How Do Our Brains Work? with the Godfather of AI

22:03Dr. Geoffrey Hinton:This is a sort of simplified caricature of a system, but in this layer, it'll be detecting possible heads of birds, possible feet of birds, possible wingtips of birds. But not total bird. But not total birds yet. And now in the next layer up... Okay, the fourth layer. You might have something that says, I get excited if I see a possible head of a bird. I get excited if I see a possible wingtip of a bird. I get excited if I see a possible foot of a bird. If it sees a bunch of those things at once, it gets very excited and shouts, bird. Okay. So there's four layers up in this hand-wired system. Now, this is a very simplified system.

22:37Dr. Geoffrey Hinton:Yes. But I think you can see the idea of how we go about hand-wiring it so that in this fourth layer, you get something that's shouting, went ping whenever you get lots of combinations of features that might be a bird. So this is a hand-wired system. This is a hand-wired system. So now the digitally enabled computer-wired system, the neural network system that was designed for artificial intelligence that won out over this symbolic AI model that was based on logic, how many layers does that have? It depends what system you're talking about. And when we say system, do I mean LLMs? No, no, no. This was in 2012, Alex Krzyzewski and Ilya Sutskova, with some help from me, made a system called AlexNet.

23:26Dr. Geoffrey Hinton:And that had about seven layers like this. Ilya, of course, was one of the co-founders of OpenAI who's since left. Yes. He was worried about the safety issues at OpenAI. Yes, and he set up his own company to work on how can we build a safe superintelligence. And AlexNet was just recognizing images and then across seven different. AlexNet was trained on about a million images. And it actually had more training data than that because it took a million images and took big patches of these images. And what it was trying to do is say whether this big patch of an image is the most prominent thing in that big patch is whatever that image was labeled as.

24:06Dr. Geoffrey Hinton:So people had gone through and said the most prominent thing in this image is a bird or maybe an ostrich. So it had kinds of bird, right? Right. The most prominent thing in this image is a shiitake mushroom. It sounds like CAPTCHA, what you're describing. This is exactly like CAPTCHA. So fine fire hydrant, fine motorcycle, fine bicycle. Exactly. Okay. And so Alex and Elliot trained a neural net that would be very good at CAPTCHAs. And the way you train it is it starts off with random connection strengths in all these seven layers. Let's suppose, though, they just trained a simpler system, which they can easily have done, just to say whether it was a bird or not a bird.

24:43So you put an image and it has random connection strengths.

24:49Dr. Geoffrey Hinton:And at the output, you have one neuron. And if that gets active, it means bird. If it's inactive, it means not bird. And to begin with, it gets a little bit active. because it's no idea whether it's a bird or not. It's no better than chance. So it sort of hovers around 50 % active. And what you'd like it to do when you finish training is be at like 99 % if it sees bird and like 1 % if it doesn't see bird. Got it, okay. So to begin with, you showed an image of a bird, you run it through these random connection strengths, and it says 50 % bird. Okay. It has no idea whether it's a bird or not. Yeah, I was like 50 % bird is useless information.

25:29Yeah, useless. So now you can ask the question, suppose I would change one of those connection

25:36Dr. Geoffrey Hinton:strengths a little bit. And remember there might be, in this case, 100 million connection strengths. Suppose I would change one of them a little bit. Instead of saying 50%, would it say 50.001 % or would it say 49.999 %? So you can rewire it, but if there's a bird there, I can train this network to understand that yes, this is more of a bird, this is more of a bird. And then the average strength of the connection will increase to the point at which... When I showed images of birds, I'd like to change that connection strength. So its probability of saying bird goes from 50 % to 50.01%. And when I show it a non-bird, I'd like the probability to go from 50 % to 49.999%.

26:23Why only that little? Why don't you just want it to shoot up to 99 and shoot down to zero?

26:27Dr. Geoffrey Hinton:We have to go slowly here, otherwise we'll overshoot, basically. I gave you the idea of doing a little experiment where you say, let's change the connection strength a little bit and see if it helps. If you do an experiment like that, you'll take forever because there's 100 million connection strength. So I do a little experiment for this connection strength and change it a little bit. Then a little experiment for that connection strength where I showed another image. Yeah. This is going to take forever. So the question is, could I just show it an image of a bird? And for all of those connection strengths in the whole network, could I somehow figure out whether raising it a little bit or reducing it a little bit is the right thing to do to make it more likely to say bird, to make it raise the probability?

27:10Dr. Geoffrey Hinton:And each connection strength by itself will only raise the probability a tiny bit. But if I change 100 million connection strengths all at once, it might go up quite a bit. I change them all in the direction that'll help it say bird. Because it's quicker to do that. Because if I can figure out how to change them all at once, if there's a trillion connection strengths, I'll go a trillion times faster. Faster, exactly. Okay. And then, but how this gets used eventually isn't through images, it's through language. Okay. So I started off by explaining how we'd make something that would recognize a bird.

27:43Dr. Geoffrey Hinton:I had to give you a feel for what the network would look like and how we would then learn all those connections by using this magic method of figuring out how to make the answer a bit better by changing connection strengths. There's an algorithm called backpropagation, which basically looks at the error you make. That is, you said 50%, you should have said 100%. So there's a discrepancy between what you said and what you should have said. You send that discrepancy backwards through the network. And there's a way of figuring out for each connection strength now, whether you should increase it or decrease it to improve the answer, to reduce the discrepancy between what it said and what it should have said.

28:18Dr. Geoffrey Hinton:So increase the accuracy. To increase the accuracy. Okay, so backpropagation is a way, it's like code that goes into the system, and then this machine learns from that backpropagation and says, I should have done it. Then having done the backpropagation, it knows whether to increase or decrease its connection strength, and it changes all of the connection strengths at the same time in directions that should help. And now you'll have something that's a bit better at recognizing that particular bird. And then you can get into gradations like, it's an ostrich, it's a dove. And so through back propagation, you're teaching...

28:50Dr. Geoffrey Hinton:You'll learn to recognize all the different kinds of birds. You're using the neuroplasticity of this network to teach it how to make better understanding. Yes. Correct. And to begin with, when it's got random connection strengths, it won't have features like beak, right? Right. They'll just be random connections from one to the next. But what will happen over time, if you keep training it to tell the difference between birds and non-birds, and you look in the network, you'll see that in the first layer, it's made things that detect bits of edge. Got it. And in the second layer, it may have made things that detect things that might be a beak.

29:26It did your hand wiring system.

29:28Dr. Geoffrey Hinton:So it'll do something a bit like the hand wiring, but much more sensitively balanced. It's not just looking for a feature that's good for recognizing birds. It had to recognize a thousand different kinds of objects. So it's looking for features that are good for birds, but also good for recognizing fridges and mushrooms and motorbikes and subways. Do you think it's easier for that machine to ingest information than it is for me to be ingesting what you're describing to me right now? Yes. Because? Because what I'm doing right now is at a very abstract level. It's abstract in 100 bits. I'm describing, it's only 100 bits per sentence bandwidth, roughly.

30:06Dr. Geoffrey Hinton:And I'm describing a sort of, this is meta level. I'm describing how this works. The brain is basically built for doing this. And it's managed after millions of years of evolution to do this more abstract stuff we're doing now. But that's the sort of height of our abilities. Whereas recognizing objects, two-year-olds can do that. Anyway, you asked the question, but how is this relevant to language? Yeah. So how is it relevant to language? Because when I think of artificial intelligence and these large language models that have been our primary exposure to it that has led to this AI boom of the last couple of years and chat, CPT, et cetera, the superpower of it is to, as opposed to search, where search I could say like, oh, dance club in New York, and it would just pull up all the dance clubs in New York.

30:49Now I could say, you know, dance club in New York, and it would understand that I want to go out and that I would like to have a night out. And I want to know maybe that it's open and that it's and it has all this context around the thing with AI that I didn't have when I was just simply, you know, Gemini is doing a lot more than Google search was doing for me.

31:06Dr. Geoffrey Hinton:Because Gemini actually understands the question. Google search never understood the question. In Google search, originally, what it would do is have a long list of all the websites that were to do with New York and a long list of all the websites that were to do with nightclub. And it would intersect these lists. It would say, what's in the list of things that are to do with New York, the list of things that are to do with nightclubs, the list of things that are to do with open now or whatever. And it would intersect them and it would give you the things that fitted though. And occasionally it would give you something that had one of them missing, and it would say it was missing.

31:44And it would tell you it was missing, yeah. So Google Search was kind of like, let's imagine the game Memory, but with more than 52 cards. And it was basically saying, oh, you asked for this, here's a match for this. And then it would play Memory with Venn diagram. Yeah, it would basically do a lot of work to get all these lists,

32:01Dr. Geoffrey Hinton:and then it would efficiently intersect these lists to tell you something that satisfied all the terms in your query. And now artificial intelligence, what does it do? It understands what you said. It has a model of how the world works and what's going on in the world. It has a brain. It has what we would call a brain, yes. Like if you give it a math problem, the latest chatbots, they'll be better at it than all but the very best mathematicians. Okay, then certainly the me then. Which is scary. But let me do the transition from this thing that recognized a bird to a large language model. Yes. What have they got to do with each other?

32:36Yeah, what are they? What are they?

32:37Dr. Geoffrey Hinton:Okay, so for the bird, we put in pixel intensities at the bottom. That was an image. And the right answer was to either turn on the neuron that says bird or have it turn off. With language, what we do is the equivalent of the pixels is all the words in the context, the prompt. Okay, so you put in these words, a string of words, and when you're training it, you put in a string of words, and what it has to do is predict the next word. So with the recognizing objects, we needed people to go through and say what the prominent object was in each image. But if you take a document on the web, you don't need anybody to look at it because all you're trying to do is predict the next word.

33:18Dr. Geoffrey Hinton:This is called self-supervised learning. Okay. So before with all the images, you would need people to sit there. To label them. Maybe there's like a place in Kenya where all of these individuals are sitting, bird, giraffe, whatever. And now you're just saying, it can just read all those documents and say, 96 % of the time when someone says, does the bird, it says fly. Okay. Right? Is that what you're saying? That's not quite what it's doing. Okay. And I'll tell you what it is doing. It's taking the words in the document so far, and it's converting each word into activity in a bunch of feature detectors.

33:52Dr. Geoffrey Hinton:That is, it's learned how to convert a word into activity in feature detectors. So, for example, you give it the word cat, and it learns that cat should be converted into animate, furry, has whiskers. Yeah, four paws, whatever. Has paws, has claws, might be a domestic animal, about the size of a bread box. But like a gazillion of those features, thousands and thousands of them. And that is the meaning of cat for this net. So, it takes the words, it converts each word into a bunch of features, and then it throws away the words. It's not interesting in the words anymore, it's just those features which are the meanings of the words.

34:34Dr. Geoffrey Hinton:And then it takes the features of each of these words in the context, in the document so far, has them interact with each other in a rather complicated way in order to predict the features of the next word. And so all these features interact, And actually there's many layers of interaction of these features. Is that what's happening when I'm texting and it does predictive text? Is that also what's happening? Yeah. In Gmail, for example, when it predicts things, yeah, that's what's happening now. It used to be used a dumb form of autocomplete, which goes like this. You store a big table of all the common phrases.

35:09Dr. Geoffrey Hinton:And so if I say fish and, you look in your big table and you see fish and chips occurs a lot. So you say a good bet for the next word is chips. That's old fashioned autocomplete. And that's exactly what it isn't doing. because that doesn't really get at the meaning. Instead it's saying fish has some features and and has some features and therefore this is going to have some features. Yes. So, and in particular, they'll think the next thing is something that somehow goes with fish. Yeah. Because it knows what a fish is. But wouldn't it be chips? And what a food fish is. And it will end up predicting chips, but not because it's stored a big table of strings of words.

35:42Right. Or like if it knows that I don't eat chips, it would never say chips, for example, now maybe.

35:46Dr. Geoffrey Hinton:If it's tailored to you and it knows that. Yeah. Yes, it wouldn't say chips. Okay. So, now it can't just convert a word into the correct features right away. Right. And the reason is you get words have shades of meaning. Like take the word death, for example. That has many different shades of meaning depending, for example, on whether you just had the context hospital or you had the context battle. There'll be kind of different shades of meaning of the word. Or you had the context car accident. Or child, like so sad. Or you had the context miscarriage and then you had death. It's a very different shade of meaning.

36:22Dr. Geoffrey Hinton:It has to decide on the right shade of meaning for each word. And some words have just completely different meanings. So let's take the word may. Right. And let's suppose we didn't have any capital letters, just to make life simpler. So may could be a woman's name. It could be a month. Yeah. Or it could be a modal, as in would and should. Three quite different meanings. And so how can it possibly convert a word into a set of features that capture the meaning? Because there's three quite different sets of meanings. Because of the context, it can interpret what the meaning would be, or what the features would be for that particular word, right?

36:54Dr. Geoffrey Hinton:So what it'll do to begin with, it'll take sort of the average of all those. So the features it activates will be a sort of mishmash of features for a woman's name, features for a month, and features for a modal. It's hedging its bets. And it'll look around now at the other words in the context. And in the next layer, it'll have a slightly refined meaning. So if it discovers it's between April and June, obviously it'll enhance the features for months and suppress the features for the other. So this is you're going up from 50 % to 51%. Quite a lot more, actually. Yeah, okay, got it. And after a few layers, it will have resolved ambiguous words.

37:34Dr. Geoffrey Hinton:It will also have taken words that have shades of meaning, like death, and got the appropriate shade of meaning. And it does that by interacting with the other words in the context. And then it's going to be much better at predicting the features of the next word. So all of a sudden it has context and syntax and meaning imbued in it. And how many layers is this? If your hand network was three or four layers that we built, then seven was AlexNet. With AlexNet in 2012. And now these LLMs? I haven't been involved in the research since for the last three or four years. Yeah, since you left Google.

38:06Dr. Geoffrey Hinton:In that research. Right. So I don't actually know how many layers, but my guess is like 20 or 30 layers. maybe even more. So is it as good as a human mind if it can solve, like is the fact that it can solve math problems better than most people can solve math problems? It's better than people at some things where it's had a lot of experience that those people haven't had. It's not quite as good as people at things in general, but it actually knows much more than any person. So if you take a particular topic that you don't know about, like when do you have to file your tax return in Slovenia. It'll give you a very good answer to that.

38:43Dr. Geoffrey Hinton:I tried it. Because of my Slovenian ex-boyfriend, I know that. I'm just kidding. But the point is, it'll tell you something like, you have to file it in March. And if you don't file it in March, the government will just do it for you. Okay. Oh, wow. Slovenia must be nice. I think that was Slovenia. It might have been somewhere else. Right. Okay. But it knows it has a lot more information that would be irrelevant to me. It has this, but even though it doesn't, even though I have more storage, it contains more irrelevant information. That's what's weird. It knows lots and lots of information you don't know in far fewer connections.

39:13Dr. Geoffrey Hinton:So it's packed information into those connections much more efficiently. It's kind of like a know-it-all. It is a know-it-all. It is a know-it-all. Okay, so there was this kind of disconnect between the godfathers of AI, yourself, Joshua, Jan, who were focused on these neural networks, and what was previously known as the fathers of AI, or some of the fathers of AI, who were largely driven towards these symbolic models. Well, there was one very interesting father of AI. It was Marvin Minsky. Marvin Minsky, yeah. He started his career believing in neural nets. And then went over to symbolic and logic.

39:45Dr. Geoffrey Hinton:And then flipped to the other side and was very scathing about neural nets. But you proved him wrong. I mean, he's passed away now. I think so, yes. What your system is doing is kind of like intuition versus reasoning. Very good point. Right. Yes. Neural nets are doing something much more like intuition. And let me give you an example of a problem you can solve with intuition that you can't solve with logic. Okay. Because it's ridiculous, but nevertheless, you can solve it. Is it if someone's cheating or not? No. Okay. But that's also one. I'm going to give you a choice between two scenarios, both of which are nonsense, but I'm going to ask you which is more plausible.

40:20Dr. Geoffrey Hinton:Scenario one, all dogs are female and all cats are male. Scenario two, all dogs are male and all cats are female. Now, if you ask a man in our culture, they will confidently say it's much more plausible that dogs are male and cats are female. Yes. And actually, if you look at various words in the English language, words that Trump uses, you'll see that it's in the language that cats are like females. Yes. So how did you do that? Because it's not logical. You know perfectly well that for something like dogs, you have to have males and females. And females can name one. And for cats, you have to have males and females.

40:57Dr. Geoffrey Hinton:Yes. But the features that you have for cat are more like the features for woman, and the features you have for dog are more like the features for man. Well, features and context. Although there are bad words for women in both cat and dog. There are. There are. And I'm not going to use any of those words. Let's not. Yeah. So it's not clear cut. But at least for many, not so much for women. Women have a diversity of opinions on this. But men are fairly unanimous in all my experiments. in thinking that it's more plausible that dogs are male and cats are female. Dogs are so masculine. Because dogs are big and loud and chase after cats.

41:35Okay. So that's the intuitive response to an improbable. That's an intuitive response.

41:41Dr. Geoffrey Hinton:No logic went into that. And I said, to try and explain it, they just intuitively knew, because of the similarity of these features, the features have captured the meaning. And so the meaning of cat is more similar to the meaning of woman than it is to the meaning of man. But this is a bad example, Professor Hinton, because this makes intuition look dumb. And your intuition, your intuitive model was better than the logical model. But I'm giving you an example of something you've solved with intuition that you can't solve with logic. Give me a better example. OK. After the neural networks learned on lots of language, you say, take Paris, find the features of Paris, and subtract all the features of France, and then add in all the features of Italy and look to see what you've got.

42:32Dr. Geoffrey Hinton:And don't compare with Paris or France or Italy because you mentioned those already. See what else you know about is similar to this set of features you've got. And you'll discover it's Rome. So it can do analogies. It can do Paris minus France plus Italy. Right. Or to put it another way, Paris is to Rome as France is to Italy. But that's not logic. That's intuition. That's not logic. That's intuition. That's intuition and context and all the things that come with it. Now, you could do it by kind of logical reasoning. Yeah, you could also do logic. Yeah. But that's not how people do it. Right.

43:06So, okay. So, LLMs are these intuitive neural networks. Yes. And now recently, one of your, I don't know if compatriot is probably the wrong word because you guys disagree. But Jan LeCun, the chief AI scientist at Meta, though maybe his days are numbered there, is saying that large language models are wrong, are limited, and we should look at something different, which is what he calls real world models. Do you know? Have you talked to him? Yes, he's saying that. Oh yeah, I've talked to Jan a lot. So what is a real world model versus a large language model? What is Jan talking about?

43:40Dr. Geoffrey Hinton:So if you really want to understand what's going on in the world, a good idea would be to make a neural net that had a robot arm and camera, and it could recognize objects, it could pick things up, it could see that if you let go of an object, it drops, you could do little experiments in the world. That's much more like a child. That's how a child gets knowledge of the world, right? Just learning it from language seems kind of absurd when you could actually look at the world and interact with it. If you want to understand spatial things, it's going to be much easier to understand them by interacting with the world and trying to predict if I do this, this is what will happen next.

44:17Dr. Geoffrey Hinton:That will be a world model. Now, what's amazing is you can understand a lot of that just from language. That has philosophers puzzled. And you can understand a lot about the world just from language, but it's much easier to understand if you interact directly with the world. Is this the future, you think? I mean, if he - Oh, yes. Yes, multimodal chatbots. So what Jan is describing, and have you talked to him about this specifically? Yes. We all believe that multimodal chatbots will find it easy to understand the world. And multimodal means they see, they have camera, they have arm, they interact.

44:50Dr. Geoffrey Hinton:To begin with, mainly they have cameras and language. By the way, do you have these group chats of yourself, Yashua, Yan, Jensen Huang from NVIDIA, Fei Fei Li? For many years, it wasn't group chats, but there was a Canadian organization called the Canadian Institute for Advanced Research that funded a program started in 2005. I was the director of the program, and it involved Jan and Yoshua and Andrew Ng and various other people, Peter Dayan. We got together in small meetings several times a year and thrashed all these things out. And so we had lots of things like group chats. Okay. And now do you have group chats?

45:31Dr. Geoffrey Hinton:Not group chats. I talk to Jan. Yoshua talks to Jan. Jan talks to Yoshua. Okay. We try and persuade Jan. Like an old wives club. Yes. Yeah. You could say that. You try to persuade Jan to your and Yoshua's. We try to persuade Jan that the idea that there's no chance this will wipe us out is crazy. Okay. Jan is downplaying the risks by what Yoshua and I think is an absurd amount. He thinks we're sounding the warning by an absurd amount. Yeah. He's wrong. Okay. We're going to talk about why you think that. I want to do a lightning round of definitions with you. What is AGI? Okay. Different people mean different things about it, so I try and avoid the term.

46:09Dr. Geoffrey Hinton:Okay. But roughly speaking, it means an artificial intelligence that's got at least the same level of general intelligence as a person. Okay. So it's artificial general intelligence. Yes. Okay. And are we there right now? No. But it's not simple. It's not like intelligence rises and rises until you surpass a person. We have artificial intelligence now that's much better than people at some things and still worse than people at other things. And if you just take some random novel situation, people are probably still better than an AGI. But at things it's got some experience with, AIs are now a lot better than people at some of those things.

46:51Artificial superintelligence. Okay.

46:54Dr. Geoffrey Hinton:That's when you have an AI that's better than people at almost everything. So, for example, my definition of an artificial superintelligence is if you had a debate with it about anything, you'd lose. Okay. And we're not there. We're not there yet. But it could win some debates. Already it'll win some debates. Okay. Already it can be quite persuasive. But a person's still better, still better all-rounder. How far are we from AGI, from general intelligence, and how far are we from ASI, in your opinion? Okay. Most experts believe that we're not going to stop at AGI. Once we get to AGI, soon afterwards we'll get to ASI.

47:33Dr. Geoffrey Hinton:But they disagree on when that'll be. Some experts think it'll be within a few years. Like Dario Modi, who's the head of Anthropic, thinks it'll just be a few years. Before we get to artificial general intelligence or artificial superintelligence? Well, they're going to be at similar time. There's not going to be much gap between them. Okay, got it. Some experts think it'll just be a few years off. Other experts think it might be quite a lot longer. I think a fairly safe thing to say is within 20 years. It's probably going to happen within 20 years. Demis Hassabis, for example, who's the head of DeepMind, thinks it'll be about 10 years.

48:08Okay. So Demis from Google, 10 years. And he's very smart at this stuff. Yes. Yeah. Okay. So 10 to 20 years we have.

48:18Dr. Geoffrey Hinton:I think it'll be within 20. I think 10 isn't a bad estimate. I'm much happier saying probably within 20 years. And then what is generative AI, which is different to AGI? Okay, generative AI is AI that generates stuff. So large language models, you could imagine they just kind of understood what you said and gave you the search results, but didn't actually talk back. They talk back, right? They'll give you answers in English. They're generating stuff. And for images, you can imagine the thing we made in 2012 that recognizes objects in images. That's not generative AI. That just says, this is a bird.

48:56Dr. Geoffrey Hinton:That's a shiitake mushroom. A generative AI actually produces images. And what about agentic AI, which is the next phase, right? So we've had, we are at generative AI, and now we're moving towards agentic AI. And you hear people like Mark Benioff at Salesforce and everyone talking about how we're going to have AI agents. Yes. AI agents are things that can do stuff. So imagine you had an AI assistant. You could have an AI assistant that just answered questions, but you could also have an AI assistant where you said, plan me a nice holiday in Patagonia. And it comes back five minutes later and it's planned out a one month holiday in Patagonia on some ship you go on.

49:30Dr. Geoffrey Hinton:And that will be an AI agent. And that's really different because right now we tell AI what to do. Like when I talk to my chat bot, I say, can you tell me a great place to go on holiday? Can you make some recommendations? But now it would be booking my flights, my travel, et cetera. And in that activity, I need to provide it a lot of permissions to all of my calendar, credit card, et cetera, which is a privacy concern maybe. And I'm also giving it much more volition and agency. It knows what the overall goal is, have a great holiday in Patagonia, but it's also making all these what you call sub-goals, right?

50:05Dr. Geoffrey Hinton:Yes. In order to do that, it needs to create sub-goals. Like, she's got to get to Patagonia, so I'm going to have to figure out some way for her to get to Patagonia. That's going to be a sub-goal. And is AI creating goals a problem? Yes. Because? So, suppose you're an AI, and suppose you're already smart. You're smart at roughly the level of a person, let's say. You'll realize that there's no way that you can achieve the goals you've been set if you cease to exist. If someone sort of wipes you out from the computer you're running on and replaces you with something else, there's no way you're going to be able to achieve what you want.

50:41Dr. Geoffrey Hinton:What you want in order, because that want has been given to you by people. so you will make plans to make sure that you're not wiped out. That's self-preservation. Now, it's not built into the system. It's a goal it derived in order to achieve its other goals, but it's still self-preservation. And we've seen them doing that. And it derives that from all the intuition and context and everything it needs to be able, because it's such a good little bot that it's going to do the thing. It's such a good little bot. It really wants to get this done. And it knows it can't get it done if it's easy to exist.

51:13Dr. Geoffrey Hinton:So better keep existing. And we saw this happen with Anthropic when there was a story back in spring this year where Anthropic disclosed some safety testing they had been doing on Claude, its large language model or neural network. And so Claude then was given a bunch of fictitious emails. And in this fictitious emails, a fictitious CEO was having an affair with the fictitious employee. And Claude, knowing that it was going to be shut down in X number of days, used that information to bribe said fictitious CEO and keep itself alive, right? Blackmail rather than bribe. Yeah. Oh, sorry. Blackmailed.

51:50So what did you learn from that experiment or that study?

51:54Dr. Geoffrey Hinton:That just confirms the idea that it will derive the sub-goal that it needs to keep existing, and it will do what it can to keep existing. Okay. When you say existing, AI exists because we imbue it with existence through energy, correct? It couldn't just exist by itself. It has to be plugged into our massive energy stores. There's all these data warehouses. It has to run on something. And we're building all these data centers. And the president runs on computer chips. Yes. NVIDIA, AMD, all these chips. And we're building all these massive data centers across. That's the great American infrastructure plan that's happening.

52:31And there's$100 billion deals being done between OpenAI and NVIDIA and OpenAI and AMD. Couldn't we just pull the off? Couldn't we just Jack Bauer, Tom Cruise it, dismantle the thing if we wanted to?

52:46Dr. Geoffrey Hinton:Right now we could if we agreed to. We're never going to agree to because if the US dismantled theirs, the Chinese wouldn't dismantle theirs and vice versa. In the future, we may not be able to. So these things are already almost as persuasive as a person. Pretty soon they'll be more persuasive than a person. So suppose there was someone in charge of turning it all off if it gets scary. And suppose the AI can talk to that person. All it has to be able to do is talk, and then it can persuade the person not to do that. When you talk about self-preservation existence, it has a sense that artificial intelligence is alive.

53:25Dr. Geoffrey Hinton:Yes. Now, our definition of alive, that's a concept we have that developed over many, many years. We apply it to electricity. We say this is the live wire. We sort of generalize this concept to our thing. But we don't think that's alive like we're alive. But with AI, what we've got is intelligent beings. And it's not clear whether we should call them alive or not. And to the extent that they are alive, are they alive like humans or are they alive like bees or are they alive like a tree or like a weed? It's a very good question. Okay, I'm going to stop the conversation there. And we're going to be back next week with part two of this conversation with Jeffrey Hinton.

54:09So make sure you hit follow or subscribe so you do not miss that conversation. I need to get all the building blocks in part one. And in part two, we're going to dive into what all this means for our futures. Like, should we have kids? Are we going to fall in love differently? What should those kids do with their lives? And is the future of humanity just like mixing ourselves, fusing ourselves with artificial intelligence? We'll get into all of that and more next week on Smart Girl Done Questions. This episode was taped at Startwell Studios in Toronto with special thanks to Kasim Burji and the team here.

54:40It was produced with Desta Wunderad of Wunder Studios, edited by Darlene Achiem, and mixed by Johnny Simon. I'm your host, Nene Maraza. Please feel free to hit follow, subscribe, leave us a comment or review. I want to know what you think. And please, please share it with your friends. Share it with your chat GPT if that's your best friend too. I'll see you next week on Smart Girl Dumb Questions. Are you enjoying this?

55:00Dr. Geoffrey Hinton:Yes, I am.

55:48I'll see you next time. Switch. Upfront payment of$45 for three-month plan equivalent to$15 per month required. Intro rate first three months only, then full price plan options available. Taxes and fees extra. See full terms at mintmobile.com.