In Our Time - Artificial Intelligence

Episode Date: December 8, 2005

Melvyn Bragg and guests discuss artificial intelligence. Can machines think? It was a question posed by the mathematician and Bletchley Park code breaker Alan Turing and it is a question still being asked today. What is the difference between men and machines and what does it mean to be human? And if we can answer that question, is it possible to build a computer that can imitate the human mind? There are those who have always had robust answers to the questions that those who seek to create artificial intelligence have posed. In 1949 the eminent neurosurgeon, Professor Geoffrey Jefferson argued that the mechanical mind could never rival a human intelligence because it could never be conscious of what it did: "Not until a machine can write a sonnet or compose a concerto because of thoughts and emotions felt", he declared "and not by the chance fall of symbols, could we agree that machine equals brain - that is, not only write it but know that it had written it." Yet the quest rolled on for machines that were bigger and better at processing symbols and calculating infinite permutations. Who were the early pioneers of artificial intelligence and what drove them to imitate the operations of the human mind? Is intelligence the defining characteristic of humanity? And how has the quest for artificial intelligence been driven by warfare and conflict in the twentieth century? With Jon Agar, Lecturer in the History and Philosophy of Science, University of Cambridge; Alison Adam, Professor of Information Systems, Salford University; Igor Aleksander, Professor of Neural Systems Engineering at Imperial College, University of London.

This BBC podcast is supported by ads outside the UK. Thanks for downloading the In Our Time podcast. For more details about In Our Time and for our terms of use, please go to BBC.co.com.uk forward slash radio 4. I hope you enjoy the programme. Hello. Can machines think? It was the question posed by the mathematicians at Bletchley Park and especially by the codebreaker Alan Turing.

And it's a question that's still being asked today. What's the difference between men and machines? and what does it mean to be human? And if you can answer that question, is it possible to build a computer that can imitate the human mind? In 1949, the eminent neurosurgeon, Professor Jeffrey Jefferson,

argued that the mechanical mind could never rival a human intelligence because it could never be conscious of what it did. He said, not until a machine can write a sonnet or compose a concerto because of thoughts and emotions felt, and not by the chance fall of symbols,

could you agree that the machine equals brain, that is, not only write it, but know that it's real, Unquote. Yet the quest rolls on for machines that are bigger and better at processing symbols and calculating infinite permutations. Who were the early pioneers of artificial intelligence and what drove them to imitate the operations of the human mind? Is intelligence the defining characteristic of humanity? And how has the quest for artificial intelligence been driven by warfare and conflict in the 20th century? With me to discuss artificial intelligence is Igor Alexander, professor of neuroscience engineering at Imperial College University of London, Alison Adam, Professor of Information Systems at Salford University,

and John Egar, lecturer in the history and philosophy of science at the University of Cambridge. John Eger, there's a fascination with boundaries that separate the artificial and the human. Let's go back to the end of the 18th century and the automata being developed then. Can you tell us about the famous Vauclausson's duck and why it was famous? My favourite object of all time. Vaucer Sond was the greatest automator maker of the 18th century, probably of all time, and he built a series of automata in the 18, so in the 1730s, which were, they could do all kinds of different things.

He built ones at the automata that could play a flute, so they could produce music, they could move, and they could almost talk. He produced a duck, most famously, which could do almost anything that a living duck could do. It could waddle, it could flap its wings, it could eat, and it could famously excrete. producing very realistic duck poo. And what was he trying to do in this?

Obviously amuse the court and entertain his own mind, but what was this supposed to prove in the way of things? Was this attempt to show that humanity could be replicated? Human mind could be replicated? I mean, the way to think about these automators, they're not toys. I mean, these are, to start with, these are extremely expensive objects. These are made out of the finest metal using the finest mechanical skills available. So these are certainly not toys.

They would have been shocking to people because they seem to be doing things which only the living or even the human can do. Things like playing music or things like eating and excreting. And these were extremely troubling. So people would look at them and be pretty. provoked philosophically. These are disturbing objects. Can you give us some notion of the disturbance, what people said, what questions were raised by this? Well, for example, if it made people reflect on what it is to be human,

if this thing in front of them is playing music or eating and excreating, yet it's a machine, then what is there left that is human? or what is it that's left that distinguishes the living from the non-living or the human from the artificial. So in that sense, these were stepping on very, very sensitive boundaries. Alison Adam, when we come to the Enlightenment period, which is later than the 70s, towards the end of the 18th century, how is this developing the notion of what the human mind is?

Because, as we've been told by John, this challenge is what the human mind is, is that they can do these things. You thought human being only could do, they can do music, they can do this, these machines. So what's developing in the atmosphere of ideas there? Well, I think what we're seeing is the development of evolutionary ideas and these ideas about the specialness or not, indeed, of the human condition, which were obviously delta blow anyway with the Copernican revolution,

but coming into, I suppose, the beginning of the 19th century, obviously before Darwin, because many evolutionary ideas were around before Charles Darwin himself, And, of course, evolution began to challenge the specialness of humanity and particularly the specialness of human cognition, which is maybe the last bit which could have been special about humans, because, of course, we have the idea that from evolution, that human intelligence arises from complexity. But let's just, we'll come to evolution later,

in the sense of Darwin. Let's just stick with the Enlightenment. I'd just like to take forward what John was saying, the ideas that are round in the Enlightenment, coming from these autonomous and coming from what Charles Babbage was doing and how this is saying things to human beings about their minds. Yes, well, I think if you take Charles Babbage as an example, of course he was the archetypal British polymath of the early 19th century,

and he sort of forms a bridge between, well, literally because he was born in 1791, a bridge between the earlier worlds of the automata, because he was influenced by these. In fact, one of his early experiences was, apparently at the age of 8 in 1800 to go along to one of these mechanical museums. I don't know if he saw the duck, but maybe burned down by then.

But he certainly saw a wonderful silver dancer. This has become quite famous, a 12-inch figurine who had a bird on her hand and this flapped its wings and so on. And then years later, of course, these museums came and went. But years later, he managed to acquire this wonderful automata in 1834. He bought it as an auction.

And he actually displayed this in his London drawing room alongside a portion of his famous difference engine. So what was you doing with the difference engine? What was he trying to do with the difference? Well, although there's that romantic influence, it's clear that there were more pragmatic influences on him as well. And we're talking about an era where, I suppose,

we're moving towards science being seen as the engine of industry, as the engine of empire, and, of course, the engine of science is mathematics. And there was a clear need for much better mathematical tables for navigation, you know, thinking of the British Empire and its huge extent. But the problem with these tables, mathematical tables, was that they were notoriously inaccurate because, of course, they had to be computed by humans.

Indeed, the first meaning of the word computer is as a human computer. You mean calculator or computer? A computer, a computer. Well, the name computer is given to people who compute things. And that's the earliest meaning. Anyway, these tables were very inaccurate. and Babbage was very exasperated by this. In fact, he famously said,

I wish to God that these calculations had been executed by steam. So he was definitely inspired to build something mechanical. So what was he doing with the difference machine? Well, he was trying to find a mechanical means of producing mathematical tables that was swift and accurate.

And he got the idea from France, from post-revolutionary France, where there was a huge project to build mathematical tables by a special method called the method of differences. Now we can bring in evolution, because how was the ideas which came out at the time? How did they influence baggage? Well, I think it's because this idea of building up something, breaking down a complex thing in something simple

and then building it up again. That was really very much what he was doing. He saw, I guess, evolution as building up a complex being from simple elements. and he wanted to build up complex calculations by mimicking the kind of division of labour that you'd have to do in producing mathematical calculations in a machine. Hugo Alexander, can we develop the Charles Babbage thing and his relationship with Ada Lovelace,

what was he doing that matters in terms of the later discussion we're going to have about artificial intelligence? What are you reaching out to, his difference machine, then his analytical machine, and how does she, the daughter of Lord Byron, she was a brilliant mathematician, how does she fit in, which she does,

into this picture. Yes, the important thing is the analytical engine, which was a much more ambitious project than the difference engine. And that moved much more into the doing of mathematics, rather than the producing of tables.

Now, here you're back to the issue about mind, because doing mathematics is supposed to be one of the highest things that you can do with your mind, according to the Greek philosophers. So he had these ideas about an analytical engine, and along came the 17-year-old Ada Lovelace, daughter of Lord Byron, as you say, a very short marriage to Anne Milbank.

And Byron left the family when Ada was about one month old. And Anne was rather disgusted with this, and so she was quite keen on getting Ada to have a very rigorous mathematical education. And nothing to do with poetry. Nothing to do with poetry as far from poetry as possible. So she, even at a very early age, she took a great interest in maths and met Babbage at the age of 17 and immediately became engaged in the project of the analytical engine. And she was truly a brilliant mathematician.

She was a brilliant mathematician. And she's often called the world's first programmer. Well, that's wrong. She was the world's first theoretical computer scientist. Because what she did in looking at a leaflet that had been written by Frenchmen about the analytical engine, which Babbage asked her to translate into English, she translated it in about 15 pages and then wrote about 40 pages of notes of her own. And within those 40 pages, she had a proof of the universality of,

of an analytical engine, which of course was never built. Right. Can you just give us the analytical engine, briefly, succinctly and brilliantly, Ego? I'm sure you can. Well, the analytical engine is about as close as you could get at that time to what we call a computer now. An important thing is that it could be automatically programmed, and what Babbage had in mind was really quite ingenious. He saw a Jacques Aloum,

which is the thing that makes textiles, and that had little programming cards on the side which put the name of the textile maker on the side of the loom. And he used that mechanism in order to provide the programming effort for the analytical engine. So the analytical engine had all the elements of a computer, including the possibility of programming it. And the theoretical end of that is that Ada Lovelace discovered that she proved that you could,

perform any mathematical function with that kind of setup. And a breakthrough of similar importance came much later in the mid-40s, the 20th century in 1940s, John von Neumann. Can you briefly tell us what he brought to the table? Well, von Neumann, by the time he got interested in computers, was a highly revered figure in the United States. He was Hungarian-born. He went to Princeton in the 19th. in 1930, and he sat on the board of the ballistics research laboratories of the Pennsylvania Moore School of Electrical Engineering. And he was asked one day to pop down to the cellar and have a look at this computer

that they built, a thing called the ENIAC, which was full of valves and what have you. And that too was producing tables largely for ballistics exercises, for pointing guns. and the builders of this were dissatisfied with the design because it had very complex valve circuits and it had to be programmed by throwing little switches on the outside of the machine. Van Neumann took one look at this or perhaps several looks at this and said,

I'll join your team and he kind of created the possibility by working with the designers of moving those programming switches into the memory of the computer. Now that was a killer move because then one started losing the distinction between what was data and what was program.

Of course the computer could sort that out, but you didn't have to have hordes of people pushing buttons on the machine in order to programming. John, can you just distinguish for us before we move on to the next stage of the program, Can you give us the distinction between computing and artificial intelligence? Yes, I mean artificial intelligence, in principle, could be demonstrated by any machine. It's what Igor mentioned about the universality of computers that gives them the edge,

which actually means that if you are, because a universal machine or a computer can do this incredible diversity and range of possible things. In principle, in fact, it can do anything that a special purpose machine could do. It could imitate anything of those simple machines that if artificial intelligence is going to be realized, it will be through a computer. So in that sense, the two are intimately connected. Does memory, does the development of a memory, is that a crucial factor? It is. Can you just explain that for us? I could give an example.

Yeah, that's better. It's when Turing was describing in his mind paper, what might be the obstacles to making a producing intelligent behavior from a machine. Well, he said, well, it probably is largely a case of storage, that once we have enough storage available to a computer, then maybe a lot of these intelligent-like behavioural functions is would be producible by a computer. Igor, can we move on to artificial intelligence straight on now?

And you believe that, I understand it, you think the first person successfully to demonstrate this was Claude Shannon. Can you briskly tell us who he was and how he demonstrated it? Yes, indeed. Claude Shannon's my hero, but putting that aside, he was probably the greatest engineer of the life. century, he's best known for having formalized the statistical theory of communication on which all communication systems are based these days, Internet, World Wide Web, the whole lot.

But he also became interested in computers because he did some work when he was a student on computation. So what he realized was that computers, as they existed at that time, which 1950, three years from the von Neumann development, he realized that they don't need to be used only for number crunching and that the numbers can be just used as symbols. They're just noughts and ones inside a computer. So he used these notes and ones to describe the state of the board of a chess playing program. and he developed some ways of looking at moves in chess, which inside the computer were numerical,

but from the outside looked like some pretty smart manipulation of ideas which were not numerical, and he was the first to write an algorithm for computer chess playing, which was still used by the IBM machine that beat Kasparov. You, John Egg, you place a similar emphasis as Igor does on Claude Channon on Alan Turing and people listening to this programme

will know about Alan Turing, Enigma and so and so forth. Can you tell us what part he played in the development of artificial intelligence and what it says about our intelligence? Can we bring that back into the equation? Sure. Just to remind people who Turing was, he was an English mathematician

born in 1912, born to parents who worked, his father worked for the Indian Civil Service. He comes from actually a family background that would know about how government works. Now, his achievements can be summarized in what he did in two papers, really. The first paper he wrote in the 1930s is an attack on an incredibly arcane problem

in the foundations of mathematics. And he goes about it in a remarkable way. He doesn't attack it directly. but in tackling this problem, he articulates this thing which we call the universal Turing machine, which he describes is a machine that can imitate any special purpose calculating machine. And he demonstrates what its properties are, what its limits are. Now, this is the imaginary machine, which is in fact equivalent to the computer.

Now, where that relates to what human intelligence is, because he sets up the description of the universal chewing machine by thinking, what can a human clerk sitting in one of the offices, like a government office, in fact, overseen by a generalist, someone who's an overseer with all the instructions, what could that kind of set up produce? What kind of intelligent behaviour? What kind of information processing could that kind of set up produce?

And he says, well, actually, this can do pretty much any kind of calculation you could ever want to do. And this is equivalent to what Babbage described actually in the analytical engine too. So in solving this very arcane logical problem in mathematics, he gives a description of an entirely new kind of machine that is equivalent to what a computer can do and it's one based on a model of intelligent human action. We're still at this stage, though, aren't we? I'm moving across to Alice now.

We're still at a stage of thinking of the mind as being that which can be described by mathematics and symbols. And this is exemplified in the famous Turing test. Can you tell listeners about the Turing test, and then let's take the argument on? Okay. Yes, now the Turing test was proposed by Turing. He called it the imitation game in a famous paper in 1950

and it was his take on the question Can a machine think which he thought was something really you shouldn't answer directly and that you could approach it more obliquely Now he proposed the imitation game which involved a man and a woman in separate rooms and an interrogator I suppose of either sex And the idea was that the interrogator would ask questions

of the man and the woman in turn Obviously no one could see anybody else so it would be via a was a teleprinter in those days. And the interrogator would try to work out who was the man, who was the woman. Now, what Turing proposed was that let's remove the man and put a computer in there, and let's see if the interrogator can decide wrongly, because the interrogator might decide wrongly, as often when the game's played like this with a computer as he did

when the game was played with two people. And that really was his Turing test. If it was about the same, then we could say that a machine could think. And he predicted that by the end of the century, at the end of the 20th century, that computers would be able to reliably pass this test at about the same level as the two people would, if you like. And this is where we're talking about thinking

the artificial intelligence is being, the dynamo behind it is the defence system. We're talking touring, Enigma, code cracking, the defence systems in America were taking up artificial intelligence the mid-20th century. How, again, did that mean, what did that mean for the direction that artificial intelligence?

What did that mean? We've got to keep this in play. What did that mean for the idea of the human mind that these people had in mind while they were doing this? Well, I think it's very much in terms of seeing the mind as something which process symbols, seeing the mind as a computer. You know, it was a very sort of historical argument in a way.

But, I mean, if you're talking about the, what happened really after the, the war now. In defence, there's no doubt about it that the Cold War fuelled a lot of developments and artificial intelligence, especially in the USA, where a lot of research money was put into AI and computing in general, and computing in general. And a number of developments came from that period. I mean, I think the thing about how AI was seen then was, I mean, clearly to get defence spending, you weren't trying to create an artificial human. They, they the couching of AI was more in terms of decision theory, command, control, things like that,

rather than seeing it as, oh, we're trying to create an artificial mind. So that's why money, they were able to get quite a lot of, a number of researchers in places like Massachusetts Institute of Technology, Stanford, were given large amounts of money to carry on working in AI. But we're still ego, we're still working on the idea that the human mind can be described in terms of mathematics, and Galilei, the book of the universe is written in the language of mathematics. So this is still the idea that this mind that we have is susceptible, most of all,

to that description, mathematics and symbols. And is that in place still, you tell me, and then can you bring in the idea of neural networks, in which you are professor of amongst many other things, in this business of artificial intelligence? I happen not to believe that the mind can be described in terms of symbols. either, but still this is what your heroes have been saying for quite a while, so I'm trying to play on your patch. They have been all the Turings and the Shannons and all that, as talking

always, you tell me if I'm wrong, please, it'll help. They're talking always in terms of we're applying mathematical formulae, we're applying symbols, these machines are responding in that way. By that, we judge their intelligence and we assess our own. That's happening up to press, and the neurons come into that, don't they, as well? You're absolutely right. The neural stand aside from this a little bit because although the history of neural networks

has to do with producing practical machines that can recognize patterns and things like that, more recently they lead to getting a better understanding of the brain. Can you tell people what, do you remind people what a neural network is?

The brain is made up of 11 billion little cells. These little cells have the property of recognizing tiny little patterns of what other cells are doing. And they're able to change their function, which is the basic way in which we say that the brain learns. Now, we can build neural networks that totally replicate the mechanisms of the brain, but then one must start distinguishing clearly. between mechanism and mind.

Mind is a total capacity of an organism to think. And if you look at the history of artificial intelligence, there has been a lot of confusion there because mind was inferred from smart behavior in a machine. And, you know, the Turing test is a bit about that, and I think the Turing test can be enormously misleading. especially the way it's interpreted these days, you know, talk to something,

is it a machine, is it a human being? If you can't tell, and it's a machine, it's like a human being. That's absolute nonsense, because you can create mechanisms that have minds of their own, and if you were to ask such a mechanism, excuse me, but are you a human being or a machine,

it would have to say, immediately, I'm a machine, and I have my own type of mind, and I process symbols, and I may do it with neural nets, or I may do it some other way, but I distinguish myself from being a human being with a mind. So all the time, Alison, we're running into the problem of, not the problem,

but there's the proposition that these methods of gauging intelligence so far employed by those looking for a mechanistic or artificial intelligence are too pure. That's a word I think I saw in your notes. Can you just elaborate on that? Well, pure... That's your pure systems. Yes.

Well, for instance, I tell you, what I think you said, is that, you know, we're talking about symbols and mathematics. We're not talking about language, the hundred and odd meanings of the word, I don't know, taking Dr. Johnson's dictionary. We're not talking about context. We're not talking about news. We're not talking about bodies. We're not talking about all the rest of the stuff. Absolutely. John, I was going to say, but that's precisely why the Turing test is very effective because it's a test about natural language. And it turns out the ability.

You think it's very effective? Well, I think it's because it's the focus is on whether this, whether a, machine can perform talking in a convincing way and that turns out to be incredibly difficult. Time and time again when

scientists and philosophers think they've got natural language cracked and understood, it turns out to be much, much more difficult than we ever expected. And the Turing test focus on communication through language over the teleprinter. It could equally be a text message.

You could run the Turing test on having two people text messaging each other. If you can do that and then convince people that your machine, if you can use language effectively and convincingly, then there's a good reason for saying that's intelligent behaviour. But it's an incredibly harsh and difficult test. I think John's picked up a very good point there. The language is one thing.

It's hard to be fooled for very long by a computer. But of course, the thing about the Turing test, I mean I agree it's not really a very effective test, is that, of course, it actually hides the bodies. So the minute you actually ask your computer and your human to walk upstairs or throw a ball, well, it's no contest, is it? It's because you've hidden that aspect that we can even have a Turing test which is in any way convincing.

So we don't seem to have gone very far, do you really? Well, just this question of language. I remember giving a talk to very young people once and a little six-year-old got up and I was talking about the difficulties that computers have in producing human-like language and doing vision and things like that. And she said, you're saying your machines are intelligent,

but my little brother can speak, and he's not very intelligent. I mean, it's easy in a way to sort of knock the machine. It's easy to say they can't get the nuance of language, they can't run up and down stairs, they can't, they can't. But the interesting thing is what they can do and how far they are towards getting to what you around this table

and those you know are trying to get them to do. How are you defining the intelligence that you want them to display? Don't misunderstand me. I'm not knocking machines. I think that there's an enormous amount of science about mind, brain, language and everything that couldn't be done without modern computers. From my point of view,

what I'd like to do is to get good simulations of the brain. I know we can make smart computers that play chess, I know we can make smart computers that reconfigure a lot of complicated situations in industry. But I think the big quest

is still to try and understand how the brain generates consciousness and mind. But then you have to move out of mathematics, don't you? You have to move into... Well, I suppose at the base of biology, you could say, with mathematics too. Well, you get a lot of data from neuroscience, which you then turn into things that are closer to mathematics

because you have to put them inside a computer. So there is this interesting interplay between computer scientists and neuroscientists that the two together have a combined intelligence, that slightly different from each. John, what sort of intelligence in your view are we looking for? I began the program with a quotation somewhere or other. Anyway, but you wouldn't believe a computer thing to create a son-over-con-share. I know it's done it.

Yeah, right. And, yes, I mean, what Jefferson was arguing there was that feeling, really knowing that you are feeling something, is crucial to be intelligent. and just creating it by the fall of symbols and artificially signaling that should not be counted as intelligent. And when Turing responded to that by saying,

well, that's a solipsistic argument. I mean, how do we know that anyone truly feels? I mean, you have to be that person to know that they feel, and therefore that argument's heading in that direction and is therefore unsatisfactory. And what I think is what I struggle to understand, it's the great historical question, is why did intelligence and become the hallmark of what it is to be human in the 20th century?

Why is artificial intelligence the troubling science? Why do you think it is? I think it's a lot to do with the fact that the 20th century was a century of conflict like no other and therefore the pressure to produce ever faster machines. was extremely important and the bottleneck in producing faster machines was automating

the control, the computation part of machines. But behind what you said, your implication is that in previous century intelligence wasn't the market. Well, yes. I mean, I was walking down a Ridley Road market near where I live. It's a typical

very busy market where they have all kinds of automata there. They have toys that move. They have a father of Christmas that bottles around and speaks. Now, I didn't find this philosophically challenging in any way. There are all kinds of things that in the 18th century people found to be disturbing because it seemed to be trespassing on what it is to be human,

that it seems to be doing things that should be human, yet it was a machine. They found that disturbing. We don't find it. Now we can see it in a street market and not be disturbed at all. But the idea of an intelligent machine, still disturbs people, which is why we're here. I think that's a very good point.

And I think some of it, I mean, as you say, we're not disturbed by some things now because we've accommodated all sorts of invasions into the body. If I can use the dreaded word cyborg. We're all cyborgs now. But I think the thing about... I think we've always been cyborists. But the thing about intelligence is it does seem to me

that we're still hung up on that. And I think... Well, why shouldn't we? Well, why shouldn't we? But why shouldn't we? But why should we be in that it's still the Cartesian split, isn't it, between mind and body or matter? We may have conquered quite a lot of the matter and body,

but we still don't know a lot about mind. And I think also, I mean, if I can return to one of the things you're hinting at about critiques before, I mean, I think one of the hints is, what you said right at the beginning when you talked about man and machine, and you didn't say man and woman in machine. It seems to me very much the... It's just a shorthand. I mean, no offences it turned.

I mean, if you have to say, man and woman he and she every time you construct a sentence, it's going to be quite long-winded. Just a human machine. That'll do. But you know, that it's very much the masculine reason and the female body earth matter. Do you think there's that distinction as glib as that, I don't? Well, I mean, I have to say it in two seconds, I suppose, but it's a much bigger story to it than that, of course. But I still think we're fascinated by that, because we see it, as I go saying, you know, mathematics being seen as

the epitome of what it is to be intelligent to reason. I think we are still hung up on that. Do you? Igor. Well, maybe we're not as hung up on it as we were in the early days of computers. I think we're beginning to recognize now that

the word intelligence is used in different ways. Computer scientists use it in a completely different way from philosophers or from psychologists. So, you know, I think just

separating these things out is important. We could call them smart computers rather than intelligent computers, and then we could look at other efforts which try and cope with understanding how the brain generates what it does. Alison, in the 1980s work began to focus on replication and trying to replicate life. Can you tell us about the innovation of A-life populations and what that means? Artificial life, E-life. Well, this is really a kind of computational artificial biology,

and there are two broadly two branches, the one that focuses on robots and the other that focuses on, I suppose, trying to mimic or to generate life in a computer. There are several important things here. It's not focusing on individual intelligence, but rather focusing on populations and growing them, and it's very much like evolutionary ideas, and indeed sociobiology, I think, comes in there as well.

And I think the key word there is the idea of emergent behavior. You know, much as when you have a flock of birds, flocking is an emergent behavior of the flock, but it's not a property of an individual bird. So it's trying to see intelligence more as, or intelligent behavior, as an emergent property. That's certainly what both in terms of the generations of populations in a machine, or indeed in terms of robotics.

And I think that's quite an interesting idea, because the thing that it, certainly if you're looking at the robots, the thing that the robots are doing is they are taking account of the idea of being in, well, some kind of body and being physically situated. Of course, many of the critiques of traditional artificial intelligence are that they are not, they're just trying to mimic minds. And of course, many people say that you can't really be intelligent without a body. You need the whole thing, not just the brain.

I mean, catching a cricket ball at high speed is an intelligent activity. Well, absolutely. But presumably that comes from the mine. I mean, you know that you've got to move at a high speed to your left and stick out your left arm. That's a mind thing. But I mean, a lot of these things, it's like, you know, the old story about riding a bicycle about the physicist

who could describe the mathematics riding a bicycle but couldn't actually ride a bike. And in fact, it's almost impossible to tell someone how to ride a bike. You've got to get on and do it. So a lot of these things are really, a lot of our intelligence seems to be to do with the body. It's not just the brain. And maybe we can't really separate them anyway.

Sorry, Ego I very much agree with Alison on this that recent efforts in robotics and in other forms of computational intelligence have recognised that these things have to be embodied they have to somehow rather be in touch with real environments in order to build up the experience of real environments that you need in order to develop intelligence.

John, can I come to you towards you? We started with you. You're a historian of science. Do you see the progress of from the famous duck right through the Babbage, through Turing and Shannon, and what Alison was just talking about, the A-Life population? Do you see there being a progress? Do you see there being a development? Do you see there being a getting towards a real description, a fuller description of intelligence? Well, I'm struck by what Igor said just at the end there,

that we seem to be moving towards a full of replication of a human mind in a human body. that to be intelligent, as we'd call a human intelligent, you have to have something that's like a human mind in something that's like a human body. So it seems like we are going full circle and we're coming back to actually producing the kinds of automata that Valcanson produced, that these were moving figurines

with human-like qualities. Well, yes, that's a good point. But I think the other thing we need, as well as the body, is we need to also be culturally situated as well, that you can't be intelligent unless you have culture. And largely the only way of getting culture is to be brought up in a culture. You can't just put it into the brain. And that's why some of the famous big AI projects have failed

because they thought you could program it in, but you have to grow it up. Igor. Yes, I wouldn't say we're trying to develop human minds in human bodies. I think we're trying to develop machine minds in machine bodies in order to understand what human minds in human bodies are. And how much, how would you describe, we can use that, the progress since the mid of the 20th century in terms of the development of these machines?

The picture is of time and time again predictions are overhyped. So Turing said we'd passed something like the Turing test in 50 years. That 50 years has gone. In 1956 they said they'd have natural language programs, there'd be computers producing aesthetically pleasing music, opposing it, all these predictions have failed. I think one sign of progress is that we treat machines with much more disdain than we did in the past, and we're very, very careful not to highlight the greatness of a machine.

Machines are just there in order for us to develop ideas, and I think that's important. Well, thank you all very much. Thanks, Igor Azanda, Alice and Adam, and John. again. Next week we'll be going back to 1819 to talk about the Peterloo Massacre in Manchester. Thank you for listening. We hope you've enjoyed this Radio 4 podcast. You can find hundreds of other programmes about history, science and philosophy at BBC.com.com. com.com.com.com.