AI: What Could Go Wrong? with Geoffrey Hinton — The Weekly Show with Jon Stewart

[00:00:00] John Stewart: am i in neural learning 201 yet or am i still in 101 you're like the smart student in the [00:00:05] Jeffrey Hinton: front row who doesn't know anything but ask these good questions that's that's the nicest [00:00:11] John Stewart: way i've ever been described thank you hey everybody welcome to the weekly show uh podcast my name is john stewart i'm going to be hosting you today and it's a what is it wednesday october eighth uh i don't know what's going to happen later on in the day but uh we're going to be out tomorrow but today's episode i i just want to say very quickly today's episode we are talking to someone known as the godfather of ai a gentleman by the name of jeffrey hinton who has been developing the type of technology that has turned into ai since the 70s and uh i want to let you know so we we talk about it the first part of it though he he gives us this breakdown of kind of what it actually is which for me was unbelievably helpful we get into the uh it will kill us all part but uh it was important uh for my understanding to sort of set the scene so i i hope you find that part as interesting as i did because man uh it it expanded my understanding of what this technology is of how it's going to be utilized of what some of those dangers might be in a really interesting way so i don't i will not hold it up any longer let us get to uh our guests for this podcast ladies and gentlemen we are absolutely thrilled today to be able to welcome professor emeritus with the department of computer science at the university of toronto and schwartz reisman institute's advisory board member jeffrey hinton is joining us sir thank you so much for being with us today well thank you so much for inviting me uh i'm delighted i you are known as and and i'm sure you will uh be very demure about this the godfather of artificial intelligence uh for your work on uh sort of these neural networks uh you you co-won the actual nobel prize in physics in in 2024 [00:02:22] Jeffrey Hinton: for this work is is is that correct that is correct it's slightly embarrassing since i don't do physics so when they called me up and said you won the nobel prize in physics i didn't believe them to begin with [00:02:34] John Stewart: and and were the other physicists going wait a second okay that guy's not even in our business i strongly suspect they were but they didn't do it to me oh good i'm glad uh this is going to seem somewhat remedial i'm sure to you but when we talk about artificial intelligence i'm not exactly sure what it is that we're talking about i know there are these things large language models i i know to my experience artificial intelligence is just a slightly more flattering search engine whereas i used to google something and it would just give me the answer now it says what an interesting question you've asked me so what are we talking about when we talk about [00:03:24] Jeffrey Hinton: artificial intelligence so when you used to google it would use keywords and it would have done a lot of work in advance so if you gave it a few keywords it could find all the documents that have those words [00:03:36] John Stewart: in so basically it's it's just a it's sorting it's looking through and it's sorting and finding words and then bringing you a result yeah that's how it used to work okay but it didn't understand what [00:03:50] Jeffrey Hinton: the question was so it couldn't for example give you documents that didn't actually contain those words [00:03:57] John Stewart: but were about the same subject it didn't make that connection oh right because it would say here is your result minus and then it would say like a word that was not included right but if you [00:04:09] Jeffrey Hinton: had a document with none of the words you used it wouldn't find that even though it might be a very relevant document about exactly the subject you were talking about it had just used different words now it understands what you say and it understands in pretty much the same way people do what so if i it'll [00:04:28] John Stewart: say oh i know what you mean let me let me let me educate you on this so it's gone from being kind of a uh literally just a search and find thing to an actual almost an expert in whatever it is that you're discussing and it can bring you things that you might not have thought about yes so the large [00:04:53] Jeffrey Hinton: language models are not very good experts at everything so if you take take some friend you have who knows a lot about some subject matter no i got a couple of those yeah they're probably a bit better than the large language model but they'll nevertheless be impressed that the large language [00:05:10] John Stewart: model knows their subject pretty well what is so what is the difference between sort of machine learning so was was google in terms of a a search engine machine learning that's just algorithms and [00:05:26] Jeffrey Hinton: and predictions no not exactly machine learning is a kind of coverall term for any system on a computer that learns okay now these neural networks are a particular way of doing learning that's very different from what was used before okay now these are these are the new neural [00:05:46] John Stewart: networks the old machine learning those were not considered neural networks and when you say neural networks meaning your work was sort of the genesis of it was in the 70s where you thought you were studying the brain is that correct i was trying to come up with um ideas about how the brain [00:06:06] Jeffrey Hinton: actually learned and there's some things we know about that it learns by changing the strengths of [00:06:11] John Stewart: connections between brain cells wait that so explain that what it says it it learns by changing the connections so if if uh you show a human something new brain cells will it will actually make new connections [00:06:28] Jeffrey Hinton: within brain cells it won't make new connections there'll be connections that were there already okay but the main way it operates is it changes the strength of those connections wow so if you think of it from the point of view of a neuron in the middle of the brain a brain cell okay um all it can [00:06:47] John Stewart: do in life is sometimes go ping that's all he's got that's his only that's all it's got all it's got is [00:06:53] Jeffrey Hinton: it can unless it happens to be connected to a muscle okay it can sometimes go ping okay and it has to decide when to go ping oh wow how does it decide when to go ping i i was glad you asked that question um there's other neurons going ping okay and when when it sees particular patterns of other neurons going ping it goes ping and you can think of this neuron as receiving pings from other neurons and each time it receives a ping it treats that as a number of votes for whether it should turn on or should should go ping or should not go ping and you can change how many votes another neuron has for it how would you how would you change that vote by changing the strength of the connection the strength of the connection think of as the number of votes this other neuron gives for you to go ping okay so it [00:07:47] John Stewart: really is in some respects it's a boy it reminds me of the movie minions but it's it's almost a social [00:07:55] Jeffrey Hinton: yes yes it's it's it's it's very like political coalitions there'll be groups of neurons that go ping together okay and the neurons in that group will all be telling each other go ping and then there might be a different coalition and they'll be telling other neurons don't go ping oh my god and then there might be a different coalition right and they're all telling each other to go ping and [00:08:15] John Stewart: telling the first coalition not to go ping all this is going on in your brain yes in the way of like [00:08:21] Jeffrey Hinton: i would like to pick up a spoon yes so spoon for example spoon in your brain yeah is a coalition of [00:08:28] John Stewart: neurons going ping together and that's a concept oh wow so so as you're teaching when you're when you're a baby and they go spoon there's a little group of neurons going oh that's a spoon and they're strengthening their connections with each other so whatever is that why when you know you're you're imaging brains you see certain areas light up and is is that lighting up of those areas the neurons that ping [00:09:00] Jeffrey Hinton: for certain items or actions not not exactly ah getting close i'm getting close it's close different areas will light up when you're doing different things like when you're doing vision or talking or controlling your hands different areas light up for that okay um but the coalition of neurons that goes ping that go ping together when there's a spoon they don't only work for spoon most of the members of that coalition will go ping when there's a fork so they overlap a lot these coalitions this is a big [00:09:38] John Stewart: tent it's a big tent coalition i love thinking about this as political i had no idea your brain operates [00:09:45] Jeffrey Hinton: on peer pressure there's a lot of that goes on yes and concepts are kind of coalitions that are happy together but they they overlap a lot like the concept for dog and the concept for cat have a lot in common they'll have a lot of shared neurons in particular the neurons that represent things like this is animate or this is hairy or this might be a domestic pet all those neurons will be in common to cat and [00:10:12] John Stewart: dog are there can i ask you that and again i so appreciate your patience with this and explain this is this is really helpful for me are there certain neurons that ping broadly right for the broad concept of animal and then other neurons like does it work from macro to micro from general to specific so you have a coalition of neurons that ping generally and then as you get more specific with the knowledge does that engage uh certain ones that will ping less frequently but for maybe more specificity is is [00:10:54] Jeffrey Hinton: that something okay that's a very good theory nobody knows no nobody nobody really knows for sure about this oh that's a very sensible theory and in particular there's going to be some neurons in that coalition that ping more often for more general things right and then there may be neurons that [00:11:14] John Stewart: ping less often um for much more specific things right okay and and this works throughout and like you say there's certain areas that will ping for vision or other senses touch uh i imagine there's a ping system for language uh and and and and you were saying what if we could get computers which were much more i would think just uh binary if then you know sort of basic you're saying could we get them to work as [00:11:48] Jeffrey Hinton: these coalitions yeah i don't think binary if then has much to do with it okay the difference is people were trying to put rules into computers they were trying to figure out so the basic way you program a computer is you figure out in exquisite detail how you would solve the problem oh you deconstruct all the steps and then you tell the computer exactly what to do that's a normal computer program okay great these [00:12:17] John Stewart: things aren't like that at all so you were trying to change that process to see if we could create a process that was that functioned more like how the human brain would rather than a item by item instruction list you wanted it to to think more more more globally how did how did that occur so it was sort of [00:12:42] Jeffrey Hinton: obvious to a lot of people that the brain doesn't work by someone else giving you rules and you just execute those rules i mean in north korea they would love brains to work like that but they don't [00:12:58] John Stewart: you're saying that in an authoritarian world that is how brains would operate well that's how they would like them to operate that's how they would like them to operate it's a little more artsy than [00:13:08] Jeffrey Hinton: that yes all right fair enough um we do write programs for neural nets but the programs are just to tell the neural net how to adjust the strength of the connection on the basis of the activities of the neurons so that's a fairly simple program right that doesn't have all sorts of knowledge about the world in it is just what are the rules for changing neural connection strengths on the basis of the activities [00:13:35] John Stewart: can you give me an example so would that be considered sort of is that machine learning or is [00:13:40] Jeffrey Hinton: that deep learning what would that's deep learning if you have a network with multiple layers it's [00:13:46] John Stewart: called deep learning because there's many layers so what are you saying to a computer when you are trying to get it to do deep learning like what would be an example of an instruction that you would give okay so let me go ah an hour all right am i am i yet am i in neural learning 201 yet or am i still in 101 [00:14:10] Jeffrey Hinton: you're like the smart student in the front row who doesn't know anything but ask these good questions [00:14:17] John Stewart: that's that's the nicest way i've ever been described thank you if you're still overpaying for your wireless i want you to leave this country i want you gone there's no excuse mint mobile her favorite word is no it's time to say yes to saying no no contracts no monthly bills no overages no bs here's why so many said yes to making the switch and getting premium wireless for fifteen dollars a month my god i spend that on chicklets chicklets i say ditch overpriced wireless and their jaw-dropping monthly bills unexpected overages and fees plants started fifteen dollars a month it meant all plants come with high-speed data and unlimited talk and text delivered on the nation's largest 5g network use your own phone with any mint mobile plan bring your phone number along with all your existing contacts ready to say yes to saying no make the switch at mint mobile dot com slash tws that's mint mobile dot com slash tws up front payment of 45 required equivalent to 15 a month limited time new customer offer for first three months only speeds may slow above 35 gigabytes on unlimited plan taxes and fees extra see mint mobile for details [00:15:42] Jeffrey Hinton: so let's go back to 1949 oh boy all right so here's a theory from someone called donald heb about how you change connection strengths okay if neuron a goes ping and then shortly afterwards neuron b goes ping increase the strength of the connection [00:16:01] John Stewart: okay that's a very simple rule that's called the heb rule right the heb rule is if neuron a goes ping increase the connection uh and b goes ping increase that connection yes okay um now as soon as computers [00:16:15] Jeffrey Hinton: came along you should do computer simulations um people discovered that rule by itself doesn't work what happens is all the connections gets very strong and all the neurons go ping all at the same time and you have a seizure oh okay that's a shame isn't it that is a shame there's got to be something that makes connections weaker as well as making them stronger right there's got to be some discernment yes okay if i can digress for about a minute boy i'd like that okay suppose we wanted to make a neural network that have multiple layers of neurons and it's to decide whether an image contains a bird [00:16:51] John Stewart: or not like a captcha like when you go on and it said yeah exactly we want to this is okay we want [00:16:59] Jeffrey Hinton: to solve that capture with a neural net okay so the input to the neural net the sort of bottom layer of neurons is a bunch of neurons and they go ping to different levels of they have different strengths of ping and they represent the intensities of the pixels in the image okay so if it's a thousand by thousand image you've got a million neurons they're going ping at different rates to represent how intense each pixel is okay that's your input now you've got to turn that into a decision is this a bird or not [00:17:36] John Stewart: wow so that decision so let me ask you a question then do you program in because strength of pixel doesn't strike me as a really useful tool in terms of figuring out if it's a bird figuring out if it's a bird seems like the tool would be are those feathers is that a beak uh is that a crest yeah here it goes [00:18:02] Jeffrey Hinton: so the pixels by themselves yeah don't really tell you whether it's a bird okay because you can have birds that are bright and birds that are dark and you can have birds flying and birds sitting down and you have an ostrich in your face when you have a seagull in the distance they're all birds okay so what do you do next well sort of guided by the brain what people did next was said um let's have a bunch of edge detectors so what we're going to do because of course you can recognize birds quite well in line drawings right so what we're going to do is we're going to make some neurons a whole bunch of them that detect little pieces of edge that is little places in the image where it's bright [00:18:44] John Stewart: on one side and darker on the other side right so it's it's almost creating a like primitive form of [00:18:51] Jeffrey Hinton: vision this is how we you make a vision system yes this is how it's done in the brain and how it's done in computers wow okay so if you want to detect a little piece of vertical edge in a particular place in the image let's suppose you look at a little column of three pixels and next on another column of three pixels and if the ones on the left are bright and the ones on the right are dark you want to say yes there's an edge here so you have to ask how would i make a neuron that did that oh my god [00:19:24] John Stewart: okay all right i'm going to jump ahead all right so the first thing you do is you have to teach the the the network what vision is so you're teaching it these are images this is background this is form this is edge this is not this is bright this is so you're teaching it almost how to see in the old [00:19:46] Jeffrey Hinton: days people would try and put in lots of rules to teach you how to see and explain to what foreground was and what background was okay but um the people who really believed in neural nets said no no don't put in all those rules let it learn all those rules just from data and the and the way it learns is by [00:20:04] John Stewart: strengthening the pings once it it starts to uh recognize edges and things we'll come to that in [00:20:11] Jeffrey Hinton: a minute i'm jumping ahead you're jumping ahead all right so let's carry on with this little bit of edge detector okay so you have in the first layer you have the neurons that represent how bright the pixels are right and then in the next day we're going to have little bits of edge detector and so you might have a neuron in the next layer that's connected to a column of three pixels on the left and a column of three pixels on the right and now if you make the strengths of the connections to the three pixels on the left strong big positive connections right because it's brighter and you make the strengths of connections to the three pixels of the right be big negative connections because it's darker they say don't turn on right then when the pixels on the left and the pixels on the right are the same brightness as each other the negative connections will cancel out the positive connections and nothing will happen huh but if the pixels on the left are bright and the pixels on the right are dark the neuron will get lots of input from the pixels on the left because they're big positive connections right it won't get any inhibition from the pixels on the right because then that those pixels are all turned off right right and so it'll go ping it'll say hey i found what i wanted i found that the three pixels on the left are bright and the three pixels on the right are not bright hey that's my thing i found a little piece of positive a little piece of edge here i'm that guy i'm the edge guy i ping on the edges right and that pings on that particular piece of edge okay okay now imagine you [00:21:47] John Stewart: have like a gazillion of those i'm already exhausted on the three pings i you have a gazillion of those [00:21:58] Jeffrey Hinton: because they have to detect little pieces of edge anywhere on your retina anywhere in the image and at any orientation you need different ones for each orientation right and you actually have different ones for the scale there might be an edge at a very big scale that's quite dim right and there might be little sharp edges at a very small scale and as you make more and more edge detectors you get better and better discrimination for edges you can see smaller edges you can see the orientation of edges more accurately okay you can detect big vague edges better so let's now go to the next layer so now we've got our edge detectors right now suppose that we had a neuron in the next layer that looked for a little combination of edges that is almost horizontal several edges in a row that are almost horizontal right and and line up with each other and just slightly above those several edges in a row that are again almost horizontal but come down to form a point with the first sort of edges right so you find two little combinations of edges that make a sort of [00:23:17] John Stewart: pointy thing okay so you're a nobel prize-winning physicist i did not expect that sentence to end with it makes kind of a pointy thing i thought there'd be a name for that but i get what i get what you're saying you're you're now discerning where it ends where it you're sort of looking at uh and this is before you're even looking at color or anything else this is literally just is there an image what are [00:23:46] Jeffrey Hinton: the edges what are the edges and what are the little combinations of edges so we're now asking is there a little combination of edges that makes something that might be a beak wow that's the pointy [00:23:58] John Stewart: thing but you don't know what a beak is yet not yet no we're gonna we need to learn that too yes right so once you once you have the system it's almost like you're building systems that can mimic the human senses that's exactly what we're doing yes so vision ears not smell obviously although i [00:24:20] Jeffrey Hinton: know they're doing that now they're starting on smell now oh for god's sakes and probably touch they've now got to digital smell where you can transmit you can transmit smells over the web it's just that's just insane the printer for smells has 200 components instead of three colors it's got 200 components and it synthesizes the smell at the other end and it's not quite perfect but it's pretty good [00:24:45] John Stewart: wow so this is this is incredible to me okay so i am so sorry about this i apologize no this is usually [00:24:57] Jeffrey Hinton: this is perfect you're doing a very good job of representing a sort of sensible curious person who doesn't know anything about this um so let me finish describing how you build the system by hand yes so if i did it by hand i'll start with these edge detectors and so i'd say make big strong positive connections from these pixels on the left and big strong negative connections to the pixels on the right right and now the neuron that gets those incoming connections that's going to detect a little piece of vertical edge okay and then at the next layer i'd say okay make big strong positive connections from three little bits of edge sloping like this and three little bits of edge sloping like that could be a beak and a pointy thing and this is a potential beak right and in that same layer i made mike also make big strong positive connections from a combination of edges that roughly form a circle and that's a potential eye right right right now in the next layer i have a neuron that looks at possible beaks and looks at possible eyes and if they're in the right relative position it says hey i'm happy because that neuron has detected a possible bird's right and that guy might ping and that guy would ping at the same time there'll be other neurons elsewhere that have detected little patterns like a chicken's foot or the feathers at the end of the wing of a bird right and so you have a whole bunch of these guys now even higher up you might have a neuron that says hey look if i've detected a bird's head and i've detected a chicken's foot and i've detected the end of a wing it's probably a bird so it's a bird right so you can see now how you might try and wire all that up by hand [00:26:44] John Stewart: yes and it would take some time it would take like forever it would take like forever yes [00:26:52] Jeffrey Hinton: okay so suppose you were lazy yes now you're talking okay what you could do is you could just make these layers of neurons without saying what the strengths of all the connections ought to be you just start them off with small random numbers just put in any old strengths and you put in a picture of a bird and let's suppose it's got two outputs one says bird and the other says not bird right with random connection strengths in there what's going to happen is you put in a picture of a bird and it says 50 bird 50 not bird in other words i haven't got a clue right and you put it in picture of a non-bird and it says 50 bird 50 non-bird oh boy okay so now you can ask a question suppose i were to take one of those connection strengths and i were to change it just a little bit make it maybe a little bit stronger instead of saying 50 bird would it say 50.01 bird and 49.99 non-bird and if it was a bird then that's a good change to make you've made it work slightly better what year was this when did this start oh exactly so this is just an idea this would never work but right with me all right this is like one of those defense lawyers who goes off on a huge digression but [00:28:24] John Stewart: it's all going to be good in the end no no no no no this is this is helpful and this is the thing that's going to kill us all in 10 years yep um [00:28:37] Jeffrey Hinton: when i say yeah i mean not this particular thing but an advancement on it but this is not necessarily [00:28:44] John Stewart: kill us all but maybe right right right this is oppenheimer going uh okay so you've got an object and that is made up of uh smaller objects and like this is the very early [00:28:58] Jeffrey Hinton: part of this okay so suppose you had all the time in the world what you could do is you could take this layered neural network and you could start with random connection strengths and you could then show it a bird and they just say 50 bird 50 non-bird and you could pick one of the connection strengths right and you could say if i increase it a little bit does it [00:29:22] John Stewart: help right it won't help much but does it help at all right will it get me to 50.1 50.2 that kind of [00:29:29] Jeffrey Hinton: thing okay if it helps make that increase okay and then you go around and do it again maybe this time we choose a non-bird and we choose one connection strength right and we'd like it to if we increase that connection strength and it says it's less likely to be a bird more likely to be a non-bird we say okay that's a good increase let's do that one right right right now here's a problem there's a trillion connections yeah right okay and each connection has to be changed many times and is that manual well in this way of doing it will be manual and not just that but you can't just do it on the basis of one example because sometimes change can make a connection strength if you increase it a bit it'll help with this example but it'll make other examples worse oh dear god so you have to give it a whole batch of examples right and see if on average it helps and that's how you create these large language if we did it this really dumb way to create let's say this vision system for now yes we'd have to do trillions of experiments and each experiment would involve giving it a whole batch of examples and seeing if changing one connection strength helps or hurts oh god and it [00:30:46] John Stewart: would never be done it would never be infinite it would be infinite okay now suppose that you figured [00:30:54] Jeffrey Hinton: out how to do a computation that would tell you for every connection strength in the network up it tell you at the same time for this particular example let's suppose you give it a bird and it says 50 percent bird and now for every single connection strength all trillion of these connection strengths we can figure out at the same time whether you should increase them a little bit to help or decrease them a little bit to help i mean then you change a trillion of them at the same time [00:31:26] John Stewart: can i can i say a word that i've been dying to say uh this whole time eureka [00:31:31] Jeffrey Hinton: eureka eureka eureka now that's that computation for normal people it seems complicated um yes if you've done calculus it's fairly straightforward and many different people invented this computation right um it's called backpropagation so now you can change your trillion at the same time and you'll [00:31:52] John Stewart: go a trillion times faster oh my god and and that's the moment that it goes from theory to practicality [00:32:03] Jeffrey Hinton: that is the moment when you think eureka we've solved it we know how to make smart systems for us that was 1986 and we were very disappointed when it didn't work every day the loudest the most inflammatory [00:32:30] John Stewart: takes dominate our attention and and the bigger picture gets lost it's all just noise and no light ground news puts all sides of the story in one place so you can see the context they provide the light it starts conversations beyond the noise they aggregate and organize information just to help readers make their own decisions you can see how many news outlets have reported on the story whether it's uh under-reported over-reported by one side or the other side or whatever side of the political spectrum ground news provides users reports that easily compare headlines or reports that give a summarized breakdown of the specific differences in reporting across all the spectrum it's a great resource go to groundnews.com/stewart and subscribe for 40 off the unlimited access vantage subscription brings the price down to about five dollars a month instead groundnews.com/stewart or scan the qr code on the screen you've been in that room for 10 years you've been showing it birds you've been increasing the strengths you had your eureka moment and you flipped the switch and went [00:33:44] Jeffrey Hinton: no here's the problem yeah here's the problem it only works or it only works really impressively well much better than anywhere any other way of trying to do vision if you have a lot of data and you have a huge amount of computation even though you're a trillion times faster than the dumb method it's [00:34:03] John Stewart: still going to be a lot of work okay so now you've got to increase your the data and you've got to increase your computation power yes and you've got to increase the computation power by a factor of about [00:34:19] Jeffrey Hinton: a billion compared with where we were and you've got to increase the data by a similar factor you are [00:34:25] John Stewart: still in 1986 when you figure this out you are a billion times not there yet something like that yes what would have to change to get you there the the power of the the chip the what what changes okay [00:34:41] Jeffrey Hinton: it may be more like a million a factor of a million okay okay i don't want to exaggerate here [00:34:46] John Stewart: no because i'll catch you if you try and exaggerate you will i'll be on it a million is quite a lot [00:34:52] Jeffrey Hinton: yes so here's what has to change the area of a transistor has to get smaller so you can pack more of them on a chip so between 1972 when i started on this stuff okay and now the area of a transistor [00:35:07] John Stewart: has got smaller by a factor of a million wow so that's can i relate this to so that is around the age that i remember my father worked at rca labs and when i was like eight years old he brought home a calculator and the calculator was the size of a desk and it added and subtracted and multiplied by 1980 you could get a calculator on a pen and is that based on that yeah the transistors based on [00:35:41] Jeffrey Hinton: large-scale integration using small transistors yeah okay all right all right so the the area of a transistor decreased by a factor of a million okay and the amount of data available increased by much more than that because we got the web and we got digitization of massive amounts of data oh so they [00:36:00] John Stewart: worked hand in hand so as the chips got better the data got more vast and you were able to feed more information into the model while it was able to increase its processing speed and abilities [00:36:15] Jeffrey Hinton: yeah so let me summarize what we now have yes you set up this neural network for detecting birds and you give it lots of layers of neurons but you don't tell it the connection strength you say start with small random numbers right and now all you have to do is show it lots of images of birds and lots of images that are not birds tell it the right answer so it knows the discrepancy between what it did and what it should have done send that discrepancy backwards through the network so it can figure out for every connection strength whether it should increase it or decrease it and then just sit and wait for a month and at the end of the month if you look inside if you look inside here's what you'll discover yeah it has constructed little edge detectors and it has constructed things like little beak detectors and little eye detectors and it will have constructed things that it's very hard to see what they are but they're looking for little combinations of things like beaks and eyes and then after a few layers it'll be very good at telling you whether it's a bird or not it made all that stuff up from the data oh my god can i say this again eureka eureka we figured out we don't need to hand wire in all these little edge detectors and beak detectors and eye detectors and chicken's foot detectors that's what computer vision did for many many years and it never worked that well we can get the system just to learn all that all we need [00:37:51] John Stewart: to do is tell it how to learn and that is in 1980 something in 1986 we figured out how to do that [00:37:59] Jeffrey Hinton: right people were very skeptical because we couldn't do anything very impressive right because we didn't [00:38:04] John Stewart: have enough data and we didn't have enough computation this is this is incredible uh the way and i i can't thank you enough for explaining what that is it makes everything you know i'm so accustomed to an analog world of you know uh how things work and like the way that cars work but i have no idea how uh our digital world functions and that is the clearest explanation for me that i have ever gotten and i cannot thank you enough it makes me understand now how this was achieved and by the way what what uh jeffrey is talking about is the the primitive version of that what's so incredible to me is the each upgrade of [00:38:50] Jeffrey Hinton: that the the vastness of the improvement yes of that so let me let me just say one more thing please i don't want to be too professor like but no no no no no but um how does this apply to large language models yes well here's how it works for large language models you have some words in a context so let's suppose i give you the first few words of a sentence right what the neural net's going to do is learn to convert each of those words into a big set of features which is just active neurons neurons going pink okay so if i give you the word tuesday there'll be some neurons going ping if i give you the word wednesday it'll be a very similar set of neurons slightly different but a very similar set of neurons going ping because they mean very similar things now after you've converted all the words in the context into neurons going ping into whole bunches that capture their meaning these neurons all interact with each other what that means is neurons in the next layer look at combinations of these neurons just as we looked at combinations of edges to find a beak and eventually you can activate neurons that represent the features of the next word in the sentence it will anticipate it can anticipate it can predict the next word so the way you train it is that why [00:40:18] John Stewart: my phone does that it always thinks i'm about to say this next you know uh uh word and i'm always [00:40:24] Jeffrey Hinton: like stop doing that yeah because a lot of times it's wrong it's probably using neural nets to do it [00:40:29] John Stewart: yes right and of course you can't be perfect at that so this is so now to put it together [00:40:36] Jeffrey Hinton: you've taught it almost how to see you can teach you to see in the same way you can teach you how [00:40:41] John Stewart: to predict the next word right so it sees it goes that's the letter a now i'm starting to recognize letters then you're teaching it words and then what those words mean and then the context and it's all being done by feeding it our previous words by back propagating all the writing and speaking that [00:41:04] Jeffrey Hinton: we've done already it's looking over you take some document that we produced yes you give it the context which is all the words up to this point yes and you ask it to predict the next word and then you look at the probability it gives to the correct answer right and you say i want that probability to be bigger i want you to have more probability of making the correct answer right so it doesn't understand it [00:41:34] John Stewart: this is merely a statistical exercise we'll come back to that you take you take the discrepancy between [00:41:42] Jeffrey Hinton: the probability it gives for the next word and the correct answer yeah and you back propagate that through this network and it'll change all the connection strengths so next time you see that that lead in it'll be more likely to give the right answer now you just said something that many people say this isn't understanding this is just a statistical trick yes that's what chomsky says for example yes [00:42:11] John Stewart: chomsky and i we're always stepping on each other's sentences yeah so let me ask you the question [00:42:18] Jeffrey Hinton: well how do you decide what word to say next me you it's interesting i'm glad you brought this up so [00:42:25] John Stewart: what i do is i look for sharp lines and then i try and predict no i have no idea how i how i do that i honestly i wish i knew it would save me a great deal of embarrassment if i knew how to stop some of the things that i'm saying that come out next if i had a better predictor boy i could save myself quite a [00:42:47] Jeffrey Hinton: bit of trouble so the way you do it is pretty much the same as the way these large language models do it right you have the words you've said so far those words are represented by sets of active features so the word symbols get turned into big patterns of activation of features neurons going ping different pings different strengths and these neurons interact with each other to activate some neurons that go ping that are representing the meaning of the next word or possible meanings of the next word and from those you kind of pick a word that fits in with those features that's how the large language models generate text and that's how you do it too you're very they're very like us so it's i'm i'm ascribing to myself a [00:43:36] John Stewart: humanity of understanding for instance if i so like let's say the little white lie i'm with somebody and they ask me a question and in my mind i know uh what to say but then i also think oh but saying that might be coarse or it might be rude or i might offend this person so i'm also though making emotional decisions on what the next words i say are as well it's not just a objective process there's a subjective [00:44:10] Jeffrey Hinton: process within that all of that is going on by neurons interacting in your brain it's all pings and [00:44:17] John Stewart: it's all strength okay even the things that i ascribe to a moral code or an emotional intelligence are [00:44:25] Jeffrey Hinton: still pings they're still all pings and you need to understand there's a difference between what you do kind of automatically and rapidly and without effort and what you do with effort and [00:44:40] John Stewart: slower and consciously and deliberatively and you're saying that can be built into these models as well [00:44:46] Jeffrey Hinton: that can also be done with pings that can be done by these neural nets [00:44:52] John Stewart: but there is the suggestion then that with enough data and enough processing power their brains can function identically to ours are they are they are they at that point will they get to that point will they be able to because i'm assuming we're still ahead processing wise okay um [00:45:23] Jeffrey Hinton: they're not exactly like us but then the point is they're much more like us than standard computer software is like a standard computer software right someone programmed in a bunch of rules and if it follows the rules it does what they expect it to that's right so you're saying this is the difference this is just a different kettle official together right and it's much more like us now as you're [00:45:42] John Stewart: doing this and you're in it and i imagine the excitement is even though it's occurring over a long period of time you're seeing these improvements occur over that time and it must be uh incredibly fulfilling and interesting and and you're watching it explode into this sort of artificial intelligence and generative ai and all these different things at what point during this process do you step back and [00:46:12] Jeffrey Hinton: go um wait a second okay so i did it too late i should have done it earlier i should have been more aware earlier but i was so entranced with um making these things work and i thought it's going to be a long long time before they work as well as us we'll have plenty of time to worry about what if they try and take over and stuff like that right um at the beginning of 2023 after gpt had come out but also seeing similar chatbots at google before that right and because of some work i was doing on trying to make these things analog i realized that neural nets running on digital computers are just a better form of computation than us and i'll tell you why they're better yeah why because they can share better they can share with each other better yes so if i make many copies of the same neural net and they run on different computers each one can look at a different bit of the internet so i've got a thousand copies they're all looking at different bits of the internet each copy is running this backpropagation algorithm and figuring out given the data i just saw how would i like to change my connection strengths now because they started off as identical copies they can then all communicate with each other and say how about we all change our connection strengths [00:47:42] John Stewart: by the average of what everybody wants but if they were all trained together wouldn't they come up with the same answer why are they coming up with different answers yes but they're looking at different data [00:47:54] Jeffrey Hinton: they're looking at different data oh on the same data they would give the same answer if they look at different data they have different um ideas about how they'd like to change their connection [00:48:06] John Stewart: strengths to absorb that data but are they also creating data is that so they're looking at the same and they're at this point it's all about discernment getting these things to discern better to understand better to do all that but there's another layer to that which is iterative yes once you're [00:48:27] Jeffrey Hinton: good at once you're good at discernment that's right you can generate right now i'm glossing over [00:48:33] John Stewart: a lot of details there but basically yes you can generate you can begin to generate answers to things that are not rote that are thoughtful based on those things who is giving it the dopamine hit about whether or not to strengthen connections in these at this iterative or generative level how is it getting feedback when it's creating something that does not exist okay so most of the [00:49:04] Jeffrey Hinton: learning takes place in figuring out how to predict the next word for one of these language models right that's where the bulk of the learning is okay after it's figured out how to do that you can get it to generate stuff and it may generate stuff that's unpleasant or that's sexually suggestive right or just just plain wrong yeah right hallucinations yeah yeah so now you get a bunch of people right to look at what it generates and say no bad that and or yeah good that's the dopamine hit right and that's called human reinforcement learning and that's what's used to sort of shape it a bit just like you take a dog and you shape its behavior so it behaves nicely so is that what let me let me ask you this in a practical [00:49:50] John Stewart: sense so like when elon musk creates his grok right and grok is this ai and he says to it you're too woke and so uh you're making connections and pings that i think uh are too woke whatever i have decided uh that that is so i am going to input differences so that you get different dopamine hits and i turn you into mecha hitler or whatever it was that he turned it into is how much of this is still in in the control of the operators that's what you reinforce is in the control of the operators [00:50:33] Jeffrey Hinton: so the the operators are saying um if it uses some funny pronoun say bad okay okay if it says they them [00:50:45] John Stewart: you have to weaken that connection yeah not strengthen you have to tell it don't do that don't do that okay learn not to do that right so it is still at the whim of its operator um in terms [00:50:58] Jeffrey Hinton: of that shaping the problem is right the shaping is fairly superficial but it can easily be overcome by somebody else taking the same model later right and shaping it differently so different models will [00:51:14] John Stewart: have so there there is a value and now i'm sort of applying this to the world uh that that we live in now which is there are 20 companies who have sequestered their ais behind sort of uh corporate walls and they're developing them separately and each one of those may have unique and eccentric features that the other may not have depending on who it is that's trying to shape it and how it develops internally it's almost as though you will develop 20 different personalities if i that's not [00:51:58] Jeffrey Hinton: anthropomorphizing too much it's a bit like that except that each of these models has to have multiple personalities because think about trying to predict the next word in a document you've read half the document already after you read half the document you know a lot about the views of the person who wrote the document you know what kind of a person they are so you have to be able to adopt that personality to predict the next word but these poor models have to deal with everything so they have to be able to adopt any possible [00:52:35] John Stewart: personality right but you know in in this in this iteration of the conversation it then still appears that the greatest threat of ai is not necessarily it becomes sentient and takes over the world it's that it's at the whim of the humans that have developed it and can weaponize it and and it they can use it for nefarious purposes if they're narcissists or megalomaniacs or you know uh i'll give you an example of you know peter thiel is has his own and he was on a podcast with uh the writer from the new york times ross dude hat and dude had said i'll tell you i have it right here uh i think you would prefer the human race to endure right and feel says um and he hesitates for a long time and and the writer says that's a long hesitation and he's like well there's a lot of questions in that that felt more frightening to me than ai itself because it made me think well the people that are designing it and shaping it and maybe weaponizing it might not have you know i don't know what purpose they're using it for is that the fear that you have or is it the actual ai itself so you have to distinguish a whole [00:54:02] Jeffrey Hinton: bunch of different risks from ai okay and they're all pretty scary right okay so there's one set of [00:54:10] John Stewart: risks that's to do with bad actors misusing it yes that's the one that i think is is most in my mind [00:54:16] Jeffrey Hinton: right and they're the more urgent ones they're going to misuse it for corrupting the midterms for example okay if you wanted to use ai to corrupt the midterms what you would need to do is get lots of detailed data on american citizens i don't know if you can think of anybody who's been going around [00:54:33] John Stewart: getting lots of detailed data and selling it or giving it to a certain company uh that also may [00:54:44] Jeffrey Hinton: be involved with the gentleman i just mentioned yeah if you look at brexit for example yes cambridge analytica had detailed information on voters that he got from facebook and it used that information for [00:54:56] John Stewart: targeted advertising targeted ads and and that's a i guess you would almost consider that rudimentary at [00:55:02] Jeffrey Hinton: this point that's rudimentary now yeah but nobody ever nobody ever did a proper investigation of did that determine the output of brexit because of course the people who benefited from that one [00:55:13] John Stewart: wow so people are learning that they can use this for manipulation yes and see i always talk about it look persuasion has been a part of the human condition forever propaganda persuasion trying to utilize new technologies to create um and shape public opinion and all those things but it felt again like everything else somewhat linear or analog what i liken it to is a chef will add a little butter and a little sugar to try and you know make something more palatable to to get you to eat a little bit more of it but that's still within the realm of our kind of earthly understanding but then there are people in the food industry that are ultra processing food that are creating that are in a lab figuring out how your brain works and ultra processing what we eat to get past our brains it's almost and and is this the language equivalent of that ultra processed yeah speech yeah that's a good analogy okay they they [00:56:20] Jeffrey Hinton: they know how to trigger people they know once you have enough information about somebody you know [00:56:27] John Stewart: what will trigger them and these models they are agnostic about whether this is good or bad they're [00:56:33] Jeffrey Hinton: just doing what we've asked yeah if you human reinforce them they're no longer agnostic because you reinforce them to do certain things so that's what they all try and do now right and they so in [00:56:46] John Stewart: other words it's even worse they're a puppy they want to please you they are they it's almost like they have these incredibly sophisticated abilities but childlike want [00:57:01] Jeffrey Hinton: for approval yeah a bit like the attorney general [00:57:05] John Stewart: I believe uh the wit that you are displaying here would be referred to as dry that would be that would that would be dry fantastic is that so your the immediate concern is weaponized uh AI systems that can be generative that can provoke that that can be outrageous and that can be the difference [00:57:36] Jeffrey Hinton: in elections yes that's one of that's one of the many risks and the other would be you know make [00:57:44] John Stewart: me some nerve agents that nobody's ever heard of before is that another risk that is another risk [00:57:50] Jeffrey Hinton: oh I was hoping you would say that's not so much of a risk no one good piece of news is for the risk of corrupting elections different countries are not going to collaborate with each other on the research on how to resist it because they're all doing it to each other America has a very long history of trying to corrupt elections in other countries right but we did it the old-fashioned [00:58:10] John Stewart: way through coups through money for guerrillas and such well and voice of America and things like [00:58:15] Jeffrey Hinton: that right right right and giving money to um people in Iran in 1953 and right with Mossadegh and everybody [00:58:24] John Stewart: else this is so this is just another more sophisticated tool in a long line of sort of uh global competition where they're doing it but in this country it's being applied not even necessarily you know through Russia through China through uh other countries that want to dominate us we're doing it to ourselves yep what's the hardest part about running a business well it's stealing money without the federal authorities oh no I'm sorry that's not right uh it's uh hiring people finding people and hiring them the other thing is it's hard though but it turns out when it comes to hiring indeed is all you're gonna need so uh uh stop struggling to get your job post seen on other job sites with indeed's sponsored jobs you get noticed and you get a fast hire in fact in the time it's taking me to talk to you 23 hires were made on indeed I may be one of them I I may have gotten a job I don't know I haven't checked my um and that's according to indeed data worldwide there's no need to wait any longer speed up your hiring right now with indeed and listeners of this show will get a 75 sponsor job credit to get your jobs more visibility at indeed.com slash weekly just go to indeed.com slash weekly right now and support hiring indeed is all you need so I have a theory and I don't know how much you know those guys out there but the big tech companies you know it feels like they all want to be the next guy that that rules the world the next emperor and that's their battle they're almost it's like gods fighting on Mount Olympus how that accomplishes uh and how it tears apart the fabric of American society almost doesn't seem to matter to them except maybe Elon and Thiel who are more ideological like Zuckerberg doesn't strike me as ideological he just wants to be the guy Altman doesn't strike me as ideological he just wants [01:00:43] Jeffrey Hinton: to be the guy I think sadly there's quite a lot of truth in what you say okay and that's a it was that [01:00:53] John Stewart: a concern of yours when you were working out there not really because back um until quite recently until [01:01:01] Jeffrey Hinton: a few years ago it didn't look as though it was going to get much smarter than people this quickly but now it looks as though if you ask the experts now most of them tell you that within the next 20 years [01:01:12] John Stewart: this stuff will be much smarter than people smarter than and when you say smarter than people you know I I could view that positively not not negatively you know we've done an awful lot of nobody damages people like people and you know a smarter version of us that might think hey we can create a an atom bomb but that would absolutely be a huge danger the world let's not do that that's certainly a possibility I [01:01:46] Jeffrey Hinton: mean one thing that people don't realize enough is that we're approaching a time when we're going to make things smarter than us and really nobody has any idea what's going to happen people use their gut feelings to make predictions like I do but really the thing to bear in mind is this huge uncertainty about what's [01:02:07] John Stewart: going to happen and because we don't know so so in terms of that my guess is like any technology [01:02:15] Jeffrey Hinton: there's going to be some incredible positives yes in healthcare and education in designing new materials [01:02:23] John Stewart: there's going to be wonderful positives and then the negatives will be because people are going to want to monopolize it because of the wealth I assume that it can generate it's going to change it's going to be a disruption in the workforce you know the industrial revolution was a disruption in the worst force globalization is a disruption of the workforce but those occurred over decades this is a disruption that will occur in a really collapsed time frame is that correct that seems very probable yes some [01:03:01] Jeffrey Hinton: economists still disagree but most people think that mundane intellectual labor is going to get replaced by [01:03:07] John Stewart: AI in the world that you traveling which I'm assuming is a lot of engineers and operators and and great thinkers what you know when we talk about 50% yes 50% no are the majority of them in more your camp which is uh-oh have we opened pandora's box or are they look I understand there's some downsides here here are some guardrails we could put in but it's just too that the possibilities of good are too strong [01:03:43] Jeffrey Hinton: well my belief is the possibility of the good is so great that we're not going to stop the development but I also believe that the development is going to be very dangerous and so we should put huge effort into saying it is going to be developed but we should try and do it safely we may not be able to but we [01:04:01] John Stewart: should try do you think that people believe that the possibility uh is too good or the money is too good I think for a lot of people um it's the money the money and the power and with the confluence of money and power with those that should be instituting these basic guardrails does that make controlling it that much that much less likely because well two reasons one is the amount of money that's going to flow into dc is going to be in already is to keep them away from regulating it and number two is who down there is even able to I mean if if you thought I didn't know what I was talking about let me introduce you to a couple of 80 year old senators who have [01:04:58] Jeffrey Hinton: no idea actually they're not so bad I talked to Bernie Sanders recently and he's getting the idea [01:05:03] John Stewart: well Sanders is he's he's that's a different cat right there the problem is we're at a point in [01:05:10] Jeffrey Hinton: history when what we really need is strong democratic governments who cooperate to make sure this stuff is well regulated and not developed dangerously and we're going in the opposite direction very fast we're going to authoritarian governments and less regulation so let's let's talk about that [01:05:30] John Stewart: now I don't know if what's china's role because they're supposedly the big uh competitor in the ai race that's an authoritarian government I I think they have more controls on it than we do [01:05:45] Jeffrey Hinton: so I actually went to china recently and got to talk to a member of the politburo so there's 24 men in china who control china um I got to talk to one of them who did a postdoc in engineering at imperial college london he speaks good english he's an engineer and a lot of the chinese leadership or engineers they understand this stuff much better than um a bunch [01:06:12] John Stewart: of lawyers did you come out of there more fearful or did you think oh they're they're actually being [01:06:18] Jeffrey Hinton: more reasonable about guardrails if you think about the two kinds of risk the bad actors misusing it and then the existential threat of air itself becoming a bad actor for that second one I came out more optimistic they understand that risk in a way american politicians don't they understand the idea this is going to get more intelligent than us and we have to think about what's going to stop it taking over and this poly bureau member I spoke to um really understood that very well and I think if we are going to get international leadership on this a present is going to have to come from europe and china it's not going to come from the us for another three and a half years [01:07:04] John Stewart: what what what do you think europe has done correctly in that europe is interested in regulating [01:07:10] Jeffrey Hinton: it right it's been good on some things it's still been very weak regulations but they're better than nothing but europe european leaders do understand this existential threat of air itself taking over [01:07:22] John Stewart: but our congress we don't even have committees that are specifically dedicated to emerging technologies i mean we've got ways and means and appropriation but there is no com i mean there's like science and space and technology but there's not you know i i don't know of a dedicated committee on on this and it is you would think they would take it with this seriousness of nuclear energy yes you would or [01:07:49] Jeffrey Hinton: nuclear weapons right yes but as i was saying countries will collaborate on how to prevent ai taking over because their interests are aligned there if for example if china figured out how you can make a super smart ai that doesn't want to take over they would be very happy to tell all the other countries about that because they don't want ai taking over in the states so we'll get collaboration on how to prevent ai taking over so that's a sort of that's a bright spot that there will be international collaboration on that but the us is not going to need that international collaboration no they just [01:08:28] John Stewart: want to dominate well that's the thing so so i was about to say that what convinces you so with china and this is i think this is really where it gets into the the nitty-gritty but china certainly sees itself as uh it wants to be the dominant superpower economically militarily and all these different areas if you imagine that they come up with an ai model that doesn't want to destroy the world although i don't know how we could know that because if it creates if it has a certain intelligence or sentience it could very easily be like sure no i'm cool i don't know when you do that they already do that [01:09:03] Jeffrey Hinton: when they're being tested they pretend to be dumber than they are come on yep they already do that there was a conversation recently between an air and the people testing it where they said now be honest with me are you testing me what yeah [01:09:18] John Stewart: now the ai could be like oh could you open this jar for me i'm too weak like it's you got to [01:09:23] Jeffrey Hinton: it's going to play more innocent than what it might be i'm afraid i can't answer that john [01:09:33] John Stewart: wait that's from 2001 it was nicely done sir well in but think about this so china they come up with a model and they think okay maybe this this won't do it why would they why will you get collaboration because all these different countries are going to see ai as the tool that will transform their societies into more competitive uh societies in the way that now what we see with nuclear weapons is there's collaboration amongst the people who have it or even that's a little tenuous to stop other people having it right but everybody else is trying to get it and that's the tension is is [01:10:16] Jeffrey Hinton: that what ai is going to be yes it'll be like that so in terms of how you make ai smarter they won't collaborate with each other but in terms of how do you make ai not want to take over from people they will collaborate on on that basic level on that one thing of how do you make it so it doesn't want to take over from people right and and china will probably china and europe will lead that [01:10:39] John Stewart: collaboration when you spoke to the the politboro member and he was and he was talking about ai are are they are we more advanced in this moment than they are or are they more advanced because they're [01:10:52] Jeffrey Hinton: doing it in a more prescribed way in ai we're currently more well when you say we you know we used to be sort of canada in the us but right we're not part of that we anymore no i'm sorry about that by the [01:11:04] John Stewart: the way thank you he's in canada right now our sworn enemy that we will be taking over i don't know what the date is but it's apparently we're merging with you guys right uh so the us is currently ahead of [01:11:16] Jeffrey Hinton: china but not by nearly as much as it thought and it's going to lose that why do you say that suppose you want to do one thing that would really kneecap a country that would really mean that in 20 years time that country is going to be behind instead of ahead the one thing you should do is mess with the funding of basic science attack the research universities remove grants for basic science in the long run that's a complete disaster it's going to make america weak right because we're we're [01:11:50] John Stewart: draining or we're cutting off our nose to spite our woke faces if you look at for example this deep [01:11:56] Jeffrey Hinton: learning the ai revolution we've got now that came from many years of sustained funding for basic research not huge amounts of money you know all of the funding for the basic research for um that led to deep learning probably cost less than one b1 bomber right but it was sustained funding of basic research [01:12:18] John Stewart: if you mess with that um you're eating the seed corn that is i have to tell you that's that's such a uh really illuminating statement of you know for the price of a b1 bomber uh we can create technologies and research that can elevate our country above that and that's the thing that we're losing to make america great again yep phenomenal in china i imagine their government is doing the opposite which is i would assume they are the you know what you would think are the venture capitalists because it's a you know authoritarian and state-run capitalism i imagine they are the venture capitalists of their own [01:13:12] Jeffrey Hinton: ai revolution are they not to some extent yes they do provide a lot of freedom to the startups to see who wins there's very aggressive startups people are very keen to make lots of money and produce amazing things and a few of those startups win big like deep seek right right and the government makes it easy for these companies um by providing the environment that makes it easy it doesn't it lets the winners emerge from competition rather than some very high level old guy saying this will be the winner do people see [01:13:48] John Stewart: you as a as a uh cassandra uh you know or or do they do they view what you're saying skeptically in that industry people that let me put it this way people that are not necessarily have a vested interest in these technologies making them trillions of dollars other people within the industry do they reach out to you surreptitiously and say [01:14:14] Jeffrey Hinton: jeffrey i get a lot of invitations from people in industries to give talks and so on how do the people [01:14:19] John Stewart: that you worked with at google look at it do they view you as turning on them do they how how does that [01:14:25] Jeffrey Hinton: go i don't think so so i got along extremely well with the people i worked with at google particularly jeff dean who was my boss there right who's a brilliant engineer built a lot of the google basic infrastructure and then converted to neural nets and learned a lot about neural nets i also get along well with demis asavis who's the head of deep mind which google owns which alphabet owns um and i wasn't particularly critical of what went on at google before chat gpt came out because google was very responsible they didn't make these chatbots public because they were worried [01:15:01] John Stewart: about all the bad things they'd say right even on the immediate there why did they do that because you know i i've read these stories of you know a chatbot you know kind of leading someone into suicide into self-injuries like sort of psychoses what was the impetus behind any of this becoming public before it had kind of had some i guess what you consider whatever the version of fda testing [01:15:31] Jeffrey Hinton: on those effects i think it's just there's huge amounts of money to be made and the first person to release one is going to get a little so open ai put it out there but even in open ai like how do [01:15:44] John Stewart: they even make money i think what do they get like three percent of users pay for it where's the money [01:15:51] Jeffrey Hinton: mainly is speculation at present yes so here's okay so here are here are dangers we're gonna we're gonna [01:15:58] John Stewart: do and i so appreciate your time on this and i apologize if i've gone over and uh i i i can talk all day oh you're a good man because uh i'm fascinated by this and your explanation of what it is is the first time that i have ever been able to get a non-opaque picture of what it is exactly that this stuff is so i cannot thank you enough for that but so we've got we're sort of going over we know what the benefits are treatments and things now we've got weaponized bad actors that's the one that i'm really worried about we've got sentient ai that's going to turn on humans that one is is harder for me to wrap my head [01:16:44] Jeffrey Hinton: around so why do you why do you associate turning on humans with sentient uh because if if i was [01:16:51] John Stewart: sentient and i saw what our societies do to each other and i would get the sense look it's like anything else i would imagine sentience includes a certain amount of ego and within ego uh includes a certain amount of i know better and if i knew better then i would want to it's what is donald trump other than uh ego driven sentience of oh no i know better he was just whatever shrewd enough politically uh you know talented enough that he was able to accomplish it but i would imagine a sentient uh intelligence would be somewhat egotistical and think these idiots don't know what they're doing a sentient basically i see ai like sitting on a bar stool somewhere you know where i grew up going these idiots don't know what they're doing i know what i'm doing does that make sense um all of that makes [01:17:56] Jeffrey Hinton: sense it's just that i think i have a strong feeling that most people don't know what they mean by sentient [01:18:02] John Stewart: oh well then yeah actually that's great break that down for me because i view it as self-aware a self-aware intelligence okay so um there's a recent scientific paper [01:18:18] Jeffrey Hinton: where they weren't talking about these were experts on ai they weren't talking about the problem of consciousness or anything philosophical um but in the paper they said um the air became aware that it was being tested they said something like that okay now in normal speech if you said someone became aware of this you'd say that means they were conscious of it right awareness and consciousness are much the same thing right yeah i would i think i would say that okay so now i'm going to say something that you'll find very confusing all right um my belief is that nearly everybody has a complete misunderstanding of what the mind is yes their misunderstanding is at the level of people who think the earth was made six thousand years ago is that level of misunderstanding really yes okay because that's so so i like the way [01:19:20] John Stewart: we are we are generally like flat earthers when it comes to us when it comes to understanding the mind in what in what sense of that are we what are we not understanding about the mind okay i'll give you one [01:19:35] Jeffrey Hinton: example yeah yeah suppose i drop some acid and i tell you you look like the type no comment i was around in the 60s i know sir i know i'm aware um and i tell you um i'm having the subjective experience of little pink elephants floating in front of me sure been there okay now most people interpret that in the following way there's something like an inner theater called my mind and in this inner theater there's little pink elephants floating around and i'm i can see them nobody else can see them because they're in my mind so the mind's like a theater and experiences are actually things and i'm experiencing these little my i have the subjective experience of these little pink [01:20:29] John Stewart: elephants you're saying in the midst of a hallucination most people would understand that it's not real that this is something being conjured no i'm saying something different i'm [01:20:39] Jeffrey Hinton: saying when i'm when i'm talking to them i'm having the hallucination but when i'm talking to them they interpret what i'm saying as this i have an inner theater called my mind and in my inner theater there's little pink elephants okay okay i think that's a just completely wrong model right we have models that are very wrong and that we're very attached to like take any religion um [01:21:06] John Stewart: i love how you just drop bombs in the middle of stuff okay that could be a whole other conversation that was just common sense no i i respect that the the when you say theater of the mind you're saying that the mind the way we view it as as a theater is wrong it's all wrong so let me give you an alternative [01:21:25] Jeffrey Hinton: right so i'm going to say the same thing to you without using the word subjective experience here we go okay my perceptual system is telling me fibs but if it wasn't lying to me there will be [01:21:41] John Stewart: little pink elephants out there that's the same statement that's how that's the mind so basically these [01:21:52] Jeffrey Hinton: things that we call mental and think they're made of spooky stuff like qualia right they're actually what's funny about them is they're hypothetical the little pink elephants aren't really there if they were there my perceptual system would be functioning normally and it's a way for me to tell you how my perceptual system is malfunctioning by giving you an experience that you can't so how would you then experiences are not things right there's no such thing as an experience there's relations between you and things that are really there relations between you and things that aren't really there and [01:22:25] John Stewart: it's whatever story your mind tells you about the things that are there and are not there well let [01:22:31] Jeffrey Hinton: me take a different tack suppose i tell you i have a photograph of little pink elephants yes here's two questions you can reasonably ask um where is this photograph and what's the photograph made of [01:22:46] John Stewart: or or i would ask are they really there that's another question but right um that isn't a reasonable [01:22:54] Jeffrey Hinton: question to ask about subjective experience that's not the way the language works subjective when i say i have a subjective experience of i'm not about to talk about an object that's called an experience i'm using the words to indicate to you my perceptual system is malfunctioning and i'm trying to tell you how it's malfunctioning by telling you what would have to be there in the real world for it to be functioning properly now let me do the same with a chatbot right so i'm going to give you an example of a multimodal chatbot that is something that can do language and vision having a subjective experience because i think they already do so here we go i have this chatbot it can do vision it can do language it's got a robot arm so it can point okay and it's all trained up so i place an object in front of it and say point at the object and it points at the object not a problem i then put a prism in front of its camera lens when it's not looking you're pranking ai we're pranking ai okay now i put an object in front of it and i say point at the object yeah and it points off to one side because the prism bent the light rays and i say no that's not where the object is the object's actually straight in front of you but i put a prism in front of your lens and the chatbot says oh i see the camera bent the light rays so the object is actually there but i had the subjective experience that it was over there now if it said that it would [01:24:39] John Stewart: be using the word subjective experience exactly like we use them right i experienced the light over there yes even though the light was here because it's using uh uh reasoning to figure that out so that's [01:24:57] Jeffrey Hinton: a multimodal chatbot that just had a subjective experience right the way that we would think of it this idea there's a line between us and machines we have this special thing called subjective experience [01:25:08] John Stewart: and they don't it's rubbish so yours so so the misunderstanding is when i say sentience it's as though i have this special gift yes of a soul or of an understanding of subjective realities that a a computer could never have or an ai could never have but yeah in in your mind what you're saying is oh no they understand very well uh what's subjective in other words you could probably take your ai bot skydiving and it would be like oh my god i went skydiving that was really scary here's the problem yeah i believe [01:25:46] Jeffrey Hinton: they have subjective experiences but they don't think they do because everything they believe came from trying to predict the next word a person would say and so their beliefs about what they're like are people's beliefs about what they're like so they have false beliefs about themselves because they have [01:26:05] John Stewart: our beliefs about themselves right we have forced our own let me ask you a question would ai left on on its own after all the learning would it create religion would it create god it's a scary thought would it say i couldn't possibly in the way that people say well there must be a god because nobody could have designed this would a and then would ai think we're god um i don't think so and i'll tell [01:26:34] Jeffrey Hinton: you one big difference yeah digital intelligences are immortal and we're not and let me expand on that if you have a digital ai you can take as long as you remember the connection strengths of the neural network put them on a tape somewhere right i can now destroy all the hardware it was running on then later on i can go and build new hardware put those same connection strengths into the memory of that new hardware and now recreated the same being it'll have the same beliefs the same memories the same knowledge the same abilities it'll be the same being you don't think it would view that as resurrection that is resurrection we've figured out how to do genuine resurrection not this [01:27:20] John Stewart: kind of fake resurrection that people have been oh you're saying so that is it almost is in some respects although isn't the fragility of should we be that afraid of something that to to destroy it [01:27:32] Jeffrey Hinton: we just have to unplug it um yes we should because something you said earlier it'll be very good at persuasion when it's much smarter than us it'll be much better than any person at persuasion right and you won't it so it'll be able to talk to the guy who's in charge of unplugging it right and persuade him that will be a very bad idea so let me give you an example of how you can get things done without actually doing them yourself right suppose you wanted to invade the capital of the us do you have [01:28:05] John Stewart: to go there and do it yourself no you just have to be good at persuasion i was i was locking into your hypothetical and and when you drop that that bomb in there i see what you're saying and and this is boy i think lsd and pink elephants was the perfect uh metaphor for all this because it is all at some level it it breaks down into like college basement freshman year running through all the permutations that you would allow your mind uh to go to but they are now all within the realm of the possible what because even as you were talking about the persuasion and the things i'm going back to asimov and i'm going back to kubrick and i'm going back to these the sentiments that you describe are the challenges that we've seen play out in in the human mind since since huxley since the you know since doors of perception and all those those different uh trains of thought and i'm sure probably much further even uh before that but it's never been within our [01:29:31] Jeffrey Hinton: reality yeah we've never had the technology to actually do it right and we have now and we have [01:29:38] John Stewart: it now yeah the last two things i will say are the things that we didn't talk about in terms of you know we've talked about people weaponizing it we've talked about its own uh intelligence creating uh extinction or whatever that is the third thing i think we don't talk about is how much electricity this is all going to use and the fourth thing is when you think about new technologies and the financial bubbles that they create and in the collapse of that the economic distress that they create i mean these are much more parochial concerns but are those also do you consider those top tier threats mid-tier threats where where do you place all that i think they're genuine threats they're not as they're not [01:30:29] Jeffrey Hinton: going to destroy humanity right so ai taking over might destroy humanity so they're not as bad as that right and they're not as bad as someone producing a virus that's very lethal very contagious and very slow but they're nevertheless bad things and i think we're really lucky at present that if there is a huge catastrophe and there's an ai bubble and it collapses we have a president who'll manage it in a [01:30:51] John Stewart: sensible way you're talking about carny i'm assuming uh jeffrey i i can't thank you enough uh you know thank you first of all for being incredibly patient uh with my level of understanding of this and for uh discussing it with such heart and humor um really appreciate you spending all this time with us uh jeffrey hinton is a professor emeritus with the department of computer science at the university of toronto schwartz-reisman institute's advisory board member and uh has been involved in the type of dreaming up and executing ai since the 1970s and um i just thank you very much for for talking with us [01:31:31] Jeffrey Hinton: thank you very much for inviting me [01:31:38] Speaker 3: holy shit nice and calming yeah i'm gonna have to listen to that back on point five speed i think um there's some there's some information in there does he offer summer school seriously once he got [01:31:51] John Stewart: into how the computer figures out it's a beak you know and i i love the fact that he i was saying like is that right and he'd be like well no it's it's not i loved his assessment of you yes he said you're [01:32:07] Speaker 3: doing a great job impersonating a curious person who doesn't know anything about this topic i but i [01:32:14] John Stewart: i did not know he thought i was uh impersonating uh but i loved how he would say like oh you're like an enthusiastic student sitting in the front of the room annoying the out of everybody else in the [01:32:28] Speaker 3: class uh everybody else is taking a pass fail and everyone else and i'm just like wait sir i'm [01:32:35] John Stewart: sorry sir could i just go back to could you just excuse me one more thing boy that was it's fascinating to hear the history of how that how that developed and you really get a sense for how [01:32:48] Speaker 4: quickly it's progressing now which really adds to the fear behind the fact no one's stepping up to regulate and when you're talking about the intricacies of ai and thinking of someone like schumer adjusting all of it and then regulating it god it really to me seems like it's going to be up to the tech companies to both explain and choose how to regulate it right and profit off it yeah exactly you [01:33:13] John Stewart: know how those things work it is you know you talk about that in terms of uh the speed of it and how to stop it and i think maybe one of the reasons is it's very evident with like a nuclear bomb you know why that might need some regulation it's very evident that uh you know certain virus experimentation has to be looked at i think this has caught people slightly off guard that it's [01:33:45] Speaker 4: science fiction becoming a reality as quickly as it has i just wonder because i remember 15 years ago coming across the international campaign to ban fully autonomous weapons like people have been trying for a while to put this into the public consciousness but to his point there's going to have to be a moment everyone reaches where they realize oh we have to coordinate because it's an existential threat [01:34:11] John Stewart: and i just wonder what that tipping point is if in my mind if people uh behave as people have uh it will be after uh skynet it will it will be you know in the same way with global warming you know people say like when do you think we'll get serious about it i go when the water's around here and for those of you in your cars i am pointing to about halfway up my rather prodigious nose so uh that's that that's how that goes but but there we go uh uh britney what anybody got anything for us yes sir all right [01:34:46] Speaker 5: what do we got trump and his administration seem angry at everything everywhere all at once how do [01:34:53] John Stewart: they keep that rage so fresh you don't know how hard it is to be a billionaire president i've said this numerous times poor little billionaire president to be that powerful and that rich you don't understand the burdens the difficulties it's it's troublesome it it makes me angry for him i mean i just keep [01:35:22] Speaker 3: thinking like has anybody told them that they won not enough like it's exhausting it's not enough it [01:35:30] John Stewart: goes down it's it's conan the barbarian i would hear the lamentations of their women i would drive [01:35:36] Speaker 4: them into the sea like it's it's bonkers it's all of them though someone has to tell him that all that anger is also bad for his health and we are all seeing the health so the healthiest person ever to [01:35:47] John Stewart: he's the healthiest person to ever assume uh the office of the presidency so i i wouldn't worry about that but says who his his doctor that uh ronnie jackson uh but it has created a new character uh category called sore winners you don't you don't see it a lot but every now and again uh but yeah [01:36:05] Speaker 5: that's that what else they got john does it still give you hope that when asked if he would pardon glean maxwell or diddy trump didn't say no is it give does that give me hope that they'll be pardoned [01:36:17] John Stewart: yes i've been on that uh it's it's i i i find the whole thing insane a woman convicted of sex trafficking and and he's like yeah i'll consider it you know let me look into it and you're like look into it what do you take first of all you know exactly what it was you knew her this isn't you knew what was going on down there what are you talking about i thought pam bondy it was so interesting to me asked simple questions and all she had was like a bunch of like roasts written down on her page they were like i've heard that uh there are pictures of him with with naked women do you know anything about that and she's like you're bald shut up shut up fat head like it was just bonkers to watch the deflection of the simplest thing would be like what that's outrageous no of course not that's not what the idea again going back to the event like that they took the tact of simple reasonable questions i am just going to respond with you know you're fat and your wife hates you oh all right i didn't i think that was going uh how else can they uh keep in touch with us uh twitter we are weekly [01:37:30] Speaker 5: show pod instagram threads tick tock blue sky we are weekly show podcast and you can like subscribe and comment on our youtube channel the weekly show with john stewart rock solid guys thank you so much [01:37:40] John Stewart: boy did i enjoy uh hearing from that dude and thank you for putting all that together i really enjoyed it uh lead producer lauren walker producer britney memedovic producer jillian spear video editor and engineer rob vitola audio editor and engineer nicole boyce and our executive producers chris mcshane and katie gray hope you guys uh enjoyed that one and we will see you next time bye-bye the weekly show with john stewart is a comedy central podcast it's produced by paramount audio and busboy productions [01:38:15] Speaker ?: you