Episode 37: Animal Intelligence and Knowledge Creation (part 1)

• 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

[00:00:10]  Bruce: Welcome to The Theory of Anything podcast. Hey, guys.

[00:00:13]  Unidentified speaker: Hello.

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

[00:00:16]  Bruce: Hey, Tracy. And we got Sadia here too, Sadia.

[00:00:20]  Sadia: Hi, Bruce.

[00:00:21]  Bruce: So this is a subject that I've been wanting to do for a while. It's something that seems like it just, it comes up multiple times in the David Deutsch fan community in particular, to a lesser degree amongst the Popper fan community, which is animal intelligence. What is animal intelligence?

How does it, how is it similar and different from human intelligence? Now, obviously humans are universal explainers, to use the David Deutsch term, and animals are not. And so that's a gigantic difference right there.

From that point though, there's lots of different opinions as to what that implies about animals. Everything from some people claiming animals are just robots and automatons, they don't feel anything, to animals are incapable of suffering, to animals are sure they create some knowledge, but it's super limited and we could probably do something similar to them using artificial intelligence today. And since I am someone who has studied, you know, at a graduate level, artificial intelligence and machine learning, I'm actually in a unique position to talk about a lot of these different claims, to kind of look at what animal intelligence is.

And I guess the short summary, this is the abstract, is that animals do all sorts of things that we can't do in artificial intelligence and that they're mysterious. We don't really understand what animal intelligence is. Never mind that we don't understand what human intelligence is, we don't even understand what animal intelligence is, although we probably are closer on that front.

[00:01:52]  Bruce: And this is one of those things where I would like to see more people interested in studying animal intelligence, because first of all, all knowledge is valuable. So the fact that animals are not universal explainers doesn't mean understanding animal intelligence won't turn out to be a valuable thing all its own. And honestly, it's just an interesting subject.

Animals are super fascinating, the things they do, the way they act, the difficulties that we have in trying to explain what they're doing and what they're thinking. Animal learning is also, it's way more general than existing machine learning or artificial intelligence algorithms. The types of things that we can teach animals to do would just be, are just completely beyond what we could do.

We could like make an artificial intelligence algorithm that could play Jenga, but we could never make a general one that you can also happen to teach to play Jenga. That's just not the way it works. You can teach an animal to play Jenga, right?

I mean, this, animals weren't evolved to play Jenga, but you can teach it to them. There's nothing like that in machine learning. Animals are doing something there that is truly mysterious when they learn to do these amazing things that we see them do and no one programmed it into their brain.

Evolution obviously has no pressure to try to teach an animal how to play Jenga directly. Therefore, the fact that they can learn to play Jenga implies that they have some sort of flexibility that goes beyond what current artificial intelligence algorithms have. Richard Byrne, David Deutsch quotes Richard Byrne.

[00:03:25]  Bruce: I don't know how to pronounce that name by the way. Maybe it's Byron or something, but I'm going to pronounce it Byrne. It's B-Y-R-N-E.

David Deutsch quotes him quite a bit in Beginning of Infinity. I became curious about him and I got one and read some of his studies and got one of his books. He's got more than one book.

I intend to read more of his books because they're really fascinating. If he is correct, now him and David Deutsch disagree on a number of things. They disagree over Richard Byrne's theory on a number of things.

They interpret Richard Byrne's theory very differently on some items. But if Richard Byrne's interpretation of his own theory is correct, then animal intelligence is a precursor to human intelligence and thus it may end up relating to AGI studies. Now, we don't know that for sure and I'm not making any strong claim here.

I'm just saying that's what Richard Byrne thinks. He could be right. He could be wrong.

And then there seems to be a great deal of interest on if animals are, quote, conscious or not. Do they feel things? Is there something it's like to be an animal?

I think all of these are questions that we can actually make progress against. Using critical rationalism, we can take a look at, looking at the studies, the observations that exist. We can actually learn about these different things.

[00:04:35]  Bruce: So now, let me tell a little bit of a story, just background on how I became interested in this. So I read Fabric of Reality, I think back in 2009. So it was before Beginning of Infinity came out, which I think came out in 2011.

Most people I know who are fans of David Deutsch, came to David Deutsch through Beginning of Infinity. And so they read that first or they had definitely read it. And he makes very specific statements about animals in Beginning of Infinity that aren't in Fabric of Reality.

So when I had just read Fabric of Reality and started to want to research the four strands that he talks about in that book, I didn't have any background that was biasing me towards a certain view of animals. So I didn't know what to think of animals. I just went out and started to look stuff up on my own about animal intelligence as part of my own studies.

Because this is still pre Beginning of Infinity where he kind of lays out a bit of a theory about animals in there. And I came across in my reading things that were just really interesting. And I didn't know what to make of them.

So for example, here's a quote from Roger Penrose from Shadows of the Mind. Shadows of the Mind was one of the books I read that was at odds with David Deutsch's theories. And I wanted to read sources that were specifically at odds with his theories to see what the criticisms of it were.

[00:05:53]  Bruce: And one of the things that he brings up is he says, I was particularly struck by a sequence on a television program in which a squirrel realized that by biting through the wire along which it was crawling, it could release a container of nuts suspended on some horizontal distance, suspended some horizontal distance away. It is hard to see how this insight could have been instinctive or any part of the squirrel's previous experience. To appreciate this positive consequence of its actions, the squirrel must have had some rudimentary understanding of the topology that was involved.

It seems to me that this was an act of genuine imagination on the part of the squirrel. OK, by the time we're done, I'll have you questioning whether Roger Penrose is right. But he's gotten the right track here, even though I don't know about squirrels.

They don't have a ton of intelligence from what I can tell. But this really is amazing. Animals do really weird things like this that is just really hard to explain without attributing to them some sort of understanding.

Where David Deutsch says that animals probably don't have any understanding. We'll get to what that means in a future episode. And then the word he uses here is the right word, insight.

Richard Byrne refers to animals. What he studies is when did animals evolve insight, which I will give you a much better definition of that in the future. But just briefly, insight would be the ability to make mental models in your head and to use them to solve problems.

[00:07:16]  Bruce: So Richard Byrne's whole studies is to try to figure out which animals are able to have insight, able to create mental models and to use them. And he has a really interesting methodology for how he tries to study that. One that's very Popperian.

He's a good scientist is what that means. Here's another one that's interesting. This comes from Nicholas Christakis.

And this is one of his books that he published, Blueprint. And he says in there, one experiment placed a pair of, I don't know how to pronounce this, capuchin monkeys in adjacent sections of a single test chamber, separated only by a mesh partition. In front of each of the monkeys was a cup holding food and a bar that when pulled would move the two cups closer.

Given the weight of the apparatus, both monkeys needed to pull the bar simultaneously and high levels of cooperation were observed. To ensure that this cooperation was not merely a chance occurrence, the researchers showed the animals performed much better when they could see each other through the partition, suggesting that the monkeys understood the other's role in the task and that they maintained their cooperation through forms of communication. Okay, again, this is really interesting.

I mean, it's not for a human, this would be, you know, very, very simple, right? But how are animals doing things like this? This goes beyond, you know, simple automatonism.

And then just an example from my sister. My sister is an animal lover. She has all sorts of animal stories for me, but she was telling me about her dog.

[00:08:47]  Bruce: This is fairly typical of dogs. It wants, if it thinks you're going to go to the car, it doesn't want to get left behind and it likes to go for a ride in the car. So it will get really super excited and nervous if it can see that people are going into the garage to go to the car.

But it feels differently if, like, you're going out the front door, you're putting on your coat to go on a walk or something like that. So the dog starts to whine and whine. And my mom, this is a particularly smart dog, I might add.

Not all dogs are this smart. My sister has stories about her previous dog that it could never do something like this. My mom comes to the dog and gives it an explanation, strange as this may sound, says, you can go in the car, but you need to first calm down and wait while we get ready.

And that dog tried to control itself and tried to stop its whining and it's just trying to do its best. Just a little tiny bit, but it's starting to calm down. And then she said, okay, now, and then it runs and jumps in the car.

Again, none of this is like super complicated intelligence compared to a human. But there's a weirdness going on there where how did this dog, was it understanding an explanation? Was it just some sort of weird classical conditioning?

You know, how do we explain this? And this is the type of thing that Richard Byrne is trying to study.

[00:10:09]  Cameo: Can I interject just a minute on the, especially on the monkey thing? Humans are super, generally are super reliant on verbal communication. And I don't know if you've ever been involved in a lot of team building stuff that's nonverbal, where you have to try and as a team accomplish tasks without being able to use verbal communication.

Those kinds of things are very challenging. The experiment that you're describing I think would be very difficult for humans. Because eventually we would get it and we would figure it out and it would create kind of a synergy between those individuals who had essentially had to overcome our dependence on language to be able to get things done.

So I just think that that's interesting because they don't have language. Those nonverbal things are probably more natural for them than they are for us.

[00:11:06]  Bruce: Interesting. Now here's another thing that I studied after reading Fabric of Reality that this one was more disappointing. Now people will laugh at me over this.

I really seriously wondered if apes that can use sign language could pass a Turing test, right? If you're familiar with the Turing test, this idea that can you tell that you're talking to a person or can you tell you're talking to a chat bot? Really nothing has ever passed the Turing test other than a human, you know, up to this point.

And I was, and you heard these stories about Koko the Ape, who was the most famous of the signing apes. And it really sounded like Koko the Ape was quite intelligent, was capable of talking. And I thought, I wonder if Koko could pass a Turing test.

So I mentioned this to my wife, and she bought me a book on Koko the Ape for Christmas, which I read. I was very disappointed, to be perfectly honest. The things that you hear in the media about Koko the Ape, they are the single best examples out of a whole lot of examples.

So, for example, they have some examples of conversations with Koko. Now, mind you, in this book, they're giving you the best conversations, like they're giving you the best conversations where Koko showed the most intelligence. OK, here's one of them.

So Barbara is one of the trainers. Barbara says, now, in the book, they put stuff in italics if they had a sign for it. And then if it's not italicized, then they said it out loud.

[00:12:37]  Bruce: So, but I'm going to just read it. And obviously, they don't sign every single word just to make that clear, but they do speak it. But then when Koko talks, it's only signs.

And so a lot of the extra words that make it into a complete sentence are missing because you don't do that in the sign language that Koko uses. So Barbara says, OK, can you tell me how gorillas talk? And Koko beats her chest, which that doesn't necessarily sound like a great answer.

But the trainers think that she would, Koko was trying to show that apes make signs, that they do things to make gestures, which we'll see later. That is exactly how apes communicate. So this could have been a good answer or maybe not.

It's hard to tell. Barbara, what do gorillas say when they're happy? Koko says, gorilla hug.

Barbara, what do you say to your baby, meaning a doll? Koko, love Koko. Barbara, what do you say to Mike when you play?

Koko, Mike, Koko love. Barbara, what scares gorillas? Koko, hat, dog.

Barbara, what do gorillas think is funny? Koko, clowns, bug. Now, these answers, you can kind of tell there's a bit of intelligence there.

I mean, this is, I don't think, I think I could probably tell Koko from a chat bot. But like, it's not even close, right? And like, why is a hat scary?

[00:14:02]  Bruce: Right? Well, can, maybe we can understand a dog. So the trainers say, oh, well, we had this incident where there was these strangers, older women with these really large hats, and they got really close to Koko and it made us nervous, and so Koko could detect that we were nervous.

So maybe that's true. Maybe Koko is repeating something from this incident. Or maybe the trainers are remembering something and somehow trying to make sense of what Koko's saying, and it's actually just a random word, right?

There's, it's so hard to tell how much Koko understands, even when you're talking sign language with her. Yes.

[00:14:36]  Sadia: I have a question here. Since I'm not familiar with how they actually, you know, how they think that Koko is understanding these things. The first thing where it says, how gorillas talk.

Like, if the gorillas really, if the gorilla talk is actually gestures, how are they able to figure out what the talk, word talk really means? And then gorilla saying that the way the gorilla talks is gestures. That makes me wonder, like, how would a gorilla understand the concept of talk if all they do is gestures?

[00:15:11]  Bruce: Yeah.

[00:15:12]  Unidentified speaker: You know what I mean?

[00:15:12]  Bruce: So, well, Koko does, has learned sign language. And that is a fairly common thing that they've done with different apes, the great apes. And they can actually, they do communicate with gestures well.

And even in the wild, they do. And they do it with intent. That's something that Richard Byrne studies have shown.

What I think what you'll find, though, is that when you teach an ape sign language, like 70% of the time, they're talking about what they want to eat. It's, they are learning new signs. There is a certain creativity there where they can actually learn new signs.

Most gorilla gestures are built in. They have a set of gestures and that's it. And they never learn any new ones.

In the wild, I'm saying. But in a lab, you can actually teach them novel gestures like American Sign Language. And they can actually learn it.

And then they can learn how to use it with intent. And this is something Richard Byrne brings out in his book. So there is a level of intelligence there in that they can learn these new gestures.

They can use it with intent. They can use it in, you know, creative new ways. But boy, is it limited.

And you're spot on, Sadia. To some degree, I have to feel like, you know, half the cool things you hear about Koko the Ape are really interpretations on the part of the researchers.

[00:16:33]  Tracy: What concerns me most is why the gorilla found the hat and not the clown scary.

[00:16:38]  Bruce: Yeah, you know, that right there proves that Koko doesn't know what she's talking about. Because clowns are scary.

[00:16:44]  Sadia: Sorry to interject, but I also wonder if the difference, you know, you wonder that why they can't have more complicated, you know, conversation or understanding. Could it be there's something to do with values that maybe, you know, in some sense, limited number of values are programmed, that the animal beyond a certain point is just not interested. So they just limit themselves.

But anyways, I just thought I just throw that.

[00:17:07]  Bruce: You know, this is an example of the mysteries of the mysteries of animals. Right. We don't understand.

And I will go into this in quite a bit more detail in in this series, in this podcast, because this is something that this is what Richard Byrne is studying. Animals do these amazing things, but they just seem to not quite get it. They're like one notch away, half a notch away from being way more intelligent than they currently are.

And they just never seem to transcend it. And I'll get into this into into future podcasts and I'll give you more specific details. It is very hard to figure out exactly what's going on with animals.

They sometimes show very strong intelligence and then sometimes they just don't. And they're seen—and Richard Byrne has a theory as to how far they can go. And I will explain his theories.

It won't be in this episode, but in a future episode. Now, another thing that I found interesting and did you guys have a chance to look at those YouTube videos that I sent out about the monkeys?

[00:18:13]  Sadia: So there was in one video, there was one young orangutan who grew up supposedly in some sort of a confinement and just didn't know because of not being with the mom. Didn't know how to hang from the trees or climb and do stuff like that. It seemed like it was really anxious and it was surprising.

I didn't know that they could cry. So the trainer who was helping that young orangutan, you know, they eventually wanted him to go back to the wild or be able to make it on his own. And it seemed like when she did try to make him, you know, hang on a rope, he started crying.

It was interesting that how there was stress and it literally was making crying noises. He didn't want to do that. So I thought that was interesting.

I had no idea.

[00:19:04]  Bruce: But did you have a chance to see the one that Frans de Waal did that about the monkey with the cucumbers and the grapes?

[00:19:12]  Unidentified speaker: I did not.

[00:19:13]  Bruce: Okay.

Let me describe that one. So there was an experiment. I actually found this from Bart.

I don't know if you guys know this, but he did a TED talk at one point. And so he put this in his TED talk. But it's a famous study that comes up all over the place.

So they have these monkeys and they're in cages. They're in captivity. They're given a task.

They're being classically conditioned. We'll talk about classical conditioning here in a second to do a certain task. Basically, if you give me a rock, I will give you a piece of food.

They got two monkeys that can see each other, but they're in separate cages. So the first monkey gives a rock and they give him a cucumber, which he gladly eats, because monkeys like cucumbers. The second monkey then gives a rock and he is given a grape, which they like grapes even more because grapes are sweet.

So then the first monkey gives a rock and he's given another cucumber. Now remember, he accepted the cucumber the first time gladly, but now he has seen that the other monkey is getting a grape in exchange for a rock. So when he's given a cucumber again, he throws it across the room and then grabs the cage and starts shaking it angrily.

You really have to go watch it, okay? And basically will not accept a cucumber from that point forward because the other monkey is receiving a grape. Now, if I were to ask you what's going on in that monkey's head, how would you describe it?

[00:20:42]  Cameo: Well, I would challenge an implied assumption that it's necessarily about the other monkey.

[00:20:51]  Sadia: So in that video, I feel like one, obviously the monkey seemed like it had the concept of being unfair or fair, right? And that seems to be a pretty complicated, you know, I think that's the complex concept.

[00:21:05]  Cameo: Well, and I will go and watch it. But I just wonder if we're personifying the monkey a little bit by trying to make it a jealousy emotion versus a knowledge that, oh, you have grapes and you're not giving, you know, I mean, maybe that counts as jealousy. But I don't, we could get into the concept of what jealousy is.

[00:21:31]  Bruce: So the word that you're looking for here is, are we anthropomorphizing? Okay. And I think that's a completely fair question.

[00:21:40]  Sadia: Well, let's try to think about it. Like, how else could we think about it? That one of them is just angry that he's not getting, I mean, I guess it would be interesting to see if there wasn't another monkey.

And did they, I wonder if they did that experiment just with one. Sometimes it gets the grape and other times it gets cucumber. It might just get mad even then.

[00:21:59]  Cameo: Right. Once it knows that a grape is an option and whoever the machine or the human or whatever, way that the item is being distributed. You know, I would love to see the exact same experiment with the first time they got a cucumber, the next time they got the grape, what happens the third time when they get the cucumber, if there weren't another monkey anywhere around.

[00:22:22]  Sadia: Just to give you an example, I have a cat who has this set time and my husband pets him and in the set place that he likes. So if someday, you know, Mark is tired, my cat literally gets upset at that time. Like he has gestures and he makes all these weird noises and he's clearly, you know, upset and then shows a lot of like almost aggressive behavior.

So in that case, that's one, you know, I mean, obviously there isn't any jealousy involved. It's just upset that it's not getting what it's expecting.

[00:22:53]  Cameo: So the response to the Pavlovian nature that the animal is conditioned to this very specific thing.

[00:22:59]  Bruce: So it's interesting that they have a similar experiment. I've got a link up that you can look at this too, where they did it with dogs and it works with dogs. In this case, one dog gets a reward, another one dog gets, you know, a pet and words of affirmation.

And the dog that was just getting pet and words of affirmation continued for a while and then just went into rebellion.

[00:23:19]  Sadia: And he just laid down, right? He would just lay on the floor, almost in defiance, like, I'm not doing this.

[00:23:26]  Bruce: Yeah. You know, it's, and this is the hard question. We have a natural tendency to anthropomorphize and Cameo’s trying to lay out an alternative theory and how to test it, which is important here.

But there are many things about animals that are just hard to explain without basically making some assumptions that they have some feelings similar to ours. Notice that even Cameo’s version of this, even though it doesn't reference jealousy, it's still referencing very human-like emotions about preferring one over the other and getting angry about it. So there's still this reference to emotions similar to humans.

But simply saying, oh, we're anthropomorphizing isn't really a sufficient explanation of anything that's going on. And this is an example of that. Maybe we could come up with some experiment that would differentiate here.

And that would be what you would do.

[00:24:18]  Sadia: Because even in that dog one, you know, one could say that, okay, so dog is going to do the trick as long as you're giving it the treat. If you don't, then just, you know, it has nothing to do with whether the other dog is involved or not.

[00:24:28]  Bruce: This is also the challenge that exists and how you go about trying to test it. How do you, how do you figure out what's really going on in? And Cameo came up with a testable case that would actually differentiate between the two theories.

Whether the animal's action is a matter of jealousy or if it's simply upset that it is not getting the reward that it would prefer. And so that is how you would want to go about this. You would not just make an assumption and then make it non-testable.

We'll come back to that. That is a tendency people have. They say, well, this is what I think is going on, and they make it non-testable.

We want to try to make it testable. Now, this is kind of the background. So I had all these things in my mind when Beginning of Infinity came out.

So I knew about numerous things about animals that were very hard to explain. And I was starting to study artificial intelligence at that point. I started to realize you can't use artificial intelligence to explain animal behavior well, that there's a gap there.

And David Deutsch, in Beginning of Infinity, he says the following, that there are two types of information that DNA and human minds evolve to store and have a property of cosmic significance. Once they are embodied, physically embodied, they tend to cause themselves to remain so. Such information, which I will call knowledge, is very unlikely to come into existence other than through the error correcting process of evolution or thought.

[00:25:49]  Bruce: So this is the quote where David Deutsch is saying, there's only two things that create knowledge, evolution and thought. Now I've challenged that in past podcasts. We talked about Campbell's and Popper's theory and how it's significantly different than this.

And then we also looked at the possibility that maybe they're talking about, they're using the word knowledge in different ways. And maybe they're both right for their definition of knowledge. But based on statements like this, people have picked up this idea that animals get their knowledge entirely from their genes.

People meaning fans of David Deutsch. I don't think there's probably anyone outside the fans of David Deutsch that strongly think that. And you'll hear this, you'll actually hear, well, animals get all their knowledge from their genes.

Because there's only two knowledge creation processes, evolution and thought. Animals don't have thought, therefore their knowledge must come from evolution. Now, I used to wonder, is this something that David Deutsch actually intended?

Well, maybe yes. Here's a quote from an interview. David Deutsch's interview with Tyler.

Tyler says, Dogs understand human social life pretty well. David says, They do not. Dogs have genes which contain knowledge, but it is fixed knowledge and it is not the kind of knowledge that constitutes understanding.

Understanding is always explanatory. And then he goes on and uses an example, which we're going to be using through the rest of this episode. Says there is the case of squirrels.

[00:27:08]  Bruce: They put a squirrel on a concrete floor. The squirrel did exactly the same behavior to try to bury nuts, even though it was having no effect whatsoever. It's just a program being enacted by its genes.

I think that it is statements like this is where the idea comes from, that animals, all their knowledge is in their genes. They have no other sources of knowledge. And this has even led to, and these are funny, by the way, but there's one gentleman that puts tweets up about, quote, buggy animal behavior.

And they're very funny. They're things like a dog. You hold it up in the air and it hears water and it starts to try to swim or something like that.

Right. Or barking at stuffed animals or something along those lines. They're often held up as proofs that animals are just robots and they have a set program and that's it.

So let's talk about this. And this is where we have to get back to the pseudo-Deutsch theory of knowledge. OK.

And the fact that honestly, it's just wrong. Knowledge in the genome is a few hundred megs. It's less than a thousand megs.

It can be, and most of it's repetitive, it can be compressed to like four megs. According to me Googling around and looking this up. So we're not talking about a ton of information that exists in the genome.

If all the knowledge the animal needed was in its genes, there would literally be no need for brains with hundreds of billions of neurons. Those brains are expensive. Brains are expensive.

[00:28:29]  Bruce: If all you needed was a little program that was in the genes, you wouldn't evolve brains. It just wouldn't happen. OK, those brains are there for something.

So let's ask the question this way. Do animal brains, so never mind the word knowledge because this seems to trip people up. Do animal brains contain adapted information that is not found in the genes, that are solutions to problems that they face in their environment, where this adapted information was created by a Popperian evolutionary epistemology?

We know the answer is yes. OK, this is not in doubt. Whether you want to call that knowledge or not, this is what's going on and this is what's filling up their brains.

Example, a dog learns to sit for a treat. That is not in its genome. It has to learn that.

OK, that's a piece of knowledge of adapted information that solves a certain problem that it learned through classical conditioning, which is a variation and selection process, right along the lines with the way Campbell understands that. So once we re situate the question, we don't worry so much about is this knowledge or not, because then we might get in a war over a definition. Once we realize we're talking about adapted information contained in the brain that solves problems, the answer is yes, that animals have sources of knowledge through learning.

Animals can learn, basically, OK? And I don't, there really is no doubt about this. This is not something— we've got tons of theories about this, really good theories that have been around for a long time.

[00:29:59]  Bruce: It had been well corroborated. So if by knowledge, you mean adapted information that are solutions to problems that are created by a Popperian evolutionary epistemology that did not come from the genes, then no, animals do not get all their knowledge from the genes. But if you want to not call that knowledge, then who knows, right?

It just depends on what you happen to be calling it, right? So this is, though, what Popper and Campbell were getting at, OK? When they were talking about this ubiquitous number of knowledge creation processes that come through variation and selection, they were including animal learning in that.

This is why they call it knowledge, is because it's part of the umbrella of Popperian evolutionary epistemology. Whether you call it knowledge or not, it still is part of that umbrella. Does that make sense?

Do you understand what I'm saying here?

[00:30:45]  Unidentified speaker: Yeah, yeah.

[00:30:47]  Bruce: OK, so let's talk about animal learning. Animals have several forms of learning available to them. I'm not even sure we know everything that animals, all the learning algorithms that animals have.

The most famous, though, is classical conditioning, which is the Pavlovian effect, right? And we talk about that one a lot. Almost every animal, you know, other than like single celled animals, have some form of classical conditioning.

It seems to go way, way, way, way down on the evolutionary chain. OK, even single celled animals have a form of learning that they call habituation, which is kind of an early form of classical conditioning. It's almost like the opposite of classical conditioning, where classical conditioning, when something positive happens and there was some sort of stimulus near it, then that stimulus becomes a predictor for the animal of what's going to follow.

You know, you're told the word sit, you sit, you know, that that means that now you're going to get the treat. And then classical conditioning can also go the other way, that if you get punished, something bad happens, then you avoid that stimulus or whatever that thing was that happened just before. That leads to some really funny circumstances, where I know of a woman who built a scratch post for her cat.

[00:31:56]  Bruce: The cat used it once, it fell on top of it, and then it would never use the scratch post again, even though she had now anchored it down. The problem was that it was now classically conditioned to avoid it, and so it couldn't overcome that fear. Temple Grandin claims that animals never overcome their fears.

That's a testable theory, I don't know if it's true or not, but that's one of her claims. Habituation is like the opposite, that something causes a stimulus if it happens too often, you stop giving the response. We're all familiar with this because we have habituation too.

You put your shirt on and the tag in the back is bothering you, but after a while the neurons stop firing and you don't notice it anymore. You forget about anything that, any stimulus is happening too regularly. That's what habituation is, and even single cell animals have a form of learning that is called habituation.

According to Richard Byrne, some animals have something called insight, which is the ability to create mental models and use them. He's trying to figure out which animals have this. This turns out to be actually a fairly difficult question.

For the most part, he limits it only to the highest animals, the ones that are known to be the most intelligent. Obviously, the great apes, elephants, interestingly, crows. Crows are as intelligent as elephants and great apes.

Even though they have very small brains. So he limits it to a very small number of animals. Although he gives examples of other animals having it.

[00:33:27]  Bruce: Even though his theory is that they don't have it. So they may have it like in some circumstances that are narrow or something like that.

[00:33:34]  Cameo: Have you ever watched the YouTube videos about the squirrel, ninja squirrel challenge?

[00:33:43]  Bruce: No, tell me about it.

[00:33:45]  Cameo: You need to watch this. There is a...

[00:33:47]  Unidentified speaker: Yeah, the one that's crazy.

[00:33:48]  Cameo: It is crazy. It is a retired NASA scientist. And the squirrels kept stealing the bird feed that he put out for his birds.

So he started coming up with and building more and more and more complex ways to keep the birds [apparently meaning squirrels] from the birds’ feed. But ultimately he built this whole obstacle course. It's a ninja obstacle course.

And you should watch it, Bruce. The ability for the squirrels to... He cannot believe how creative the squirrels are to break through all these barriers that he gives them.

[00:34:25]  Bruce: Interesting. Yes. By the way, I put an article up on Twitter.

Animals show a form of creativity, right? They come up with interesting things that could not possibly have been anticipated by their genes and that they're able to still come up with creative solutions to those. Obviously made out of movements that do come out of the genes.

So there's kind of a dividing line there that's difficult to explain.

[00:34:48]  Sadia: I guess to me, it seems like they definitely have creativity, but it's limited to solving only specific problems. For example, like my cat quite often experiences boredom, but it's only willing to entertain certain things when it comes to alleviating it. You know, either play with me or I kill some other animal, some chasing game, you know, something like that.

But why not? If I just turn on the TV and there's something going on, why is it that my cat is incapable of, you know, there is something going on there, but it doesn't work for my cat.

[00:35:20]  Bruce: Right, right. Okay. So the learning algorithms that we just, oh, then the one last one, a very few animals really only confirmed in great apes, although there could be other animals that have it that we just haven't confirmed it in, can do something called behavior parsing.

This is the thing that David Deutsch brings up in his book, Beginning of Infinity. Behavior parsing is a memetic transfer of whole programs of action that can be used flexibly to accomplish desired goals. It's very impressive what animals can do with, the animals that can do behavior parsing, what they can do with it.

And I will give you some examples of this. On the other hand, behavior parsing has a very strong mechanical element to it in terms of learning the steps in the program. And David Deutsch uses that to suggest that animals aren't very intelligent.

And he also, I'll give you a quote here in a second. He gives examples of where they do it, even though it doesn't make sense. They'll learn some gesture and then they'll do it, even though their thumb's missing or something like that.

It's not going to do anything for them. They'll still do it. So it also demonstrates, on the one hand, it demonstrates a great deal of intelligence.

On the other hand, it also demonstrates a great deal of lack of understanding. So we'll talk, this is Richard Byrne's kind of main theory is behavior parsing. So we'll spend some time on that in one of the future episodes.

[00:36:35]  Sadia: Could I throw out an example here? Because again, observing my cat, so whenever cats use a litter box or something outside, they like to cover up their poop, so they'll dig it up or and then pee or poop, they'll just cover it up. So a lot of times, I've noticed sometimes my cat will go into the litter box.

The poop is in the litter box, but my cat is too dumb to understand that if it's scratching the floor, that's not covering up the poop. It'll look at the poop and it's scratching the floor. And then it just eventually just walks away.

It's like unable to see the connection that what I'm scratching here is not covering the poop. So it seems like it's very mechanical at that.

[00:37:16]  Bruce: Yes, it is. Who has seen a dog try to bury its food with its nose? I mean, like I've seen dogs try to do that.

They're doing it on a linoleum floor with a bowl, right? And they're making this gesture like they want to bury their food. So yeah, there's things like this definitely show a really strong lack of understanding on the part of animals.

[00:37:38]  Sadia: But the squirrel—that ninja track, the squirrels have this amazing ability that given different types of variations of problem, they actually do figure out novel ways of getting around it.

[00:37:51]  Bruce: Okay. And this is the weird thing about animals is that they show gigantic lack of understanding and gigantic creativity, but just in different circumstances. It's very hard to figure out what's going on with animals.

Right. There's a great deal of mystery with animals. And like I said, you can't look to AI to help you here because animals are doing things AI can't.

Right. We don't have any. If you were to, if I were to go create an AI that could do classical conditioning like a dog, I'd be a Nobel Prize winner.

Right. I mean, like it's that big a deal. Like we just can't do it.

Now, these algorithms that we're talking about, all of them have a variation selection process involved. It's most obvious in the case of classical conditioning, repeat variations that keep happening near a positive event or the opposite of that. Insight allows you to do mental variations to a problem in your head.

So if animals have insight, then that's what they're doing. They've got some ability to try out mental variation, just like a human would. OK, although humans obviously do this much, much, much, much better.

[00:38:51]  Bruce: All of these learning algorithms are variation selection processes and therefore fall under the Popperian understanding of knowledge that they're created through evolutionary epistemology. OK, that's the way Popper looked at the word knowledge, different than maybe the way David Deutsch looks at it. So if we're sticking with this idea that knowledge only comes from two sources, which David Deutsch has advanced, then we would have to say that animals learn and what they learn is adaptive information that gives them solutions to their problems.

It's stored in the brain. It uses a Popperian evolutionary epistemology, but we're just not going to call it knowledge. This is at a minimum.

I mean, like you can have a word have as many meanings as you want, but at a minimum, this may not—it's a little confusing to try to not call it knowledge. So this is something that Bart in particular, but Sadia also kind of brought this up to me.

So this is probably closer to the way Bart brought it up to me, Sadia. But you did say something similar to this to me once. The argument was David Deutsch is reserving the word knowledge to mean specifically open ended knowledge creation that can produce true novelty or some variant of that.

This is actually not an unreasonable idea. It seems likely to me that maybe he is thinking something like that. Now, mind you, he's never said that.

[00:40:12]  Bruce: Okay, so we are guessing. If he ever actually came out and said, look, I'm using the word knowledge to specifically refer to the open ended sorts of knowledge creation or something along those lines. That would make things a lot more clear as to what he's getting at.

He's never actually said that. But I don't think this is necessarily a terrible guess, right? That's kind of what he's thinking.

[00:40:32]  Sadia: I think the problem there seems to be that he doesn't want to bring in any mention of induction. And as you kind of talked about this too, like, you know, a lot of people working in, you know, an AI and stuff, they do talk about some sort of inductive reasoning, right? But with David Deutsch, he'll always just say either you have knowledge, induction is a myth, there's nothing there in between, right?

Either. But then the problem is, even if you have, you know, there was some, not to get off topic, but there was a program that generated a bunch of theorems, right?

[00:41:07]  Bruce: Yes, the Theorem Prover.

[00:41:08]  Sadia: So didn't David Deutsch said that, yeah, the knowledge was already there in it, but I think there's something missing there, right? Because we have to run it on a computer and the computer, you know, the algorithm and some process had to happen for some, you know, those theorems to be generated. So what exactly is that then, right?

[00:41:27]  Bruce: So the theorem prover was like one of the very first AI algorithms ever written, AI programs ever written before the word AI existed. And it led, it was brought up in a conference where where the term got coined for the first time. And it created a proof that was novel, that no one in the world had ever seen before.

It was for something we already knew how to prove in some other way, but it came up with a shorter novel proof for that thing. I don't remember exactly what it was, which was really kind of amazing. OK, and it does work through a variation selection process.

It tries out different variants. It has heuristics that it uses to try to narrow the search scope, which did come from a human. OK, now that theorem prover is one of the specific examples Donald Campbell in his paper uses as an example of something that does create knowledge.

OK, so this is this is why I originally raised the issue that the Popper-Campbell view of knowledge of necessity must be something very different than the Deutsch way. The way Deutsch is trying to use the term.

[00:42:32]  Sadia: OK, because something has to unfold, right? Or, you know, there is something that that is on top of the algorithm.

[00:42:40]  Unidentified speaker: Yes.

[00:42:40]  Sadia: You have to run the algorithm and it has to try that, you know, there has to be a variation selection.

So there's a process involved.

[00:42:48]  Bruce: OK, so let's let's pretend like let's take this idea that that Deutsch is trying to reserve the word the word for the more open ended kind of knowledge creation. We can always allow multiple definitions. It's always OK.

You never need to argue with somebody over what a word means. You can say, well, let's have two definitions and that's always available to you. OK, so knowledge one, we're going to call that the Popper's definition of knowledge.

It is adapted information about solutions to problems, possibly from the product of improvements that come from variation selection. He would have actually said that comes from the improvements of variation and selection. But I'm leaving a little bit open ended just in case there's other other sources we don't know about.

Not saying that there are, but I'm trying to leave it a little bit open ended. Therefore follows Popper's evolutionary epistemology. That would be knowledge one.

Knowledge one is this adapted information. It's the Campbell version of knowledge. It's what you get out of improvements from any variation selection process that's trying to solve a problem.

Then we'll call knowledge two Deutsch's definition. It's the subset of knowledge one that comes from open ended variation and selection processes. And therefore, because they're open ended, can create extra novel results.

This is OK. We could think of it this way. We could understand both men as having legitimate definitions of knowledge that even have a relationship between them.

[00:44:09]  Bruce: Where Deutsch's definition is a subset of Popper's definition, but it's a special subset. If we were looking at it in this way, then the statement that there's only two sources of knowledge, which will now take to mean knowledge two: genes and human ideas and memes, that's now a true statement. OK, this is how we might go about reconciling David Deutsch's statements with a certain understanding of knowledge, specifically the more open ended variety creation process.

And I suspect this is what he has in mind. And that's also just to make this a bit more clear. I doubt he would tell you animals don't learn, right?

I really doubt he would tell you that because everybody knows animals can learn. Everyone knows you can teach a dog to sit for a treat. So let's look at this.

What would be the advantages of looking at this? What we're now going to look at is Deutsch's definition, knowledge two. Well, there really is something special about open-ended knowledge creation, because it really does produce novelty in a way narrow knowledge creation can't.

Just to put this in perspective, the immune system. David Deutsch, in an interview with Eli Tyre, says, I think we know for sure the immune system doesn't create knowledge. We're now going to assume, now, according to Campbell, it does, right?

According to Popper, it does because it's a variation selection process. It creates antibodies for a disease that has never been known before in the history of the world. That's the way it works.

[00:45:33]  Bruce: That's why it uses a variation selection process to try to come up with antibodies. And it keeps the ones that are actually working. It throws away the variants that don't work.

It's a straightforward variation selection process that is knowledge creation in the knowledge one Popperian sense, but isn't knowledge creation in the David Deutsch knowledge two sense. You can see that what he might be getting at here. The immune system is never going to produce a jet, right?

I mean, it's whatever it's doing, you know, it's creating knowledge, it's creating adapted information, but it is so narrow. It is going to do it specifically to make antibodies, and that's it. And it's repertoire is set to creating antibodies.

[00:46:16]  Sadia: So what comes to my mind then is that it seems like in the case of humans, the knowledge creation involves even creating new values. Whereas in the case of animals, the values might be set like,

[00:46:30]  Bruce: oh, interesting,

[00:46:31]  Sadia: right? An animal [can/can’t unclear] create new values.

So it is definitely there is a clear difference. There is something quite different about humans in that sense.

[00:46:39]  Bruce: That's a really interesting theory, Sadia. I actually think there could be something to that from my reading.

[00:46:46]  Sadia: And then what you could say is that in that sense, where are those values come from? Those were programmed through genetic evolution. But then once those are programmed in an animal, then the variation selection can be kind of like an optimization process.

It's always constrained by the values that are programmed. But with humans, there is no such limit to what values we can create.

[00:47:08]  Bruce: You know, that's a really good theory. We need to give that theory more thought. I don't think I've ever heard anyone vocalize that theory before, Sadia.

[00:47:15]  Sadia: Well, that's behind my theory of morality, too, but that would be a different topic.

[00:47:20]  Bruce: Okay, we'll have to bring that up again. And I need to think about that more. But everything, initially, my thought is that's a very interesting theory.

We need to look at that more. Maybe what we could say is that the genes have a set values for animals, but then they're able to create knowledge in ways to how they're going to solve for those values in a limited capacity. But humans can actually create entirely new values.

[00:47:44]  Sadia: So can I quickly quote Popper in his book Unended Quest? He said that values actually, values are created with problems. And then he also says all life is problem solving.

So if we say that, like, we can see that the problems that a cat or a dog solves are limited. And so are the values. They go hand in hand.

Right. But something in the humans actually, you know, the unended problem solving, we don't even know what the future problems are going to be like. Right.

[00:48:16]  Bruce: Right.

[00:48:17]  Unidentified speaker: Anyway, that's what I just mentioned.

[00:48:20]  Bruce: And let me let me just say that knowledge one creation processes, the ones we currently know about, the the ones that we can actually program, they really aren't that impressive in a lot of ways. I mean, sometimes they do some pretty impressive things. AlphaGo is very impressive, but even AlphaGo is never going to create a jet.

Right. I mean, like it is still super narrow in its domain. Right.

[00:48:40]  Sadia: And it did seem like because I listened to your podcast and I even watched that video on AlphaGo. It seemed like they were nonstop teaching it. They gave the AlphaGo had access to all the games that had been played out there by different players.

So it was learning from those. Right.

[00:48:57]  Bruce: Initially, they eventually the current AlphaGo does not use any human games. It just plays itself.

[00:49:04]  Unidentified speaker: And that's actually that knowledge is in there, right? Somewhere.

[00:49:08]  Bruce: No, no, it's already. They started from scratch in AlphaZero. That's why there's the zero in it.

So they don't actually they actually found that that worked better to not use human games. But initially, when they didn't know what they were just getting it off the ground, human games were very useful. But once they kind of figured out the algorithm, they found that it actually was better just to let it play itself.

But the human games, they thought that might have been part of the reason why it would, like, [delude?] itself. But anyhow, that's a different story. So and the paramecium example that we use a lot that comes from Campbell, it literally has a set number of predetermined directions that are already known.

So some knowledge, some knowledge one creation processes are so limited that their repertoire is literally set. Now, that's not always true. Genetic programming, for example, has an infinite repertoire.

In theory, it could discover anything. It never does in practice. That's a why, you know, we don't know.

That's part of the problems that need to get solved. But many knowledge creation processes in the knowledge one category, the Popperian category, are super unimpressive to the point that you can understand why people maybe are even hesitant to call it knowledge. So in some ways, the fact that Campbell was willing to call it knowledge, willing to say, look, the paramecium is a variation selection process, and it is coming up with them within the domain it was meant to do, a novel solution to a problem.

[00:50:29]  Bruce: It's blocked. It needs to find its way out. It has no way of knowing which is the way out.

It is novel within that sense. Right. But it does it with preset determined, predetermined outcomes.

And you can see why some people would say, you know, that's not quite what I meant by knowledge. So in some ways, the but but we can also see that it does follow a Popperian evolutionary epistemology, which is why Campbell and Popper wanted to include it under the umbrella of knowledge creation.

[00:50:57]  Sadia: So let me ask you this question of what you said. Can you go back one slide with the genetic, the genetic programming you said has an infinite repertoire. Do you mean by that, like all the different possibilities, the landscape of possibility?

[00:51:11]  Bruce: It uses programming languages and programming languages are known to be, you know, Turing complete. So the repertoire would be everything. If you buy computational universality, which maybe you don't.

But if you do, the repertoire would be universal.

[00:51:27]  Sadia: I guess I'm still not getting it. Universal in what sense? Like, oh, so anything that is programmed that you can possibly imagine.

[00:51:34]  Bruce: It can discover, it can in theory discover any algorithm.

Yes. Any Turing algorithm.

[00:51:41]  Unidentified speaker: Got it, got it.

[00:51:43]  Bruce: Okay. So now we've talked about why the Deutsch definition of knowledge.

What we're assuming is the Deutsch definition of knowledge actually does make some sense. Now let's talk about why maybe it doesn't make some sense and why the Popper Campbell one might be better. And of course, it depends on your circumstance, right?

And it's okay to have two. But let's be realistic about this. There really is a meaningful difference between knowledge created that's in the genes and the genes having a learning algorithm that then in turn creates knowledge based on its environment and then puts it, stores it in the brain or the nervous system.

Those are just not the same thing. It's really important to understand that they are not the same thing. Now, we talked about this in Episode 20, Evolution Outside the Genome.

I received comments. So in Evolution Outside the Genome, we had it where they had the Picasso frogs that a scientist grabs in the tadpole its eyes or its, you know, other sense organs, pulls them into an incorrect position. And then when it transforms into a frog, the eye moves into the right spot, even though it's starting in the wrong place compared to what the genes would have known about that.

And then it does it through error correction. It does it by correcting itself until it gets it into the right spot. Well, that sounds like a learning algorithm.

Now, we don't know what the algorithm is. And so we're making some guesses here and everything is a conjecture anyhow. But that sounds like a learning algorithm.

[00:53:12]  Bruce: It sounds like it's doing some sort of variation selection process under the Campbellian sense. And it's moving the eyes into place, even though they're starting in the wrong spot. So I got comments from people on this.

And it makes sense. These comments do make sense. I understand where they're coming from.

They would say, you claim the knowledge of where to place the eyes in the Picasso frog is determined by a learning algorithm in the cells rather than being in the genes. But didn't the knowledge for how to do that come from the genes? So doesn't that mean the knowledge was in the genes?

OK, this is what we might call the credit assignment problem. That since the knowledge for how to do it came from the genes, we're going to say the knowledge was in the genes, not in the cells. OK, this argument is equivalent to saying this is obviously an extreme version.

You claim human beings create knowledge via their creative algorithm, but didn't the knowledge of the creative algorithm come from the genes? So isn't all creativity really just knowledge in the genes? Well, that's a stupid argument.

I mean, obviously that's a stupid argument, but it's the same argument. And it misses the point for the same reason. If you're going to try to assign credit like that, everything has to go back to the genes at some point.

And we're not really interested in assigning credit. What we're really trying to understand is how do variation selection processes, how does evolutionary epistemology work with animals, with algorithms? How does it work?

[00:54:36]  Bruce: That's the question we're really trying to answer. We're not that interested in who gets the credit. And really think about it, the genes could never have anticipated a problem like a scientist decides to move the eyes in the tadpole to the wrong place.

But it didn't have to. And the reason why is because it already had realized that having a flexible error correcting learning algorithm that moves the eyes in the right place is less brittle than trying to just move it in some direction, right? That is more flexible.

So, yes, the genes created the knowledge that's in the learning algorithm. The learning algorithm in turn creates knowledge also. Okay.

Maybe you want to say less. So I don't care. You know, the genes have more knowledge, more important knowledge, more open-ended knowledge.

We can say this in a lot of different ways, but it's important to understand that there is a difference there. Okay. And that's what that's the point I'm trying to get across.

Also, this really does lead to confusion. People, even if Deutsch really does intend this as, look, I'm reserving the word knowledge for this open-ended, more novel kind of knowledge creation, people are instead interpreting it as the pseudo-Deutsch theory of knowledge, which is wrong. I don't know how else to say it.

So you hear things like learning algorithms are misnamed. They don't create knowledge at all. All knowledge came from the programmers, so there's nothing special about them.

[00:55:52]  Bruce: This is false. This is literally just a false statement that is a misunderstanding where you've taken what Deutsch said and you've turned it into the pseudo-Deutsch theory of knowledge, and this is just wrong. “Animals do not learn.”

I've heard that before. “Animals have all their actions, reactions pre-built into their brains from the genes.” It's literally impossible for the reasons that we gave.

The genes don't have enough information to do that. One response that would be a very good one would be if somebody said to me, I can see that Popper and Deutsch define knowledge differently, and they are both right under their respective definitions. I have had two people say that to me so far, Sadia and Bart.

Everybody else I've talked to about this will tell me, and I'm not kidding you, Popper is wrong, Deutsch is right. They're not seeing it as a different way of looking at knowledge. They're seeing it as there's only one way to look at knowledge.

They're falling into what I call word essentialism. They're missing the fact that there are things outside what Deutsch calls knowledge that follow evolutionary epistemology. That's just the truth.

That is what Campbell's whole theory is about.

[00:56:59]  Sadia: I think to me, it seems like what they might be missing. DNA has a memory where it stores information. But I think the fact that now in animals, there's also a memory associated with brain.

That in itself tells us that, okay, if it can store knowledge, it can also update that memory. And hence, why can't we see that there could be a variation selection at that level going on? But I guess the problem that those people see is that they somehow feel like overall genetics offer the biggest constraint.

Genetics eventually just constrain everything. And I guess the question would be, is it possible that to a certain amount that there could be some knowledge created in the brain of an animal that could almost have, I hate to use the word top-down causation, but somewhere where some constraints now created there could now, sort of like an epigenetic phenomena, where now that certain knowledge that's created in the brain starts affecting your DNA, right?

[00:58:05]  Bruce: I think that that is where this is going, right?

[00:58:09]  Sadia: Because in humans we know that there is such a thing, right? The epigenetics, I guess that would be epigenetics, right?

[00:58:16]  Unidentified speaker: Yes.

[00:58:17]  Sadia: Like if you're mentally, if you're stressed out and stuff, that can actually affect at the level of your genes. So why couldn't that be the case with animals too?

[00:58:28]  Bruce: Yes. And then here's my next argument here. I need a word.

I can't keep saying adapted solutions that are solutions to unanticipated problems that use a Popperian evolutionary epistemology process, a variation selection that are stored in the brain. It's just too long, right? I need a word to refer to whatever that is.

The immune system.

[00:58:51]  Unidentified speaker: That's good.

[00:58:54]  Bruce: So the immune system, what the immune system does, it does something.

It creates something and it does it through variation and selection algorithm. There's a word for that. There is actually a word in the English language for what the immune system creates.

The word is knowledge. That's the word. It's been the word for a very long time.

And this is kind of the problem with utilizing an existing word in a special way, which Deutsch likes to do. And sometimes that's the best thing to do. I'm not trying to say it's bad.

I'm just trying to say there's a danger with it that you have to accept, which is that it will cause confusion with what the word originally meant. Okay. Which might still be a useful thing, a useful concept, which in this case, it is a useful concept because it's part of what we want to study as part of Popper's evolutionary epistemology.

Refusing to call that knowledge doesn't change the fact that something is going on there that needs a word and it's distinct from the genes. And that this something is following Popper's epistemology. This is my real argument and why I'm going to use the word knowledge.

I'm not trying to say the other word, other definition is wrong. I don't believe that. I don't believe in word essentialism.

I think it's fine for David Deutsch to use the word knowledge in a special way. But people need to understand why I use this other word is because I just don't have another choice. There is no other word for it in the English language.

[01:00:18]  Bruce: So Popper's, here's another thing. Popper's definition, go back to the way I defined them. The Popper definition is about what knowledge is.

It's a solution to a problem. Deutsch's definition is about what process created it, came through an open-ended knowledge creation process. Now, there's some precedent for this, like industrial diamonds versus natural diamonds.

They're physically the same thing. But one came through one process and one came through a different process. So we value one more than the other.

So this knowledge two, defining knowledge in terms of the process that creates it, isn't necessarily wrong or bad. It might make sense, especially since we don't know.

[01:00:58]  Cameo: If you're using that as an example, the human choice to value the one over the other is silly and is artificially created by people who want to monetize it. And the fact that they're able to trick people into valuing the one over the other, there is no difference. Why would you possibly value one over the other?

[01:01:23]  Bruce: You make a good argument that I am not going to try to challenge. So, yeah, I mean, like, if I could, I would probably, you know, want to buy an industrial diamond for cheaper and use that for a wedding ring. You know, why not?

Looks just as pretty. You know, it's no one can tell the difference.

[01:01:39]  Cameo: So [a gemologist] couldn't tell the difference. And so why? So I'm going to if that's the example, then we should not value the way something is created.

We should only value the outcome of what it does for us.

[01:01:56]  Bruce: This is probably a good spot to remind us, though, that this is not necessarily Deutsch's definition. We're pretending like it is based on an argument that Bart had made to me that maybe this is what Deutsch had in mind. But we don't know that for sure.

Let me make an argument here. Okay. So let's imagine a scenario.

And this is probably in the future because I don't think we have knowledge to do this yet. But let's say that we think about like some pandemic happens and you have a bunch of scientists in the lab that are going to engineer and structure an antibody that will then fight off this virus or bacteria or whatever it is. Okay.

So they use their creativity and their knowledge of how antibodies work and how diseases work. And they use explanations, scientific explanations, and they work up, we need this antibody to be like this. And they instantiate it and then they duplicate it and they create these antibodies that you can inject into a person and it makes the disease go away.

Okay. We don't do that today. We do, we use the immune system itself to create it.

But we can imagine in the future scientists being able to do this. They just, they just used knowledge two, the Deutschian version of knowledge, because it's the open ended creation process. So the antibodies that they created now embody knowledge in the sense of knowledge two.

[01:03:22]  Bruce: Now imagine the immune system just on its own hypermutates, uses DNA to hypermutate according to the Nobles, so it's not really different than the genome. And it comes up with the exact same antibodies just by using random variation and selection. Okay.

These antibodies are identical in every way to the ones created by the scientists that use creativity to bring it together. The only difference was that it was created by a quote mechanical process of the immune system using variation and selection. We now have this weird circumstance where they have identical antibodies, literally identical antibodies.

And one set we say embodies knowledge, meaning knowledge two, and the other is just pre-existing knowledge in the genes. Even though they're the same, it doesn't make sense, right? It leads you into weird circumstances like that if you try to define things in terms of the process that creates it rather than what it is.

And that may be fine. It may still be okay to talk that way. Cameo's, I acknowledge Cameo's argument here, but someone else may feel differently.

Maybe they're very much against industrial diamonds. And this would be my argument for why maybe it's not the best to try to use the word knowledge in this way to only refer to certain open-ended processes. And regardless of what we call it, the only reason why, when we say, oh, these antibodies, they embody knowledge, knowledge two, but the others were just pre-existing knowledge in the genes.

[01:04:52]  Bruce: Well, the only sense in which that's true is in that the genes had a learning algorithm, the immune system, that then created knowledge one, okay, that was novel. So I don't think this is the best way to go about trying to understand knowledge. I guess at the end of the day, I propose that we use the existing word for animal learning.

Animal learning creates knowledge, period, end of story. You can think of it as knowledge one versus the Deutschian knowledge two. That's fine.

But it's the only word I really have to describe it. Now, I also want to make this is my final argument on this front. I believe that trying to define it in a special way like this leads to confusion even for David Deutsch himself.

And here's my argument for why I believe that's the case. In Fabric of Reality, David Deutsch defines knowledge and he defined, he's had more than one definition. He's like five or six different definitions of knowledge, he claims.

In the fabric of reality, it was convergence across the many worlds. So the example he uses was DNA. And he says, look, you've got some string of DNA that's a gene and it contains knowledge.

And maybe you have the exact same string inside of the junk DNA, which doesn't contain knowledge. How do you tell those apart? He says, well, you could look across the many worlds if you could.

[01:06:02]  Bruce: And you could see that the junk DNA varies across the worlds, whereas the knowledge-bearing genes, DNA, would be the same because they're hard to vary. I actually think that's a very good definition of knowledge. And you don't actually need many worlds to do this.

You could go look at just multiple variants of the same species and you would discover the same thing, that the genes, the knowledge-bearing sequences, they are the same across members of the species, whereas the junk DNA, even if it was the exact same sequence, would vary across the species. You can do this with computer programs, right? If I run a machine learning algorithm and it's a neural net and it's trying to find the right set of parameters that are going to work, it will, there's only a small set of different, there's the set of possible parameters is much larger than the set that's good, right?

Because good parameters, even for heuristics like this, are hard to vary. And so it's going, you would find across multiple runs that it tended to find, you know, maybe, there may be hundreds instead of, you know, the infinite number, but you'd probably find there's a hundred or so different good configurations and it always finds one of those in that hundred. So under that definition of knowledge that he used in Fabric of Reality, he is describing knowledge one, the Popper definition of knowledge, not the Deutsch definition of knowledge.

[01:07:23]  Bruce: Can you see that that's the case? And think about the immune system as the example. OK, the DNA that it hyper mutates would would converge across the many worlds to the right sequence to kill that bacteria, to create the antibody that kills that bacteria.

So in its DNA, in this case, according to the Nobles’ paper, so that definition of knowledge from Fabric of Reality, that is the Popper definition of knowledge. Now, you might argue with me here, oh, but he changed the definition in Beginning of Infinity. He defines knowledge as information that once physically embodied tends to cause themselves to remain so.

This is really a very novel definition of knowledge. And by the way, I think it's a correct one. Let's take a look at the immune system again.

Is that definition of knowledge, does it fit the immune system? Sadia, that's probably a question for you.

[01:08:10]  Sadia: I'm going to pass on that.

[01:08:12]  Bruce: It's the same thing, right? The immune system has to hyper mutate to find the sequence of DNA that creates the antibody that actually works against the disease. The variants of the antibodies, it tries lots of variants, it kills off the ones that didn't work, and it keeps the ones that were actually effective.

So they stay embodied. So under the Deutsch definition of knowledge from beginning of infinity, the immune system creates knowledge. So that definition of knowledge also fits the Popper definition of knowledge, not the Deutsch definition of knowledge.

I think this is the problem, is that if you try to define it in terms of a process, you bump into problems. So he's trying to come up with a more substantive definition, but we don't know enough about open-ended processes. So any definition he comes up with will happen to always be the Popper definition.

[01:08:59]  Sadia: Actually, one thing that I was thinking about, why I kind of drifted off, was I was actually thinking about the whole importance of the concept of a process in the definition of knowledge. Because, I mean, even if it is open-ended, right? I mean, if we just focus that if as long as there's a process of variation and selection, it's a process of knowledge.

Like, that's the way I'm kind of thinking about it, right? In one case, okay, so maybe it keeps going. In the other, maybe it just becomes restricted.

But I don't think—I think in that sense, there would be kind of a unity in both that in both cases, we see that as knowledge.

[01:09:33]  Unidentified speaker: Yes.

[01:09:34]  Sadia: The knowledge is the process. It's not like a book sitting on a table just has knowledge. The book doesn't have knowledge till somebody is actually doing something with it, right?

That there is a mind that's engaged with it and knows how to decipher what's in the book. I mean, if the entire human race died out, and would that book still have knowledge? Because Deutsch does say, right?

[01:09:57]  Bruce: Actually, Popper says it does. He says that that's world three. Because if an alien found it, they could decipher it and it would contain knowledge.

So he believes that the book, in some sense, contains objective knowledge, which he sees as separate from subjective knowledge, which would be what's in the mind.

[01:10:13]  Sadia: Maybe we're going off track. But one of the things I've always, it's always fascinated me is that if there was a book with a bunch of, there were no pictures or anything. It's just like, let's say, a book written in English.

And there was no book, nothing else that showed how to decipher it. Could an alien actually figure out what that has?

[01:10:33]  Bruce: Yeah, they may not be able to.

You're actually right.

Think about us trying to decipher ancient languages. Without the Rosetta Stone, we can't do it.

[01:10:44]  Sadia: Or not just that. I've always wondered that if some alien communicated with us and if there wasn't any type of a visual thing involved, because a visual might still help us. But if it just came in a bunch of symbols, and even if they were patterns, without knowing how to decipher them, we wouldn't really know, right?

So in that sense, from our viewpoint, we wouldn't even see it as knowledge. So for something to have knowledge, there has to be a process and there has to be some understanding between whatever is, you know, like if it's happening in our brain, that when we're reading the book, you know, that there is a way, like meaning is important in a sense.

[01:11:19]  Bruce: Yes. The way Popper explains that is he says that world three can only interact with world one, world one being the physical realm, world three being the realm of objective knowledge, through world two, which would be minds. So that is actually pretty similar to what you just said.

I just happened to read that recently, by the way. So that's interesting that you brought that up. Let's get back to animals now and how this all applies to animals.

There's actually one more really interesting point that needs to be brought up in terms of knowledge creation in animals and genes. So let's go back to the squirrel example. A squirrel has a little program in the brain that when it has food, it attempts to dig into the ground.

It attempts to put the food into the ground, it attempts to use its nose to pat the food down into the ground, and then it buries it. Okay, so that it can hide the food. And it will try to do that on concrete, right?

Just like your cat will try to bury stuff on linoleum. So here we have this algorithm that is very mechanical, shows really no real understanding. And so that's why Deutsch is using this example.

So he would say then, well, that's knowledge in the genes. It's enacting a program that's in its genes. And which is true, by the way, that's correct.

[01:12:37]  Bruce: There's something missing, though, from this story. Namely, how does an algorithm like that evolve in the first place? So now you think about it, this is the exact same problem that all evolution has.

Let's say an animal evolved, the ability to dig once, you know, does one attempt to dig. You know, that wouldn't be useful. Let's say it maybe does it multiple times, it digs a hole.

That's not a survival, it has no survival value. It's a useless program. Let's say it evolved the ability to pat down the food.

That maneuver on its own without having first dug has no survival value. It would never evolve it. This complex set of steps that are in the algorithm, each one has to work under Neo-Darwinian evolution.

Each one has to on its own be valuable. It looks like they're not valuable except as a collection. When we claim that animals just evolve these algorithms, and we don't go any further and don't explain anything any further, we're actually being Lamarckian, that this animal needed this algorithm so it evolved it.

If you want to understand how to square this with Neo-Darwinian evolution, there's actually a well-known explanation that exists. Campbell points this out. It's called the Baldwin Effect.

Quoting Campbell, he says, Complex adaptive instincts, instincts are the mechanical processes that the genes have. This is separate from what they learn, according to Campbell. Adaptive instincts typically involve multiple movements and must inevitably involve a multiplicity of mutations, at least as great in number as the obvious movement segments.

[01:14:09]  Bruce: Furthermore, it is typical that the fragmentary movement segments or the effects of a single component mutation would represent no adaptive gain at all apart from the remainder of the total sequence. The joint likelihood of simultaneous occurrence of the adaptive form of the many mutations involved is so infinitesimal that the blind mutation and selective retention model seems inadequate. Okay, in plain English, it's not possible under Neo-Darwinian evolution to evolve complex algorithms like we see animals do.

It literally violates Neo-Darwinian evolution. Unless, this is the next quote from Campbell, the adaptive pattern being thus piloted by learning, any mutation that accelerated the learning made it more certain to occur or predispose the animal to certain component responses would be adaptive and selected no matter which component or in what order affected. The habit thus provided a selective template around which the instinctive components could be assembled.

In plain English, it's not a problem because evolution evolved learning algorithms first. If you were to go look at that squirrel and you were to go back in time, maybe even to something that was an ancestor prior to the squirrel, you would find that the original animals learned to dig and bury the nut and pat it down with their nose basically through regular learning processes. And then they memetically transferred it to other animals.

[01:15:34]  Bruce: It turned into a survival advantage to have learned this algorithm that's not built into the genes. Once that has moved through the animal populace through memes, animals do have memes, we're going to talk about this, then there is a survival advantage to it evolving into the genes. Because if the animal evolves to pat down automatically once it's buried, it's learned how to dig a hole.

And then it now has a gene that says once the hole is dug, just pat it down with your nose. Now it doesn't matter that it doesn't by itself have survival value because it's happening within the habitat that includes learning algorithms.

So if you want to understand the whole concept of learning things, or sorry, of having algorithms that exist in the genes, which certainly exist, you have to understand that the learning algorithm has to come first, or neo-Darwinian evolution doesn't even work.

[01:16:26]  Sadia: That's actually really interesting. So let me just first of all make sure I get it. So you're saying that, you know, with the animals, a lot of that behavior that we talk about instinctual behavior, right?

That knowledge in the genes to be, you know, actually came because they were trying certain things out and whatever works now.

[01:16:45]  Bruce: That's correct. This makes it more impressive. Yes, it's true.

Animals have algorithms in their genes that they follow and they don't understand. That is absolutely true. Once you realize that some animal in the past had to learn it, it's more impressive, right?

[01:17:00]  Sadia: In other words, the explanation of that is not in their genes, but actually in their learning.

[01:17:06]  Bruce: Yes, it did not get into the genes until they learned it. They had to learn it first. It had to pass memetically.

Nicholas Christakis, I won't read this whole quote here, but in Blueprint he talks about this as well. He talks about how over time certain behaviors that were previously learned may become genetic. They call this the Baldwin Effect.

Baldwin was an early person who studied this. He gives an example of a bird where the bird has to learn the song and there's survival value to attract mates if you can get the song right. So there’d be a survival advantage to have parts of the song start to be part of the genetics and eventually the whole song becomes genetic.

At that point, it's just an automatic response that they do. This is a very important thing to understand. And thus, when we talk about buggy animal behavior, we need to go back and we need to understand buggy animal behavior in light of what we actually know with the Baldwin effect and with knowledge in the Popperian sense.

So let's talk about the squirrel example where it tries to dig on concrete and tries to bury the nut on concrete. What does this actually demonstrate now that we have better context? Well, it does not suggest that if some animal behavior is innate, that all must be innate.

[01:18:20]  Bruce: We know that can't be the fact. The Baldwin effect requires that they have to learn it first. Unless, I mean, like you could make a case against the Baldwin effect in certain cases where each individual step was by itself a defensive, like maybe fight maneuvers or something like that, right?

But when we're talking about the squirrel example, you have to reference the Baldwin effect, or you're basically supporting Lamarckian evolution. It does not demonstrate that animals get all their knowledge from the genes. This is literally impossible knowledge in the Popperian sense, not the Deutsch sense.

However, they do get all their open ended knowledge creation from regular genetic evolution. So there's a sense in which we might say all the knowledge is in the genes, but it's only the Deutschian definition, which I've given my reasons for why I don't prefer. It does not demonstrate that humans never do equivalent things.

So most of you probably know the pot roast principle. But if I say that the pot roast principle, do you know what I'm talking about? Maybe you've never heard it called that.

[01:19:17]  Unidentified speaker: Okay.

[01:19:18]  Bruce: Can somebody explain the pot roast principle?

[01:19:20]  Cameo: The pot roast principles is a story about a mom putting the pot roast in. And the daughter said, why do you always cut the ends off? And she said, well, my mom always cut the ends off.

And so they asked grandma, why did you always cut the ends off? And she said, that's how big of a pan I had.

[01:19:37]  Bruce: Right. So they were carrying on this meme that you're supposed to cut off the end of the pot roast and the reason for it wasn't there anymore. And then humans also have automatic reflexes.

Like we have a ton of automatic reflexes and responses that are very much like animals having an automatic algorithm, even if we're just talking about like the fact that you'll pull your hand away when the snake strikes, even though it's behind glass or something like that. So humans do have equivalent things just like animals do. Although we don't have anything quite as complicated as the squirrel example, I think.

Or maybe we do. I just don't know what it is. But I don't know of any examples like that.

It does not demonstrate that animals are not conscious or that they are automatons. Those are not the same thing, by the way. We'll talk about that in the later podcast.

What does it demonstrate? It does demonstrate that animals have really severe inabilities to understand things like a human does. A human, even if they had an automatic response to try to bury their nuts and pat it down with their nose, they're not going to do it on concrete because they have an explanation for why that's the wrong thing to do.

And animals just don't have that, right? And this really is a big giant gap, which is part of the reason why animals sometimes come across so intelligent and sometimes come across so stupid, right? They just, they do not have access to human level explanations.

[01:20:59]  Bruce: We will talk about Byrne's theory. He thinks they have access to a certain kind of explanation, very limited. And we'll get into that.

And then only some animals. I do think the squirrel example is a very strong indicator of the lack of understanding that animals tend to have. Now, you could probably train an animal to not make these mistakes.

Like you could classically condition an animal to not dig on concrete, right? Because they still have a learning algorithm that can override those, those built-in algorithms. But with a human, you wouldn't need to classically condition them, right?

Once the pot roast principle, once they talk to their mother and they find out, OK, here's the explanation for why she did it, and it doesn't apply to us. They change their behavior immediately, right? You don't need to go classically condition them to change their behavior.

You just explain it to them and the behavior changes. And I also think Deutsch would argue here that humans make up explanations. That if you were to ask these people who were cutting off the end of the pot roast, you know, why do you do that?

They may say something like, well, I think maybe, you know, my mom used to always do that. I thought maybe it was to make it juicier, you know, when it's cooking. Humans will make up explanations.

[01:22:11]  Bruce: Now, we're going to see later, I'm going to challenge whether those explanations are actually the reasons why we do things or if we're confabulating them on the spot. There's actually some really interesting studies around this. But I do think that's a difference, right?

At a minimum, we can give you explanations, even if we're making them up on the spot. We can give you explanations for why we behaved the way we did. That's usually at least semi-reasonable, where animals can't do that at all.

With this in mind, animal learning really does amaze me. Okay, so Carlos E. Perez, who's a machine learning expert that's trying to study AGI.

He says, Have you seen a machine with the autonomy of a honeybee? I did a post on Jenga playing cats and dogs. The fact that a dog can be trained in a few examples to play Jenga.

A dog uses its mouth, a cat just uses its paw. It will actually go try to find something to grab. And at one point in one of the videos that I have, it tries to cheat.

It doesn't want to have to experience knocking over the pile. So it tries to convince the owner that the piece that he pulled out is its piece. So, I mean, those clever things that these dogs and cats can come up with.

[01:23:17]  Bruce: If I were to try to do that via machine learning using reinforcement learning, it would be, I mean, coming up with a robot and training it from the ground up to do Jenga. It would require thousands, tens of thousands, hundreds of thousands of runs. And you would have to have it program in a world space first.

You know, I mean, like it's it's nowhere near. We don't know if I were to take any existing robot out there, the best built robot that wasn't built specifically to play Jenga and wasn't given specific knowledge about how to play Jenga by a programmer. And then it said, I'm going to classically condition this robot to play Jenga.

You could not do it. There's none in existence that could do it. Animals are amazing.

Classical conditioning as an algorithm. It seems so simple. You ring the bell and the dog salivates because it's used to getting the food when it hears the bell ring.

Right. It sounds like we understand what we're talking about when we talk about classical conditioning. Go try to program it.

Okay. And this is where the rubber meets the road. This is why I believe that you don't fully understand it if you can't program it.

Okay. That comes from Richard Feynman. Go try to program classical conditioning.

No one knows how to program it. And I'm serious when I say you would win a Nobel Prize if you could program classical conditioning. It is amazing what animals can learn and learn to do.

[01:24:39]  Bruce: Clearly, it's something way shy of AGI. And yet even it is beyond our current understanding. It's mysterious to us.

So this is kind of the background for the rest of what we're going to talk about is on the one hand, I'm going to emphasize animals are brilliant. On the other hand, I'm going to emphasize animals are stupid. On the other hand, I'm going to emphasize animals are amazing how well they can learn.

On the other hand, I'm going to emphasize animals are so stupid compared to humans on how we learn. And these are just all simultaneously true. And trying to figure out exactly why that is, is what Richard Byrne tries to study.

And it is and it is hard. You read one of his books and he's constantly contradicting himself because there's almost always counter examples to every theory about animals. And he goes through and he just takes you through it.

It's one of the things I really love. The book I'm talking about in particular is Evolving Insight by Richard Byrne. I also have The Thinking Ape.

I haven't read that one yet. It's a fascinating book. The next podcast that we do, I'm going to start to take you through his theory.

And I'll compare it to Deutsch's interpretation of his theory. Both of them are the same on most points, but they definitely have a different slant on what they think the significance of the theory is. Deutsch emphasizing how stupid animals are.

[01:25:58]  Bruce: Byrne emphasizing how smart animals are. And I think you end up getting kind of a more rounded view by looking at both views simultaneously.

[01:26:05]  Sadia: By the way, the humans can also be so stupid at times that you see that a certain part of the population, they really, you know, will say that a certain part of the population should just be treated as automatons or not. You know, you see the same argument, even humans using it towards other humans.

[01:26:24]  Bruce: Well, there's actually an interesting point there. You're talking about still universal explainers, but there are humans that are not universal explainers, right? I mean, there are there are humans that exist that have only a brain stem.

And that's it. These humans with only a brain stem, they still have preferences. They like things and dislike things.

They have favorite songs. They get happy when you give them their baby sister and get excited. There's a lot that these humans will do even though they're not universal explainers.

And they certainly act like they have a degree of consciousness. These humans, they are probably more equivalent to an animal, right? They may even be less intelligent than some animals because many animals have a neocortex.

Many mammals have a neocortex and these humans don't. There's a great deal of mystery around all of this. And these questions aren't just about animals because there are humans that, what we're talking about, also this applies.

David Deutsch in his books, he proposes defining a person as a universal explainer. And I'm actually not against that definition, although I do think it's another case of a specialized definition. I would not consider a person that isn't a universal explainer, a human that isn't a universal explainer, as not a person.

They're not a universal explainer, they're not a person in that sense, but they are still a person in some sense, some legitimate moral sense. That is something that we have to deal with, is there's questions around this. How much do these people feel?

[01:27:54]  Bruce: How much do they really experience? How much do they understand when they have no neural cortex at all? There's all sorts of mysteries around them, too.

Well, maybe we can even cover that in some future podcasts. I probably won't get into the humans version of this for the Animal Intelligence Podcasts. But there's a number of interesting mysteries around what is it like to be a human that doesn't have intelligence, that doesn't have the ability to explain things.

All right, that's it for today.

[01:28:21]  Sadia: I think it would be interesting to get into even just the whole thing that why in humans, certain people, even coming from the same circumstances, why certain people throughout their childhood even are more open to change, which I would say to exercising their free will, while the others just pretty much just whatever, whether it's a culture or they just remain, they like to remain conditioned, no matter how hard you try somehow. It's a tough one. I mean, I really don't know.

At least we could maybe discuss possible reasons. Why certain humans are more open to being rational than others?

[01:29:00]  Bruce: Yes. There's interesting questions. Well, what is autism?

Right? Why Temple Grandin argues, and you know, maybe this is true, maybe it's not. She argues that autistic people are in between humans and animals, that they have some of the perceptions of animals because they have a reduced frontal lobe.

Based on that understanding on her part, whether it's right or wrong, she felt that she understood animals better than most, than neurotypicals. And she built a career out of the fact that she experienced life more like an animal so she could figure out what was causing the animals to become afraid. You know, like there's something waving, a little flag waving, and so the cows won't go this direction.

And no neurotypical can figure it out. And she sees the flag wave and it immediately causes fear for her. She goes, oh, the animals are going to be feeling fear over that waving flag.

So she'll go tell them you need to get rid of that flag. And then suddenly the animals will all start traveling the path that they wanted them to. They're no longer afraid to.

And she gives tons of examples of this in her books where she's able to kind of capitalize on her autism as an advantage over neurotypicals because it allows her to understand where the animals coming from.

[01:30:16]  Unidentified speaker: All right.

[01:30:16]  Bruce: Well, thank you guys. This has been a fun episode.

[01:30:19]  Unidentified speaker: Yeah. Yeah. All right.

[01:30:20]  Unidentified speaker: Let's do it again soon.

[01:30:21]  Unidentified speaker: Okay. All right. Bye. Bye. Bye.

[01:30:29]  Bruce: The Theory of Anything podcast could use your help. We have a small but loyal audience and we'd like to get the word out about the podcast to others so others can enjoy it as well. To the best of our knowledge, we're the only podcast that covers all four strands of David Deutsch's philosophy as well as other interesting subjects.

If you're enjoying this podcast, please give us a five-star rating on Apple Podcasts. This can usually be done right inside your podcast player or you can Google The Theory of Anything podcast Apple or something like that. Some players have their own rating system and giving us a five-star rating on any rating system would be helpful.

If you enjoy a particular episode, please consider tweeting about us or linking to us on Facebook or other social media to help get the word out. If you are interested in financially supporting the podcast, we have two ways to do that. The first is via our podcast host site, Anchor.

Just go to anchor.fm slash four dash strands, F-O-U-R dash S-T-R-A-N-D-S. There's a support button available that allows you to do reoccurring donations. If you want to make a one-time donation, go to our blog, which is four strands dot org.

There is a donation button there that uses PayPal. Thank you.