying to catch the most used ones that are kind of got their own little halos, and it's just a way of trying to help convey what's really a far more complex idea, okay? So we can see this as the word induction might point to a concept generalized using specific observations, in which case induction is real, okay? Or we might be using the word induction to point to a different concept of inductive logic, in which case induction is false and does not exist, and it is a myth, okay? Both of these can be true at once, as long as you understand that it's just a word, and it's the concepts we care about. What we should be asking is not, is machine learning induction? That's a stupid question, okay? We should be asking, does machine learning generalize using specifics and how does it do it?

[00:22:05]  Unidentified speaker: Okay?

[00:22:05]  Unidentified speaker: And what are the limitations or constraints around that?

[00:22:09]  Bruce: Or we might ask the question, doesn't machine learning use inductive logic?

[00:22:13]  Unidentified speaker: Okay?

[00:22:14]  Bruce: If you get more specific, if you concentrate and you use just a few more words and don't argue over the word induction and instead argue over the concept, okay? You'll find that you can make progress much further, okay? And much quicker and much better and a lot of confusion will disappear. So let's take free will as an example. So we've got this idea of free will and I'm putting the up on the screen, I'm putting the part of the word cloud or the halo up there. So free will kind of near the center, not too close to the center of the halo, but close enough to the center of the halo. We've got this idea that you make a choice based on your own preferences versus being coerced into it, okay? Did you sign this contract of your own free will and choice? And that's a very straightforward understanding of the concept of free will, okay? But then you've got this other concept that's also kind of near the center of the halo, at least when we start, that free will is somehow this thing that's neither deterministic nor stochastic, and it overrides physics in some way. So if you were to roll back time and you were to do things over, then you would be able to just on your own accord, using your free will, be able to make a different choice, and you're unimpacted by what came in the past. And this is an important concept supposedly by people who believe in it, that it determines that you deserve your punishment. It's the basis for just deserts. Now I've crossed out several of these. I left the one about signing a contract of your own free will and choice.

[00:23:58]  Bruce: That's a very real, very existent form of free will. Okay, this other one that we're talking about, overriding physics, it's neither deterministic nor stochastic, and it's the basis for deserving judgment and having just deserts. All of that's been disproven at this point, at least assuming that our current best theories hold. We have to always assume that I mean that. Those literally are now impossible. They're completely at odds with our current best theories. So those have all been effectively proven false. They're false under the current theories, is the way you'd probably should technically say it.

[00:24:32]  Unidentified speaker: Okay.

[00:24:33]  Bruce: So then you're going to have people, and they're going to start arguing. So Sam Harris is going to say, there is no such thing as free will, and he's adamant about it.

[00:24:43]  Unidentified speaker: Okay.

[00:24:43]  Bruce: And then you have Daniel Dennett, and he is saying, yes, there is such thing as free will, and he is equally adamant about it.

[00:24:53]  Unidentified speaker: Okay.

[00:24:53]  Bruce: And you're looking at it, and you're going, well, is there free will or isn't there free will?

[00:24:58]  Unidentified speaker: Okay.

[00:24:59]  Bruce: And because you're trying to force all these different sub-concepts as if they're a single concept, and they aren't.

[00:25:05]  Unidentified speaker: Okay.

[00:25:06]  Bruce: And then because a word can change over time, it's possible that the sub-concepts of free will might change over time. So we have this idea amongst the crit rat community, the Deutschian crit rats, that free will has something to do with this idea that you've got lots of degrees of freedom in your choices, and you can even invent new choices due to your creativity. You'll hear Deutschian crit rats say that's what free will really is.

[00:25:37]  Unidentified speaker: Okay.

[00:25:37]  Bruce: Well, you know, that concept, that's a really interesting concept. And I can see why you might slap the label free will on it, even though it's clearly not the same concept as the original concepts we were talking about.

[00:25:52]  Unidentified speaker: Okay.

[00:25:53]  Bruce: Now, so David Deutsch is effectively making up his own entirely new definition of free will. And there's nothing necessarily wrong with that. Okay. Words have to change over time. They evolve over time. And Popper, there's even a Popperian history of this, where Popper talks about, just use the word, use an old word to point to a new concept and ignore the old concept, right?

[00:26:19]  Unidentified speaker: And there's maybe even a basis for doing this.

[00:26:21]  Bruce: Because, I mean, did we really expect David Deutsch to make up a totally new term? Or is he going to have to just use the closest term that exists in the English language? Now, here's the problem, though. How long will it take for the followers of David Deutsch, or maybe even David Deutsch himself, to very quickly start to confuse this new concept that he is labeling free will and the old concepts? Like, for example, I've seen crit rats who will first say, free will is this ability to have degrees of freedom and use your creativity to make new options. And therefore, you deserve punishment. Well, those have got nothing to do with each other. You're now commingling two really unrelated concepts that both happen to go by the term free will, one of which is false and one of which is true and actually exists. And you're trying to somehow grab a part of the halo, the part that's been disproven, unfortunately, and you're trying to yank it in to your new definition, okay, just by virtue of the fact that you happen to use the same label for both, okay? And this is where a lot of the confusion comes from, and this is why you're going to see me going, what the? And like, I can't even look at these debates over free will without having hives break out, because all they need to do is Daniel Dennett should say, look, when I'm using the term free will, I'm just talking about the fact you're not being [coerced?], you get to make your own decisions. And Sam Harris should say, you know what? If that's what you mean by free will, then yeah, of course that exists, right?

[00:27:50]  Bruce: And you're right, that we've got a long history of using the word free will to mean that. So it's kind of dumb for me to, you know, this is the intelligent Sam Harris that I wish he was sometimes. It's really, you're right, that type of free will, that's a real thing, that is part of its history of what we mean by it. Okay, but what I really mean is, is that somehow this idea of free will got caught up into some supernatural ideas that have now been disproven, right? I mean, like, it seems like you could move past these debates in under 30 seconds if you just accepted that words and concepts are different, and then use a few more words to point to which concept, sub-concept you were actually talking about. Okay, so let's now apply this to the idea of probability. Do we have a similar issue with the word probability? Okay, so David Deutsch insists that randomness does not exist. In fact, he goes to great lengths to remove all probabilistic language from his speech, as Peter, as you were pointing out. He'll say things like he won't say, oh, that's more probable, he'll say that's more plausible, okay? Or he'll say, you know, you don't need beliefs, you know, beliefs are unnecessary, you just go with the best theory. And he's got this whole thing going on where he's trying to remove probabilistic language, okay? Now, here's the problem, of course, in the English language, the word plausible literally just means seeming reasonable or probable, right?

[00:29:19]  Bruce: So I mean, as much effort as he's going through to try to remove the word probable from his language and use plausible instead, those are just synonyms, like they're literally just synonyms, okay? So it's unclear what we're accomplishing when we do this, okay?

[00:29:34]  Peter: I can't remember who exactly was pointing this out, but someone was pointing out recently how difficult it is for even him to do that.

[00:29:42]  Bruce: Yeah, it's Increments guys on their podcast where they're reviewing [recall?].

[00:29:47]  Unidentified speaker: Yes, that's right.

[00:29:48]  Bruce: So they end up having to use the terms because they're actually useful terms, okay, that point to actually existing real concepts, okay? So what I'm asking is, what does Deutsch even mean when he says randomness does not exist? And if he's right about that, is he right in a way that actually matters or is it kind of a superficial word game? Okay, and I'm really not clear on this point, okay? So now let's actually back up a little bit and let's talk about, is, do we have a problem with the word probability that's kind of like this? Well, we sort of do, okay? It's more nuanced than say the example with induction. But like frequentists, that's one of the schools of philosophy around probability theory, they use probability as a synonym for frequencies, okay? Popper, he came up with his own philosophy of probability, the propensity theory. So to him, the word probability means more like a propensity to change in specific ways, okay? Now, Bayesians, they use the word probability as a synonym for plausibility, okay? Which, interestingly, might make Deutsch's wording crypto-Bayesian, okay? Now, they're more likely to use a term like credences or beliefs or something like that. But if you really look at what they're saying, they're trying to use the word probability in exactly the same way that Deutsch is trying to use, or arguably in exactly the same way that Deutsch is trying to use the word plausible, is how well criticized is this theory, okay?

[00:31:22]  Unidentified speaker: And compared to its competitors, all right?

[00:31:27]  Bruce: So, there has been literally centuries of war over what probability means, the word probability means. And it's a war, is this a war over concept, or is it over words, or is it somehow both?

[00:31:42]  Unidentified speaker: Okay.

[00:31:43]  Bruce: And how might we tell the difference? So, let me tell a story here. When I, after I did the interview with David Deutsch, I was trying to push him on, you know, he kept insisting there was no randomness, and I kept saying, well, look, there's quantum randomness, can't we call that randomness? And he would push back. And I was kind of expecting him to say, because he's a Popperian, I was kind of expecting him to understand this idea of concepts versus labels, and to say, well, yeah, of course, we can use the word randomness to refer to quantum randomness as something distinct from pseudo-randomness.

[00:32:22]  Unidentified speaker: Okay.

[00:32:22]  Bruce: And I was hoping he would say that. And I pushed him a couple of different times on this.

[00:32:26]  Unidentified speaker: And he wouldn't say that.

[00:32:29]  Bruce: So I came away feeling a little confused. And at this point, I had only really listened to his discussion with the Increments podcast. And I had had this chance to ask him a few questions that left me feeling a little confused. I hadn't yet gone out and listened to his video on physics without probability, which I'll get to in a second here. So I went off and I wanted to talk with members of the CritRat community and try to understand why is he making such a big deal about this. So I came across two members of the CritRat community that were willing to talk with me about this. I'm showing them here as if they're talking with me together, but they weren't. This is like separate people at separate times that don't know about each other. So I said, hey guys, I don't understand. What do you mean, since you accept Deutsch's point here, what do you mean by randomness does not exist? They're like, okay, no problem, we can explain. They were total sweethearts about it. One of them literally called me, and we did a Zoom session for, I don't know, two or three hours, and he intently listened to me, and did not give me any sorts of guff, and he would listen to each of my concerns, and he would let me lay it out, and then he would really do his best to try to help me through it. And then I would give him problems, and he might go, I'm not sure on that one. I'm going to think about that one.

[00:33:53]  Unidentified speaker: I'm going to get back to you.

[00:33:56]  Bruce: Totally fit critical rationalism very well.

[00:33:59]  Peter: Sounds like we need to have him on the podcast.

[00:34:01]  Bruce: Yes, we probably do.

[00:34:07]  Bruce: Here are key arguments that they used with me. And these are all familiar arguments that you've probably heard Deutsch make, or at least hint at. But here's the way that I guess the CritRat community is interpreting David Deutsch here, okay? So there's this idea that the scientific community, and really beyond the scientific community, generally misunderstands probability. And probability is often misunderstood due to these misunderstandings, these physical misunderstandings in physics as to what probability really is. In fact, there is no such thing as randomness or stochasticity. There is only pseudorandomness. So the argument is that the fact that everything that we call randomness is actually pseudorandomness. That is the general misunderstanding that leads to all sorts of various other misunderstandings in probability theory. Now, let me just say that I agree that there are lots of misunderstandings in probability theory. And I've heard different explanations as for why this is. And this is actually an interesting question in and of itself.

[00:35:15]  Unidentified speaker: And this is an interesting interpretation.

[00:35:17]  Bruce: They're saying these misunderstandings stem from the fact that there's no such thing as real randomness or stochasticity, and it's all pseudorandomness.

[00:35:27]  Unidentified speaker: Okay.

[00:35:28]  Bruce: So that's their explanation, this crit rat explanation. I don't know if this is what Deutsch intends or not. I'm going to show you in a second that it's unclear if the crit rats are correctly understanding Deutsch or not. But this is what they're interpreting Deutsch as saying, okay? So now I hear them out on this, and it's not making sense to me. And I go, wait, I'm confused. Pseudorandomness and randomness, they are often physically entirely different, right? There's a reason why we split things into randomness, true randomness and pseudorandomness. They aren't the same thing in real life, right? And both crit rats at different times, they're like, well, no, that's not right. There is only pseudorandomness and that's it. And that is the source of confusion in the scientific community.

[00:36:16]  Unidentified speaker: Okay.

[00:36:18]  Bruce: So then certain key arguments get used here. And by the way, I'm taking most of these as paraphrases from David Deutsch himself from episode 100.

[00:36:27]  Unidentified speaker: Okay.

[00:36:28]  Bruce: So arguments might include things like this. If you're playing poker, which chair you were sitting in determines the outcome of the game.

[00:36:35]  Unidentified speaker: Okay.

[00:36:36]  Bruce: Because it's pseudo-random. If you're using a pseudo-random generator, then who wins the game is going to be determined by which chair you're in.

[00:36:44]  Unidentified speaker: Right?

[00:36:46]  Bruce: So this is one of the arguments about how it's different and etc.

[00:36:51]  Unidentified speaker: Okay.

[00:36:51]  Bruce: Quantum physics proves probability theory is sometimes just outright wrong. Because there are cases where there are two probabilities that add up to a non-probability. I've, Deutsch used this argument in the podcast to me.

[00:37:04]  Unidentified speaker: Okay.

[00:37:05]  Bruce: And then he says, what really matters is that you don't know the odds of a die throw. It might be that the sides aren't all one sixth.

[00:37:13]  Unidentified speaker: Okay.

[00:37:13]  Bruce: It might be like maybe the sixth is one fifth or something like that. But if nobody knows that that's the case, and Deutsch actually used the example of nobody takes the die and rolls it a bunch of times to test.

[00:37:24]  Unidentified speaker: Okay.

[00:37:25]  Bruce: Then it doesn't matter that you don't know because you have ignorance as to which side is favored. So everybody's still on even keel and it's still a fair game. And therefore, so goes the claim. What you really want is not randomness, which does not exist, but just a fair process that's unpredictable.

[00:37:45]  Unidentified speaker: Okay.

[00:37:46]  Bruce: So I'm looking at this explanation, and I'm even more confused now because I just know that there's something wrong with every single one of those arguments. So I suddenly go, you know what?

[00:37:57]  Unidentified speaker: That doesn't make sense to me.

[00:37:59]  Bruce: Las Vegas very carefully tests their dice to make sure they are very close, indistinguishably close to 1 sixth, right? They weight them, they balance them, they use their explanatory knowledge, and then after they've done all that, then they go and they have an automated roller.

[00:38:16]  Unidentified speaker: Like I saw this on the news, right?

[00:38:17]  Bruce: I saw the automated roller for myself. And they roll it and then they make sure that it's within certain bounds, and if it doesn't happen to fall within those bounds, they jettison the die and they won't use it, because they absolutely do not want you to accidentally be able to find out that that die, in fact, does favor the six, and therefore you use that die and you can win more often. That's a bad thing if someone can ever figure that out. As Deutsch points out, if you just go roll the die a bunch of times, you can figure that out if that's the case. So they go to great lengths to make sure that's impossible. And then this idea that we use random process, quantum processes, we in fact, like Las Vegas literally uses quantum processes to ensure true randomness on their slot machines. They do not want to use, there are slot machines that use pseudo-randomness, but those are not the high quality slot machines because they may result in a pattern. So what they do is they actually come up with a way to measure quantum processes, use a heat source and measure the atoms coming off the heat source, which is determined by quantum randomness or something along those lines. And that's a very big deal in Las Vegas, that they try to make sure that they're using true randomness on their slot machines and not pseudo-randomness. And then there are no quantum processes, I'm certainly not an expert in quantum physics like Deutsch is, but like I've studied quantum computation, and I know there are no quantum processes that don't follow probability theory.

[00:39:56]  Bruce: There's a distinction that we need to make, right? There's the quantum wave function that until you observe, before you observe it, which is entirely deterministic. And then there's what happens when you observe it and the quote unquote wave function collapses, okay? And then you apply the Born rule, and the Born rule takes the wave function, and it then determines a probability by taking the square of it, okay? Now, it is true that you can set up an experiment. The double split experiment is an example of this. You can set up an experiment where you can make the wave, the deterministic wave function create an interference pattern such that the Born rule is going to tell you that it's going to be a 100% success, that it's going to go that you will get this wave pattern, or the photon will come out this slit or something along those lines. I would have to show this in more detail to show exactly what I'm getting at. But it's true that the quantum mechanical processes are not the same as classical processes, and that they can add up to this thing where the Born rule will assign it 100%. But once it assigns it 100%, even though it's 100%, that is still totally regular probability theory. Probability theory entirely can encompass the idea of 100%.

[00:41:18]  Unidentified speaker: Okay?

[00:41:18]  Bruce: There's no actual difference. The probability theory doesn't come into play until you apply the Born rule. And what there, what Deutsch is talking about is what happens before the Born rule, while you're still in the deterministic process. So I don't understand why this is being offered up as a difference from probability theory. This follows probability theory exactly, right? So at this point, both crit rats at different times said, you know, I think you're mistaken. I think you've misunderstood something about how this actually works.

[00:41:47]  Unidentified speaker: Okay.

[00:41:49]  Bruce: So I tried to propose to both of them at different times that maybe we need three definitions, because I feel like a lot of confusion is coming from the fact that we're trying to collapse this down to two definitions when really we're talking about three definitions. So here's my modest proposal, to avoid confusion. Let's declare there to be three kinds of randomness.

[00:42:10]  Unidentified speaker: Okay.

[00:42:11]  Bruce: So the first, randomness type 1, aka true randomness. This is the kind of stochastic event that would exist if we didn't live in a multiverse. This kind of randomness does not exist, and I agree it does not exist. So I'm on board as long as by true randomness, you mean this type 1.

[00:42:31]  Unidentified speaker: Then let's have a type 2.

[00:42:33]  Bruce: Randomness type 2, aka apparent randomness. The kind of stochastic event from the point of view of an observer in a single universe that actually does exist. And it's now known to split the multiverse, and we know that at the level of the multiverse is deterministic. But from the point of view of an observer, it's a real stochastic event.

[00:42:54]  Unidentified speaker: Okay.

[00:42:56]  Bruce: Note also that when I say from the point of view of the multiverse, there are no such things as observers at the level of the multiverse. Observers at the level of the multiverse do not exist. This is kind of important, right? So we're making up a fiction of someone who observes how the wave function develops without being inside the wave function. And then we're saying for that person, it would be entirely deterministic. And they would see that actually, when you roll the dice or whatever, it splits the multiverse into multiple versions of itself with different outcomes, okay? But that person does not exist.

[00:43:38]  Unidentified speaker: It's a pure fiction, okay?

[00:43:40]  Bruce: Which is why I would say we call this apparent randomness. Then we've got type 3, aka pseudo-randomness, which is distinct from the other two. It's an unpredictable process that is determined at the level of a single universe as opposed to type 2, which is determined only at the level of the multiverse, but totally undetermined at the level of a single universe. Okay, and this type 3 pseudo-randomness, it mimics stochasticity by being unpredictable and by having a fair amount that comes out in different ways and things like that. If you're doing 1 to 10, then you get an equal number of 1 through 10, something along those lines, okay? And I put this out there and I'm like, will you guys accept that really there's 3 things we're talking about here? Type 1, type 2, type 1, true randomness.

[00:44:31]  Unidentified speaker: And notice I'm even calling it true randomness.

[00:44:33]  Bruce: I'm allowing you your terms for it, okay? I just want to make sure that we're clear that there's an additional separation that needs to be made, okay? So type 1, true randomness, type 2, apparent randomness, or maybe we can call it subjective randomness, I don't know. Or type 3, pseudo-randomness, which is not randomness, but behaves similar to randomness because you can't predict it, okay? By the way, you can predict it. You can roll it back and try it again, it'll do exactly the same results.

[00:45:02]  Unidentified speaker: But the first time, you can't predict it, okay?

[00:45:06]  Unidentified speaker: So here's the thing.

[00:45:08]  Bruce: If I can get them to accept that, then of course, my next thing is I'm going to say, well, let's not call it type 2, apparent randomness since it's the only real type. Let's call it actual randomness.

[00:45:21]  Unidentified speaker: And boom, okay?

[00:45:22]  Bruce: We've got this true randomness, actual randomness and pseudo randomness. Well, of course, this was a bridge too far, because now we're bridging on me outrightly saying that there's something wrong with Deutsch's point of view.

[00:45:35]  Unidentified speaker: Okay.

[00:45:36]  Bruce: So I get back from both of them. Well, no, there are only two categories of randomness. There's the true but non-existent kind, and there's pseudo randomness. And both of them were initially very insistent that this was the case, and that there was no need for this extra category of apparent or actual randomness.

[00:45:56]  Unidentified speaker: Okay.

[00:45:57]  Bruce: So I started thinking about this. How in the world can I resolve a factual disagreement like this? Like, I know for sure I'm right about this. This is a huge part of computer science, which is my background, right? It has to do with how we define Turing machines, the fact that we do or don't need stochasticity. And there's all sorts of things that fall out of this. And so how would I convince them that there's an actual need for this additional distinction? So I thought of something, because somewhere in the back of my mind, I remembered this from like school.

[00:46:29]  Unidentified speaker: Okay.

[00:46:30]  Bruce: So take a look at the screen here. What you've got here is an example of true randomness.

[00:46:35]  Unidentified speaker: Okay.

[00:46:36]  Bruce: So you've got this, basically, you just hit, you just place random dots somewhere on a 2D space, and it just looks like white noise, right? Like it's just a bunch of grains, and there's no discernible pattern.

[00:46:49]  Unidentified speaker: Okay.

[00:46:50]  Unidentified speaker: This is true randomness.

[00:46:53]  Bruce: When you are using true randomness, or what I'm calling actual randomness, you will get a pattern like this, or no pattern, right? Here is pseudo-randomness, okay? This is the same algorithm, exactly the same algorithm, but one, the first one's using true or actual randomness, and the other one's using pseudo-randomness. You can immediately see the pattern.

[00:47:20]  Unidentified speaker: Okay.

[00:47:20]  Bruce: This is why people don't like pseudo-randomness in Las Vegas, right? Because then you've got, you're opening yourselves up to the possibility that someone can figure you out and cheat.

[00:47:30]  Unidentified speaker: Okay.

[00:47:31]  Bruce: So, and now in this case, this is like super obvious. It's not always so obvious.

[00:47:37]  Unidentified speaker: Okay.

[00:47:37]  Bruce: By chance, the guy who did this, who by the way is Bo Allen, boallen.com/random-numbers.html. This is where I'm getting this from. He found this on his first try. It may be that a pseudo-random generator would look more like the first one, but the pattern might exist at a different level. Like you would have to do more of them before the pattern could be found. You can never be sure there isn't a pattern.

[00:48:06]  Unidentified speaker: Okay.

[00:48:07]  Bruce: There is no, in computer science, there is no general test for lack of a pattern. So when you're using pseudo-randomness, you were using it at your own risk that someone might exploit a pattern.

[00:48:22]  Unidentified speaker: Okay.

[00:48:22]  Bruce: Which is why Las Vegas doesn't use pseudo-randomness.

[00:48:27]  Unidentified speaker: Okay.

[00:48:28]  Unidentified speaker: Now, I showed this to them, and this did change things.

[00:48:31]  Bruce: They had to stop and think about it. They went back and forth with me a few times. They're like, okay, I guess I see your point. There's actually a physical difference between what you're calling actual randomness and what you're calling pseudo-randomness. So at this point, they bought into this idea that maybe we could accept that there were three categories.

[00:48:51]  Unidentified speaker: Okay.

[00:48:52]  Bruce: Now, let me actually show where this came from.

[00:48:54]  Unidentified speaker: Okay.

[00:48:55]  Bruce: Where did we actually get on a computer that random pattern? It came from www.random.org, which it is an API that exists out on the Internet that allows you to avoid using pseudo-randomness in your software, so that you can be guaranteed to not have a pattern. This is obviously for some applications really useful, because we don't want to risk the existence of a pattern. So it's like figuring out which chair in a poker game is most likely to win. Someone might figure that out if you're using pseudo-randomness. If you're using actual randomness, then it's impossible for there to be one chair that favors winning.

[00:49:40]  Unidentified speaker: Okay.

[00:49:40]  Peter: Is there kind of a gray area though, between pseudo-randomness and randomness? Is it just a matter of what lens you're looking at this?

[00:49:50]  Unidentified speaker: Yes.

[00:49:51]  Unidentified speaker: Good question.

[00:49:51]  Unidentified speaker: Good question.

[00:49:52]  Bruce: Let me explain that in just a second. Let me finish this slide off before I lose my momentum here.

[00:49:57]  Unidentified speaker: Okay.

[00:49:58]  Bruce: So from random.org, here's what it says. It says random.org offers true random numbers to anyone on the internet. The randomness comes from atmospheric noise, which for many purposes is better than pseudo-random number algorithms typically used in computer programs. People use random.org for holding drawings, lotteries and sweepstakes to drive online games for scientific applications for art and music.

[00:50:28]  Unidentified speaker: Okay.

[00:50:29]  Bruce: So please note that atmospheric noise is determined by quantum effects.

[00:50:36]  Unidentified speaker: Okay.

[00:50:36]  Bruce: So they are literally using quantum effects to create true randomness that is physically different than pseudo-randomness, which may create a pattern. And if you want to avoid that, then you're going to use random.org, their API.

[00:50:51]  Unidentified speaker: Okay.

[00:50:52]  Bruce: Now, Peter, let's go back to the question you just asked. The answer is yes.

[00:50:59]  Unidentified speaker: Okay.

[00:50:59]  Bruce: And this is part of the problem, is that words don't have individual meanings. They've got word clouds, they've got concept clouds, they've got halos.

[00:51:09]  Unidentified speaker: Okay.

[00:51:10]  Bruce: If someone were to say to me, this algorithm is stochastic. They probably don't literally mean they're using atmospheric effects to create true randomness. They may well just mean I'm using pseudo randomness.

[00:51:25]  Unidentified speaker: Okay.

[00:51:25]  Bruce: And it's not a fully, it's not a determined process. The word random does get applied in place of pseudo randomness if in context you're talking about something where it doesn't matter.

[00:51:39]  Unidentified speaker: Okay.

[00:51:40]  Bruce: And this is the way words actually get used in real life.

[00:51:43]  Unidentified speaker: Okay.

[00:51:43]  Bruce: You might refer to a pseudo random generator as stochastic.

[00:51:49]  Unidentified speaker: And that's fine.

[00:51:51]  Bruce: But then someone might say, well, it's not really stochastic. And they'll go, yeah, no, of course. I just was in context. That didn't matter. So I was calling it stochastic.

[00:51:59]  Unidentified speaker: And this is how we do things.

[00:52:01]  Bruce: We choose the lens based on what the problem is we're trying to solve and what we're actually talking about. And the word can get reused in both ways.

[00:52:09]  Unidentified speaker: Okay.

[00:52:10]  Bruce: Which is why I'm so confused that we're making such a big deal about randomness not existing.

[00:52:16]  Unidentified speaker: Okay.

[00:52:17]  Bruce: This just isn't the way words get really used in real life. So let me actually take, go back to our three types of randomness here. So randomness type one, true randomness. Randomness type two, apparent or actual randomness, actual in the sense that it actually exists. Randomness type three, pseudorandomness. Now, it's not too hard to see that if we're going to accept these three definitions as valid, which I've now got the crit rats I'm talking to to accept, that we might as well just get rid of the first type, because it doesn't exist anyhow. We're right back to two types, true randomness, which now that's the new, what we used to call apparent randomness. We're going to now call that true randomness, which by the way, is what random.org meant by the term true randomness. Then we'll call the other one pseudorandomness. Then we don't need that other type, like maybe in certain weird circumstances, we can talk about this other type of non-existent randomness. But it doesn't matter in the vast majority of contexts and situations, okay, which is why I don't understand why we're doing this. Like, it just seems so strange. And just another thing to point out, Deutsch himself actually in his book Fabric of Reality admits that there is a physical difference between what I'm calling apparent randomness and pseudo-randomness, okay?

[00:53:43]  Unidentified speaker: So here's a quote.

[00:53:45]  Bruce: He says, it is perhaps worth stressing the distinction between unpredictability and intractability. Unpredictability has nothing to do with the available computational resources. Classical systems are unpredictable, or would be if they existed, because of their sensitivity to initial conditions. Quantum systems do not have that sensitivity, but are unpredictable because they behave differently in different universes, and so appear random in most universes. And this is really the point I'm trying to get at, right? And this is what I was actually trying to push David Deutsch on. I don't understand why we don't just call quantum effects, even if I accept that they're actually splitting the multiverse, even if I accept that at the level of the multiverse, they are deterministic. Why not just call that true randomness? Because that's what the word really means. Like that's how it gets used in real life, right?

[00:54:40]  Unidentified speaker: That's what people mean.

[00:54:41]  Bruce: If somebody says true randomness, they mean something like a quantum system that isn't pseudo-random, okay? And even Deutsch is admitting there's a difference. So I don't understand why this matters, right?

[00:54:53]  Unidentified speaker: It seems like it's very confusing.

[00:54:56]  Bruce: So now I get an argument back, something like this. But what you're calling true randomness now, it's not really randomness because it comes from a determined process, just like a pseudo-randomness comes from a determined process. And in fact, this was the answer Deutsch gave me when we interviewed him, okay? Now here's my response. Yes, but there's a huge physical difference between coming from a deterministic process in a single universe, pseudo-randomness, or coming from a deterministic process from the level of the multiverse, which is true randomness, okay? Because one splits the multiverse and one doesn't, and that's why they are physically different. That's why one's completely unpredictable and the other one is potentially predictable, okay? So it seems like this answer, well, it comes from a determined process. I understand the motivation here, but it seems like it's missing what really mattered, right? You're trying to say, well, if it comes from a determined process, then the essential definition of randomness is that it doesn't come from a determined process. And I'm saying, no, wait, why don't we say, if you're going to do that, first of all, that's essentialism, you shouldn't do that. But if you're going to do it, why not instead say, well, whether it's random or not isn't if it comes from a determined process; it’s if it comes from a determined process within a single universe, okay, then you've still got the distinction correct, right? And you've eliminated all sorts of confusion.

[00:56:35]  Bruce: So I simply don't understand why we're insisting on making up a whole new category of randomness only to then declare it not real, when there is a real kind of random process that actually does exist. So what I'm looking for is not a label change based on a choice of definitions, because that is a meaningless thing to me, okay? I'm trying to understand how Deutsch's non-existent true randomness, and what I'm calling apparent randomness, are in any practical sense different from each other. Obviously, they are different at the level of the multiverse, okay? But in terms of me as someone who lives inside of a single universe and only lives inside of a single universe, will they ever behave differently? And it seems like this is the right question, because we're questioning probability theory, which is a mathematical formalism that's meant to be isometric to random processes, which actually do exist, right? The quantum processes actually are random, okay? And probability theory actually is fully isomorphic to it, alright? Whereas pseudo-randomness is not fully isomorphic to probability theory.

[00:57:50]  Unidentified speaker: There are differences, okay?

[00:57:53]  Bruce: So I want to see a new theory of probability that takes this difference of true randomness and apparent randomness into consideration, and then makes different predictions, better predictions, and then we can falsify the old one.

[00:58:06]  Unidentified speaker: This is critical rationalism.

[00:58:07]  Bruce: Of course, that's what I want to see. I could care less what we call it, right? I just want to make sure we don't eliminate and end up with entirely removing the fact that quantum processes are actually random compared to a pseudo-random process, and those are different things. Let's not collapse those as if they're one, because they're not, okay? So if all this is just an essentialist claim, that we shouldn't call it randomness because quote unquote true randomness doesn't quote come from a determined process, a deterministic process, but yeah, it behaves exactly in every conceivable way, exactly like a stochastic process.

[00:58:50]  Unidentified speaker: Okay, then I have a proposal.

[00:58:51]  Bruce: Let's just say true randomness is something that is not deterministic at the level of a single universe, and pseudo-randomness is something that is deterministic at the level of a single universe.

[00:59:00]  Unidentified speaker: And let's be done.

[00:59:01]  Bruce: No need for a third non-existent category that clearly has led to confusion. Notice that the crit rats I talked to, they were convinced the correct interpretation of Deutsch was that everything was pseudo-random. Okay, and I had to struggle to get them to see that there was this other category that existed because the way he's chosen to word it collapses those two together by emphasizing that they're both quote unquote deterministic and ignoring the difference that mattered, that one was deterministic at the level of the multiverse, and one was deterministic at the level of a universe, and that is a physical difference.

[00:59:35]  Unidentified speaker: So, and doesn't this seem reasonable?

[00:59:37]  Bruce: And this is the thing I keep trying to bring up with the CritRat community over this. Wasn't the word randomness as a word literally invented to point to the concept we're now calling apparent randomness? Okay, presumably the word existed to be a label for a real concept for use by observers that happen to live within a single universe, as opposed to non-existent observers that don't exist in the multiverse. Okay, so of course, it must be that if you're really trying to get down to what's the center of the halo, okay, on the word randomness, okay, what's the original use, the original meaning, okay, if you want, of the term, that would be it. Like quantum processes, that's what they meant. They meant when they threw out there the word stochasticity in its purest form, and they weren't using it as a synonym for pseudo-randomness, which they sometimes do, they meant something like quantum processes that are wholly unpredictable.

[01:00:35]  Unidentified speaker: Okay, that's what they meant.

[01:00:37]  Bruce: That's what the word means, if you will. Okay, I don't understand why we need to say, ah, but it's not true randomness because at the level of the multiverse, it's deterministic, right?

[01:00:48]  Unidentified speaker: I don't see what we're adding by doing that.

[01:00:52]  Bruce: And when you do that, what you'll end up with is exactly what you did end up with, a collapse of two distinct concepts down to one less useful concept that then really makes it harder for you to think clearly. Okay, so my rule is this, don't ever do away with a useful word that points to a real concept. That just removes your ability to think clearly. By the way, I would put that out there for anyone that's in the free will debate, you know, wars. Okay, stop, like, don't, don't try to get rid of the word free will if it actually points to a useful concept. Instead, talk about how we've come to understand free will differently, and what parts of it turned out to be supernatural and not related to the real concept, right? Like, this is really a much better way to speak about things, because it allows you to think more clearly. So, after going through all this, I realized, you know what, I've never watched Deutsch's video, Physics Without Probability. I'm going to need to really dig into this deeper. And as you probably know from the podcast, I've been trying to make sense of Bayesian epistemology, and I had Ivan on the show, and I've been reading his book, I'm almost done with his book. And these are kind of similar, but similar problems, but coming from a different angle, right? Is you've got, on the one hand, you've got the Deutschian crit rats, and not just the Deutschian crit rats. This is something that's more general. Like, David Miller had some really negative things to say about certain aspects of Bayesian reasoning that to me don't make sense, right?

[01:02:24]  Bruce: So there's this kind of open hostility towards probability theory that comes out of the crit rat community. And then there's kind of this Bayesian view that just everything should be put in terms of probability theory. Is one of them right? Is one of them more right? Are they both wrong? Like, I don't know, right? I really need to dig into this a lot deeper. So I did go off and I watched the video, Physics Without Probability, and it was eye-opening. It was not what I was expecting. And in this video, I'm not going to get into my critique of it. I need some time to actually tease out what he says, compare it to different sources out there. But the question on my mind is this, was this all just the essentialist mistake, assuming that randomness as a word had to mean, did not come out of a deterministic process, even at the unobservable level of the multiverse? Or was there an actual difference being offered that would change probability theory by improving on it in some way?

[01:03:26]  Unidentified speaker: So, inquiring minds want to know.

[01:03:27]  Bruce: That's what I want to know, okay? And this podcast, we're done with this podcast.

[01:03:32]  Unidentified speaker: This is my setup.

[01:03:34]  Bruce: And that's why I wanted to put this in terms of words versus concepts. I want people to think about this differently, right? Is this a word game that we're talking about? Or is this a concept difference that we're talking about?

[01:03:46]  Unidentified speaker: Or could it be both?

[01:03:47]  Unidentified speaker: Could there be confusion between the two of them?

[01:03:49]  Bruce: Okay, where sometimes it's a word game, but there's some sort of actual conceptual difference that we need to discuss. And this is where I'm going to be going from here. This isn't going to necessarily be like the next podcast or something. Think of this as yet another thread I'm going to pull that we're going to do over time in future podcasts. All right, Peter,

[01:04:08]  Peter: I think this is a fun, punchy episode. And I really liked how you framed that about the word clouds at the beginning. That was just so cool.

[01:04:20]  Bruce: And that was the part I really needed a visual, right?

[01:04:23]  Unidentified speaker: Yeah,

[01:04:24]  Bruce: to like you can you can understand what I'm saying, kind of, but when you see it, that you go, oh.

[01:04:29]  Peter: It really just brought home what I've thought about a lot of with these essentialist word games. And I think it's everything you said there seemed very true. I guess maybe one quick question. Do you have an example of how this really matters in life?

[01:04:53]  Bruce: Well, I just gave the example of Las Vegas, right?

[01:04:56]  Unidentified speaker: Okay.

[01:04:56]  Bruce: So maybe you consider that an unimportant example. Okay, but it doesn't have to be important. It just has to be that there's an actual difference, right? So because pseudo-randomness does create patterns, and theoretically you can roll them back and try it again and it'll do the same thing. Pseudo-randomness is not quite the same as true randomness. Whereas quantum randomness is exactly the same as true randomness, right? And so that's why we will use quantum processes to generate true randomness in some cases, right? Where, I mean, you gotta think about this from the point of view of Las Vegas. Maybe you don't care very much about if they're using pseudo-random or true random, but it's their money on the line if somebody hacks the system and figures it out, right? So making sure that they're using true randomness is something that they really are motivated to do because that difference really does matter in that case.

[01:05:57]  Unidentified speaker: Okay.

[01:05:58]  Peter: So when I use my Universe Splitter app on my phone, do you have one of those?

[01:06:04]  Unidentified speaker: No, I don't.

[01:06:05]  Peter: Well, it's just an app that takes a quantum reading from a lab in Sweden or something like that, and it will give you an A or a B, and presumably it splits the Universe, and if you make a decision on it, you can split the Universe. I tell people that my app does that and they think I'm crazy.

[01:06:29]  Unidentified speaker: But I think it, I mean.

[01:06:33]  Bruce: If MWY is true, and I would consider that a best explanation.

[01:06:37]  Unidentified speaker: Really does work.

[01:06:38]  Bruce: It really does split the Universe.

[01:06:40]  Peter: And so it's random from the perspective of a single universe, but not from the perspective of the multiverse.

[01:06:49]  Unidentified speaker: That's right.

[01:06:50]  Peter: I mean, that just makes perfect sense. I don't even know why that's so controversial, but I don't think it is.

[01:06:57]  Bruce: Let me give you a couple other examples. Like in machine learning, they will use true randomness. Keep in mind that anytime I, for me, true randomness means what is quantum events, right?

[01:07:08]  Unidentified speaker: Okay.

[01:07:09]  Bruce: So they will use true randomness, and they can improve the machine learning algorithm because it avoids patterns, right? So it will actually give a better result, it will explore the space better.

[01:07:21]  Unidentified speaker: Okay.

[01:07:22]  Bruce: Now, there are no cases that are known where the difference is so substantial that there's certain algorithms. Like we could imagine a universe where the laws of physics were such, where if you didn't have a quantum true random generator attached to your Turing machine, that there would be certain kinds of algorithms that it couldn't run. Like that could have been true. And at one point, they wondered if that was true. But what they've really found was, is that we don't know of any algorithms where you can't plug a pseudo random generator in and the algorithm won't work. Now, it may not work as well. And it may have limitations or constraints, such as the fact that you can run it back and have it give the exact same result again.

[01:08:09]  Unidentified speaker: But that's useful.

[01:08:10]  Bruce: Like a lot of times when you're playing with these things in real life, you set the seed on your pseudo random generator so you can get the exact same result every single time, because you don't want to have to keep dealing with it. I don't know if this was a random difference or not. And so because of that, they don't consider the universal computer as requiring a random generator.

[01:08:30]  Unidentified speaker: Okay.

[01:08:31]  Bruce: And because they do it based on which algorithms you can run, and you can run the same set of algorithms on a deterministic Turing machine, or on a Turing machine that has a random number generator, true random number generator. But that doesn't mean you can't get additional efficiencies. It doesn't mean that there aren't real life constraints that can impact you still. And that was why I wanted to make sure we hadn't collapsed these two concepts into one.

[01:09:00]  Unidentified speaker: Okay.

[01:09:00]  Unidentified speaker: Well, this is a record.

[01:09:02]  Peter: I think we finished in less than an hour.

[01:09:04]  Unidentified speaker: Yeah.

[01:09:05]  Peter: And well done, Bruce. And I really will look forward to your episode on the physics without probability lecture. That's one that I've watched it a couple of times. I can't really say that I have. It's probably one that is a little bit confusing to me still, but yeah, you know what?

[01:09:28]  Bruce: I haven't figured it out entirely either. I'm actually studying. I've bought some textbooks on probability theory and machine learning, Bayesian reasoning in particular. And I am trying to study probability theory. I've done it before, like when I was in school, but it's mostly forgotten. And I'm trying to learn it at a deeper level. And I'm hoping to accomplish two things, like really at the same time. I'm hoping to simultaneously figure out, is there anything to what Deutsch is saying? And if so, what is it? Like clearly, the way he's currently saying it is leading to confusion, right? So is there some nut there that I need to hold on to and say, yeah, this is what he really means? Or is this something we can ignore? Like I need to really determine that for myself. But at the same time, I really want to be able to understand Bayesian reasoning so that I can understand why it is Bayesian epistemologists try to base everything on Bayesian reasoning. And then I want to understand, how does that differ or is the same as critical rationalism? Is Bayesian epistemology really just isomorphic to critical rationalism? Do they have heavy overlap, but they differ in some places? Or is one of them superior to the other in some way? And I'd really like to know the answer to that question. I think I have to actually figure this out and put some time into learning the mathematics. Or I don't know that I'm going to be capable of answering these questions.

[01:10:53]  Unidentified speaker: I would just be guessing.

[01:10:54]  Bruce: I'd be armchair philosophizing.

[01:10:56]  Unidentified speaker: I want to actually know the answer, if that makes any sense.

[01:11:00]  Peter: Well, you're doing important work and asking hard questions. And I've enjoyed listening to you at least, Bruce.

[01:11:06]  Unidentified speaker: All right.

[01:11:07]  Bruce: Thank you very much.

[01:11:08]  Unidentified speaker: Have a great day.

[01:11:09]  Unidentified speaker: All right.

[01:11:09]  Unidentified speaker: You too.

[01:11:11]  Unidentified speaker: Okay.

[01:11:11]  Unidentified speaker: Well, that was cool.

[01:11:20]  Unidentified speaker: Hello again.

[01:11:21]  Peter: If you've made it this far, please consider giving us a nice rating on whatever platform you use, or even making a financial contribution through the link provided in the show notes. As you probably know, we are a podcast loosely tied together by the Popper Deutsch Theory of Knowledge. We believe David Deutsch's Four Strands tie everything together. So we discuss science, knowledge, computation, politics, art, and especially the search for artificial general intelligence. Also, please consider connecting with Bruce on X at BNielsen01. Also, please consider joining the Facebook group The Many Worlds of David Deutsch, where Bruce and I first started connecting.

[01:12:04]  Unidentified speaker: Thank you.