Episode 40: Byrne vs Deutsch on Animal Intelligence

• 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

[00:00:13]  Bruce: Welcome to The Theory of Anything podcast. How's it going, Cameo?

[00:00:16]  Cameo: It's great, Bruce, how are you?

[00:00:18]  Bruce: Good. We don't have Tracy or Sadia today, but that's why we have lots of people, is so that somebody shows up. So we're gonna talk, this is the final episode on animal intelligence. Hopefully, if I can get through it all in one session, we may revisit, may or may not revisit animal sentience at some point. That's way more speculative than animal intelligence. We've got really good theories around animal intelligence from what I can see, Richard Byrne's theories and the ones he quotes. Where I don't think there's anything equivalent in animal sentience. There is some theory there and it's even somewhat testable. So it's not completely nothing, but you're definitely speculating a lot more in the animal sentience realm. Okay, before we get into this, did you by any chance watch the video of the dolphins that I sent you?

[00:01:04]  Unidentified speaker: No, I didn't.

[00:01:08]  Bruce: Well, let me tell you about it. Okay, it was really interesting. After our discussions, I went to my sister's. My sister's really into animals, and she's a pet lover. Of course, she's biased towards her animals, but she's really interested in animal intelligence as a pet lover. I was talking with her about, I'm studying animal intelligence, and she immediately started offering up examples. One of the ones she offered, she said, I could give you a link to one about dolphins. I'm like, oh, that's cool. We don't know that much about dolphin intelligence, but we think they probably have insight. She gives me the link and it was amazing. And I don't want to overstate it because of course it's only one incident. Anytime you have with animals one incident, you never know for sure what's going on. But so I don't want to overstate it. I would want to see this like replicated.

[00:01:54]  Bruce: So these are two dolphins that have been taught a sign language. That's one of the ways they communicate with dolphins is they got the signs they use. Obviously the dolphins can't sign back, but they can tell the dolphin, do this trick or create your own trick. Okay, so they taught them the concept, create your own trick. So they would say that and the dolphins would come up with some trick that they had never been taught. Okay, so they're clearly understanding the concept of create your own trick. Then they told them for the very first time, right on the video, they said, create your own trick, but do it together. So the dolphins swim to the bottom of the pool in the video, you can hear them talking for a moment, and then they come up, jump out of the water simultaneously, and then lay on their back and flip their tails up in the air simultaneously, a trick that they have never been taught before.

[00:02:43]  Cameo: They made up their own trick and then executed the trick.

[00:02:48]  Bruce: It was amazing. For them to be able to do that, unless it was some weird coincidence that for them to be able to do that, they had to have the concepts and the language down of doing a trick, doing it together. They had to have understood it even though they've never been trained in it. They had to then somehow communicate. I don't know that dolphins, these dolphins have been taught some language. They may have more language skills than a normal dolphin does, but they had some ability to be able to communicate with each other so that they could synchronize what they were going to do. It's stunning. Way beyond anything I would have thought dolphins were capable of doing.

[00:03:20]  Unidentified speaker: Interesting.

[00:03:22]  Bruce: Even based on reading Richard Byrne's stuff, I would not have guessed this was possible. Now, my sister gave me some of her own experiences too. She has a dog named Sophie, and Sophie is the smartest dog she's ever had. She says most of my dogs were nowhere near as smart as Sophie. That does happen. Sometimes, you've got an Einstein dog that just somehow is much smarter than a regular dog. Sophie, she took her into the bathroom, and my sister was playing with the scale, and Sophie sees that there's lights turning on on the scale. She's interested in it. My sister teaches her that the command tap tap means to hit the scale so that you can turn the light on. She says tap tap, she hits it, she sees the light come on, she gets excited that she made the light come on. That's regular trial and error learning so far, nothing out of the ordinary.

[00:04:11]  Bruce: Then Sophie has this thing where she comes downstairs, and the door is actually ajar, but she can't tell. She doesn't have any concept of a door being ajar. As far as she's concerned, the door is closed, she can't get to my sister. She's whining and whining, let me in, let me in. My sister can see that the door is actually ajar. All she has to do is nudge it and it'll open up. So she says to Sophie, tap, tap. And at first, Sophie doesn't get it. And she's like, whine, whine, whine. She goes, tap, tap. And I think she tried this a couple of times with Sophie failing at first and then tried it again later. And then finally, Sophie stops for a second and then tap, tap on the door. The door swings open and she comes in. Now this is a new trick that she's learned. She had somehow abstracted the idea of tap, tap on the scale to using it on the door.

[00:05:02]  Bruce: And it wasn't easy. She had to, it took her a few tries to be able to realize that's the right thing to do.

[00:05:08]  Cameo: She had to be helped through the abstraction, but she was still capable of understanding that that abstraction was there after some prompting.

[00:05:18]  Bruce: Yeah. Now Sophie had another one that was impressive. They had a game they played with her where they would call the name of a person in the room. And if Sophie went to that person, she would get a treat. And at first she had no clue. She would just try randomly going to people. And then at one point she suddenly, according to my sister, jumped up in the air excited, then went to the right person. And then from that point forward, never failed to go to the right person ever again.

[00:05:42]  Unidentified speaker: Oh, interesting.

[00:05:43]  Bruce: So now I don't think that, I mean, like this is pretty smart for a dog, but I don't think we're outside the bounds of what Richard Byrne is saying at all, right? This trial and error learning, I'm not even sure it's true insight in the way he uses it. Now, this next one is a lot more impressive. So my sister has a cockatoo named Moonlight, and when Moonlight would touch her nose, with Moonlight would use her claw to touch my sister's nose. And she would, my sister would go beep anytime she touched her nose. So she kind of trained Moonlight to know that she's gonna get the reward of having her say beep if she touches her nose. So one day, my sister as a joke touched the bird's nose and the bird goes beep, never having been trained to do it. And so she was like shocked. She's like, wow, she made the connection between a bird's nose and a human's nose.

[00:06:34]  Cameo: Yeah, that's pretty impressive.

[00:06:36]  Bruce: So then later she has a fish in a bowl and the fish swims up to the edge of the bowl to feed. And the bird comes over, taps the, taps where the fish's nose is and goes beep.

[00:06:45]  Unidentified speaker: Oh, crazy.

[00:06:47]  Bruce: Then later she does it to the hamster. Here's what's interesting. Moonlight is a cockatoo, which is I think the same family as parrots, which is one of the birds that Richard Byrne thinks might have insight.

[00:06:59]  Unidentified speaker: Right.

[00:06:59]  Bruce: Well, this certainly seems like, this is a level of abstraction that's shocking, right?

[00:07:04]  Unidentified speaker: It is.

[00:07:04]  Bruce: So I think that might have required insight for them to be able to do that, for her to be able to make this connection between, you know, four different species' nose. And even be able to reverse my sister saying beep and the bird saying beep.

[00:07:18]  Cameo: Yeah, yeah. It is pretty impressive insight.

[00:07:21]  Bruce: So now let's go back and we'll talk about animal understanding. So we were just talking about certain levels of animal understanding, and we've been talking about that in past episodes. Now in an interview, in David Deutsch when he was doing an interview with Tyler, Tyler says, dogs understand social life pretty well. And David Deutsch says, they do not. Dogs have genes which contain knowledge, but it is fixed knowledge. Now I'm cutting off the rest of what he says. I'll get back to it in just a second. If I stop right here, it really seems like he's saying something something very, very radical. It seems like he's literally claiming that animals understand nothing at all and are nothing more than computer programs with entirely genetically pre-programmed automatic behavior. And as we've talked about, many fans of David Deutsch believe that's what he's saying and believe that themselves.

[00:08:07]  Bruce: So maybe that's not so weird, but that is certainly a radical position and drastically at odds with everything we've been talking about in Richard Byrne's theory. Here's the problem though. David Deutsch then goes on to say, and it is not the kind of knowledge that constitutes understanding. Understanding is always explanatory. Well, that's okay. I mean, like if that's what he really means, of course animals don't have human level explanations, right? If these-

[00:08:31]  Unidentified speaker: They can't explain things to us.

[00:08:33]  Bruce: Right, and probably can't in their heads really either, right? They don't have, as I'll go through, Richard Byrne claims they don't really have strong knowledge of physics or mechanism or anything like that, okay? We've given a few examples where they clearly do, but those are rare. They're almost counter examples to some degree. Now at this point, if you take David Deutsch as simply meaning, oh, they don't have human level explanation, then what he's saying is really trivially true, right? It no longer comes across as radical. So I was asking myself, you know, which one did he really mean? By the way, I just want to clarify, the word understanding can have multiple meanings. Now, David Deutsch is saying, I'm taking it to mean human level explanations, but that isn't necessarily the way Richard Byrne's gonna use the term, okay? So clearly the fans of David Deutsch think he means the more radical view.

[00:09:18]  Bruce: I think they're right. And the reason why I think they're right, and I'm open to the fact I might be wrong, but the reason why I think they're right is because he cuts off Tyler so strongly in the interview. He says, they do not. If all he's saying is, is they don't understand like a human does, it's not clear Tyler was saying, oh, they have human level explanations. I mean, it doesn't seem like at all Tyler meant that, right?

[00:09:40]  Cameo: Right, well, and that would be a fairly silly thing for anybody to say.

[00:09:44]  Bruce: Right, right.

[00:09:46]  Cameo: Even your sister probably doesn't believe that.

[00:09:48]  Bruce: Right, so I think David Deutsch was, I think the fans of David Deutsch have him right that at least speculatively, he believes that animals are just computer programs that are automatic and there's no real learning. They actually have fixed knowledge in their genes. We've talked about how that can't possibly be true, that they have to be able to pick up adapted information from learning, we might as well just call that knowledge. But this, I think does clarify what his view was. Here's the more important thing though, is that this is a different understanding of the word understanding than Byrne. Let me give you a quote from Byrne. I've already think I've given this quote before, explaining why he uses the word insight to describe animal mentalizing. Insight remains in everyday usage as a down-to-earth lay term for deep shrewd or discerning kind of understanding.

[00:10:32]  Bruce: Let's not see them necessarily as completely at odds with each other. If you wanna read David Deutsch in the more trivially true sense, then they're not in contradiction. But if you wanna read him in the more radical sense, then yes, they are in contradiction and they have completely different theories. Now, let's kind of answer the question, which one's right and how do we know? Okay, well, by now it should be obvious from previous episodes, I think Richard Byrne's right. I think Richard Byrne's got this right. He's done way more research into it, but let me explain how he looks at it all the way. So he defines insight as the ability to inspect and manipulate a mental representation [of] some part of the world when away from any opportunity to see or in any way perceive it directly. That probably doesn't sound that impressive at first. Like maybe you think all animals can do that, but what Richard Byrne is saying is that most animals can't.

[00:11:21]  Cameo: Okay, they have a concept and hold a concept in their mind even when they're not able to see it.

[00:11:28]  Bruce: Correct. So Richard Byrne says, my view that there's a few animals that can do this and most can't is gets people upset. And he talks about cognitive psychologists. We're not talking about people who work with animals, but human cognitive psychologists. He says, cognitive psychologists study humans and they don't much bother with animals. The assumption is that the use of mental representations is unavailable to animals, most likely because mental representations are causally related to possession of language, which animal species lack. The suggestion that the capacity for insight evolved earlier than language and so it may be shared with other living species would be regarded as a pretty dangerous one by most cognitive and comparative psychologists. It's disturbing to comparative psychologists because it suggests that non-human animals may be able to think after all.

[00:12:14]  Bruce: This is from page three of Evolving Insight. Okay, so he's first explaining why there are many scientists in the human cognitive psychologist field that would be very bothered by his findings, okay? Then he goes on and he says, actually, I offend both sides. And he says, it simply never occurs to most non-psychologist people, including most professional zoologists that animals have anything qualitatively different inside their heads than humans do. If people compute with mental representations, then presumably all animals must also do. So there is a risk I may offend them too by arguing in this book that the most animals lack insight into how things work. So I just thought that was really funny, but this is, he's walking a difficult line here. The prevailing theory amongst human psychologists is animals have no insight, and the prevailing theory amongst, say, zoologists is, oh yeah, of course they do, right?

[00:13:08]  Bruce: Or pet owners, or lay people, or basically anybody else. Byrne's whole career is to figure out when insight evolved in animals, and this proves very difficult because the evidence is mixed. Animals that have no insight may, under some narrow circumstance, show something similar to it. And like our squirrel example from the very first episode where he was able to figure out how to cut the line to drop the food, I don't think there's any real chance squirrels have insight. And yet clearly in that one circumstance, it acted as if it did. So this is why it's so hard to study. That was why I emphasize Byrne's methodology as being repetitive, where you have to repeat over and over again. And then if you consistently find when dogs act like they have insight, we can actually explain it through trial and error learning. But when chimps do it, we can't.

[00:13:49]  Bruce: There's examples where we can't. Only then does he conclude, okay, chimps must have insight. Very good methodology. Okay, so now what is insight? Here's the canonical example, all right? Sultan the Chimp had learned to use sticks to get food that was out of reach. So one day, they decide to give Sultan two short sticks, neither long enough to reach the food. So he plays around with each stick, he can't get to the food, he gives up and he goes somewhere else, still playing with the sticks or whatever. Then this is the quote from Byrne, then insight dawned, while fiddling with the sticks for no apparent purpose, Sultan happened to push two sticks together and they held, making a longer stick. Sultan suddenly became animated, took the sticks across to the out of reach food and used his new combined tool to reach it with immediate success. Now, what Byrne emphasizes here is that Sultan had to return to the site of the food to use the sticks.

[00:14:43]  Bruce: If this had all happened and it was all right in front of the food, we would have assumed it was trial and error learning, that just by chance he happened to [hook?] them together, and then just by chance happened to use it. But because he had to put them together and then had to have a mental representation, oh, I can use this to go get the food, then had to go back to get the food, that's why he believes this is an example of insight. Sultan was able to think about his problem even when not engaging in the task, because he has a mental representation of its structure in his mind. He saw the significance, the longer stick was relevant to the unsolved food problem. He computed a solution to the problem in his mind.

[00:15:21]  Unidentified speaker: Okay. Yeah, I can get behind that.

[00:15:23]  Bruce: Now, that was actually, I think, the first experiment, if I remember from Byrne's book correctly, that had ever happened where they started to wonder if chimps had insight. Okay. That had led to a lot of the studies that Byrne later goes on to do. Here's another one that I'm taking from Nicholas Christakis' book Blueprint. He says, by the way, this is almost the same as the monkey example I used in the very first episode, but it's more impressive. Frans de Waal took this same idea of the cooperating monkeys from the first episode, and tried it with elephants. This time, careful controls were put in place to see if the elephants really understood that cooperation was required. Remember, they have to cooperate to get the food.

[00:16:01]  Unidentified speaker: Yeah.

[00:16:02]  Bruce: The experiment was to put two elephants in different lanes and offer them a rope connected to a table. If they both pull on the rope, the table would move food towards them until they could eat. The elephants quickly learned to cooperate. But was it possible that they just learned to pull the rope, and they were unaware they were cooperating? So they released the elephants at different times. If the elephant started pulling without the second elephant, then they'd know the cooperation was just coincidence. But the first elephant would wait for the second elephant before starting to pull. But what if the elephant was just learning something like, pull rope when near another elephant? So de Waal cleverly tried releasing both elephants, but one of the two ropes coiled up visible but out of reach. In this case, neither elephant made any attempt to pull the rope because they apparently knew it wasn't possible to get food without cooperation of the other pulling the rope.

[00:16:49]  Cameo: Oh, interesting.

[00:16:51]  Bruce: Now, unlike the Sultan example, which is just one example, this is a repeatable experiment.

[00:16:55]  Unidentified speaker: Yeah.

[00:16:56]  Bruce: This is really try to come up with an alternate explanation besides insight for this. I don't know how you would do it.

[00:17:01]  Cameo: I don't either. It seems pretty obvious that they're using insights to understand what they have to do.

[00:17:08]  Unidentified speaker: Yes.

[00:17:09]  Cameo: They're not getting lucky.

[00:17:11]  Bruce: Yes. Okay. Now, this leads us to the concept of behavior parsing. And you asked a question in the last episode, why does David Deutsch disagree with Richard Byrne on this? So I'm going to actually get into this. And what we're going to see is that Deutsch never really says anything wrong. He's like, right about everything he says, but he's missing the bigger picture, if that makes any sense. So behavior parsing, Deutsch uses it, Richard Byrne’s studies into behavior parsing as proof that animals don't understand things. And Richard Byrne does explain behavior parsing in very mechanical terms intentionally. Now remember, Richard Byrne's whole purpose is to start with the assumption that anything that can be explained mechanically will be explained mechanically. So Richard Byrne, based on his studies, there's these complex maneuvers that you have to do, that the chimps have to do, to be able to take food with nettles and then remove the nettles so it can be eaten and it doesn't hurt.

[00:18:07]  Bruce: And it can't possibly be just a mechanical set of trained maneuvers because you have to, it's a whole program. You have to do this a certain number of times until this happens. Then you have to link in the next part of the program. And then you may have to go back to the other ones. Like a whole flow chart has to exist between the mechanical maneuvers. So Richard Byrne, based on his studies, he believes that the individual maneuvers, the mechanical maneuvers that are done individually, are learned entirely by just mimicry. That the infant sees the mother and possibly the alpha male, the other chimps will never eat around another chimp. So he only has two examples. He gets to see it hundreds of times before he has to do it for his first time. So Richard Byrne believes that the right explanation there is that they're just mechanically learning to see lots of different maneuvers that accomplish something.

[00:18:59]  Bruce: And then they're statistically pulling out what the correct maneuver is for themselves. Okay. And he even produces evidence of this by showing that they pick up the idiosyncrasies of the two chimps that they see and not of any of the other chimps. On the other hand, it's not complete mimicry. They usually have their own kind of way of going about it. There's their own idiosyncrasies also that are involved. Deutsch takes the fact that this is a mechanical process and says, see, apes don't understand anything. So let's go ahead and read what Deutsch actually says. Apes are capable of recognizing a much larger set of possible meanings. Some of them are so complex that aping has often been misinterpreted as evidence of human-like understanding. Note the word human-like understanding there. This is something that Deutsch really always emphasizes human-like understanding, which is why it's sometimes hard to know if he's making a radical statement or a trivially true statement.

[00:19:50]  Bruce: For example, when an ape learns a new method of cracking nuts by hitting them with rocks, it does not then play the movements required to crack the nut sequence like a parrot does. The movements required to crack the nut are never the same twice. The ape has to aim the rock at the nut, it must have to chase the nut and fetch it back if it rolls away, has to keep hitting it until it cracks rather than a fixed number of times and so on. By the way, I think it's interesting that he attributes dumb mimicry to parrots when we've actually seen that some parrots actually understand the concept of numbers. Parrots are one of the animals that very likely does have insight. Such activities may seem to depend on explanation, on understanding how and why each action within the complex behavior has to fit in with the other actions in order to achieve the overall purpose.

[00:20:37]  Bruce: But recent discoveries have been revealed how apes are able to imitate such behaviors without ever creating any explanatory knowledge. In a remarkable series of observational and theoretical studies, the evolutionary psychologist and animal behavior researcher, Richard Byrne, has shown how they achieve this by a process that he calls behavior parsing, which is analogous to the grammatical analysis of parsing of human speech or computer programs. Okay, he goes on to emphasize that this is a very slow process compared to humans. Okay, that because it's a mechanical process, they have to generalize from hundreds of examples. Whereas with the human, I could explain to you what I'm doing and maybe show you once or twice and you'd get it because you would then form an explanation of what I meant in your head. It would likely be correct or you would ask questions till it was corrected.

[00:21:23]  Bruce: And then you would be able to do it in your own way. You wouldn't have to mimic me. Okay, so far so good. I don't think there's necessarily any disagreement between Byrne and Deutsch here unless, because he's specifically talking about human-like understanding, right? Byrne doesn't believe they have human-like understanding. Okay, so now let me give you an example of programming imitation, one that Byrne uses from his book, Evolving Insight. That's very interesting. So we have Supinah the Orangutan. So Supinah has a goal. I believe it's a she, I'm going to say she, to use soap and laundry for fun. Okay, so Supinah has seen the camp staff wash clothes with soap and do laundry, and wants to go wash the laundry for herself, because she's seen this and it looked fun. Okay?

[00:22:12]  Unidentified speaker: Sure.

[00:22:13]  Bruce: They wash soap and laundry at the camp dock, and there's a guard protecting them because the staff is afraid of the orangutan. So Supinah is afraid of the guard. Supinah comes up with a way to bypass the guard, and here is the plan she comes up with using program imitation. Supinah steals a canoe full of water by untying it. She stops to go check on the guard, and seeing that the guard is still there, continues. She rocks the canoe to get the water out. She reorients the canoe and then propels it forward. When she arrives to the dock, when she arrives, the staff jumps into the water to escape, and then she goes about using the soap to clean the laundry, rubs the soap on the wet clothes, brushes with soap. She scrubs the clothes with brushing. She wrings the clothes, the wet clothes out, et cetera. This is a program level imitation. Now, notice that there is no human that has done this.

[00:23:06]  Bruce: Okay, there is true novelty in the program that Supinah has put together. No human would have to go steal a boat to go be able to wash clothes. So Supinah was able to start with a goal, I want to wash the laundry and soap, and she's not just mimicking how to wash laundry and soap, she's actually stringing together actions she's seen elsewhere, but in a novel new way that she's never seen before, so that she can go accomplish a goal. This is what behavior parsing and program level imitation really is. The behavior parsing part is mechanical, the program level imitation requires insight.

[00:23:44]  Cameo: I agree.

[00:23:44]  Bruce: And this is the point that Byrne is trying to make. Also, I want to point out that, whereas Deutsch emphasized that you need hundreds of examples for a great ape to be able to mimic a behavior, a mechanical behavior, I kind of doubt Supinah the Ape had watched hundreds of people rock a canoe to get water out. So when we talk about it requiring hundreds of examples, they have hundreds of examples. That's really Byrne's point. And he's saying that's sufficient. We know that's sufficient, right? Because there's hundreds of examples. We're not necessarily saying they have to have hundreds of examples. That's an open question. So that is another difference where I feel Deutsch may have slightly misunderstood Byrne's purpose there. They normally probably do require hundreds of examples if they're actually trying to learn to eat in the wild, but Supinah clearly was picking up something in a far fewer number of examples.

[00:24:40]  Bruce: Supinah the orangutan could have learned each step by imitation. Every single one of the steps, untying the boat, rocking the boat to get the water out, each of the points, each of the steps in terms of trying to wash things, going across the river. All of those could be learned by imitation, but the flexible intelligence is at the level of the program, not at the individual action level. It does not require human level explanations to be able to do what she's doing. It does require some sort of mentalizing, but it doesn't require that you understand, say, why rocking the boat gets water out. Okay, you just have to know, rock the boat, the water will go away. Okay, do you understand the difference there?

[00:25:18]  Unidentified speaker: Yeah, yeah, I do.

[00:25:19]  Bruce: So behavior parsing can thereby be understood, this is from Byrne's paper, Imitation is Behavior Parsing. Behavior can therefore be understood statistically in terms of its correlations, circumstances of use, effects on the environment without understanding of intentions beyond the two mentioned, of like the goals and trying to accomplish a goal, without the intentions or the everyday physics of cause and effect. So he is agreeing that behavior parsing requires, does not require any sort of true cause and effect understanding of why rocking the boat gets the water out or something like that. The theory is that apes and humans can learn certain actions, by the way, Byrne is the one who says and humans, Deutsch doesn't agree with that. So this is, I'm giving you Byrne's point of view here, can learn that certain actions statistically create certain outcomes and then they can flexibly paste these actions together into programs to accomplish something.

[00:26:12]  Bruce: So animals can generalize from a few hundred examples and mimic the individual steps. How do they do that? Well, he says it statistically. What we're talking about is induction, right? And this is where I feel like there's still lots of open questions and Deutsch is maybe too quick to just assume we understand what's going on here. Although I don't really disagree with his overall point. How do apes watch, let's even say a few hundred examples and then understand what it is they need to do? Okay, or in the case of Supinah with the boat, maybe just a dozen examples. The idea is that they're inductively like machine learning. They're not the [Baconian?] induction, but they're coming up with what is the, you know, what are the steps I need to do? How do I go about this? And they're coming up with this maneuver that they then know how to do, okay.

[00:27:00]  Bruce: Byrne is certainly acting like it's just obvious that it's possible to do that, but we don't really know how to program that. Like we could do it if we knew exactly, okay, I'm going to program to rock the boat to get the water out. But again, animals can set their own decisions on what they're going to learn, right? So there's something more going on there. And we don't have example of an algorithm that could explain it, if that makes any sense.

[00:27:23]  Cameo: It does, it does.

[00:27:25]  Bruce: Okay, now for people listening, you're not gonna be able to see this, but Cameo will be able to see it. This is a flow chart and I'll like put it up on my blog so there's a link in the show notes. And this is the flow chart of what a typical program, real life in the wild program imitation would be like. And it's a giant flow chart, it's like a program, right? First you pull something into range, then you grip the stems loosely near the base, slide up. Is that enough? If no, then do it again. If yes, then grip the base of the leaf. I won't read the whole thing, but there is a whole bunch of decision points and repeats and flow from here to there, wait until you reach a certain point and then flow there. Okay, this is program level imitation and it is what is required for chimps in the wild to be able to eat. They would die if they couldn't pick up program level imitation from their mother and the alpha chimp.

[00:28:14]  Bruce: Because the food that's available to them, there's like lots of different food available to them, but most of the food monkeys can go take. They've got to be able to, and can get there faster because they're smaller. So chimps have to be able to get at food sources that monkeys can't, don't have enough insight to be able to get. Does that make sense? And Byrne believes this is what led to the evolution of insight. Well, one of the things that led to the evolution of insight was the fact that to be able to survive, they needed to develop that kind of program level imitation. Okay, so Deutsch's point, animals do not have human level understanding or full explanatory power. Like I said, we learned from far fewer examples. We could have learned a lot of this in far fewer examples than Supinah the Orangutan could. However, according to Byrne, insight is a sort of very simple explanation, but it's based solely on correlations.

[00:29:01]  Bruce: So here's again, quoting Byrne from his book is Evolving Insight, Behavior parsing gives the ability to see below the surface of behavior and detect the logical organization that produced it. Behavior parsing picks out the correlation structure of a changing environment quite well. Cause as correlation is the valuable everyday way of representing reality. A reliable correlation of this kind might be described as a pretty good cause, and only physicists or philosophers dealing with the fundamentals of matter may need to go much beyond it. So Byrne's conclusion is that animals have insight [or] mentalizing, and behavior parsing is not to be understood as a purely mechanical behavior. Just the steps in it are to be understood as purely mechanical behavior. And he sees what the apes, what the Supinah of the ape is doing, or the orangutan is doing, is she's learned this maneuver produces this result.

[00:29:53]  Bruce: I do an input, I do an out, and it gets an output. His assumption is they're not getting anything else, that it is literally just cause as correlation. This is what his theory is anyhow. So they don't have true causes beyond, I have a goal I need to accomplish, I have maneuvers I can do that have an input and an output, and I've got the ability to string them together through mentalizing to form a program that will accomplish my goal.

[00:30:19]  Unidentified speaker: Yeah.

[00:30:19]  Bruce: Okay, now going back to Deutsch in the beginning of infinity, an ape parses a continuous stream of behavior that it witnesses into individual elements, each of which is already known genetically how to imitate. Okay, I'm going to do fact check now on the things that Deutsch says. We're going to find he for the most part gets it right, but sometimes maybe doesn't quite get the full context. So what he just said here is strictly true, but Byrne points out, so chimps in the wild, they believe that the gestures that chimps use are all genetically known. Okay, so they have this set of gestures, they only, in the wild, they only use those gestures, and so it's always the same ones, and even if they're not raised by each other, they can still learn the same ones, or they still have the same ones. So they believe it's basically a genetic program, that they can have the set of gestures, they can choose the gesture, but they don't have to ever actually learn the individual gestures.

[00:31:13]  Bruce: So what Deutsch is saying here is strictly true. However, Byrne points out that apes in ASL studies, so not in the wild, can and do learn how to use gestures that they don't already genetically know. They just don't do this in the wild. So apes are not limited to genetic gestures and can learn new ones. They just don't, except when humans make them, basically. This is actually an important point, because if they only had genetic gestures, that really would be, that would really challenge the idea of them having insight. But if they do have insight, then they should be able to learn new ones, even if they don't, and in fact, they do. And this is something that he points out, is like chimps do a whole bunch of things in the wild that show insight, where most other great apes don't. But the moment you take them into captivity, the great apes can all do the same things a chimp can do.

[00:32:01]  Bruce: They just don't happen to do it in the wild. And it is, it is fascinating. Okay, he talks about how chimps and great apes seem to have excess intelligence for their environment. They don't need, they're able to do things that they don't need to do. So continuing with David Deutsch, it turns out that in every known case of complex behavior in non-humans, the necessary information can be obtained merely by watching the behavior many times and looking out for simple statistical patterns. It is a very inefficient method requiring a lot of watching behavior that a human could mimic almost immediately by understanding their purpose. Okay, so fact check. So think of Byrne's example of Sultan the Chimp, having insight dawning and putting the sticks together, or Supinah going out and washing clothes, okay? These are examples where it did not require a few hundred examples.

[00:32:48]  Bruce: So this is why I say, Deutsch is assuming it requires a few hundred examples, where what Byrne's really saying is, they usually have a few hundred examples, so that's probably sufficient. But he's not necessarily saying it requires it. Also, these are not mechanical examples, right? The individual gestures are, but the actual goal is novel. So Deutsch is correct for the individual set of actions that they have to be learned by watching and learning patterns, but the apes show real insight to figure out how to string those actions together into a set of actions, often needing to repeat some of them, to accomplish both an overall goal and the sub-goals that work towards that goal, okay? They have to be able to see, as Byrne put it, see beneath the surface of the individual moves and come to an understanding, oh, I have to do this one to get that output, I have to do this one to get that output, and then figure out how to string those together, okay?

[00:33:42]  Bruce: Apes are unable to imitate sound. This is continuing with Deutsch. They cannot even parrot sounds, repeat them blindly, despite having a complex inborn repertoire of calls that they can make, recognize, and act upon in genetically predetermined ways. Okay, so this specific example is true, but it seems to me like this is a misleading statement. Now, David Deutsch did run his summary of Byrne's studies across Byrne, okay? And Byrne okayed them. So for whatever reason, Byrne felt like this one was okay. But let me just explain why I feel like it's a little misleading to the average reader. So Byrne at length points out that apes have no ability to learn audible calls, but he also at length points out that they can learn new communication gestures through ASL and that they use them with intent. So it's true that the inborn repertoire of calls of apes is genetically predetermined and even that their gestures in the wild, they're genetically predetermined, but Deutsch is using it to show they don't understand when in fact, they can go beyond that.

[00:34:42]  Bruce: And we know that from studies, right? Which suggests that they do understand something, that they're not just completely mechanical. Then Deutsch continues, the ape avoids infinite ambiguity and copying by already knowing inexplicitly the meaning of every action that it is capable of copying. And it is only capable of associating one meaning with each action that it can copy. Okay, so I have to, this one I have to break down a little bit. So I am not aware of anywhere Byrne claims that the meaning of each action is genetically already known. In fact, he claims that the animals have to learn which gestures imply which meanings. So like young apes, they have these genetically predetermined gestures, but they don't know which one means which. So a young ape will try every gesture and then over time start to figure out which ones mean what. And then eventually they stop trying every gesture and they try just the ones that convey their actual intent to the other ape.

[00:35:34]  Bruce: Okay, so to me that seems different than what it seems like Deutsch is saying here. And then I also didn't find anywhere where he claims that they only associate one meaning with each action.

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

[00:35:45]  Bruce: But that one seems reasonable to me. Even though I never found Byrne saying that specifically, that seems to fit the idea of correlation as causation. You've got one input, you've got one output. So my guess is that Deutsch has this one right, okay? So Byrne's theory is that animals learn something like this. Perform behavior A will cause B. So a simple correlative cause and effect. I must repeat C until sufficiently D, repetition of behavior until a goal is reached, or I must string B and D in order to accomplish E, program level imitation. This is what Byrne believes animals are capable of doing. Nothing really beyond that. So now, from Interview with Tyler. Tyler, do chimpanzees understand in your view? David Deutsch, no one knows, but they show virtually no sign of understanding anything. So then he covers Richard Byrne's studies with wild gorillas, and this is again, quoting David Deutsch.

[00:36:36]  Bruce: Byrne did some ingenious experiments, or rather observations to try to determine whether they understand, apes understand, gorillas understand, why they are doing each particular action. Currently, these gorillas are prone to certain injury, which disables their thumb. When you've disabled your thumb, one of those motions learned via memes becomes irrelevant, and the others become less effective. But the gorillas that have learned how to do the thing will make the motion, the inefficient motion, again and again, every single time. Okay, does this show a lack of understanding? Absolutely, okay. It is, I understand exactly why Deutsch is raising this as an example of gorillas just don't seem to understand much. Okay, however, this is actually consistent with Byrne's theory, and that's the point I'm trying to make, okay, is that Byrne does not believe that they understand the meaning of why the individual gestures work, but rather they understand how to string them together intelligently.

[00:37:31]  Bruce: So Byrne, recall that he said, behavior can thereby be understood statistically in terms of its correlations, circumstances of use, effects on the environment, without understanding the intentions beyond the goal [or?] the everyday physics of cause and effect. So Deutsch isn't saying anything here that is at odds with Byrne's theory. It's what you might call a non-differentiator. Okay, he's creating a narrative of, see, they don't understand this, therefore I'm kind of inductively concluding that they just don't understand things. And he's missing the fact that Byrne agrees with him. They don't understand the meaning of the individual gesture, but they do in some sense understand, I do this gesture, I get this output, I can string these together, I can set goals. Their insight operates at a different level. Okay, we've talked about Byrne's methodology.

[00:38:18]  Bruce: It's incredibly rigorous. Literally anything that can be explained as not needing insight is just assumed to not need insight, okay? I think this is what really attracted Deutsch to Richard Byrne's theory, because Richard Byrne spends considerable time debunking that animals understand things. I've read the quotes, we've seen that he does this, okay? But his main point is that insight is required to string them together into a program level imitation, that insight is required to do a lot of the things that they do, it's only a smaller group, it's not most animals. This is Byrne's main point, all right? Sure, he's emphasizing the lack of understanding in most cases, but he does that because that's his methodology. He's trying to find the examples that don't fit that explanation, and he's finding them. He's finding examples where some animals must understand things to some degree beyond simple trial and error learning.

[00:39:14]  Bruce: I think this is how the two of them see it different. Deutsch sees behavior parsing as evidence that animals have no understanding, while Byrne sees it as evidence that apes do have mental models that they use. So Tyler asked David why animals and human intelligence isn't a continuum. So David Deutsch had just barely said that animals don't understand anything, but humans have universal understandings. And Deutsch says, I don't think it could be a discrete break because evolution would happen gradually. My best guess, I think what happened is that the capacity of the brain to store memes, to store programs in the brain rather than the genes, increased for some reason very fast because of some reason these memes were very valuable. So the capacity for memes increased rapidly. Okay, just as a side note, so far, this is what Byrne says. Byrne's trying to say, to be able to learn how to eat and to be able to feed themselves, they had to develop these more complex memes that were at the program level instead of just the individual gesture level.

[00:40:12]  Bruce: Continuing with Deutsch, once memes go beyond a certain complexity that cannot be copied, all we can do is look at the behavior and guess what the purpose was. Complex memes have to be transmitted like that rather than by aping. And then there came a moment when our species was capable of explanatory knowledge, but they never used it for further tens or hundreds of thousands of years. Okay, so the thing I really want to kind of emphasize here is first of all, David Deutsch is correct. According to Byrne's theory, this is what happened, is that they started needing these more complex memes that couldn't be mimicked entirely. So, but here's the difference. Byrne is saying, so they learned to mimic at the program level and then to understand how to string things together and that required insight. This is what Byrne is saying. So Byrne is creating an in-between.

[00:40:57]  Bruce: Whereas David Deutsch thinks he's got dumb animals and humans with explanations, Byrne is specifically claiming there's an in-between point. There's another jump in-between that had to happen first and that was required before human intelligence could evolve. Now, what are problems with this theory at this point? Let's kind of just be honest about this. So Byrne never really explains how behavior parsing actually works. He just sort of assumes that if you see a few hundred examples, that ought to be enough for an animal to statistically figure out which parts of the behavior matter and to be able to mimic it from there. Because induction works, I guess. I don't think we've got a good explanation there. I think that that's a more computation, that's a more computer scientist type question. It's just not what Byrne's interested in. He doesn't address how Supinah learned to rock a boat to get water out in very few examples.

[00:41:46]  Bruce: There must be something more going on there. So we don't really have a truly good explanation, even with Byrne's explanation. There's more that we could dig into and try to understand this better. And as I mentioned, the current machine learning techniques would require hundreds, require thousands, or even hundreds of thousands of observations to generalize. So even when we are talking about animals being stupid, that they need hundreds of examples to learn something, that's still super impressive. It's like beyond impressive compared to the algorithms that we know how to make. So there's also the mystery of how do animals, even if it does require a few hundred examples, how do animals do that? Because we don't know how. We don't know of any algorithm that allows that to happen. So Deutsch in Beginning of Infinity now says this, my guess is that every AI is a person, a general purpose explainer.

[00:42:34]  Bruce: It is conceivable that there are other levels of universality between AI and universal explainer, AI meaning robots that don't know anything, and perhaps separate levels of those associated attributes like consciousness. But those attributes all seem to have arrived in one jump to universality in humans. And although we have little explanation of any of them, I know of no plausible argument that they are at, there are different levels or can be achieved independently of each other. So I tentatively assume they cannot. In any case, we should expect AI, AGI is what he really means here, to be achieved in a jump to universality, starting with something, starting from something much less powerful. This is the quote that has led people to believe animals aren't sentient, those that are fans of David Deutsch, because he basically here says, for in humans, we saw this jump all at once of consciousness, and so what's associated with consciousness and universal explainer in a single jump.

[00:43:33]  Bruce: What I'm going to suggest is that David Deutsch, he says he doesn't know of a plausible argument. Well, I think Byrne's is a plausible argument. I think David Deutsch is just not aware of what Byrne's full argument is. I think this is an open question at this point. We really have to question everything at this point and look into this more deeply. Byrne says, I propose that insight evolved twice, and that both kinds of insight are available to adult modern humans. Non-human great apes, with their intentions limited to expected results, can do very well at non-verbal experimental tests of insight. Byrne is saying there are two kinds of insight, animal insight and human insight. Humans have both. You have to have the animal one before you can gain the human one. It had to come first. It was to use an analogy from computational theory, you have the finite state machine, you've got the pushdown automata, and then you've got the Turing machine.

[00:44:30]  Bruce: Each of those is a different level of universality with the Turing machine being the one that finally encompasses every possible algorithm. He is claiming that normal animals are finite state machines. Byrne is claiming normal animals are finite state machines. Animals with insight like great apes are pushdown automata and humans are Turing machines. That is what Byrne is claiming. That's an analogy obviously that I'm making up. Byrne makes it clear that he sees behavior parsing as a form of insight and a necessary precursor to human level explanations. He also points out that his theory makes testable predictions. Now, this is what I find interesting. He says all nonverbal tests of theory of mind will eventually be passed by nonhuman great apes, and all nonverbal tests of cause and effect logic will eventually be passed by nonhuman great apes.

[00:45:17]  Bruce: So his theory has testable consequences that we can actually eventually see if he's right or not, basically. Does that make sense?

[00:45:25]  Cameo: Yeah, totally makes sense.

[00:45:27]  Bruce: As of yet, they haven't passed all these tests yet. So he's claiming that they eventually will once we figure out how to, it's sometimes hard with animals to figure out how to test them because they have their own interests and they may choose to ignore you and things like that.

[00:45:39]  Cameo: Some of what they do, I think, can be easily misinterpreted.

[00:45:42]  Bruce: Yeah. So now here's another one. Human language is itself a strong example of behavior parsing, or at least maybe. We do not, for the most part, teach children how to speak via explanation. We don't say this word means this, and we don't explain it to them. They really just learn by hearing others speak and generalizing. In fact, in exactly the same way we're talking about behavior parsing. They hear it lots of different times and statistically they pick up what it means. Inductively, I guess, I don't know. We don't know how, right? So behavior parsing, one could argue, is required to learn language and is therefore a precursor to universal explanatory abilities. And I actually, though Byrne never specifically says what I just said, I think this is what he's getting at. On the other hand, to defend the Deutsch view, one might argue that we learn language by explaining words to ourselves.

[00:46:30]  Bruce: Now, is this possible before we—language is acquired? I don't know. Maybe if explanations aren't, if you can do explanations without language, maybe you could still fit Deutsch's view to what we actually see, and it doesn't have to be Byrne's view. I would consider this two competitive theories that need to stay competitive for the moment. We don't want to get one over the other at the moment. Now, keep in mind, though, that we do apparently use language to explain things in our own head. If children used lingual explanations to learn language, that would be an infinite regress. So we would be claiming that explanations are different than lingual explanations in some sense. If you're into AGI and you agree with Deutsch on this, that would be a hint now about AGI. On the other hand, if Deutsch is wrong, that would be a mislead on AGI. This is why it's important that we actually take all the theories seriously and try to work things out.

[00:47:20]  Bruce: So it's possible that behavior parsing may be necessary, but insufficient to learn language, because otherwise apes could learn language, right? So we know it must be insufficient. Now let's talk about, I haven't gotten into this one, but let's talk about self-awareness, which is something that Byrne talks about, and it seems to be a big interest of Deutsch's also, although they're completely different on this one. So Deutsch has criticized the mirror test a number of times. I even responded to him on Twitter when he was talking about, someone ought to just make a cell phone that recognizes itself in the mirror, and we can prove that the fact that you can pass the mirror test does not prove that you have self-awareness. He says, yeah, it would be trivial for software to pass, so he doubts that the fact that some animals can pass the mirror test shows us really anything.

[00:48:09]  Bruce: Now, I brought up on Twitter, on the place where he said this, I said, you know, the only problem, I agree with what you're saying, but the only problem with what you're saying is, is that if someone programs a piece of software specifically to recognize itself in the mirror, then there's a human programming it to recognize itself in the mirror, and then they have an explanation of why it can do it. We're kind of starting with the assumption that there's no evolutionary purpose to recognize yourself in a mirror. And if that's the case, that animals that can recognize themselves in a mirror, they're doing it by some other means completely different than the way a smartphone would do it.

[00:48:47]  Unidentified speaker: Yes, right.

[00:48:49]  Cameo: There is not an agent that is teaching them how to do the thing, or training them to do the thing.

[00:48:55]  Bruce: Yes, and they don't have mirrors in the wild. I mean, mirrors don't exist in the wild, so there would be no evolutionary purpose for them to recognize themselves in the mirror. Now, somebody came back and gave a decent criticism. They said, um, what if, like, they needed to learn to recognize themselves in the mirror because they would otherwise get spooked at their reflection in the water? And I said, that's actually a good point, but that's testable. We can't just assume you've now taken care of things. We should now find the implication of what you just said, if you take it seriously as a theory, is that we should find that all animals that have to drink from pools of water will pass the mirror test. Otherwise, your explanation doesn't make sense. Nobody responded to me after that, and that conversation was kind of done.

[00:49:40]  Cameo: And there are not very many animals that can recognize themselves in water.

[00:49:45]  Bruce: Yeah, spoiler, that is correct. Okay, so what I'm going to do is I'm going to refer to the Deutsch mirror test as the smartphone mirror test. Okay, that would be you program a smartphone to recognize itself in the mirror. It's trivial, it's not that hard, and he's right that we could do that, and that would require absolutely no insight at all to be able to do that. Therefore, it is at least possible to pass the mirror test using the smartphone mirror test approach, okay? Now, question, how many animals are known to be able to pass the smartphone mirror test? Okay, where they just recognize themselves in a mirror on their first try, which is part of the smartphone mirror test. So, here's Byrne's answer. When confronted with a mirror for the first time, most animals treat the image as an unfamiliar animal of the same species, conspecific. They make inappropriate social responses and often try to find where the other animal is by looking behind the glass.

[00:50:47]  Bruce: Domestic cats show these reactions. Young children show them. And adult humans who have never in their lives experience with mirrors also show them.

[00:50:57]  Unidentified speaker: Meaning even humans?

[00:50:59]  Bruce: Even humans! No animal passes the smartphone mirror test. None. With experience, however, reactions change. All humans come to realize that the image is of themselves, a result of reflection. At this point, they begin to use the mirror in distinctive new ways, examining parts of their face they cannot reach. Cats, on the other hand, do not. Their social response may habituate, but they simply ignore the mirror and show no signs of understanding the reflected images. But why is this so hard anyway, especially for an animal with paws or hands, that they can see in action? There is continual opportunity to see that the paw in the mirror appears the same as their own. There was therefore great excitement when it was found that a few species behave like humans, apparently eventually understanding the image as themselves. In other words, we're not trying to explain, we're not starting with the idea that, oh, just because they can see something in the mirror that shows they have self-awareness, we're actually showing the opposite, that the vast majority of animals can't pass the mirror test and therefore have no sense of self.

[00:52:07]  Bruce: That's actually what we're trying to explain with the mirror test. Now, if that's true, that's a conjecture. We don't know exactly why most animals can't pass the mirror test. But the conjecture, they have no sense of self, that would explain it. If animals literally just lack the concept that I'm a self, I'm a cat just like the other cats in my species, then they would never be able to pass the mirror test. It would be impossible for them to. The lack of insight would make it impossible.

[00:52:38]  Cameo: All right, so can I ask a question?

[00:52:41]  Unidentified speaker: Yes.

[00:52:43]  Cameo: Do you think that the sense of self would have, do you need a sense of self to be able to have intelligence?

[00:52:53]  Bruce: Okay, that's a good question. Now, it just so happens that the animals that pass the mirror test are for the most part the same group of animals that Byrne by experiment has demonstrated have insight. However, there are some exceptions. I've heard that some ants can pass the mirror test. I don't believe for an instant that ants have a sense of self.

[00:53:14]  Cameo: I agree.

[00:53:15]  Bruce: That may actually be a different sort of case in the case of ants. I haven't looked into those experiments. I've just had people tell me that. So I don't think the mirror test is meant as any sort of foolproof method. Let's determine if this animal has a sense of self. I think it was really intended exactly the opposite. It was trying to demonstrate that most animals do not have a sense of self. And then by implication, that would mean that an animal that does have a sense of self would have to be in the category of the ones that passed the mirror test. And then the media misrepresents. I will give Deutsch complete credit on that point. The media doesn't even come close to representing it correctly. So in many ways, Deutsch's criticisms of it would be completely correct for the media version of the mirror test, but not for the Byrne version of the mirror test, if that makes any sense.

[00:54:04]  Unidentified speaker: Okay.

[00:54:05]  Cameo: Yeah, because they're very different.

[00:54:08]  Bruce: Now, one of the things that Byrne talks about, and this story doesn't quite fit, he says that one of the theories is that you have to develop a sense of self if you're a larger animal like an ape. Like monkeys, they just swing through the trees, and they don't have to worry about how to engineer their maneuvers because they're too light. When a chimpanzee swings through the trees, it has to think a lot about its own body and what its body is doing. And so one of the theories Byrne suggests is that that may have been what caused some animals to gain a sense of self.

[00:54:45]  Unidentified speaker: Okay.

[00:54:46]  Bruce: That it was really just a solution to an evolutionary problem, they have to be able to swing through trees. Now, how good is this explanation?

[00:54:52]  Cameo: I think it's a crummy explanation, I gotta tell you.

[00:54:56]  Bruce: I mean, like, we're just making conjectures, right? But like, why do parrots then have a sense of self, you know? I mean, I don't know, right?

[00:55:04]  Cameo: And if, you know, when you're talking about apes that seem to have communication at the, that's built into their system, why, they wouldn't, the thing, the thing any animal needs to think the very least about is how it moves through the world, right? Like, we don't think about running, we don't think about, it's a very questionable concept.

[00:55:30]  Bruce: All right, fair enough. I wasn't sure I bought it, and even Byrne doesn't buy the explanation, he criticizes it also, but he throws it out there as one possibility and then he criticizes it. So I'm telling you, Byrne's really a good scientist. He really tries to criticize each of the theories and try to figure out what survives. One other thing I should mention is the way they test for the mirror test is they'll do something like, they'll put a dot on their hand, an animal's hand, and then on their forehead, but while they don't notice that it's happening. When they see the dot on their hand, they'll like remove it, but they won't remove the one on their forehead because they can't see it. Then if they go and look in the mirror and see it, then the ones that pass the mirror test will remove the dot because they know it's them in the mirror.

[00:56:13]  Bruce: They know it's their reflection. In the case of like dolphins, which do pass the mirror test, they can't do that test quite so easily, but they have seen dolphins like go up to mirrors and open up their mouth to try to look inside their own mouth. They do the exact same sorts of things that humans do when they come across a mirror for the first time. So anyhow, just wanted to explain that. Okay, so now I'm just about done. Let me just wrap up here. So we've seen that apes have an amazing level, the greater apes have an amazing level of intelligence. But they also have an amazing level of ignorance. They just don't get how powerful some of their ideas are. This is, this is Byrne's point. So quoting from Byrne, apes may build novel, hierarchical organized actions, but seem to have no concept of the power of hierarchical organization more generally.

[00:57:05]  Bruce: Example, apes can learn ASL in captivity, but they never in the wild use anything but genetically predetermined gestures. They use gestures to intentionally influence others, but never comprehend how ASL might allow them to communicate their intentions to others. They never propose or introduce new labels, because the idea of reference itself is missing. They can use reference, they can learn a language, a basic language like we talked about ASL, or the symbols, or the things we've talked about in the past episodes, but the idea of reference itself is missing. They can't figure out, oh, I can do other things through symbolic reference. They learn to deceive others and understand complex ideas like if they can't see it, if the other animal can't see it, they have no knowledge of it. But they never show any knowledge of the idea that they themselves might have been deceived.

[00:57:58]  Bruce: Well, we actually did see a couple examples of that, but I think he means in the lab. Those were like in the lab.

[00:58:03]  Cameo: But only very borderline.

[00:58:05]  Bruce: Very borderline, yes. Like the example he gives is a female ape hides the male ape that she wants to have a mate with behind a rock because the alpha can't see, can see her but not the other male ape. And then she's able to get away with mating and the alpha doesn't break it up. The alpha never, so that's deception. She's showing a degree of insight and deception. He can't see the other ape, therefore he doesn't have the knowledge, “I’m mating with it.” That's intelligent. But the alpha ape, there's never been a case of an alpha ape finding evidence later that he's been cheated on and getting upset at that point. That is just beyond what they can do with deception. They can pass a mirror test proving some concepts, some concept of themselves as an entity, but they never become embarrassed or prideful at their appearance. They are capable of pedagogy, and we talked about how they, really rarely, but some chimps have modeled behavior slowly for their infants in past episodes, but they use it so rarely they can't possibly understand how valuable pedagogy is.

[00:59:14]  Bruce: In short, great apes seem to be able to only model explanations at a single level, one input, one statistical output. They lack any sort of deeper insight beyond that, like humans have. And then let me just, this is just Byrne from The Thinking Ape, I'm going to end on this, this is his own words describing his own understanding. Monkeys and lesser apes develop tactics in the social arena which are impressively intelligent, but there is usually no evidence to suggest that the tactics are acquired by qualitatively different mechanisms than those of other animals, i.e. trial and error learning. Instead monkeys and lesser apes are quicker at learning, especially socially, but it's not different. They're just, they can pick it up.

[00:59:57]  Cameo: They're faster at trial and error than in other animals.

[01:00:01]  Unidentified speaker: Yes.

[01:00:02]  Bruce: But there is also a sharp discontinuity between one type of primate and the other, a Rubicon of cognitive capacity. This intellectual watershed lies between monkey and great apes. The great apes, especially chimpanzees, give abundant evidence of a greater depth of intelligence than possessed by any monkey, the ability that we call insight in humans. One possible way of looking at the insight of great apes is to say that apes seem to have the ability to think, plan, and compute, although admittedly only in a limited way. This is a very good summary of Byrne's theory.

[01:00:38]  Cameo: I like this as a wrap up. One thing, well, actually two concepts I had kind of written down just to touch on really briefly here. When we were talking about the ape that wanted to go do the laundry, I mean the chimpanzee, I'm very curious about the idea of fun in animals in general. Because, you know, fun is kind of this super abstract thing anyway. Like, how does, you know, this chimpanzee watching the people do the laundry and says, oh, that looks fun. But I've also seen, you know, videos of, just this last week I saw this video of a dog that somehow figured out he could stand on this, this piece of metal and slide down this little teeny hill on snow. And he's doing it over and over again. He's pushing the piece of metal up to the top of the hill with his nose, climbing on it and sliding down over and over.

[01:01:39]  Unidentified speaker: You know, he's playing.

[01:01:41]  Cameo: So in Byrne's findings and Byrne's discussion, does he talk about the concept of fun as it relates to intelligence?

[01:01:50]  Bruce: No. So the word fun was my word. I don't think he doubts that animals do things for fun. And I was actually just watching a video on a Ted Talk on YouTube where they were talking about crows and how crows will actually take a little piece of wood and then fly up in the air and surf in the air on the piece of wood. Even though they can do this with their wings, they choose to do it on the piece of wood because they enjoy it just like a human would surfing.

[01:02:17]  Unidentified speaker: Okay, crazy, but okay.

[01:02:21]  Bruce: Animals have behavior, we're getting a little bit into sentience, which I've been trying to avoid, but animals have behaviors that I don't think there is any way currently to explain by any known theory other than the assumption that they have very similar feelings to ours. And that's the way we explain it. We explain it that way because it's the only explanation we currently have. It could be wrong, right? It could be that animals feel nothing, okay? But when you see a dog sliding down and you say, or an orangutan wanting to wash clothes, which clearly has no, any sort of survival value for it.

[01:02:54]  Unidentified speaker: Sure, none at all.

[01:02:55]  Bruce: It's completely natural and correct to say, oh, the animal's having fun. It's not that you know for sure that's the case. It's just the only explanation you happen to have available.

[01:03:07]  Cameo: Well, and typically the things they might be doing, you can think, oh, I can see where that's fun. Like little kids love laundry because it's fun to have your hands in water and squish in the things. You know, we recognize it as fun because those are the same things we might do to have fun.

[01:03:23]  Bruce: That's right. Okay. They're, they're, and if I ever do an animal sentience one, it won't be the next episode. We'll, we'll wait for a while if I ever decide to do it. That is the point I'm gonna make, right? Is that there's kind of this need to explain animal behavior in terms of them having valences, feelings, things like that. And they have to be at least somewhat similar to our own. I know that fans of David Deutsch are bothered by that. And they think they have an alternative explanation that it's all robotic. There is no alternative explanation that explains all this through robotic algorithms. Period. End of story. So they don't have a competing explanation. That's actually, in a nutshell, my entire view on animal sentient. It's not that I think I know animals are sentient or not. It's that we have one explanation. I don't know how good it is, but there's no competitor at the moment.

[01:04:12]  Bruce: I can't explain why a monkey acts like it's jealous because it doesn't get paid as much as another monkey without referencing the word jealous. There's no non-feeling explanation available where I can say, well, it's trying to signal, and I mean, I could, but it wouldn't be a good explanation. It would be so clearly just something I made up to try to explain it away. Let's say we had a theory of philosophical zombies. Some humans actually were philosophical zombies, and we knew it because they told us. It's unlike the philosophical zombies where you can't tell the difference. They go, oh no, I can't feel any pain. I can just tell that my hand has damage, so I pull it away. Let's say humans were, there was a group of humans like that. Then maybe that would make sense. We would study that. We would develop a theory around how that worked, how it was different than when you had feelings.

[01:05:07]  Bruce: It would become testable between the two groups. But there are no such humans, right? We don't know if philosophical zombies are even physically possible at this point, right? So there's no reason, you can't use that as an explanation for animals because it's a non-explanation. Whereas doing things because it's trying to have fun is an explanation that is a good explanation that we already know, that it's already part of our own evolution. We understand it directly. It fits with evolutionary theory. It ties into all sorts of things. We can pinpoint where in the brain the feelings come from, and it's in the brain stem for the most part. Other than maybe suffering, they all come from the brain stem. So we know it's early in the evolution of animals that they started to feel things. We understand, we have a theory that's developed about why they have these feelings.

[01:05:55]  Bruce: It's actually curiosity. Curiosity is the explore–exploit trade-off. To be able to talk about animals intelligently and not come across like you're just completely ad-hoc-ing, you have to reference animal feelings today because it is the sole explanation that actually explains things. That's it. There's nothing special going on. I'm not trying to advocate for, therefore I know animals feel things. It's just the only way I know how to go about explaining things because it's the only explanation available. And I think that's fine. I think we should, because that is a good hard-to-vary explanation to assume they have feelings similar to ourselves. And that's an explanation that works for us. It's a good explanation for humans, so of course it's a good explanation for animals. And it's a good explanation for humans evolutionarily. It's a good explanation for humans how it ties into our evolution, how we understand our brain structure, how we understand where the feelings come from.

[01:06:49]  Bruce: Everything we know, everything we've studied, it makes perfect sense to assume other humans feel things just like you do. That's the only explanation. There is no alternative explanation. And the same is true for animals. You just apply it over for animals. You can see they're having fun, so you call it them having fun.

[01:07:05]  Unidentified speaker: Sure.

[01:07:06]  Bruce: I don't know what else to say on that subject, to be honest. I don't think Byrne would strongly in any way disagree with me on this. He's not studying that. He would accept, though, and he does talk about the valences of animals, the fact that they get the reward system. How do you explain reinforcement learning if they don't have valences that are rewards and punishments for them? You know, I mean like trial and error learning assumes the existence of valences of some sort.

[01:07:31]  Unidentified speaker: I agree.

[01:07:32]  Cameo: I agree. Well, let's put a pin in that one then, and then when we finally get back around to doing sentience, we can talk about fun and feelings as part of that.

[01:07:42]  Bruce: All right. Sounds good.

[01:07:44]  Unidentified speaker: All right.

[01:07:45]  Bruce: That wraps up our four-part episode on Richard Byrne's theories about animals. And in this episode, we went into depth about how Byrne and Deutsch understand these theories differently. The way this all started, I read what was in Beginning of Infinity, and I even have a blog post where I talked about how behavior parsing explains away many things that seem like explanations in humans. But then I pointed to the existence of like the elephant example where there was no behavior parsing because there was no imitation happening hundreds of times. And yet animals still seem to have some sort of explanation or insight going on. And I said, I don't know if behavior parsing can explain that or not. I'm going to now go study Richard Byrne's behavior parsing so that I can understand it better. And it was only then that I realized, ah, Richard Byrne has a very different understanding of his own theory.

[01:08:36]  Bruce: And that was what led to making these podcasts and trying to get the other point of view out.

[01:08:43]  Cameo: I think it's a great point of view. It'll be interesting to see if you can get people to engage with it and respond to it. Because as you've mentioned in some of the earlier episodes specific to this, the Deutschians are passionately against this. And have, if anything, it seems like, kind of picked it up as a key part of the philosophy.

[01:09:05]  Bruce: Yes, at least some of them. I mean, there's kind of two groups. I'm in one group and a bunch are in the other. But yeah, there's definitely a group that feels very passionately about this. And you know what? I don't even want to talk them out of it. Right? I mean, if they feel very passionately about that, that's fine. I don't know that they're wrong. I'm not claiming they are wrong. Right? I'm just trying to say, look, I'm just curious. I'm just trying to figure out what do we really know? How do we think we know it? I want to understand that. I'm not interested in digging into one position or the other. I've done that. It looks to me like Richard Byrne has thought this through. And I think he's right that there is a intermediate jump to universality. Now, that could be wrong. But let's get back to AGI studies. That would be something worth knowing if it were true.

[01:09:52]  Bruce: And it would be worth knowing if it wasn't true.

[01:09:54]  Unidentified speaker: Yeah, absolutely.

[01:09:55]  Bruce: Right? If, in fact, there had to be a sub-jump to universality of animal-level insight before you could get to human-level insight, if those two are related in some way but you can separate them, that's really something important to know. And the fact that he has a theory about what it is, that it's correlation as causation, that would be a really interesting starting point for trying to figure out what animal intelligence is as a step towards figuring out what AGI is.

[01:10:26]  Cameo: If he's right. But it certainly is worthy of more thought and discussion since nobody really knows how to create an AGI yet.

[01:10:35]  Bruce: Right, right. And I was talking to a PhD friend who's interested in David Deutsch, and I was talking about my interests in AGI, and I said, you know, my podcast, if you really pay attention, like 60% of the episodes are secretly about AGI. I hide it, but the reason why I'm doing animal intelligence is because I'm trying to figure out if there's something there I should learn first, right? And I've come away with the feeling, maybe I'm wrong, I've come with the feeling that maybe I should look into animal intelligence first, because it may be a stepping stone to AGI. I think there's a good theory that exists, Richard Byrne's theory, that suggests that that is the case. So that might be worth—now, presumably it's easier to figure out animal intelligence than human intelligence. So that would be a good stepping stone then. And if I'm wrong, well then animal intelligence is still in and of itself an interesting subject that we need to come to understand.

[01:11:34]  Cameo: I've known this about [you] for years, just for the record. I've known that that's what all of this is about.

[01:11:41]  Bruce: All this is about, yeah. So you really look at each episode, I'm exploring different concepts related to AGI, but I dig into each concept. The only real exceptions are like when we do Marvel or Star Wars or you know, Hindu dance or something like that.

[01:11:59]  Cameo: Even those, I think they're loosely connected.

[01:12:03]  Bruce: They're connected by nerdiness. That's certainly true.

[01:12:07]  Unidentified speaker: All right.

[01:12:07]  Cameo: Well, this was a great episode.

[01:12:09]  Bruce: All right. Thank you, Cameo.

[01:12:10]  Cameo: Thank you.

[01:12:11]  Bruce: Bye-bye. 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.

[01:13:01]  Bruce: 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 fourstrands.org. There is a donation button there that uses PayPal. Thank you.