Models

Speaker A: 00:00:00

Hello and welcome to the Cognitive Engineering Podcast produced by me, Fraser McGruer for Aleph. In this series of podcasts, we take a look at interesting topics and discuss what we think they tell us about analysis and decision making. I'm here with Nick Hare and Peter Coghill of Aleph and this week we're discussing models. Nick, go for it.

Speaker B: 00:00:24

, the first discovered in the: 1920

Speaker A: 00:01:30

What, the kites are the models of the kites?

Speaker B: 00:01:32

Well, the kites, the kites, the desert kite is the name of this structure. Got it. There are near some of these desert kites. They found engravings of the kites, i.e. which are small plans. Right. And these things are massive. They're models. They're little models, basically, in order to facilitate planning the construction of the actual real thing. So they basically make little engravings, pictures, plans of the desert kite that they're going to build. And then they go out and build it. Now, these things are massive. You can only see them from the air. So they're like hundreds of feet long. What it shows is that the idea of making a little version of the big thing in order to play with it is thousands of years old. Yeah. Right. So it feels like making models. And that's a sizable proportion of and possibly a lot older. But there are those sort of little model horses and model people, aren't they? That are sort of tens of thousands of years old. So really, it's very similar to when we, in the data science world, talk about models, making a little version of some bigger thing that you want to understand or play with, and then using it to advance your understanding. So we've realized, I think, that we've never really done a podcast episode about models. But they're very important. So that's what this is about. Models.

Speaker A: 00:03:12

Brilliant. Look, I do apologize and I don't want to distract us. But I'm having trouble visualizing this model. And I know this is not really what we want to talk about. But I'm there now. But I don't understand what this kite thing is. How is it a kite that moves animals, that corrals animals or something?

Speaker B: 00:03:32

The thing I have described to you is called a desert kite. Imagine a big V-shaped wall. But hold on. Does this go up in the sky? Fraser, stop thinking about other stuff. Does it go into the sky? No, it doesn't. Imagine a big V-shaped wall. Yes, yes. Imagine the kite where it's just the bottom half of the kite.

Speaker A: 00:03:51

By the way, I don't like your tone. But keep going. But not with that tone.

Speaker B: 00:03:54

Big V-shaped wall. Animals come in the wide end. They get funneled towards the narrow end where they are killed by humans waiting for them. This structure, whether you like it or not, is called a desert kite. Is it anything to do with the kites in the air? Yes, because of the shape. Reminiscent in shape. Got it. Because of that V. Got it. OK. And so were you to draw a little plan of this, which is... So the desert kite, just to reiterate, is a large V-shaped wall in the desert. It's not a kite on a string or anything like that. Big V-shaped wall in the desert. And imagine you were a Stone Age person making an engraving of that, right? It would slightly resemble what you think of as a kite. But the point is that's what we're talking about.

Speaker A: 00:04:44

OK. Brilliant. Just to say, though, is although they didn't have the writing or indeed the letter V, because they probably didn't have an alphabet, I don't think that's very complicated. And I can't really see why it needed a model, because some bloke or lady could just hold their hands up in that sort of shape.

Speaker B: 00:05:03

Well, it's things like, where are we going to put this in relation to the hill? Oh, they would frame it within the geography. Draw a little bit of... Oh, OK. And then you work out...

Speaker C: 00:05:11

It's disgusting, because it's not one person building this. It's going to be a big team of people. They'll be able to say, this leg here is sort of like 100 paces. Yeah, got it. Got it.

Speaker A: 00:05:23

Nice. OK. OK. So apologies to you, too. And apologies to our listeners for that. Great models. Peter, what have you got? You should be all about this. Oh, big on models. Yeah.

Speaker C: 00:05:35

Yeah. So Nick mentioned that models are a reduction in scale, but they're not just that. It's a reduction in complexity as well. Models are useful when you want to take a phenomenon in the real world and you want to sort of be able to replicate the real world sufficient to play around with that phenomena. But you can't replicate the whole world. So you need to trim bits off. So you trim off the bits that aren't important. So if you're building an economic model to predict the price of eggs, you don't really care about gravity and things like that. So you can get rid of the physical parts. Yeah.

Speaker B: 00:06:10

And you're not going to try and model every single egg.

Speaker C: 00:06:12

And you're not going to try and model every single egg. So you take, you downscale the scale of it in the idea of the sort of scale models that the desert kite is, but also complexity. So you can have a less frequent time step or less accurate numbers in it. Or you can have fewer numbers, fewer variables that you're dealing with.

Speaker A: 00:06:34

Is it correct, I mean, to use a word that you guys use a lot, and I kind of understand. Is it fair to call a model a heuristic?

Speaker B: 00:06:47

I can see what you mean. Yeah. I feel like a heuristic is like a model of a complex decision process in a way. It's like, okay, actually, if you're going to make this decision properly, you need to think about 200 things. But it turns out the only thing you need to worry about is, you know, when's the next bus coming? Yeah. And so like, we'll do that thing if the next bus is 10 minutes or more away. You know, it's like a simple heuristic.

Speaker C: 00:07:10

And I was just thinking when Nick was saying these models have been around a while. I think they're probably as old as thought. I think your brain, like intelligence is partly about sort of building models for things, which are, you could say, heuristics, which is sort of a simple. I can't model everything that you're thinking, Fraser, or you as a person in my head. But I've got an idea of what you're likely to do or how you're likely to behave in certain situations. So I have got some form of model in my head about you that helps me understand you, helps me empathise with you, et cetera, et cetera. Thank goodness. Not that I ever turn it on, but it's there. So I think you kind of think in models as well. Yeah, yeah, yeah. So models are as old as thought.

Speaker A: 00:07:58

Yeah. Okay, model. We want to understand something that's a bit complicated, maybe, but also for our own understanding, but for the understanding of others, right? Is that reasonable to say that sort of thing?

Speaker B: 00:08:08

Yeah. Although, I mean, I think because it comes so naturally to us, I think it's actually quite hard to explain what we mean by a model of something. So I think the sort of definitions you see tend to say, well, it's like a representation of something that is meant to be informative in some way. Right, there you go. But that then kind of boils us down to another problem, which is what does it mean for something to represent something else? In what sense does, say, a model car, which is made out of porcelain, what's it got in common with a big car?

Speaker A: 00:08:48

Because it needs to transmit some information, but there's a balance because there's no point in transmitting exactly the same amount of information because it's the same thing and it ceases to be a model, right? But it's got to be enough that it is informative. Right. So there's a bit of a balance there.

Speaker B: 00:09:03

Yeah, but if we think of it as basically a model share as something that shares some features with the thing it's supposed to represent. And the idea is that, well, it just doesn't share as many features. We just have these things that we have not bothered to put in our model that we know are in the real thing, but because for some reason, we don't care about them. So if you were trying to work out if a car would fit in your garage and you had a model of your garage, you'd want a model of the car that was the same proportion, the same size and shape. You wouldn't really care about the mass. But if you wanted to know how a car is going to perform in a collision, then you actually need a totally different set of parameters about the real car. Maybe its size and shape aren't as important as the materials it's made of and so on. So it's like, well, actually, you know, there's an infinite number of possible models you could have of any system. And really, it comes down to what you want to use it for. Okay.

Speaker A: 00:10:04

Models. Right. We figured out what they are. By the way, actually. Go on. One of the distracting things in what I'm sure was an excellent example of the desert kite. Sorry to hark back on this. It just made me think of birds as well. Of like, you know, because I can't help but think of the red kite. Real birds or model birds? Well, I think they're the same thing. Well, hold on. Shall we just start calling them the desert V-shaped wall for you first? Yes. See, this is a good example of where the model has fallen down for me. You know, maybe some people need simpler models than others. But anyway, you know, it makes me think, I'm sure there's some, I bet there's a bird called a desert kite. Anyway, I can't help but feel I'm taking us in the wrong direction.

Speaker B: 00:10:46

No, no, I like it. I suppose it would be interesting to ask what makes a good model and how we can know whether a model is going to be any good for what we want to use it for. But, you know, then the flip side of that would be what makes a bad model. But so, I mean, I think there's what?

Speaker A: 00:11:02

I've got to interrupt you. Sorry. Go on. One of my favourite, because one of my favourite moments in film, Spinal Tap. This is Spinal Tap and Stonehenge. But anyway, I'll just leave that out there. We should put the clip. Yeah, exactly.

Speaker B: 00:11:16

There's some benighted soul who hasn't seen Spinal Tap. We'll put the clip out there.

Speaker A: 00:11:20

And sorry, Nick, I interrupted you.

Speaker B: 00:11:22

Yeah, because it seems like, I mean, going back to what Peter was talking about earlier, there's this trade-off between kind of accuracy and computational complexity that, you know, you. So there's, I mean, a very famous sort of example stroke short story by, well, there is one version of this kind of idea, which is in a very short, short story by our favourite author, Jorge Borges.

Speaker A: 00:11:46

Ah, yes, yes, yes.

Speaker B: 00:11:47

Who wrote one paragraph short story called On Exactitude in Science. And it basically describes a province where they made increasingly detailed maps until they eventually they made one which was the exact same size as the country itself. And, you know, so you sort of think, well, that's in one dimension, the perfect model, because it's got every single thing that the real world has. But in the dimension of actually being useful, it's failed. So it's like, how do you know how much? When do you stop taking stuff out? You know, and so and you face that problem, you know, any kind of model you're going to build where you think, well, you know, I can take this bit out and that bit out, and it's still producing kind of the right results. But, you know, but it's computationally more straightforward.

Speaker A: 00:12:38

And this is not surprisingly feels very close to this question of information and processing. Right. Yeah. Which is right at the heart of, well, we all know what it's at the heart of. Aleph, right?

Speaker B: 00:12:52

Exactly. But but it is like the problem is, you know, if you're trying, if you're building a model to find something out to find it, there is a kind of I mean, essentially, what you're doing is you're kind of saying, well, I'm going to find a model which is simple and straightforward and produces the right the right output. Right. For kind of the what I know. So you've got some things which you know, and you know that the model has to produce those outputs. And the idea is then you use the model to find out about stuff which you don't know yet. Right. So you subject the model to kind of conditions which aren't stuff you've seen already. That's the point. Because if you'd seen it already, you wouldn't need to build the model. You're building and making the model to explore situations you haven't seen. So there's that thing of like, well, you know, the model needs to behave in the same way that the real thing would behave under those circumstances. But actually, how do you know?

Speaker C: 00:13:46

Yeah. And that's one of the failure mechanisms of a model is if there are certain input conditions which would force some unforeseen behavior that your model doesn't account for because it's missing key machinery for dealing with that. Yeah. Then that's one way you could fail.

Speaker A: 00:14:04

I mean, a model can only ever take you so far. But that's the point of models in a way. Right. Because otherwise it goes back to being that perfect map. Yeah. It will always. It has to have some limitations. It has to. All right. I mean, you guys must have done this so many. You must do this all the time. This must be bread and butter for you. Right.

Speaker B: 00:14:23

Well, I make, you know, statistical models of things which generate the right numbers. So that's the that's the idea of, you know, kind of. Me too. Yeah. Wow. I didn't know that. Yeah. That's what we know about each other. Yeah, yeah, yeah. We share that. You know, where you've got a bunch of numbers you might have is, you know, here's the geographical location. Here's a time. Here's, you know, some other variable. And then there's an output. Maybe it's a house price or something. And you want to predict all of those house prices based on the variables you've got. You make a thing that basically a set of equations that produce the right that produces the right numbers. Right. So when you when you crank the handle with those inputs, you get that output more or less. You go, well, that'll do now. Yeah. Does it. But does it tell you does then looking at the model, does it tell you something interesting? Just because it produces the right output doesn't actually mean that it's going to going to be correct or that the structure of the model is telling you something true about the real world. Because there might be any number of ways to produce, you know, the real world outputs, but with a different design.

Speaker A: 00:15:34

But also, I think what I find fascinating about this is, as we said right at the beginning with your example of that, that bird that flies around the desert is that this stuff has been around for thousands of years. This concept is maybe intrinsic to being a human and trying to understand stuff and get stuff done. And I was wondering what advances there might have been in modeling over the last however many millennia. Because the concept is maybe can't move. And the only thing that can have changed or developed is this question maybe of processing power and information. Right. Yeah. That just occurred to me. I don't know if that's a springboard for something or. Yeah. Where do we go? What do we want to say? What more can we do here in our discussion?

Speaker B: 00:16:23

Well, I mean, the point you're making, I mean, I think that's people might not think of it this way. But essentially, the process of designing artificial intelligence using machine learning is doing exactly that. You know, you've got a bunch of behavior, which is sort of correct. And you have a kind of empty, complicated mathematical structure. And you tweak that structure until the thing starts behaving in the way you want it to behave. And you just keep tweaking it. Now, 50 years ago, we had to use fairly analog methods like linear regression, which was just basically running an algorithm on the numbers to get a line which would kind of work as your model. Before that, I think and Peter probably knows a lot more about these than I do. But there were like analog computers, so things to predict the tides, for example, which would be like various wheels that would all get cranked with one another and, you know, and produce the right the right output for when the tides would be. And yeah, but but I think now it's it's we're getting to the point where the complexity of the system being represented is is almost arbitrary because you can just represent anything with, you know, a big enough, say, neural network.

Speaker C: 00:17:38

And that's because computers just dropped in price because of computers, and then ever greater sized computers. And also, there's a bit of a sort of shift in the way that science is done. There's like there's technology things like the computing and the sensing and we cover the globe in temperature sensors and everything gives loads of data about how everything works in, say, for example, climate things. But there's also a bit of a shift away from some people working in a laboratory doing very, very careful experiments and very, very carefully measuring things and to well, we'll just run an experiment lots of times, collect lots of data and crunch that data. And then the computer will spit out some equation which tells us what the model is. So if you think back to how Principia Mathematica, how Newton did his experiments, he was very meticulous in how he went about his measurements of star positions, planet positions, and did lots of sort of very hand, very tough hand calculations to work out the relationship of how the tides went around.

Speaker B: 00:18:42

I thought he just sat in a garden and an apple dropped on his head.

Speaker C: 00:18:45

Yeah, and lots of very kind of intricate thinking. Whereas now we could just sort of we could point to if we didn't have the laws that govern how things move around in the solar system, we could just point things every single star we can sort of see, put that through a computer and the computer could then crunch out how things work.

Speaker B: 00:19:03

Yeah, and almost like that would be a better model in terms of performance, but it wouldn't give us any insight. And I think that's that's one of the big limitations for, you know, machine learning is that you get these, you know, you can train a thing that will do a perfect job at, you know, finding a cat in a picture. I think in the past, so the way that we would have done that 30, 40 years ago is humans to go, right, if you want to know what cat looks like, it's got pointy ears. So we're going to have a designer thing that finds pointy ears in photos and it's got whiskers. So we're going to look for, you know, lines at this angle and you design a kind of cat finding mechanism by hand. And, you know, and in that process might tell you a bit about how cats work. You work out, oh, the big discriminator is actually the whiskers, you know. And whereas now, you know, the structure of the network you've designed to do that, to identify those things, it's incomprehensible. It doesn't tell you anything about cats. It's just a big load of numbers. And when you look at the thing, the patterns they're looking for, they don't mean anything to us by and large. So, you know, you can get a thing that is a better model in terms of output, but a worse model in terms of the understanding it gives you.

Speaker A: 00:20:19

So as in, not surprisingly, as in kind of all areas of technological progression, you're talking about cutting edge. Yeah, well, of course, is, you know, but it's that sort of literal, sorry, it's that saving of cost in whatever way you want to interpret that. Right. Okay, so where do we want to go? Where can we go from here? Peter?

Speaker C: 00:20:39

What might be interesting is exploring, like, what are models that people don't think are necessarily models? Because everyone does, everybody does modelling, you just do it, you do it in your head. You say, I gave an example of how you think about somebody you know, but you do all the time when you, I want to walk through that door. Your brain has got a model of how a door works that lets you then reach out to the handle, turn the handle. And you kind of, you can, that model is sufficiently abstract and general that you can sort of apply to other doors. I don't have to work out every single door as I come to, I don't have to think about it very hard because I've seen lots of doors. So I've got a door model that I can then apply to other doors in my head.

Speaker A: 00:21:21

But the problem with that, that kind of inflation almost ceases, it worries me that that ceases, the word model starts to become meaningless, it feels like, if it's just, well, everything's a model. Or do you know what I mean? It starts to, it's in this sort of, what do you call it? In a continuum? Yeah, exactly. Where at one end of it starts, I definitely know what a model is. At the other end, you're going, well, that's a model as well.

Speaker B: 00:21:49

When does a model become an analogy? Yeah, anyway, I think they're the same. I mean, I think Peter's right. It's incredibly fundamental to the way that we think and the way that we learn. Because, you know, when we're learning, we are, well, what we're saying is, okay, if I make a generalization about a door, right? So the fact that you say that this thing is a door is, and that it has some sort of similarity with other things called doors, is because you're comparing the features of the two things. And you're saying that this thing shares enough features with this other thing, all of these other things called doors. So I know this is a door. And then that enables you to bring all of the learnings, the deep learnings that you've had about how doors work, to manipulating this one, right? So you know that story which we've discussed in the past, Funes the Memorius, the Borges story, to bring, so Borges gets a double shout out on this podcast. I think triple, actually. That's, you know, that for Funes, every single object was totally unique. So he had no ability to say, well, this door is as different to that door as it is to a cat, right? They're completely unique objects and is completely crippling. You can't think if you have that. And I think so. I think this kind of is absolutely fundamental to the way that we think. Any time you're predicting what the impact of something you do does, it's based on modelling, which is essentially analogies. It's essentially saying this situation I'm in is similar to some other situation. And these objects are similar to some other objects. And that enables me to predict how they're going to behave. So, yeah, I think it is really fundamental.

Speaker C: 00:23:29

Other models that people use day to day. So Nick mentioned maps. So, yeah, I think that's a great example of, you may have heard of Google Maps or Waze, I suppose, that's a pictorial representation of the world around you. It's a model. And it's a model. So and it's got kind of the right amount of information to be useful without having too much to be unuseful. And dynamic digital maps are great because they bring in and filter out information as you go to different levels of Zoom. So that's a model. This is classic London tube map area, isn't it? This kind of thing. Tube map, that's a really good model. The kind of topographical model just shows you what connects to what and where and in what order. Doesn't show you actually where things physically are. That's a very good model. You've heard of photographs? Yeah, I have.

Speaker B: 00:24:19

Yeah, I mean, according to the very general definition. Are they models? Yeah, yeah, exactly. I mean, they've got, they share features with the thing that they're representing. I mean, and you can use a photograph to learn about something. You can look at a photograph and learn about the thing it's photographing. We don't notice that. Well, hold on, let's turn this around for a moment.

Speaker A: 00:24:36

I'm a bit worried now because. Are you a model? And if so, what are you? Yeah, yeah, I'm still trying to find that one out, right? Well, no, but yeah, but there is a difference between, although a photograph can be used as a model, and it definitely is on one level, in another sense, it's got nothing to do with a model. Because if a purpose of a model is to explain something. But even then, does it, right? Even then.

Speaker B: 00:25:00

I mean, that's what I think you'll say is that, you know, and the fact that you might interact with a photo in a different way to how you'd interact with the person might mean that you get some understanding out of it that you wouldn't get with the real life interaction. But I think this is, I mean, I think we've touched on this is a very fundamental idea. That's been kind of, you know, I suppose we've raised it up and we pointed at it and said, look, oh, we've discovered there's this thing which is models and it's useful to have that concept, right? And I guess it's so natural. It's one of those things we don't notice we're using all the time. But I suppose one thing I want to bring up is that obviously, as humans, we really enjoy, and I suspect because it's useful, we really like models, right? We like actually like playing with toys, toy cars, flipping brilliant, love toy cars, doll houses, you know, computer games like SimCity and Civilization. So I suppose I want to bring that up. Like, why is it such fun?

Speaker A: 00:26:06

Hair. Why is playing with models such fun? You've hit upon something here, haven't you? Because I'm sure all of us were into models when we were kids, right? And I love the way that until this moment, we've not made any obvious jokes about models, right? But that moment's been gone now.

Speaker B: 00:26:23

Yeah, classic Fraser.

Speaker A: 00:26:25

I've messed it up, yeah. But anyway, I remember when I was a kid, yeah, lead models, Warhammer, all that business. Loved it. Toy soldiers, airfit, all this kind of stuff. We all loved it. And what is it about models that we love? And it must be so closely connected to what we're talking about, which is it's almost the real thing. It is the real thing. And yet it's got the quality that you can have it. It's accessible to you, right? And yeah. And also, what's interesting, I suppose, is we don't make models of, I'm trying to think what we don't make models of for kids. Fire? I don't know. But I think you can get Lego fire. Yeah, you can. Yeah. But it's that way of interacting with the world and conjuring up your, or feeding, nurturing your imagination with this stuff.

Speaker B: 00:27:20

Yeah, I mean, that's it. That you can tell a story with your model. And you could tell a story in your head. But the model, it's like it's externalizing and doing a lot of the work that your brain would have to do. You know, because it has things like which the real world has, like continuity and physical place. And you can say, right, this now the little man's going to go out into the garden. And now you don't need to remember that he's not in the room with those other people, you know? And it's like it really just it means you can act the thing out. And as I said, yeah, I but I think that is it's not something we learn to do. It's so intuitive that children get it straight away. The idea of why. But why is it fun? What is why is it fun playing with little models of things?

Speaker A: 00:28:06

Well, I think it is that it's that it's the the the only boundaries you're a kind of imagination. It sort of allows you to lock into that. Why is that fun? Oh, well, when why is when we could keep going? Why? You know, but but but also why would we? Yeah. But it reminds me of a few years ago. One of the first documentary films I ever made was and the fascinating people was Puppeteers. Right. And I was saying, you know, I was interviewing them. What is it about puppets, you know, about marionettes and why this and not say acting on stage or in film or whatever? And they said, oh, well, the wonderful thing about marionettes, about puppets is that it gives you that there are no boundaries, you know, that we would have to do exactly what they're told. Yeah, but sort of. Yeah. And yeah, it was that. It was that it just lays open this wonderful, this wonderful kind of valley of exploration.

Speaker B: 00:29:03

Yeah. And there's a real joy in, you know, when you go to a museum and in the museum, there's a little model of the museum. I love that. And in the model, there's a model. Well, there is there is a yeah, there is a model village, isn't there somewhere where they have a model of the model village in the model village? And and then they have a little model. Yeah. They have a little model of the model village in the model. We'll dig it out and put it in the podcast. But but that gives me just an inexplicable joy. Just, you know, it's just joyous having a little model of something.

Speaker C: 00:29:38

Yeah. Yeah. And I think it's to do with the it gives you because the physical model filling in for your you mentioned it being like a memory. The little guy goes out in the garden. You don't need to remember the guy being in the house anymore. That is a sort of multiplier in your capacity for playing things out. Yeah. So I think the joy is it going to having models for something lets you unlock the door to a wider search space of things you can do. And hence new challenges and new new stories you can tell yourself. That's where the joy comes from, I think, is in this new place you can go and play rather than just being in your own head.

Speaker B: 00:30:19

Yeah. I mean, I but I suspect, you know, that kind of that urge to to pick them up and play with them is ultimately because they use they were learning something. And I think, you know, when I when I first started learning economics and, you know, you learn these kind of relationships, I would say almost hidden relationships between things in the real world, like, you know, the the cost of oil goes up and somehow the price of beans rises mysteriously in the shops. And if you know, if you don't, if you never bothered learning about economics or whatever, it's just a vague idea. But then economics gives you these tools to trace these effects, you know, through the system. And I remember thinking it's almost like my brain is going this. I can use this to somehow optimize the world. I'm learning how to control things in the world and what the effects of, you know, making these little changes would be. And it's not that dissimilar from, you know, from picking from picking up making a car out of Lego and driving it into a wall and saying, you know, it's like now I can make these things happen in the real world. You've suddenly got the power to learn about and do things in the real world. And having the little model of it is 90 percent of the way there.

Speaker A: 00:31:33

Well, look, congratulations, because you've almost made economics sound exciting and interesting and fun. So maybe that explains a little bit about the way why you are the way you are. Yeah, thanks. But unfortunately, also explains why I'm like I am. And look, I need to finish this off. I can't tell you how much I enjoyed that. I thought I was good. I thought that was brilliant. I really enjoyed that. However, yes, to just to bring us back down to earth, kites.

Speaker B: 00:32:05

Oh, God, I thought you were going to say, what's your favorite model that you've made, that you've ever made or something? Yeah, OK. What's your favorite kite? Ironically, a model kite would be a really bad model of a kite, because it turns out that the kite, in order to behave like a kite, needs to be actually the size it is because of something to do with air pressure and surface area and stuff like that.

Speaker A: 00:32:26

Yeah, yeah, yeah. Yeah, no. So I can't leave this thing about kites now. To be honest, although you guys have been chatting away for the last half hour, all I've been all I've been thinking about is kites. So what I would like from you, and it's too obvious to ask, what's your favorite model and stuff? Yeah, favorite kite, definitely. You know, what's your favorite kite experience or thought? And that can be interpreted that however you like. It can be your personal experience of a physical kite, but it could even just be a reference in art or literature or whatever, you know. So, sorry, kites. Yeah. As in the flying stuff.

Speaker B: 00:32:58

Right. Favorite literary reference to a kite. Well, I've got my answer straight away. Go on then. Let's all fly a kite from Mary Poppins. No, you stop. Well, that was silly of you, wasn't it? It was, wasn't it? Well, you could just choose from one of the many hundreds of other fictional kite references. But it could be a factual one, as in the kite flyer of people or whatever it was. No, but it's a really joyous song, actually. Really. I mean, I love Mary Poppins anyway, but it's really, you know, a kind of your soul. It takes your soul up with it.

Speaker A: 00:33:28

It's almost like it lifts you up like a balloon. Yeah, like a balloon. But yeah, once again, Nick, we've discovered, you know, not only do we have our deep interest and love of economics in common, right, and board gaming, but also music. Musical theater. Yeah, and actually, I don't really like musical theater, but I do love Mary Poppins. And I know we're meant to be talking about kites on this podcast. But actually, my favorite Mary Poppins one is Feed the Bird. Oh, it's so beautiful. And it's a little bit sentimental. And I love playing on the piano. It's a proper tear jerker, that one. Yeah, beautiful. But yeah, that's my favorite. So we can share that. We don't have to have a different one, you know. Peter, you can come and join us in the whole kite, you know, let's go fly a kite.

Speaker C: 00:34:12

I do like Mary Poppins. But my favorite kite experience was I was on holiday, and I can't remember where, somewhere in the UK. Somewhere windy, but not too windy, I hope. Probably, it's probably somewhere like the Lake District or Peak District. I was walking. Anyway, we stopped with a friend and we stopped at this nature reserve place. And it was a red kite reserve. It was before red kites became big. Hold on, you're taking us in the wrong diner. We've gone totally off. That is totally not what I meant. That's not a kite. I know it's not, that's why I thought of it. So if you tied a string to its leg, though, it could be a bit unlikely. Stretched down with sticks to, yeah. But it was, yeah, so some red kite reserve. And this was 20 years ago, and red kites are everywhere now in the UK, back to sort of pre, yeah, numbers they used to be. Pre-18th century levels. So you didn't see many of them. Anyway, there's a little pond. And in the pond, the middle of the pond, it was quite a big pond, maybe 50 metres, 100 metres around. And in the middle was a little island. And there was kite feeding time, where somebody came out with a big bucket full of cadavers of rabbits and things, chucked it on. Of red squirrels.

Speaker A: 00:35:32

This is your treasured memory. This is your treasure.

Speaker C: 00:35:35

And they chucked it on this mound, and nothing happened for a few minutes. And then a few birds started circling around. And then within a few minutes, there were hundreds of these enormous red kite birds just flying around. And soon after that, when they reached a sort of critical mass, they started swooping down and grabbing these bits of carved up rodent off the mound. It was very impressive just to watch these very beautiful birds.

Speaker A: 00:36:08

Yeah, nice. Okay. Well, there you go, listeners. Everything you ever wanted to know about kites, but were too afraid to ask. All right, let's stop there. Wonderful. Thank you, as always, for listening to the Cognitive Engineering podcast. I'm Fraser McGruer, we've been here with model and kite specialist Nick Hare and Peter Coghill of Aleph. Until next time, goodbye.