 Right, so it's now my pleasure to introduce a friend and colleague of mine Gavin Brown from the School of Computer Science at the University of Manchester Who's going to give us a very interesting talk as I've already seen it once on how to make computers think Thank you very much. All right, so before we begin Do you remember right back at the beginning of the day? Professor Jim Miles was here. He was the head of school and you gave him a big welcome saying you're enjoying yourselves You remember? Yeah. Yeah, so I want to beat him. He's my boss. I want to beat him. So, oh, I've just noticed he's over there So even though there's only half of you here, are you all having a good time? Oh, yes I'm sorry. I did that in one there. All right Okay, so I'm going to talk to you today about something I find fascinating This subject making computers think this is what I do for a living every single day, and I think it's brilliant Now I don't actually connect people's brains up to computers like this So most of you will be leaving with your brains in your heads. Don't worry and I don't really build Wally We can't actually build Wally properly yet We don't know how to do all the amazing things he can do because he's actually just a cartoon But we can do lots of pretty impressive things This is what we're going to talk about today. But first of all, I Want to tell you all that you are amazing every single one of you is absolutely incredible Whoo, exactly give yourselves a whoo All right, I want to prove it to you though Even though you seem to believe it. I like everyone to stand up. I would like everyone to stand on one leg Finger on the nose and hop up and down All right, and now swap legs All right. Thank you very much. Give yourselves a round of applause and sit down right so Hopping hopping up and down how many of you how many of you Used your brain right then Did anyone use their brain? So it required thought to hop up and down Did it he's thought about it a little bit. Yeah, maybe you didn't want to bump into the gentleman next to you Maybe a little bit. So if you were thinking a little bit What about um this next guy? He's doing what you were doing He's not going finger on his nose He's not swapping legs, but he's popping So is he thinking when he's doing that? Who thinks he's thinking? Yes or no? Yes, yes Put your hands up. You think yes, he's thinking Hands over you think no, he's not thinking Okay, most people seem to think no. He's not thinking. That's very interesting This robot is pretty famous. He's called as a mo and he's been in development for a very long time 33 years this first robot back here at the left-hand side was built in 1980 33 years ago now that's almost as old as some of your teachers almost Sorry teachers So he's so amazing isn't he can hop up and down just like you were doing So impressive Maybe not so clever as we thought Has anyone else fallen upstairs or fallen downstairs when you you just take two steps and then fall over? Yeah, so sometimes you do as well But you probably didn't fall over in the same way that he did just kind of go like that You probably put your hands out to stop you Because you're smart you're clever. So what do computers find difficult? Clearly the computer inside the robot here was finding that task of walking upstairs pretty difficult and Hopping up and down took 33 years to make it do that Now there's a computer inside your phone if you have a phone there's a computer right here There's a computer here. There are computers everywhere in our lives now and some things they find difficult and some things They find not so difficult So which way round is it what's difficult and what's not so difficult for this? I'm gonna need a couple of volunteers So I'm gonna need a student and a teacher so if you're with your teacher point to your teacher and we will try to identify somebody Who have we got here? Oh, we got a lot of okay? All right, sir. I think you have to come on up and you have to bring who you're gonna bring with you Who would you like to embarrass? Oh? James Okay, give them a round of applause as they come up Mr. Hughes, okay, so we're gonna come over here so mr. Hughes and James so James I'm gonna give you this and If you just come over here so everyone can see you so on the screen can I ask what you teach mr. Hughes You teach I see to that's fantastic. Have you ever done any maths? Little bits of maths from school. Okay, so on the screen behind you. I'm gonna put up some sums And I'd like you to do them using this computer in your head and James gonna do it with this computer a different type of computer So are we ready and if anyone in the audience knows the answer feel free to shout out as well, okay? We ready Here's one. Oh, I heard 29. I heard 29 here as well So James 29 29, okay? Okay, so that seemed to work out quite well Let's try another one. Are we ready James? Are you ready? Yeah? Yeah, okay? Everyone remember you can shout out if you want to Mr. Hughes you're ready 169 did you cheat did you hear that? Yes, okay, so somebody here. Let's check James. What do you think? 169 okay, well done. Oh, we're doing okay. We're doing okay one more. We ready mr. Hughes 15 15 I think it's a bit bigger than 15. Yeah, are we ready? What do we think we get the microphone? It's orange the answer Uh It's a big number apparently. Yeah, it's um 400 so 400 four million three 347 thousand five hundred and eight. Yes. Well done. Okay. Thank you very much Give them a round of applause as they go back to their seats mr. Hughes Would you just catch this for me, please? That's very good. Well done. Would you throw it back for me, please? Was that difficult? Okay, well, thank you very much. Give me a round of applause you guys back to his chair so some things there Mr. Hughes seemed to find a bit difficult some things like the ball Catching the ball seemed to find easy. Was it easy to catch the ball and how about the sums? It got a bit more difficult how about some quadratic equations Would you like to do some quadratic equations in your head for us? Not today. Okay. Well Yeah, so there are some things that computers find hard and some things that computers find easy and some things that humans find hard and some things that humans find easy so it seemed That for mr. Hughes and most humans doing maths doing all these hard sums It's pretty hard. You know, it's difficult and Doing things like catching a ball That seems to be quite easy doesn't it? You know you catch ball catch a ball every day of your life probably But for computers for a robot to do Remember a robot is just a computer on legs it's the other way around and Virtually everything that Computers find hard we find easy and vice versa even this thing in the middle chess We find pretty tough, but computers Not so tough nowadays. There are some pretty good chess playing computers out there. So let's talk about that Chess who plays chess? Yeah, I get quite a few of you Do you think when you play chess? Yeah So it needs thoughts does it it needs thought to play chess who thinks yes Yes, thought play chess. Okay, great. So does that mean if it needs thought to play chess? That my phone is thinking When it plays chess with me it's following patterns. So does that mean it's not thinking then It's following detailed instructions. Yes, we have a genius in the front row here so Yes, it's following some kind of pattern. So maybe it doesn't require as much thought as we thought This is Gary Kasparov this picture was taken about 16 years ago and He played chess Against a computer the thing about Gary is he was one of the best chess players in the entire world at the time He's what's called a grand master at chess and he played chess against a computer called deep blue and this computer Searched through every possible part of its memory banks on how to play chess against Gary and it beat him And it was the first time in history that a computer had beaten a human a really proper really good human at playing chess And that was a bit of a revolution People suddenly started to think well maybe What we think of as thought what we consider thought isn't actually thought because computers seem to be able to do it and play chess So Maybe we need to reconsider The idea of thought how would we even know if a computer was thinking when it was playing chess? Does anyone know who this is? Alan Turing. Yeah, so this is Alan Turing He is was a very smart man He actually laid the foundations for everything we know as computer science nowadays And he worked in various places But one of the places he actually worked was here at the University of Manchester and in 1950 He asked a very particular question he thought could a computer ever think That you know in 1950 the whole idea of computers wasn't really very widely known They'd only just really been invented so for Alan to suddenly go. Well, I wonder if it could think That was a pretty big question to ask and to consider So he tried to figure out how he would know if the computers they were building at the time And he was working on the very first type of computer that could store programs in its memory as we think of today He was working on that here at the University of Manchester and he imagined a scenario where he would test whether a computer was actually thinking and Nowadays we know it as the Turing test and his test Had a room or two rooms In one of those rooms there was a person and In the other room there was a computer, but it was a clever computer in some way and the person in the people or the person in the room could type and The computer could send messages as well But they could only communicate with Alan via email or messages So he couldn't speak to them or see them. He would just send emails or text back and forward And Alan considered well, if the only way I could communicate was by seeing messages from inside these rooms How would I know which one's the computer? And you might think well, you know, of course, I'll be able to tell which one's the computer Because it will just do computer-y things. Well, let's try so this is a genuine transcript from what's called a clever bot And a clever bot is something that can communicate in this sort of way doing the Turing test and in one of the rooms Is a clever bot a computer talking and in one of the in the other room there is a human So I say first of all hello to the thing in room one And I am the thing in room one replies and says hey, how can I help you you a computer? So I just ask it straight away No, no, no, my name's Sarah, of course Well, maybe your computer wouldn't lie. So maybe it is of a person And I change the subject I say well my car's broken To see how the thing in room one responds And I hear well sorry to hear that so seems to be able to know that you know, that's a bad thing In room two I say hi. How you doing? How are you and the thing in room two replies? I'm all right. How are you? like that and Then I say your computer again, just ask Does that matter and then I reply well, yeah cuz cuz I'd win the game then and I could win the Turing test Well, my name's Tom and I'm a person so I'm a person on this side and Over here that the person that the thing has said no my name's Sarah. So both rooms have claimed that they're people So which one do we think is? Actually a person which one's actually a computer shout out room one if you think it's if you think the computer is room one And if you think it's room two computer Okay, it's about balance there maybe yeah, so what's the answer? ah So it was about balanced, but it turned out that the thing on the right room two was actually the computer and This little trick that might have fooled you the how are you up there? Well, all we've done there is replace the word a are e are With the letter R and that's a pretty simple thing for a computer to do just search and replace in text You can do that with Microsoft Word every day So all it does is have these little rules which make it look like it's a human look like it's a human using text Speak or missing out punctuation in the I'm part there, but it's pretty simple to get a computer to do that nowadays The thing is here. You just saw six lines of communication with these rooms To pass the Turing test a computer has got to convince four judges Over 20 minutes having conversations about any subject at all and no computer has ever passed this test very very difficult There's actually a million dollar prize if you can make a computer pass this test Put up by a man a millionaire called Hugh Loebner in America, but no computers ever done it yet Now the Turing test Alan thought well If I ever would meet a computer that would pass my test and I would see it communicating It looks like it's thinking so maybe it is Maybe because I think that it's thinking maybe it is actually thinking that's a bit of a weird thing to say But how do I know that you're actually thinking? How do I know that you're thinking? How do I know that you're not robots? How do you know you're not just a computer inside a body? And Alan thought How do I really know that the people around me are thinking We don't really know and This is the principle of the Turing test that if a computer could behave Like it's clever like it's intelligent like it's thinking maybe that's enough. Maybe it is doing everything we would want from it So But what about all the different things you can do you can do a lot of stuff So You can see things you're amazing. I told you earlier on you can see things you can recognize things you can move around You can learn and you can listen and talk and communicate you can do some pretty impressive things So which one of these needs the most thought who thinks seeing and recognizing things needs the most thought One two three four a few people who thinks moving around and jumping and moving around needs the most thought No, okay, who thinks listening and talking requires the most thought Okay a few and who thinks learning requires the most thought ah So most people think learning requires a lot of thought so does that mean you're all thinking in class when you're learning Yeah, if you if you say that here that means you're promising that you're going to be thinking next time You go into class and your teachers are listening so Learning maybe well, let's take that one later, but let's just talk about seeing things and recognizing things How would a computer see you Or recognize you Anyone know what that is? What's that? It's an Xbox Kinect. Yeah So an Xbox Kinect is a pretty impressive device You can play computer games in your living room without a little controller And you can just dance around Move your limbs around and it detects where you are in the room and allows you to play games. That's a pretty cool thing How it must be able to see you if it's doing this so This is what it actually sees When you're moving around Dancing around like this lady here what it sees is these blobs and lines connecting them So it sees joints like this and it sometimes it sees blocks for your body And it can see you in several different ways like this Sometimes it can detect where you're stepping you can see the little pulses on the floor where it's heavier Or sometimes it will see you like this when it can figure out where the different parts of your body are and how far They are away from you It's a pretty cool thing, but how is it seeing you so if you're in your room in your living room dancing around What the Kinect device does is shoot out a little beam of infrared light So on my hand here from my little pointer. I've got red light, but infrared light is Even further down the spectrum of light and it's actually invisible light It's a much longer wavelength of light as it's called and it's invisible to the naked eye but We can actually see it if we use a computer. So down here. I have a little Kinect device Could I ask you to stand up and kneel down here? You can Can we see all the little dots on the gentleman's face? So if you just move your shoulders like this Move you saw it. Can you see all the little dots on his shirt? So this is what the Kinect sees when it sees you it can see all the little dots and when the then dots move in a particular pattern Moving in this way or that way it means that your shoulders pointing this way or that way and it can figure out From all those dots moving all the time because it's a very fast computer it can figure out where you're standing Okay, thank you very much. Okay, give them a round of applause Okay, so The infrared light comes out and it it projects these dots onto you This is in your living room genuinely when you're playing like that The light reflects From you just like normal light reflects off a mirror But because the dots are in a particular configuration when you're doing this or that The Kinect can go aha right the dots are in that particular pattern That means the hands pointing upwards or the shoulder or the knee or something else is happening And it's because it's got a big memory it can figure out given that pattern that means they're doing this And it does this the computer just asked the Kinect What's out there in the world the Kinect uses that dot pattern to figure out which bits you because you You are moving your dots move whereas the dots in the background of the furniture don't move So you can figure out which bits you and which bits the background From that it can turn you into a little stick man, and it can figure out where your joints are and From that it can figure out which bit of the joint is you know your wrist or your elbow or your head So it has these little rules that say well, you know the knee bones connected to the thigh bones So if I detect the knee bone, I know the thigh bones just above it So it has these little rules and it goes round and round and round just doing this consist Continually detecting where you are and because of these patterns that detects you can play the games So it's pretty impressive it can see you It can also See several people at once and it can track where they are So here it is tracking three different people giving them different colors yellow blue and green and it can figure out which person is which That's pretty impressive. I can look around and I can see which person is which in this room and So can the Kinect so does that mean it's thinking? well, we talked about all these things earlier on and thinking Seems to be not such it's so easily definable now. What about this next one listening? We talked about moving and jumping around and we saw that the robot earlier on could move around We talked about seeing and it seems that the Kinect can see and so on so maybe neither of those actually needs thought Listening and talking well, what does a computer perceive when you talk to it? Computers as you know, they just work on numbers So when you speak like into this microphone the signal from the microphone gets converted into electricity Goes into the computer and what the computer perceives is little pulses of electricity going up and down up and down when I speak louder Pulses go higher when I speak softer or in a different tone of voice Maybe loud deep or for high-pitched the pulses change configuration and again because of this pattern finding ability The computer can figure out that that particular pattern or things close to that mean how are you? And we can do this Even in a phone nowadays So you can talk to computers and they can respond to you Let's see how this one does So this is Google. I don't know if any of you have tried this yet, but If you see if you go on to Google you can see a little button over on the right-hand side and that says search by voice So let's try it. Tell me about Manchester United Manchester United is playing the game athletic on the 11th of August at 1500 hours So it seems pretty clever. It knows the man United's schedule and it could figure out that I meant the football team from that That's pretty clever. Let's try on something else like a sum we had from earlier on What's the square root of 17? Square root 17 is four point one two So again, it seems pretty clever. It's figuring out what I mean from these patterns in the the signals going up and down It's converting that into figuring out that I want a particular thing from it once it wants It knows that I want it to do a calculation Then it's giving me the answer. That's pretty clever. So is it thinking who thinks Google is thinking hands up a little bit, okay, okay, so It seems to be pretty clever So I'm I'm logged in there as you can see you can see my name at the top left-hand corner That's inside Google and you can see a little picture of me up there. So, you know, you know, so I am What's my name? Oh No, did you forget your name? So Maybe it's not as clever as we thought it's got my name up there on the screen It's got a photo of me. I've been using it for years and it doesn't know what my name is Maybe it's not as clever as we thought So it's not so clever, but sometimes it seems to be clever. Oh Well Right well, let's think about the next one on the list learning Now everyone at the beginning said that learning requires the most thought and the others. Well, no, maybe they don't need thought So let's consider this I'm going to need a Volunteer again this time. It's going to be a teacher and I'm going to embarrass them So if you'd like to volunteer your teacher shout out the name of your teacher What can I hear? Mr Okay, give him a round of applause as he comes up Thank you very much. So what's your name? Mr. James. Mr. Jones. Mr. James. Okay. Okay. All right Would you like to come up on stage? Okay, so up on the screen just behind you Well, you can read them from here if you want up on the screen on the screen They're going to be some instructions. So here's I've got my teacher So they're going to be some instructions and I'll let you to follow the instructions. Okay. You ready? ready one Position the right shoulder joint at 270 degrees public ordinance with a protrusion of five degrees beyond the body plane Can we not do that? Oh, okay. All right 270 degree. Okay, right. Yes. Are we ready? Okay? That took a little bit. Okay next one. Oh Come on. Mr. James. Oh, okay oscillate. Okay. All right next one next one next one. Okay. Come on Come on. Oh, okay. Oh, that's really close. Okay. Thank you very much. Give me a round of applause as he goes back to his desk so Was this hard to do sir? Mr. James was this was this hard to do? It was quite hard to do. So I gave you a very precise list of instructions So why was it so hard to do? Well, let's try something else I'm going to get this robot to follow those instructions and see if it can do it better than you did but I'd like everybody To just copy what the robot does Are we ready? Yeah, everyone ready. I want full audience participation here. Okay So just copy what the robot does whatever it does Okay, so it's like it's standing up Everybody stand up Yes Okay, what's that? It's dang very well. Gangnam. Are you ready? One, two, three. Okay, thank you very much. Give yourselves a round of applause. That was very impressive so That was really impressive How but how did the robot do it? I gave the instructions To the gentleman, but it couldn't quite do it But robots can follow instructions computers can follow precise instructions very very well So I just gave the equations Describing where he should move his arms and he could do it just by articulating his joints And when it comes to dancing well, we just take Psy We write computer code equations and instructions that describe How we want the robot to dance and it can just do it because we tell it how to do it but you You just saw how to do it You saw a maybe youtube video or something like that and you just copied you learned from examples of how To dance like that that was pretty impressive And you are pretty impressive. You're amazing as I said, you can just learn How do you do these things? You've learned how to stack blocks when you were younger You're hopefully learning something today. You're learning lots of things all the time But could a computer ever learn who thinks a computer could learn? Hands up. Okay a few. Okay, so as it turns out Yes, anyone with the hands up was right and Though we can't quite do as impressive things as walley yet. We can do something called machine learning And this is the subject that I work in So this subject called machine learning I'm now going to describe to you roughly how it works So we can take this a computer and We show the computer Examples of what we want it to do just like I showed you I showed you the robot doing this and you could just copy You knew you saw what we wanted and you could do it Inside the computer is what's called a mathematical model But don't worry about the word mathematical there. It's just sums and equations things that you are either learning now Or you will learn in a couple of years time. It's just things like algebra and probabilities They're all what make up a mathematical model. Maybe slightly more advanced things like calculus That forms a big part of mathematical models But those are things that you'll learn when you do gcse's or a levels and even more things if you come to university So we show these examples to the computer And then a human supervisor kind of gives it the thumbs up or the thumbs down says yeah, well done, you know, you're doing very well And inside the computer's head inside the computer's memory It's reconfiguring itself. It's trying lots of different configurations of the equations and the coefficients in there Changing them around doing sums trying to find a situation that the that the human will be happy with But it's doing all this automatically Eventually When the human is happy gives a smile and says yes, okay, that's it You're done the computer can stop in a particular configuration And spit out this finished model And this model will be able to be used for for example a dancing robot But it was never programmed explicitly saying This is where you should move your arm Or you should do this or you should do that It was just given examples saying this is what I mean by dancing And it tried to do it and then the human said Good bad, okay, maybe in the middle and it adjusted itself like you would now you might think that Um dancing robots are not so useful in society But things like this your camera phone or camera This is actually designed by computers learning as well so All of these children in these photographs They've not been brought into a big laboratory somewhere and pictures taken of their faces So we know what they look like so we can draw little boxes around their faces when they look into a camera next time And you know if a baby is born and you point a camera at it You can still the camera will still pick out the face So it hasn't been explicitly programmed with This is the face you should pick out Instead what happened is that the computer program in the laboratory somewhere was shown lots of examples of faces Saying this is a face. This is a face. This is a face. This is not a face Maybe showing a picture of a wall or a football and Just from those examples, it was able to pick out patterns general patterns that meant This is a face. Maybe with eyes and ears and thing, you know eyes a particular kind of Angle maybe ears a particular position on the head But it could generalize as well and kind of figure out the general structure of a face is this and that's what is in your phone nowadays It's not just phones and drawing little boxes around faces If you ever shopped online if you ever used amazon you Have been the human supervisor giving the thumbs up or the thumbs down If you've ever clicked fix this recommendation on amazon or given a four out of five stars or five out of five stars Saying yes, I really like that book algorithms in a nutshell If you click that and say yes, I really like that Behind the scenes inside the big computers amazon The computer program is adjusting itself changing all of its equations around and figuring out the general patterns of stuff You like Such that next time it can recommend a good book like maybe simply javascript something else that you might like But it does all this automatically just from the feedback that you give it and yes, you might think well, this is just shopping but There's another fib I told you earlier on I said that the connect Had little rules inside it to say this is an elbow and so on in fact The connect learned how to see you So they didn't just in the same way that we don't have people walking into labs and taking you know checking what their faces look like What they did to build the connect was they got a few people. I think it was a couple of thousand people And they looked At them and they sort of labeled their knees and said right to the computer This is a knee. This is an elbow. This is a shoulder and so on and then the connect figured out the general patterns of which bit is a knee in Anybody so all these people have different body shapes He couldn't possibly see every possible body shape or every possible angle at which I could move my arms and legs But it figured out general patterns, but it did that by learning There are other things that are not even on the market yet, but are in research laboratories This is the google self-driving car And you might think wow, you know that's science fiction. Does anyone know who this up here is at the top right? Who's that? It's kit. Yeah, that's david hasselhoff more more well known nowadays for being on there. Is it pop idol or Something like that. I think so But a long time ago that was science fiction That was like, you know You couldn't imagine a car that could speak to you or recognize your voice and drive around by itself But it's happening now. This is what the google self-driving car sees In the same way that the connect beams out the little patterns of light The google car does it in a similar sort of way using a different using a laser instead of the infrared And the green bits are things that it's been taught Not to bump into those are people Or you know walls and things the blue bits are things that it's been taught Saying right you can drive on that that's called a road, but all the while it was taught in the research laboratory You can drive on that you can't drive on that figure out the general pattern and it learned the rest of it It didn't have to be shown every possible road in the world It was taught general patterns and it figured out the rest itself You can see the person at the top right here Just driving along you can see how fast they're moving because there's a Motorway and there's a fast car up here And they just hands off the wheel and it can steer it can change lanes It can do anything that mostly anything that human can do So this is pretty impressive But the thing is was that car thinking Was google thinking Was the robot thinking Was the connect thinking Oh, maybe we have to reconsider what we consider thought now The thing is all of these things You can all do and remember what I said at the very beginning You are amazing You can do all these things from when you're this age from when you're very young You can recognize faces a baby can look up and see where its mother is straight away And he can learn things and it can see and hear and all these things you can do all these things in one package So how do you do it? Well, you Have one of these a brain And inside your brain. I've got a little model here Inside your brain are lots of little neurons as they're called cells And these little neuron cells in your brain if I break open the brain here Right deep within Your brain and all over it these little neurons are firing tiny little pulses of electricity around And when you were very young and your mum said yes, well done or gave you a big smile When you said your first word like mama That kind of changed the connections inside your brain connections between particular neurons Started to get stronger and stronger and stronger And that's the process of learning You were able to learn by all these neurons Flushing electricity and getting stronger connections between them and weaker connections the ones that weren't necessary were Killed off. They were just rubbed away And they weren't needed. So you figured out Inside here what neural connections you needed and these are just tiny little cells inside They're just smaller than you can imagine and you have a hundred billion of these cells inside your brain That's a lot But the thing is how could we ever get computers to do this? How could we get computers to learn in the same way that you do by having neurons? Well Maybe we don't need to What we can in fact do is make Simulations of neurons. So don't be worried by the equations here. This is just this is kind of the language Which computer scientists use sometimes and it's just a language of Communicating what we mean by these mathematical models But again Your teachers will tell you these things are all things that you would learn in Maybe late gcse's or early a level maths. These are called derivatives And it's just a way of writing stuff down And you'll learn all this sort of stuff about how These equations mean this later on in your school life and then maybe a bit more when you come to university But what we were doing here is we can actually have these kind of models of neurons Kind of simulations that are not exactly like the neurons or as clever as the neurons in your head But a bit like them So if the neurons in your head are kind of like a formula one car doing all these incredible things The neurons that we can simulate that we can make and have pretend neurons inside the computer here They're kind of like this Kind of a shopping trolley Maybe rolling down a hill not so good or maybe a model car at best So we can't we can't like you know invent terminators and robots that can do everything that you can do Yeah, so I want to tell you a true story Genuinely true story So It's 1986 Cast your mind back teachers. You might remember that time In a university laboratory somewhere in America There was a man His name was terry I won't try to pronounce his second name Sinovsky But his he was working in the lab working on a very old computer The computer there is Hundreds of times less powerful than the computer in your phone Nowadays, so a very slow computer compared to what you've got And what terry was doing was using these models of neurons using these kind of mathematical things Which simulated the way the tiny little neurons in your head behaved and there was Kind of artificial electricity flowing around the connections inside his neuron this network of neurons and what terry wanted to do Was to show Examples of syllables pronouncing words to this simulation So when you see the letters c and h next to each other, you know that means ch And when you see the letters a and h next to each other, you know that means ah And from different patterns of letters, you know what sounds Come from each of those because you know how to read What terry wanted to do Was try to get a computer to do the same thing to learn how to read So he wanted from showing it to syllables or showing it words and showing it some examples of well when you see c and h together That means ch just like teaching a baby to talk So he wanted it to come out and be able to say words so It was late at night in this laboratory And terry was getting tired and he thought it's time to go home. I've been working hard enough all day So as he was leaving the office, he pressed the keyboard and pressed enter and he heard He heard that So that didn't sound much like talking It's just Not really even human sounding sounds But a few hours later after these Artificial neurons had a chance to kind of connect to each other in different ways and see the examples that terry had provided Where we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're we're And the next morning when terry came in The weather today looks good outside. It is sunny The computer had learned how to read And he had just shown it newspapers And given it the weather report and it had been able to read it to him in the morning He didn't ever program into it the rules. This is how to read and so on He'd shown it examples and it had learned how to do it by getting rid of the connections That it doesn't need if you remember this network was much bigger when I first started and it's pruned away the bits that it doesn't need so terry's Simulation This kind of model of the way the little cells in the brain behaves had 300 neurons inside it Now that's about enough for a tiny little worm Tiny little worm just floating around in the water So it doesn't mean that worms are going to suddenly start speaking to us But if these neurons are programmed in a particular way, it means We can actually get that many neurons to learn things But the thing is this was 1986 So many years later This is what we're doing So this is professor steve ferber He works here at the university of manchester and he is working Not just on these mathematical models of how neurons Communicate and so on but on making microchips Computers that can behave like that and he's not just able to simulate 300 of these little cells inside your brain But a billion of them I'll take some questions at the end So 2014 That's when the computer is going to be finished the one that he's building And that's when we think it will have these simulations of all these neurons and It's a pretty impressive thing. It's not suddenly going to start talking or living or communicating with us This is about enough for a worm. This actually turns out to be the same number of neurons that are in a cat Or 10 mice That doesn't mean that we're suddenly going to have a robot cat running around everywhere because Even programming something as simple as a cat is Incredibly difficult Cats are pretty clever things. They can do lots of stuff and though we know how to make simulations of all these neurons That are you know, really really big. We can't make something as clever as a cat yet I want to tell you one more true story now Does anyone recognize who this is? Anyone recognize who's the geeky kid? Who is it? It's me. It is. Yes. So this is me. This was me age 13 long time ago When I was 13 I read a magazine article In a pc magazine And inside the article was the story that I've just told you about this Network of things that could learn how to speak and I loved it I love the whole idea of this idea that computers could learn So I became fascinated by this And I went to university to study computer science Which was the thing the area that I needed to know in order to know how to do this myself And as the years went by I studied harder and harder And now I work here and I'm a researcher in what's called artificial intelligence Which is the name for all the areas that I've been talking to you about today I thought it was amazing and I get to do this cool stuff every single day Now maybe In a few years You'll be doing stuff that I hope you'll be doing stuff that you think is incredibly cool Maybe you'll be building the new self-driving cars of tomorrow Maybe you'll be building hover cars that can fly to where they want to go to and have to navigate by themselves Maybe you'll be inspired by something else Something in the movies that you really think is cool and you want to make a computer do those things Maybe something like iron man here the clever computer there Or maybe you'll think of something that isn't even in the public eye yet So this is a kind of artists mock-up This is just a picture of a robot astronaut That nasa The space agency hopes that they might one day be able to build completely and send out to explore the solar system But in order to do that it would have to be a pretty clever robot It would have to do all the things that we've been talking about talking about today in one package and more So it's a long long way off Maybe you'll be doing stuff like that But whatever you choose your future to be I hope you're inspired by today and by lots of other things You've seen in your studies and here today at Manchester And have fun. Enjoy it So thank you very much I'll finish there