Latest / Strachey Lectures / Strachey Lecture: Steps Towards Super Intelligence

Transcript

Strachey Lecture: Steps Towards Super Intelligence · Strachey Lectures

• 0:02
  20. Yeah, I think everybody.
• 0:20
  And welcome to the 2018 Michaelmas term strategy lecture organised by the Computer Science Department in Oxford.
• 0:28
  This series of distinguished lectures is named after Christopher Street XI, the first professor of computation at Oxford University.
• 0:36
  The straight she lectures are generously supported by Oxford Asset Management.
• 0:42
  I'm Peter Jevons.
• 0:44
  I'm acting head of the Department of Computer Science and it's my job to I pleasure to welcome Speaker today Professor Rodney Brooks.
• 0:54
  Mr. Brooks is the Panasonic professor of robotics emeritus at M.I.T., where he was for ten years,
• 1:02
  the director of the MIT Artificial Intelligence Lab, and then the Computer Science and Artificial Intelligence Love.
• 1:08
  After I merged with C.S.
• 1:13
  He was born in Adelaide, Australia and received his bachelor's and master's degree in mathematics before joining Stanford University for his Ph.D.
• 1:22
  He was a research scientist at Carnegie Mellon University and at MIT on the faculty at Stanford before joining the MIT faculty in 1984.
• 1:34
  Facebook's research has been in computer vision, robotics, artificial life, artificial intelligence.
• 1:41
  He develops behaviour based robotics, which has been deployed in robots on the surface of Mars.
• 1:48
  In thousands of bomb disposal robots in Afghanistan and Iraq inside Fukushima Daiichi after
• 1:54
  in the intimate aftermath of the 2011 tsunami in thousands of factories around the world.
• 2:02
  And in tens of millions of homes in the form of robot vacuum cleaners.
• 2:09
  He was co-founder, chairman and CTO of iRobot and of Rethink Robotics.
• 2:17
  He's won numerous scientific awards.
• 2:19
  Too many to mention, and also starred as himself in the 1997 movie Fast, Cheap and Out of Control that was named after one of his scientific papers.
• 2:35
  But these days, he describes himself as trying to inject realism into public discourse about the AI boom.
• 2:44
  We may be getting some of that this afternoon and also working on the question of how order can arise from disorder.
• 2:52
  So there's much more I could say, but I think it's better to let Ronnie speak for himself.
• 2:56
  So I ask you to welcome my speaker, Professor Rodney Brooks. We see people talk about artificial intelligence that's been around for a long time.
• 3:10
  Recently, there's been a movement that says it's called artificial general intelligence,
• 3:16
  which tries to distinguish itself from artificial intelligence to say they want to build a complete, intelligent entity.
• 3:22
  And I'm going to show you that I think that's a little bit of over marketing,
• 3:26
  because that's been the goal of the original AI people from from day one.
• 3:32
  And then there's been the idea of artificial superintelligence, which is rather poorly defined.
• 3:41
  But is the idea that with soon going to have artificial intelligence,
• 3:46
  which is much better and smarter than people, and so somehow it's going to kill us all.
• 3:55
  Nick Bostrom, is Nick here, by the way? Nick's not here.
• 3:59
  Okay. Well. Nick Bostrom, you know, did a survey a little while ago when we're going to get artificial general intelligence.
• 4:07
  And then there was another sort of extra survey around that.
• 4:12
  And the claim is that the median estimates from people in the field is that we'll get
• 4:18
  artificial general intelligence in 2040 and an artificial superintelligence in 2060.
• 4:23
  And once we get artificial superintelligence, everything's going to go crazy.
• 4:28
  I think both those numbers are very optimistic.
• 4:32
  And Nick has had his book on superintelligence, which the press has picked up.
• 4:38
  And I think, you know, because he's associated with Oxford and it says,
• 4:43
  you know, there's a real chance that superintelligence is going to kill us all.
• 4:47
  The press really picked up on that. But if you go to Nick's Web page, these are some of his featured papers.
• 4:54
  Where are they? Why? I hope the search for extraterrestrial terrestrial life finds nothing.
• 4:58
  Particularly he even mentions Mars. I hope there's no artificial, you know, extraterrestrial life there because it might come and kill us all.
• 5:07
  And the vulnerable world hypothesis is about all the ways that things could kill us all.
• 5:11
  And the how unlikely the doomsday. Doomsday catastrophe is about how experiments with the physics super colliders could kill us all.
• 5:22
  It's really and so on. Even the a typology of potential harms is about knowledge.
• 5:28
  That is true, but that probably we shouldn't know because it might kill us all.
• 5:33
  But the only thing missing is, is talking about AI that, you know,
• 5:40
  in a way that it might kill us all and then damping down AI and that will kill us all.
• 5:43
  But that's maybe a bit self-referential. So I feel sorry for Nick actually, about what it must be like to be him,
• 5:51
  because he's really afraid of a lot of stuff, but it hasn't stopped other people.
• 5:56
  Martin Rees is a dear friend of mine, the former astronomer,
• 6:00
  or maybe still the Astronomer Royal has talked about how I may, you know, end up killing us all.
• 6:10
  Stephen Hawking is worried about it. Oh, now finally we've got an MIT professor, but he must know about AI right now.
• 6:18
  He's a professor of physics, and he's never been he's never given a talk in AI or computer science at MIT.
• 6:24
  Well, then there's Elon Musk. He must know about AI, and he thinks it is going to kill us all.
• 6:31
  It's going to be an apocalypse now. He thought that I and robotics was going to help build cars, but it turned out that was a little beyond that.
• 6:41
  And Stuart, I haven't mentioned you and I'm sorry I left you out.
• 6:48
  So Alan Turing either. I will.
• 6:51
  I will. Yeah. But here's the here's the point.
• 6:54
  Maybe maybe I'm just a grumpy old guy. And, you know, maybe they were right and I'm, you know, get off my lawn.
• 7:01
  So I decided, well, okay, let's let's rethink this, you know, am I being unfair?
• 7:08
  And I go back to Marvin Minsky's, 1961 paper, which, for those who haven't seen it, is an amazingly good paper,
• 7:16
  well over 100 references where he talks about how to build artificial intelligence and he breaks it down in five sections.
• 7:26
  The search pattern recognition, which most closely corresponds, I think today to two deep learning sort of things.
• 7:33
  And then there's learning, planning, induction, and he he frames each of those as methods of controlling search and that's his view.
• 7:41
  So I thought, well, you know, Marvin did that for for artificial intelligence.
• 7:45
  If I want to, you know, really not be just a grumpy old guy, maybe I should rethink things and think about steps towards superintelligence.
• 7:54
  What would it take to get there? And so that's what this talk is about.
• 7:57
  What would it take to get to superintelligence? And I'm going to go through these sections, starting with a brief history of AI.
• 8:06
  I'm sure Alan Turing, too. Well, very well known papers, uncountable numbers in 1936 and his 1950 computing machine intelligence,
• 8:16
  which is where he would get the Imitation Game and the Turing Test from.
• 8:21
  But he had another paper in 1948, which was not published until 1970.
• 8:26
  His boss, whose name was Sir Charles Darwin, was grandson of another Charles Darwin wouldn't let him publish it,
• 8:36
  and so that was only published posthumously. And in that he starts off and you can see echoes of this paper in his 1950 paper.
• 8:46
  The possible ways in which machinery might be made to show intelligent behaviour are discussed.
• 8:51
  The analogy with the human brain is used as a guiding principle, so I think he was perhaps the first person to talk about.
• 8:58
  Computing machinery, and he was talking about computing machinery, being able to emulate human intelligence.
• 9:07
  He specifically talks about discrete controlling machinery by which he means essentially a digital computer.
• 9:14
  If you read the you know, this was 1948, the words went around and he says brains very nearly fall into this class.
• 9:22
  And so he talks about in this paper ways that machines could learn.
• 9:27
  And then in 1950, he published he did publish this paper computing machinery and intelligence.
• 9:34
  I proposed to consider the question, can machines think?
• 9:37
  And this is where he started talking about what he called The Imitation Game, which started out as Could I, man.
• 9:46
  And could you tell whether it was a man or a woman? If you were just getting questions back and forth to them, if they were trying to fool you.
• 9:55
  And then he goes on to use that to say, Well, what if it was a machine that could fool you about whether it was a person or not?
• 10:03
  Then surely it could be said to be thinking is essentially his argument.
• 10:07
  I think I think, Aaron, you may disagree with me a little bit on this, but I come back to this a little later.
• 10:13
  I go back and forth on what he meant.
• 10:17
  He does say there's no convincing arguments of a positive nature to support his views, which is an honest statement from him.
• 10:23
  And he also talks about ESP as being being a proven thing and that ESP should be perhaps considered.
• 10:29
  So telepathy, etc. So it's a little strange by today's standards.
• 10:35
  But he says that, you know, if we look at The Imitation Game and a computer can fool a human observer 70% of the time.
• 10:47
  That's a good substitute for can a machine think? And he says, I believe that in about 50 years time,
• 10:53
  we be possible to program computers with a storage capacity of about ten to the ninth to make them play The Imitation Game and win 70% of the time.
• 11:03
  Where does he get that ten to the ninth from? Well, he he says that the Encyclopaedia Britannica 11th edition has two by ten to the ninth bits in it.
• 11:13
  So somehow that turns into ten to the ninth probably being enough.
• 11:17
  And but 10th of life is the program space because in the computers fast enough,
• 11:22
  he thinks because then he says he can produce about a thousand digits or bits.
• 11:27
  He's referring to bits there, a program a day and that's before assemblers, remember.
• 11:31
  So it was really arduous to write code. And so he says 60 workers working steadily through 50 years might accomplish this job.
• 11:38
  And if you multiply it out and make them work 333 days a year each, you get exactly ten to the ninth bits.
• 11:45
  So somehow the two by ten, the ninth bits in the 11th edition of Encyclopaedia Britannica becomes a program of length ten to the ninth bits.
• 11:54
  By the way, I'm not sure anyone's ever checked this, but I looked around my my living room and that was, you know, 30 year old robot.
• 12:01
  But over in that bottom right hand corner. What's that? Oh, it's Encyclopaedia Britannica.
• 12:05
  Britannica, the 11th edition. I happened to have it, so I open to a random page counted, and it comes out to about two by the end of the night.
• 12:14
  So he was accurate. Now, people remember the term artificial intelligence coming from the 1956 A.I. Workshop at Dartmouth.
• 12:27
  And this is the proposal was only about 12 pages long and the first few pages were written by John McCarthy.
• 12:33
  And he just goes right in and starts using the words artificial intelligence.
• 12:37
  And this is, as far as we can tell, the first use of that term.
• 12:41
  But he doesn't explain it. We propose a two month ten man study of artificial intelligence and the conjecture that every aspect of learning or
• 12:50
  any other feature of intelligence can in principle be so precisely described that a machine could be made simulated.
• 12:57
  So I think despite what the artificial general intelligence people say that, oh, it's a new thing to go after all aspects of intelligence.
• 13:04
  Certainly John McCarthy, Marvin Minsky, etc. thought they were going after all aspects of intelligence back in the fifties.
• 13:12
  John does say, well, maybe the speeds as as the Turing.
• 13:16
  Turing said we may need more memory, but not more speed.
• 13:19
  John says more speed is probably necessary, but the major obstacle was just knowing how to do the programming.
• 13:26
  And later, you know, into the sixties, seventies,
• 13:29
  you saw John and Marvin and other people thinking that we were going to have human level intelligence within, you know, a decade sort of thing.
• 13:37
  So that idea has been around for a long time. Now I'm going to give four approaches to artificial intelligence that have come about.
• 13:44
  And these are cartoon total cartoons.
• 13:48
  They're very cartoonish. What I'm going to describe here there just to give a general flavour to it, please hold back.
• 13:55
  You know, it's a cartoon. It's not a. It's not your book.
• 13:59
  You know, it's it's just a cartoon.
• 14:02
  So the first one is the symbolic approach, which definitely John McCarthy was, was was it was a leader.
• 14:10
  And here basically you have symbols and these are these what they call bolded words.
• 14:20
  And those symbols are sort of indivisible things.
• 14:23
  But you can talk about relationships between them and make and you can talk about rules of inference and how the rules of inference work.
• 14:30
  And. So you can have some thing resembling intelligence discussing objects in the world as symbols.
• 14:41
  The problem has been how you ground those symbols in actuality, in real perception.
• 14:46
  And that will come in a minute when we get to deep learning and other things.
• 14:50
  But these symbols. To the machine that's looking at them don't have the meaning that we associate with and read.
• 14:58
  You know, we read a lot in the cap. But to these symbolic systems, you know, it could as well being zero five, three seven.
• 15:07
  They're all instances of g0083. So I just did substitutions for symbols there.
• 15:13
  Oh. In those relationships, instance of that has a lot of meaning to us,
• 15:18
  but they are symbolic sorts of things and there are strict rules of inference, etc.
• 15:25
  And so, you know, really it's something more like this is how things are speaking.
• 15:32
  There's a great advantage to these symbols though.
• 15:34
  You can compose symbols and compose things from different subsystems by using the symbols, and that isn't true for some of the other approaches.
• 15:43
  Second approach I want to talk about is neural networks cartoon again.
• 15:47
  And it's, you know, 2.0, 2.1. It's been rediscovered and rebuilt many times.
• 15:52
  Marvin Minsky, in his 1954 thesis, was doing neural networks at Princeton.
• 15:59
  But then, you know, it came again, late fifties, came again, sixties, etc., etc.
• 16:03
  And just for those who who don't know, again, a cartoon neural networks,
• 16:09
  roughly the idea is you have these things which are in some really abstracted way,
• 16:16
  modelled after neurones, starting with a 1943 paper by McCulloch and Pitts.
• 16:23
  They're not really like real neurones in any way, but the press likes to pick up on, Oh, they're modelled after the brain.
• 16:30
  It's a very far thing from the brain.
• 16:32
  The idea is you have inputs which come into the neurones on the left and a feedforward network where the neurones do something.
• 16:39
  I'll tell you what that is in a minute and push it onto the ones further to the right.
• 16:44
  And then and the output your these the ones on the right are some sort of classes.
• 16:50
  And and the one that is most activated and we'll talk about that in a second is the the winning class of what the inputs are looking at.
• 16:58
  And each of those neurones is typically a linear weighted sum of the inputs.
• 17:04
  The inputs are between zero and one X, one through X, and the weight them and the weights are what I learned the numbers, just a bunch of numbers.
• 17:13
  This multiplied in sum to get a number that could be quite big, negative or positive.
• 17:19
  So you put it through a function which squashes it down to zero one again, and that's the individual neurone.
• 17:25
  So you have inputs from some sort of sensory apparatus and it goes through these hidden layers, feeds forward.
• 17:33
  And then on the right, maybe, you know, one of the if the if the one labelled cat has the highest output score,
• 17:39
  it says this cat in the image, the dog in the image is a car in the image, but it's feedforward.
• 17:45
  And we know that the brain doesn't work like that at all in in the human visual system.
• 17:49
  And we'll come back to this in a bit.
• 17:51
  There's always things going the other direction from the output towards the input in I think it's V v2 one part of the visual cortex,
• 17:58
  there's ten times as many connections going backward as there are forward.
• 18:02
  These are all feedforward and that has some implications. And then the big revolution about nine years ago was something called deep learning,
• 18:09
  where we went from two or three layers to 12 layers, and now you can find hundreds of layers and some versions of it.
• 18:19
  And this has been the revolution in A.I., which I think has got everyone thinking important again, just like it was with expert systems decades ago.
• 18:30
  And it is what is driving everything about AI today.
• 18:35
  It's exploitation of this particular algorithm. And it was nine years ago that that this was was done.
• 18:42
  I think people tend to think, oh, every three weeks as a new revolutionary concept coming out in AI, but it's not true.
• 18:49
  It's it's fairly slow. The revolutionary concepts, this was a big one.
• 18:53
  And one of the things that happened was that previously with these, you know,
• 18:59
  just two or three led networks, there'd been some algorithms developed by people that's a person,
• 19:05
  you know, a person programming, not analysing the picture of the car,
• 19:09
  but writing the operators that look at the images, maybe find circles, maybe find straight lines, etc.
• 19:15
  And they're the inputs, the network out of which comes classification in deep learning.
• 19:22
  People got rid of that handcrafted input processing.
• 19:25
  And that, by the way, is why speech understanding has gotten so good.
• 19:29
  You know, five years ago, speech understanding systems were still pretty bad.
• 19:33
  But Alexa, the Amazon Echo, if any of you used that, it's pretty damn good transliterated speech to text,
• 19:43
  not talking about understanding, but getting what the words correspond.
• 19:46
  And we had accents don't matter too much.
• 19:51
  Our noise in the room doesn't matter too much.
• 19:53
  And that's because the deep learning got rid of the human built early operators on the speech signal and learnt much better ones.
• 20:03
  People have overgeneralise that to say, well, we should never have the human doing any part of the job.
• 20:08
  A lot of learning. People say that, and I think that's a mistake in many ways.
• 20:13
  And the world came to know about this in a in a 2014 article in The New York Times by John Markoff, where this image was showed.
• 20:22
  And I Google a couple of networks. Google labelled this a group of young people playing a game of Frisbee,
• 20:29
  and I think most A.I. researchers were surprised that was pretty damn good and much better than.
• 20:33
  Were you surprised, Stuart? Yeah. Yeah, I was surprised.
• 20:37
  It was surprising. This was a of an eye opener that they could do this and.
• 20:46
  Uh, so that, that, that really got the world talking about, I think around 2014.
• 20:53
  In neural networks, I'm going to include something called reinforcement learning.
• 20:57
  I'm going to just lump that together. This was Donald Mickey, who was a colleague of Alan Turing at Bletchley Park.
• 21:05
  He did this in 1961 at the University of Edinburgh. I think it was a professor of surgery, so he wasn't allowed to use a computer.
• 21:11
  So he built the set of matchboxes with with coloured beads in them and and had reinforcement learning.
• 21:17
  Learn to play the game. Tic tac toe. Traditional robotics is the third approach I want to talk about.
• 21:25
  I think this, you know, got started with the. Uh oh.
• 21:31
  Just lost his name. Someone at MIT. Who?
• 21:37
  Larry Roberts. Yes. Thank you. Larry Roberts. Who showed how to take images of Polyhedra and get the lines out of now.
• 21:44
  This time, taking an image was a was a job.
• 21:47
  You took a picture with a camera. You got film, and then you scanned the film.
• 21:53
  Know bit by bit mechanically to get the image.
• 21:56
  And he showed that you could get polyhedra, simple polyhedra from actual images.
• 22:01
  And then I went off and saw, okay, what about line drawings?
• 22:05
  Can we figure out the three dimensional structure from that?
• 22:07
  And that worked out fairly well with shadows in there, etc. And it was then used for robotics experiments as the cube stacker at MIT,
• 22:16
  which looked at a stack of cubes and tried to model them, and the computer copied the A.I. program, copying the stack of blocks.
• 22:26
  Unfortunately, this is the only photo I can find of it, which doesn't really show you much much.
• 22:32
  Swri Stanford Research Institute. Now, of course, these are right next door to Stanford.
• 22:37
  In 1970, they built the robot shakey and it lived in a world of polyhedra and with each side of each polyhedron painted in a matte colour,
• 22:48
  and it built 3D models of the Polyhedron planned how to get around.
• 22:53
  This is Hans. This is a photo I took in 1979.
• 22:56
  That's Hans Moravec up there filming his robot, the cop out.
• 23:01
  And it's only ever outdoor run because it needed the whole mainframe and no one was going to let him use the computer.
• 23:08
  But this was the last day before the lab closed down to move the campus.
• 23:12
  So he got the computer to run all day, this one outdoor run.
• 23:17
  And you can see he built Polyhedra for the obstacles.
• 23:21
  But it took 15 minutes of mainframe processing to to to process the images, to plan a motion of one metre.
• 23:30
  And the shadows moved a lot in that 15 minutes, and it really screwed up the algorithms.
• 23:37
  But traditional robotics then was in a sense, the world built a complete world model and then plan what to do.
• 23:46
  This is the robot Freddie at University of Edinburgh,
• 23:50
  with a fairly simple polyhedron modelled by the way, that hand is about a metre wide and a metre high.
• 23:56
  That is not a miniature camera there next to it. That's about a 30 centimetre long camera.
• 24:03
  You know, we all today think cameras are tiny and plentiful and cheap.
• 24:06
  They were not for most of the history of I.
• 24:10
  And then there's behaviour based robotics which I attribute originally to Grey Walter, who was at Bristol at the American at Bristol.
• 24:19
  This is a 1950 paper he wrote on his tortoise's tortoise as on the bottom left there.
• 24:26
  This particular one had a circuit with two vacuum tubes or valves.
• 24:33
  They're the you know, the those that's a vacuum tube there.
• 24:39
  There's a filament which heats up, spits up electrons and anode there, which collects them.
• 24:44
  And these plates in between can modulate it. And that's how that's how we that's how people used to do electronics.
• 24:50
  It's only got two modules, but it could demonstrate all sorts of behaviours.
• 24:56
  And this is from that Scientific American four page article.
• 25:00
  The next year he published an another one on a machine that could learn there.
• 25:05
  It had grown to seven vacuum tubes or valves for learning and it was very much like Pavlov.
• 25:11
  The 3000 cycles there is about a whistle that it could hear and and do Pavlov type experiments.
• 25:17
  So there that was behaviour generated in a sort of holistic way.
• 25:22
  I took that and changed that a little bit for digital, where I connected sensors to actuate is instead of building one central model,
• 25:30
  maybe had lots of partial models all happening at once and each layer generating different behaviours and having the
• 25:35
  actuate it sort of figure out what to do with the conflicting demands and put them together in a into networks.
• 25:44
  This is a robot called Genghis. On the left are these little finite state machines which make up the layers.
• 25:50
  And there's actually six copies of many of them for each leg, one for each leg, 57 finite state machines and 12 layers there on the left.
• 25:59
  And I got these robots to walk around and do interesting things.
• 26:02
  But they they did have implications. Sojourner, which went to Mars, landed in 1997 for the primary mission of seven Sol, seven days.
• 26:13
  And the secondary mission for a further 21 Sols was was operating from the ground.
• 26:18
  But at the Sol 28 the behaviour system was turned on and there's this was taken on Sol 72.
• 26:27
  There's Sojourner off in the distance from the lander, wandering around exploring Mars by itself.
• 26:34
  And then, as Pete mentioned, this has been the basis of the Roomba.
• 26:39
  20 million of those in homes, a lot of robots in Afghanistan and Iraq dealing with roadside bombs in Fukushima.
• 26:48
  We got there a week after it happened, helped shut down.
• 26:52
  And really it was essential for the shutdown of the reactors that were still operating during the cold shutdown.
• 26:58
  And then the top middle layer, you see one of the larger robots, a 200 kilowatt robot with a suction tube.
• 27:05
  That's a really big Roomba that was used to clean up radioactive stuff.
• 27:08
  And when I was last in Fukushima in 2015, these robots were still operating and many more.
• 27:14
  And the cleanup is not going to be finished till 2050, by the way.
• 27:20
  And then more recently have put robots in factories using this behaviour based system.
• 27:26
  Around the year 2000, Damian Esler and Bruce Blumberg took my finite state machines,
• 27:33
  which sort of mix logic and behaviour together and split them into something separate called behaviour trees.
• 27:39
  And those behaviour trees have become quite popular at MIT, at Rethink Robotics, my company.
• 27:47
  When someone shows the robot what to do, it automatically builds a behaviour tree on the left there and you can go and edit it.
• 27:52
  But more importantly, perhaps or more interestingly, about two thirds of all video games are now programmed with behaviour trees.
• 28:00
  This is one of the frameworks, unity, there's a whole bunch of them.
• 28:04
  And so people all around the world are programming that AI little characters in video games using behaviour trees.
• 28:12
  So it's, you know, if you count the, the actual number of instances, you know,
• 28:16
  every one of the horde of those little creatures in the video game that are coming to attack you is,
• 28:20
  is running a separate behaviour tree or running an instance of a behaviour tree.
• 28:24
  The raw numbers is most AI systems in the world.
• 28:27
  Sorry, are behaviour based. I have forgiven Stuart for in his book saying there's been no known application of behaviour based robots.
• 28:41
  Anyway, with first edition, this is my summary of the four approaches to AI.
• 28:49
  Symbolic is very deliberative. Traditional robotics is very deliberative.
• 28:54
  Behaviour with behaviour trees can be reactive and deliberative and neural networks.
• 28:58
  I see a cat, it doesn't really know what it's doing.
• 29:02
  It labels and here I get a really scientific study that's a joke of what are the strengths of the different approaches for composition?
• 29:14
  Symbolic is the best for composition in on a scale of 1 to 3, it's symbolic.
• 29:20
  It's best composition because those symbols let you patch different things together via the symbols.
• 29:26
  But symbolic is pretty bad at being grounded, whereas the neural systems really ground in sensor data to the symbols that are the labels.
• 29:35
  So they're good at grounding. And then, you know, just for fun, I again very scientifically compared this to a child and I added cognition,
• 29:48
  remember, on a scale of 1 to 3 and these are the numbers that I that I came up with.
• 29:54
  I think we're very, very far away from what even a child can do.
• 29:58
  And I'll talk about that in some detail a bit later. Now, I think a lot of people have been very bad at predicting the future of AI.
• 30:06
  And I. I had a thinking technology review last year about the seven ways I see people getting their predictions wrong,
• 30:15
  including treating a sufficiently advanced stuff as magic and thereby attributing
• 30:22
  any quality you want for the purpose of an argument about what it's capable of.
• 30:27
  I'm just going to talk about two of these cases performance versus competence and suitcase words.
• 30:34
  So performance versus competence. When we see a person perform some tasks at some level,
• 30:39
  we know we have a good intuition for what else they know around that performance in that general area.
• 30:47
  So if we see a person describe this image as a group of young people playing a game of Frisbee,
• 30:53
  we'd expect to be able to, you know, ask the person, what's the shape of a Frisbee?
• 30:57
  We'd expect them to know that. We'd expect them to say no.
• 31:00
  Whether a person can eat a Frisbee, can a three month old person play Frisbee, how young are they?
• 31:07
  You know, kind a person. But the labelling systems know nothing about this.
• 31:12
  They not have any general competence around these symbols, which they label the images with.
• 31:19
  They have a performance, but not a competence.
• 31:21
  In the same way that deep blue Kasparov at chess, but Deep Blue couldn't be a coach in chess in any shape or form.
• 31:31
  Whereas a human who be Kasparov could probably even teach me to play a little bit better chess than not really lousy chess that I play.
• 31:42
  And it would be weird if a person could label those images and didn't have these more general competence.
• 31:47
  So I think people hear about the performance of a system.
• 31:50
  It beat the world champion and go, Wow, it must be really intelligent.
• 31:54
  It must be able to do just about anything. But no, it's very narrow.
• 31:58
  And why are these labelling?
• 32:01
  You know, why have these, you know, neural nets that label images based on probabilities when it comes down to it and you know,
• 32:10
  yeah, that's a 90% chance, that's a person, a 60% chance, it's a person.
• 32:14
  But we would never make the mistake. Oh, it's 20% chance.
• 32:17
  That piece of tree is a person. But they get put together and come up with something.
• 32:23
  His his adversarial attacks.
• 32:27
  On on on on. Deep learning has become very popular.
• 32:31
  This is one of the earliest ones. His a deep learning network. It says with 9090 9%, that's a guitar, 100% that's a penguin.
• 32:39
  And then a program goes and plays around with images and tries to create images and does hill climbing and a genetic space.
• 32:47
  And it comes up with that. There's this image here, literally this image that says at 100%, that's a guitar and that's a penguin.
• 32:56
  And also it's found, you know, a set of pixels that provoke the early stages.
• 33:02
  And that's sort of weird. And and so you get these weird classifications.
• 33:09
  My favourite there is the school bus. You can see the school buses, shortness of it.
• 33:13
  For an American, you know, that's what school buses sort of look like with a yellow.
• 33:16
  But it's clear that there's no real spatial understanding that convolutional neural networks actually do lack coherence of spatial input.
• 33:27
  And I'll come back to this later, because it actually has significant.
• 33:32
  Problems for us. Suitcase words.
• 33:35
  This is a term that Marvin Minsky came up with where he pointed out that many words we use have many, many meanings.
• 33:43
  Now what happens is an AI researcher comes up with something that can do a little
• 33:51
  bit of classification or a little bit of recognising or a little bit of reading.
• 33:55
  The Institutional Press Office says, We got to we got to get a press release about this.
• 34:01
  And before you know it, it gets turned into, you know, a local local professor has built up an AI program that can hallucinate.
• 34:12
  For instance, we've seen hallucinations o. AI systems going on loosely.
• 34:15
  This is really wild. But these are suitcase words.
• 34:19
  Take. Take the word learn. For instance, learn to ask means many, many different things.
• 34:24
  Learning to play tennis is related to, but rather different from learning to ride a bicycle.
• 34:30
  It's certainly a very different sort of process than learning ancient Latin.
• 34:36
  And even though they're both in both blackboards, learning Latin again is a very different process from learning algebra.
• 34:42
  I'm good at algebra. I was really lousy at Latin and I'm really lousy chess.
• 34:48
  Learning to play chess is a different sort of skill again, learning to play music again.
• 34:52
  These are all very different skills, but we use that same suitcase word and in my last company I'd have the VCs call me up and say,
• 35:01
  So the carrier is is doing just come in.
• 35:04
  In the sounds in the Bay Area, the carriers is learning about robots.
• 35:08
  The robots have learning. They've got to have learning. They'll be beaten.
• 35:13
  And that learning is it also shows my love for VCs and learning learning away around a new city is a very different sort of process.
• 35:21
  So these are suitcase words. And when we hear, you know, the machine can do a little aspect of it,
• 35:26
  it often gets generalised that big aspect and it's a much bigger set of skills on a code, no cheering.
• 35:35
  And this is getting back, Aaron, to what he really meant in The Imitation Game.
• 35:42
  Can you tell whether it's a computer or a person answering your questions?
• 35:46
  And I sometimes thought that he was using it as a rhetorical device to say, well, if you can't tell,
• 35:53
  then you can't say anything more about thinking because the machine, you know, can fool you.
• 35:58
  So you can't say a machine can't think because you just don't know.
• 36:02
  And I don't know what the step from man versus woman.
• 36:07
  To Machine versus persons doesn't quite fit in here, but with what it was doing it's a little strange.
• 36:14
  Then can it be said to be thinking?
• 36:18
  And so some people and I at various times, various versions of me over the decades have thought that it was just a rhetorical device,
• 36:24
  a thought experiment to show in principle, we can't rule out that a machine can think.
• 36:29
  But then as I reread his papers, now I see that he talks about how to construct a program for such a machine.
• 36:35
  He thinks 3000 person years that 50 years or 60 years or 50 programmers was 50 years of 60 programmers was way too onerous.
• 36:44
  So he suggests having the program learn like a like a child and then play the game.
• 36:50
  And so at some level, he does seem to treat it as a real test.
• 36:53
  And so it's become, you know, a benchmark for A.I., supposedly, but it's something that can easily be hacked.
• 37:00
  So all the Turing Test competitions are won by stupid programs which are not intelligent at all.
• 37:07
  So I suggest that we get rid of Turing's imitation game and think about better tests, get machines to do real tasks in the world.
• 37:16
  And I'll just give you a couple of examples. If we could do these and we we do not have a clue how to do these.
• 37:23
  I mean, you could you could give, you know, the best AI company in the world, you know, 5000 people devoted to these tasks.
• 37:31
  And they wouldn't get very far in the next five years. Certainly if they were really lucky, maybe they'd get somewhere in the next ten years.
• 37:38
  One is an elder care worker, which is an embodied task.
• 37:42
  So a living care provider for a for an elderly person over decades.
• 37:48
  And we will need these, by the way, as as the baby boomers get older.
• 37:52
  And it's got to understand human relationships, expectations in a household provide physical help the person,
• 37:57
  including manipulate the whole body as they get weaker, understand their degrading language.
• 38:02
  You know, Alexa can only understand pretty good language, but as a person gets a little hard of hearing and can't get nouns so well,
• 38:12
  they start pointing, they start nodding, they start grunting, understand, provide for human needs, etc., etc. a whole bunch of things it has to do.
• 38:21
  A service logistics planner would be a disembodied system design implement perhaps a new dialysis ward in an existing space from scratch.
• 38:32
  This is just an example. Now people can do this.
• 38:35
  So if we're going to have superintelligent superintelligence better be able to do all these things, because otherwise it isn't superintelligence.
• 38:42
  And to do that, it's got to do all sorts of geometric reasoning, quantitative geometric reason, quantitative physical simulation,
• 38:49
  understand human needs and fears, understand how family members will feel and act in dialysis for all sorts of stuff.
• 38:55
  So the elder care worker, these are happy pictures that you get from Google.
• 39:00
  When you Google elder care. The reality is not this happy but, you know, give physical help to people.
• 39:08
  If you if they at this stage it's really not so happy because they normally wear a diaper.
• 39:13
  And, you know, figuring out how to help people at that stage is going to be very difficult for a robotic elder care worker,
• 39:20
  but help them with all sorts of physical tasks. The service logistics planner I base on the idea of an army colonel.
• 39:27
  When the US military goes into some new area, Army colonel is given tasks,
• 39:32
  set up a hospital, set this up, set up schooling, set up all sorts of stuff.
• 39:36
  Well, set up a dialysis ward amount and Colonel will be expected to do it.
• 39:41
  And they have to figure out, you know, how the patients are going to sit or lie,
• 39:45
  how many places you need in dialysis ward, what the flow of people is going to be through it,
• 39:53
  what the the nurses or other attendants need to do with the people, what sort of information they need to give,
• 40:00
  what sort of feedback, what the layout of the space for dialysis needs to be, what for waiting room?
• 40:06
  How people get from public transportation in the city or cars in into an out of whatever
• 40:12
  facilities whole bunch of problems which we cannot have an automatic system do at the moment.
• 40:18
  And if you say, oh, deep learning will do it, well, you've got to have an awful lot of dialysis machine examples.
• 40:25
  That's going to be pretty hard data to get in in the in the quantities that you need.
• 40:30
  So I think these sorts of challenges are the sorts of trouble.
• 40:35
  If we if we can't do these sorts of things, we're not getting towards human level intelligence.
• 40:41
  So what's hard the day I'm going to my blog, I've got just seven random ones.
• 40:47
  I'm going to briefly talk about four things that are hard today that we that we have no idea of.
• 40:52
  But that superintelligence proponents sort of assume that we're pretty good at first.
• 40:57
  Is real perception as distinct from labelling those images?
• 41:01
  Here's some other examples. No. One, these are adversely generated.
• 41:06
  No one would get that third one from the left and top row and say that's an armadillo.
• 41:11
  But this particular network does well. That's sort of fun and games, but it does have some real implications.
• 41:17
  So. This is from a Senate hearing in the US Senate from two or three months ago where there's
• 41:23
  a stop sign that an automatic driving vision system labelled as a 45 mile an hour sign.
• 41:31
  Now it's got four pieces of tape attached to it. It was produced.
• 41:35
  Purposely to fold it.
• 41:36
  And if you look around the S and the T with those two pieces, it's sort of like a four and over the O and P, maybe it's sort of like a five.
• 41:45
  You can sort of see maybe how it got, but how a stop sign is red.
• 41:51
  Why didn't it get? Why didn't it get that?
• 41:55
  That couldn't possibly be a speed limit sign. It's read. Well, it turns out the colour is not colour.
• 42:02
  We think colour is just from the colour of the pixels and.
• 42:08
  When you use deep learning and you don't use a human designed input system, it actually doesn't get colour constancy.
• 42:14
  This is from Ted Adelson. And they might be. What do we see that anyone?
• 42:20
  Want to look for in check? Yeah. How do you know it's a checkerboard?
• 42:23
  Because it's black and white. Oh, it's black and white. See those two squares there?
• 42:27
  They're the same colour. And there I've expanded the pixels.
• 42:35
  Now, how could they be the same colour? Look, black and white.
• 42:39
  Remember all those things going downward instead of upward?
• 42:43
  It turns out we compensate for their shadow. There's a shadow there.
• 42:46
  We compensate for that. Right. And that's how we know that it's black and white, which it really is.
• 42:51
  But if you just look at the pixel value and don't label the individual parts, then you get it wrong.
• 42:58
  And what does this. And what colour are the strawberries?
• 43:04
  Red. Yeah, they're red.
• 43:06
  So of the three quarters of a million pixel, there are only 122 pixels where the red component is bigger than both the green and the blue.
• 43:14
  They're the ones with the the biggest red component where they're both bigger than blue and green,
• 43:21
  and they're the three pixels which just have the biggest red in them.
• 43:26
  They're not very red. And when you get rid of the strawberries, the any any if when the strawberries are there,
• 43:31
  you sort of see a little red in the grey right next to it that's flowing over from you seeing the strawberries.
• 43:37
  And when you get rid of it, it becomes less red.
• 43:40
  Some some may or may not see that, but that's where you know, you're using are these are strawberries.
• 43:45
  They must be red. Your reconstructing everything. We do that all the time.
• 43:49
  And colour constancy is just one of many, many visual tricks that we use which make good sense.
• 43:55
  If you can't label the colours and teach the deep learning that what it doesn't learn them.
• 44:01
  So the Deep Learning Network never realises that stop signs are red because in pixels space a lot of them aren't red.
• 44:08
  And so it's not a feature and we don't have colour constancy.
• 44:12
  So, um, realism in real problems is, is, is, is different from what I call tech rows, you know, with the myriads of them in the Silicon Valley,
• 44:26
  you know, I'm going to apply machine learning to X, I'm going to apply machine learning to I just need data.
• 44:30
  But they don't necessarily know what the data they need to get the robustness if
• 44:35
  we're going to have an elder care worker better work in human homes at a human home.
• 44:38
  Right. This is from Aaron editing his thesis. This is what his home looks like, his kitchen.
• 44:43
  That's the sort of world we have to deal in. But we're pretty good, you know?
• 44:48
  Anyone know what that is? Yeah.
• 44:51
  It's a tow container. And how do we know that? Well, maybe that's the.
• 44:55
  The salt shaker is priming. That's the salt. But we sort of know that that's.
• 44:59
  But if you just saw it by itself or you saw that by itself, it's a rice cooker.
• 45:05
  You know, you can get that because, you know, it's a kitchen. I'm going to do a little experiment now.
• 45:11
  Does anyone know what steampunk is? Sure. Does anyone not know what steampunk is?
• 45:17
  Oh, good. You're my test subjects. This is theme.
• 45:21
  Okay, that is a style. Steampunk is more steampunk.
• 45:28
  Here is this these are usually make up is there's more steampunk.
• 45:32
  Okay, now we do the test. You've had three examples. Is that steampunk?
• 45:36
  There are two maker faire. No, that's not steampunk.
• 45:41
  She has goggles. Not that steampunk wear is goggles.
• 45:47
  The others had goggles while they around his neck. Okay. They thought we had goggles.
• 45:51
  Steampunk. No steampunk. Yeah.
• 45:55
  I think that's debateable. I think it's sort of lousy steampunk now.
• 46:00
  Steampunk? What about that? Yeah. It's got nothing to do with goggles.
• 46:04
  You know, this is the examples. It wasn't.
• 46:06
  We had robot arms, but you, you know, some of you, they were able to learn that category from three examples.
• 46:12
  And you got it pretty, pretty well. When we use a deep learning, we show it millions of examples, hundreds of thousands of times each very different.
• 46:22
  So we don't have real perception. We don't have real manipulation.
• 46:26
  This is the air lab at Stanford in 1978.
• 46:31
  And there you see on the right of the image, the blue arm. We also had a gold arm which is off of the camera.
• 46:38
  I don't have a picture of it. How do I know that gold arm is there? Because that's me.
• 46:45
  So here's the gold arm in the lobby of the computer science department at Stanford today.
• 46:50
  And you might remember these arms, Steve.
• 46:54
  And you see the gripper. There is a parallel jaw, gripper and worm screws where the two fingers come together in parallel, back and forth.
• 47:02
  That's 1978. Here's what my company was selling for grippers in Nice in 2018.
• 47:08
  Same thing. Not much improvement in the hands.
• 47:13
  We actually sold kits for parallel grippers and kits for suction grippers.
• 47:18
  Here's the shunt catalogues thousands of pages of different size of parallel jaw grippers.
• 47:24
  And that's what people use for robot today.
• 47:26
  We are lousy at manipulation, but you know, humans can do all sorts of manipulation and superintelligence should be able to do them too.
• 47:37
  Superintelligence should be out of play or run.
• 47:44
  You know, if superintelligence is going to, you know, take over and kill all the people, it better be able to.
• 47:50
  Well, it doesn't have to cook, but it better be able to do the all the manufacturing that people currently do with their hands.
• 47:55
  That's Julia hit. Hit again, doesn't have to be cooking, but this just shows the dexterity of humans.
• 48:02
  This is a sushi chef where they're using the force and changing the angle of the knife so they don't pull apart the pieces of the fish.
• 48:11
  And we just we can just figure out how to do these things with our hands.
• 48:16
  We cannot get Steve, can we can we get a robot to do any of these?
• 48:20
  No, he says. Steve says, no, it must be true. We can't we can't get a robot.
• 48:25
  And all our manufacturing relies on these sorts of dexterity with sloppy floppy objects, etc.
• 48:32
  You know, some people are experimenting with manipulation for people and, you know, putting clothes on people and stuff, but we can't.
• 48:39
  This is from my home state of South Australia and if Nick was here I should tell him, you don't have to worry.
• 48:45
  We can't give the robots the knives. They can't cut up rib cages yet.
• 48:51
  We're a long way from that.
• 48:52
  We can't do most of the manipulation that people can do the very simple pick and place, even after working on it for 40 years.
• 49:01
  So the thing read a book. Why read a book? A lot of human knowledge is in books.
• 49:05
  If we're going to have superintelligence, maybe it should read just like we read to get a lot of knowledge.
• 49:10
  And you know, every so often there's a scare a bit humans at reading.
• 49:14
  Good headline or maybe not.
• 49:16
  This was this is where a little while ago people thought, oh, suddenly this AI system can read a book when examined, closed, not so well.
• 49:26
  And here's some examples of it. So these are the Winograd schemas from NYU.
• 49:31
  Alice tried frantically to stop her daughter from chatting at the party, leaving us to wonder why she was behaving so strangely.
• 49:36
  Who's who? She. Well, Alice.
• 49:40
  Yeah, but if we change one word, the dog was barking.
• 49:45
  The delivery truck zoomed by the school bus because it was going so fast.
• 49:49
  What was going fast in the truck? Yeah. The delivery trucks went by the school bus because it was going so slow.
• 49:54
  Now it's the school bus. Sam pulled up a chair to the piano, but it was broken.
• 49:59
  So we had the stand and said the chair was broken. But if instead he's seeing it was the piano that was broken.
• 50:05
  Now, to answer all those questions, you're doing all sorts of different,
• 50:08
  different sorts of simulations for each question in your head about what's going on.
• 50:12
  And we can't do those sorts of sort of common sense gross simulations at the moment.
• 50:21
  And writers have written books. Just assume we know a lot of stuff.
• 50:24
  You know, you all know that Prince William is taller than his son, Prince George.
• 50:28
  But you don't know and you know that you don't know whether that's going to be true in 20 years, which is an interesting thing.
• 50:34
  You just know that I didn't even have to get trained on that. You just know that, you know, down the bottom, dolphins eat raw fish.
• 50:40
  So to some humans, dolphins don't usually cook food.
• 50:45
  Most humans do. We just know this stuff. And it's background and it's assumed in all writing that we know the stuff.
• 50:52
  Our air systems have very little common sense. John McCarthy wrote his first paper on this topic in 1958.
• 50:58
  We're still away from it. And Tampa just announced a $2 billion research program on common sense.
• 51:07
  And, you know, there's a couple of sorts of it. Here's one of the pieces.
• 51:10
  Know in the main objects, agents, places. They're saying this is what we want you to do.
• 51:17
  Do all these things for objects, agents and places. One of those months on the right then not months for the research.
• 51:23
  They're the age of babies when they can do them. These are all lives spell qui Susan Carey.
• 51:29
  So it's a really it's a multibillion dollar project on getting the common sense of an 18 month old.
• 51:35
  That's what the funding agency believes is still hot. Not quite at superintelligence yet.
• 51:40
  And I'm going to skip over writing or debugging a program.
• 51:43
  But some people think that, you know, we're going to have artificial intelligence going to rewrite its code from scratch.
• 51:51
  Basically, my argument here is that people keep saying this, you know, we cannot we cannot have.
• 51:58
  And then they say, what's superintelligence going to rewrite the rules of physics, too?
• 52:01
  By the way, even Max TEGMARK may not believe that he's a cosmologist.
• 52:09
  But, you know, we we don't have anything that can understand this program.
• 52:14
  And I was going to go through and explain how I just glance at it.
• 52:17
  And I know all sorts of stuff about this program. And none of those deductions, anything that we can have any sort of system do at the moment.
• 52:25
  So what do we work on now? I'm running out of time. I think we can't work on those.
• 52:30
  You know, the health care worker directly, the older care worker.
• 52:34
  We can't work on the systems logistics planet directly. It's too hard.
• 52:38
  But these are sorts of goals that we could work towards.
• 52:41
  And if we're making progress towards any of these goals, our AI systems would have more common sense and be more robust.
• 52:48
  A two year old has colour constancy and what colour categories can map form to function can figure out that they can sit on this.
• 52:56
  This can act as a chair even though it's not a chair. So and then form and function.
• 53:01
  They have object classes and can categorise objects that look totally new at the pixel level.
• 53:05
  Unlike the deep learning they can do one shot subclass learning.
• 53:09
  You take, you take a kid to the zoo and they see a giraffe for the first time in their life.
• 53:14
  You don't say, Oh, by the way, that's an animal. They know it's an animal.
• 53:18
  At age two and after they've seen a zoo for a few seconds, you go home and they open a book.
• 53:24
  They know that's a giraffe. One shot learning. Four year old can talk and listen when they come in about turn taking understanding cues.
• 53:34
  They know when they're in a conversation with someone and one of the participants changes they if multiple
• 53:40
  participants in the conversation they know how to get attention and direct their remarks to someone,
• 53:45
  they can tune in a lot of conversations. They know when someone is suddenly speaking, we speaking differently from normal.
• 53:51
  So they know a lot of stuff. And currently Alexa knows none of this stuff, for instance.
• 53:57
  And Alexa is a fantastic push forward, by the way.
• 54:00
  I'm not belittling it. It's a lot more. Six year old can estimate from vision how to pick up many objects.
• 54:08
  Whether it's going to be one hand, wants to handle what's going to wrap their arms around it, use the whole body,
• 54:13
  they pre shape their hands for a grasp they can apply and control force appropriate for a task they can use.
• 54:19
  Chopsticks. A six year old who's grown up in a chopstick house can use chopsticks as a complex thing.
• 54:24
  They can do all sorts of tasks. None of our industrial robots can do any of these.
• 54:31
  They can even pick up. This is a step towards elder care.
• 54:34
  They can pick up cats and dogs and pet them and they can wipe their own bums, which is important for elder care, too.
• 54:44
  An eight year old can articulate their beliefs, desires and intentions, and they understand that different people, other people have different.
• 54:52
  Beliefs, desires and intentions. Some of this might be a nine year old, but eight year olds are fit the pattern better.
• 54:58
  I apologise and they can deduce many of these things by observing others.
• 55:04
  And we don't have systems that can do that today.
• 55:06
  But if we made progress on any of these four things, it would help our systems operate in the world where things go wrong.
• 55:14
  Oh, the superintelligence destroying us. Thanks, Nick. But here's where it goes wrong.
• 55:20
  It goes wrong in the datasets getting putting bias into systems.
• 55:25
  And some of the headlines get a little hysterical, but there's real issues there.
• 55:31
  Watson Health You know IBM tried to push Watson too fast the in the health care and it was a bit of
• 55:37
  a big failure but not why they are pushing it into into digital marketing now and that's a really.
• 55:45
  Oh well, anyway, forget it. But what if we're completely off track?
• 55:52
  You know, we think building intelligent machines is like building copies of ourselves ultimately.
• 56:02
  Okay. Well, maybe we can do that. What if we see these two dolphins, A and B, and then we look more closely and we notice that B is a robot.
• 56:11
  Are we going to conclude that a bunch of A's got together and built the B?
• 56:16
  Did the dolphins build the robot dolphin?
• 56:21
  We don't think they're capable of it. Okay.
• 56:25
  Are we capable of it? Are we capable of building human level intelligence?
• 56:29
  We'd like to think we are. We're pretty arrogant about it.
• 56:34
  But maybe we're not. Maybe we were just not good enough. Just like we think the dolphins are just not good enough.
• 56:39
  Maybe the aliens up there are looking down on us. Look at those little humans trying to kind of make copies of themselves.
• 56:45
  But we're just not smart enough. But maybe we are smarter, but we're going about it the wrong way.
• 56:51
  And I want to use flight as an example here. You know, people always say, well, you know, we don't we don't we don't make things fly by copying birds.
• 56:59
  And Lord Kelvin was wrong just a few years before heavier than air flight, although it's not clear exactly what he was saying.
• 57:06
  But. No. The Wright brothers. One out.
• 57:12
  You know, earlier people had said, well, we need flapping, flapping wings and that wasn't helpful.
• 57:16
  And then Lilienthal maybe mispronouncing his name, really understood that it was a static wing in airflow that was important for gliding.
• 57:24
  He he did over 2000 glides before he died doing what he loved gliding.
• 57:31
  And he wrote a book saying in the German title, says, It's a human flight inspired by birds.
• 57:37
  And Wilbur Wright was certainly a knew about that book read that book and he observed
• 57:43
  that birds use a change of shape of their wings in order to roll left and right.
• 57:49
  And that was they did have better engines.
• 57:51
  But the key innovation for for Wilbur and Orville Wright was realising that control of the wing shape in the airflow was important.
• 58:02
  Up until that point, people hadn't been thinking about control.
• 58:05
  That turned out to be the the critical thing. We use our computers.
• 58:10
  We've had, you know, Moore's Law just delivering us better and better cocaine for a long time and computation and certain metaphors.
• 58:21
  But maybe we're not thinking about things the right way. Maybe.
• 58:24
  And maybe it's you know, maybe it's going to be 150 years before someone figures that out.
• 58:30
  Some things take a long time. We do not know how long it will take.
• 58:34
  But I want to end with something that Alan Turing said in 1950.
• 58:39
  We can only see a short distance ahead, but we can see plenty that needs to be done in AI and there's plenty for us all to do.
• 58:46
  So thank you. And I'm sorry I went a bit long. Thanks.