WEBVTT 00:00:00.000 --> 00:00:03.517 The following is a conversation with Yann LeCun, his second time on the 00:00:03.600 --> 00:00:06.757 podcast. He is the chief AI scientist at 00:00:06.840 --> 00:00:10.357 Meta, formerly Facebook, professor at 00:00:10.440 --> 00:00:15.637 NYU, Turing Award winner, one of the seminal figures in the history of 00:00:15.720 --> 00:00:19.757 machine learning and artificial intelligence, and someone who is 00:00:19.840 --> 00:00:24.557 brilliant and opinionated in the best kind of way, and so is always fun to 00:00:24.640 --> 00:00:28.197 talk to. And now a quick few second mention of 00:00:28.280 --> 00:00:31.717 each sponsor. Check them out in the description. It's the best way to 00:00:31.800 --> 00:00:34.477 support this podcast. First is Public 00:00:34.560 --> 00:00:36.677 Goods, an online shop I use for 00:00:36.760 --> 00:00:40.157 household products. Second is Indeed, a 00:00:40.240 --> 00:00:43.117 hiring website. Third is Roka, my 00:00:43.200 --> 00:00:47.397 favorite sunglasses and prescription glasses. Fourth is NetSuite, business 00:00:47.480 --> 00:00:49.957 software for managing HR, financials, 00:00:50.040 --> 00:00:53.117 and other details. And fifth is Magic 00:00:53.200 --> 00:00:56.077 Spoon, low-carb, keto-friendly cereal. 00:00:56.160 --> 00:00:59.917 So, the choice is business, health, or style. 00:01:00.000 --> 00:01:04.637 Choose wisely, my friends. And now onto the full ads. As always, no 00:01:04.720 --> 00:01:08.037 ads in the middle. I try to make this interesting, but if you skip them, 00:01:08.120 --> 00:01:12.797 please still check out our sponsors. I enjoy their stuff, maybe you will, too. 00:01:12.880 --> 00:01:14.837 This show is brought to you by Public 00:01:14.920 --> 00:01:17.997 Goods, the one-stop shop for affordable, 00:01:18.080 --> 00:01:21.197 sustainable, healthy household products. 00:01:21.280 --> 00:01:26.557 I use their hand soap, toothpaste, and toothbrush. I think I use a bunch of 00:01:26.640 --> 00:01:30.077 other stuff, too, but that's that's what comes to mind. 00:01:30.160 --> 00:01:33.037 The products often have this minimalist 00:01:33.120 --> 00:01:37.597 black and white design that I just absolutely find beautiful. I love it. I 00:01:37.680 --> 00:01:41.757 love minimalism in design. It doesn't go over the top. It doesn't doesn't have 00:01:41.840 --> 00:01:43.837 all these extra things and features that 00:01:43.920 --> 00:01:46.877 you don't need, just the essentials. 00:01:46.960 --> 00:01:51.077 I think it's hard to explain. But there's something about 00:01:51.160 --> 00:01:55.637 the absence of things that that take up your 00:01:55.720 --> 00:01:59.437 attention that allows you to truly be attentive to 00:01:59.520 --> 00:02:03.597 what matters in life. Anyway, go to publicgoods.com/lex 00:02:03.680 --> 00:02:08.357 or use code Lex at checkout to get $15 off your first order, plus you will 00:02:08.440 --> 00:02:11.917 receive your choice of either free pack of bamboo straws 00:02:12.000 --> 00:02:14.557 or reusable food storage wraps. 00:02:14.640 --> 00:02:18.997 Visit publicgoods.com/lex or use code Lex at checkout. 00:02:19.080 --> 00:02:20.757 This show is also brought to you by 00:02:20.840 --> 00:02:23.957 Indeed, a hiring website. I've used them 00:02:24.040 --> 00:02:28.237 as part of many hiring efforts I've done for the teams I've led in the past. Most 00:02:28.320 --> 00:02:32.277 of those for was for engineering, for research efforts. 00:02:32.360 --> 00:02:36.037 They have tools like Indeed instant match giving you quality candidates 00:02:36.120 --> 00:02:40.237 whose resumes on Indeed fit your job description immediately. For the past 00:02:40.320 --> 00:02:44.877 few months I've been going through this process of um building up a team of uh 00:02:44.960 --> 00:02:48.517 folks to help me. I've been doing quite a bit of hiring. 00:02:48.600 --> 00:02:51.637 It's a treacherous 00:02:51.720 --> 00:02:56.837 and an exciting process because you get to meet some friends. So, it's a it's a 00:02:56.920 --> 00:03:01.157 beautiful process, but I think it's one of the most important processes in life. 00:03:01.240 --> 00:03:05.277 It's selecting the group of people with whom you spend your days with. 00:03:05.360 --> 00:03:08.877 And so, you got to use the best tools for the job. Indeed, I think is an 00:03:08.960 --> 00:03:13.197 excellent tool. Right now, you can get a free $75 00:03:13.280 --> 00:03:17.477 sponsored job credit to upgrade your job post at indeed.com/lex. 00:03:17.560 --> 00:03:23.040 Terms and conditions apply. Go to indeed.com/lex. 00:03:23.480 --> 00:03:25.037 This show is also brought to you by 00:03:25.120 --> 00:03:27.997 Roka, the makers of glasses and 00:03:28.080 --> 00:03:32.717 sunglasses that I love wearing for their design, feel, and 00:03:32.800 --> 00:03:36.477 innovation on material, optics, and grip. Roka was started by two 00:03:36.560 --> 00:03:40.837 all-American swimmers from Stanford, and it was born out of an obsession with 00:03:40.920 --> 00:03:43.397 performance. I like the way they feel. I 00:03:43.480 --> 00:03:45.357 like the way they look. 00:03:45.440 --> 00:03:48.277 Whether I'm uh running like a fast-paced 00:03:48.360 --> 00:03:50.637 run, we're talking about 8-minute mile 00:03:50.720 --> 00:03:52.757 or faster, or if we're doing a slow pace 00:03:52.840 --> 00:03:57.357 run, 9 10-minute mile along the river in 00:03:57.440 --> 00:04:00.837 the heat or in the cold, or if I'm just 00:04:00.920 --> 00:04:02.757 wearing my suit 00:04:02.840 --> 00:04:05.437 out on the town, however that expression 00:04:05.520 --> 00:04:10.877 goes, I'm not sure, but they look classy with a suit, they look badass in running 00:04:10.960 --> 00:04:15.397 gear. It's just uh my go-to sunglasses. 00:04:15.480 --> 00:04:19.477 Check them out for both prescription glasses and sunglasses at roka.com and 00:04:19.560 --> 00:04:24.837 enter code Lex to save 20% on your first order. That's roka.com and enter code 00:04:24.920 --> 00:04:30.197 Lex. This show is also brought to you by NetSuite. NetSuite allows you to manage 00:04:30.280 --> 00:04:34.997 financials, human resources, inventory, e-commerce, 00:04:35.080 --> 00:04:38.997 and uh many more business-related details all in one place. I'm not sure 00:04:39.080 --> 00:04:42.357 why I was doing upspeak on that sentence. 00:04:42.440 --> 00:04:46.597 Maybe because I'm very excited about NetSuite. Anyway, there's a lot of messy 00:04:46.680 --> 00:04:49.237 things you have to get right when running a company. 00:04:49.320 --> 00:04:52.037 If you're a business owner, if you're an entrepreneur, 00:04:52.120 --> 00:04:56.637 if you're a founder of a startup, this is something 00:04:56.720 --> 00:05:00.877 you have to think about. You have to use the best tools for the job to make sure 00:05:00.960 --> 00:05:04.997 all the messy things required to run a business are taken care of for you so 00:05:05.080 --> 00:05:07.117 you can focus on the things that you're 00:05:07.200 --> 00:05:09.837 best at, that where your brilliance 00:05:09.920 --> 00:05:14.917 shines. If you are starting a business, I wish you the best of luck. It's a 00:05:15.000 --> 00:05:19.277 difficult journey, but it's worth it. 00:05:19.360 --> 00:05:23.797 Anyway, right now special financing is back. Head to netsuite.com/lex 00:05:23.880 --> 00:05:25.757 to get their one-of-a-kind financing 00:05:25.840 --> 00:05:29.477 program. That's netsuite.com/lex. 00:05:29.560 --> 00:05:34.517 netsuite.com/lex This episode is also brought to you by 00:05:34.600 --> 00:05:38.837 Magic Spoon. The OG, not quite OG, but really 00:05:38.920 --> 00:05:42.877 old-school sponsor of this podcast that I love. It's a low-carb, keto-friendly 00:05:42.960 --> 00:05:45.357 cereal. It has 0 g of sugar. It's 00:05:45.440 --> 00:05:50.237 delicious. I don't say that enough. It's really is delicious. Given that it's 0 g 00:05:50.320 --> 00:05:51.917 of sugar, it's very surprising how 00:05:52.000 --> 00:05:55.117 delicious it is. 13 to 14 g of protein, 00:05:55.200 --> 00:05:58.477 only 4 net g of carbs, and 140 calories 00:05:58.560 --> 00:06:00.877 in in each serving. You can build your 00:06:00.960 --> 00:06:05.797 own box or get a variety pack with available flavors of cocoa, fruity, 00:06:05.880 --> 00:06:09.037 frosted, peanut butter, blueberry, and cinnamon. 00:06:09.120 --> 00:06:12.437 Cocoa is my favorite. It's the flavor of champions. I don't know why I keep 00:06:12.520 --> 00:06:14.477 saying that, but it seems to be true. 00:06:14.560 --> 00:06:17.837 Anyway, Magic Spoon has a 100% happiness 00:06:17.920 --> 00:06:22.677 guarantee. So, if you don't like it, they will refund it. Who else 00:06:22.760 --> 00:06:25.637 will give you a 100% happiness guarantee? 00:06:25.720 --> 00:06:30.477 Go to magicspoon.com/lex and use code Lex at checkout to save $5 00:06:30.560 --> 00:06:33.797 off your order. That's magicspoon.com/lex 00:06:33.880 --> 00:06:35.837 and use code Lex. 00:06:35.920 --> 00:06:38.477 This is the Lex Fridman podcast, and 00:06:38.560 --> 00:06:43.600 here's my conversation with Yann LeCun. 00:06:53.960 --> 00:06:55.980 >> [music] 00:06:58.160 --> 00:07:00.317 >> You co-wrote the article Self-Supervised 00:07:00.400 --> 00:07:04.117 Learning: The Dark Matter of Intelligence. Great title, by the way. 00:07:04.200 --> 00:07:06.717 With Ishan Misra. So, let me ask, what 00:07:06.800 --> 00:07:09.277 is self-supervised learning and why is 00:07:09.360 --> 00:07:11.837 it the dark matter of intelligence? 00:07:11.920 --> 00:07:14.277 >> I'll start by the dark matter part. 00:07:14.360 --> 00:07:17.557 Uh There is obviously a kind of learning 00:07:17.640 --> 00:07:20.437 that humans and animals are 00:07:20.520 --> 00:07:23.677 uh are doing that we currently are not 00:07:23.760 --> 00:07:27.597 reproducing properly with machines or with AI, right? So, the most popular 00:07:27.680 --> 00:07:31.797 approaches to machine learning today are or paradigms, I should say, are 00:07:31.880 --> 00:07:34.637 supervised learning and reinforcement learning. 00:07:34.720 --> 00:07:37.037 And they are extremely efficient. 00:07:37.120 --> 00:07:41.677 Supervised learning requires many samples for learning anything. 00:07:41.760 --> 00:07:46.357 And reinforcement learning requires a ridiculously large number of trial and 00:07:46.440 --> 00:07:49.677 errors to for, you know, a system to learn anything. 00:07:49.760 --> 00:07:54.147 Um And that's why we don't have self-driving cars. 00:07:54.230 --> 00:07:56.677 >> [laughter] >> That's a big leap from one to the other. 00:07:56.760 --> 00:07:59.357 Okay, so that to solve difficult 00:07:59.440 --> 00:08:02.877 problems, you have to have a lot of uh 00:08:02.960 --> 00:08:07.077 human annotation for supervised learning to work. And to solve those difficult 00:08:07.160 --> 00:08:09.917 problems with reinforcement learning, you have to have 00:08:10.000 --> 00:08:13.637 some way to maybe simulate that problem such that you can do that large-scale 00:08:13.720 --> 00:08:16.317 kind of learning that reinforcement learning requires. 00:08:16.400 --> 00:08:19.197 >> Right. So, how is it that, you know, 00:08:19.280 --> 00:08:20.837 most teenagers can learn to drive a car 00:08:20.920 --> 00:08:24.157 in about 20 hours of practice? 00:08:24.240 --> 00:08:26.917 Whereas uh even with millions of hours 00:08:27.000 --> 00:08:31.957 of simulated practice, a self-driving car can't actually learn to drive itself 00:08:32.040 --> 00:08:35.117 properly. Um and so, obviously, we're missing 00:08:35.200 --> 00:08:38.597 something, right? And And it's quite obvious for a lot of people that, you 00:08:38.680 --> 00:08:42.517 know, the immediate response you get from many people is, well, you know, humans 00:08:42.600 --> 00:08:47.997 use their background knowledge to learn faster. And they're right. Now, 00:08:48.080 --> 00:08:51.997 how was that background knowledge acquired? And that's to be questioned. 00:08:52.080 --> 00:08:54.477 So, now you have to ask, you know, how 00:08:54.560 --> 00:08:59.357 do babies in the first few months of life learn how the world works? Mostly 00:08:59.440 --> 00:09:02.837 by observation because they can hardly act in the world. 00:09:02.920 --> 00:09:05.757 Um and they learn an enormous amount of background knowledge about the world 00:09:05.840 --> 00:09:09.797 that may be the the basis of what we call common sense. 00:09:09.880 --> 00:09:15.117 Uh this type of learning is not learning a task. It's not being reinforced for 00:09:15.200 --> 00:09:20.237 anything. It's just observing the world and figuring out how how it works. 00:09:20.320 --> 00:09:24.877 Building world models. Learning world models. Um how do we do this? And how do 00:09:24.960 --> 00:09:28.197 we reproduce this in in machines? So, self-supervised learning 00:09:28.280 --> 00:09:32.877 is, you know, one instance or one attempt at trying to 00:09:32.960 --> 00:09:34.957 reproduce this kind of learning. 00:09:35.040 --> 00:09:39.477 >> Okay, so you're looking at just observation. so not even the interacting 00:09:39.560 --> 00:09:43.677 part of a child. It's just sitting there watching mom and dad walk around, pick 00:09:43.760 --> 00:09:47.357 up stuff, all of that. That's the That's what you mean about background knowledge. 00:09:47.440 --> 00:09:51.877 >> Perhaps not even watching mom and dad, just, you know, watching the world go by. 00:09:51.960 --> 00:09:55.077 >> Just having eyes open or having eyes closed or the very act of opening and 00:09:55.160 --> 00:09:58.917 closing eyes that the world appears and disappears, all of that basic 00:09:59.000 --> 00:10:03.277 information. And you're saying in in order to learn 00:10:03.360 --> 00:10:06.677 to drive like the reason humans are able to learn 00:10:06.760 --> 00:10:10.277 to drive quickly, some faster than others, is because of the background 00:10:10.360 --> 00:10:13.597 knowledge they were able to watch cars operate in the world 00:10:13.680 --> 00:10:17.517 in the many years leading up to it, the physics of basic objects, all that kind 00:10:17.600 --> 00:10:18.837 of stuff. >> That's right. I mean, the basic physics 00:10:18.920 --> 00:10:21.917 of objects, you don't even know you you don't even need to know, you know, how a 00:10:22.000 --> 00:10:24.597 car works, right? Because that you can learn fairly quickly. I mean, the 00:10:24.680 --> 00:10:28.477 example I use very often is you're driving next to a cliff. 00:10:28.560 --> 00:10:33.277 And you know in advance because of your, you know, understanding of intuitive 00:10:33.360 --> 00:10:36.557 physics that if you turn the wheel to the right, the car will veer to the 00:10:36.640 --> 00:10:39.997 right, will run off the cliff, fall off the cliff, and nothing good 00:10:40.080 --> 00:10:44.277 will come out of this, right? Um but if you are a sort of, 00:10:44.360 --> 00:10:47.877 you know, tabular rasa reinforcement learning system that doesn't have a 00:10:47.960 --> 00:10:51.837 model of the world, uh you have to repeat falling off this 00:10:51.920 --> 00:10:55.157 cliff thousands of times before you figure out it's a bad idea. And then a 00:10:55.240 --> 00:10:59.277 few more thousand times before you figure out how to not do it. And then a 00:10:59.360 --> 00:11:02.557 few more million times before you figure out how to not do it in every situation 00:11:02.640 --> 00:11:04.397 you ever encounter. 00:11:04.480 --> 00:11:06.797 >> So, self-supervised learning still has 00:11:06.880 --> 00:11:11.957 to have some source of truth being told to it by somebody. 00:11:12.040 --> 00:11:14.797 >> And is >> So, you have to figure out a way without 00:11:14.880 --> 00:11:17.317 human assistance or without significant 00:11:17.400 --> 00:11:22.277 amount of human assistance to get that truth from the world. So, the mystery 00:11:22.360 --> 00:11:26.517 there is um how much signal is there, how much truth 00:11:26.600 --> 00:11:30.357 is there that the world gives you, whether it's the human world, like you 00:11:30.440 --> 00:11:35.117 watch YouTube or something like that, or it's the more natural world. So, how 00:11:35.200 --> 00:11:39.197 much signal is there? >> So, here's the trick. There is way more 00:11:39.280 --> 00:11:42.517 signal in sort of a self-supervised setting 00:11:42.600 --> 00:11:46.357 than there is in either a supervised or reinforcement setting. 00:11:46.440 --> 00:11:50.477 And this is going to my, you know, analogy of the cake. 00:11:50.560 --> 00:11:53.437 >> Yes. [laughter] >> Uh the, you know, the cake as someone has 00:11:53.520 --> 00:11:56.357 called it, where when you try to figure 00:11:56.440 --> 00:12:00.597 out how much information you ask the machine to predict and how much feedback 00:12:00.680 --> 00:12:02.797 you give the machine at every trial. 00:12:02.880 --> 00:12:05.557 In reinforcement learning, you give the machine a single scalar. You tell the 00:12:05.640 --> 00:12:09.277 machine you did good, you did bad. And you and you and you only 00:12:09.360 --> 00:12:11.437 tell this to the machine once in a while. 00:12:11.520 --> 00:12:15.917 When I say you, it could be the the universe telling the machine, right? Um 00:12:16.000 --> 00:12:19.637 but it's just one scalar. And so, as a consequence, there is you you cannot 00:12:19.720 --> 00:12:22.437 possibly learn something very complicated without many, many, many 00:12:22.520 --> 00:12:25.197 trials where you get many, many feedbacks 00:12:25.280 --> 00:12:27.917 uh of this type. Supervised learning, 00:12:28.000 --> 00:12:30.437 you you give a few bits to the machine 00:12:30.520 --> 00:12:33.037 uh at every every sample. 00:12:33.120 --> 00:12:37.557 Let's say you're training a system on, you know, recognizing images 00:12:37.640 --> 00:12:39.517 on ImageNet. There is 1,000 categories. 00:12:39.600 --> 00:12:43.957 That's a little less than 10 bits of information per sample. 00:12:44.040 --> 00:12:47.437 But, self-supervised learning, here is the setting. You Ideally, we don't know 00:12:47.520 --> 00:12:49.157 how to do this yet, but ideally, you 00:12:49.240 --> 00:12:52.517 would show a machine a segment of a 00:12:52.600 --> 00:12:56.397 video and then stop the video and ask ask the machine to predict what's going 00:12:56.480 --> 00:12:59.320 to happen next. 00:12:59.440 --> 00:13:00.677 And so, you let the machine predict, and 00:13:00.760 --> 00:13:04.357 then you let time go by and show the 00:13:04.440 --> 00:13:09.277 machine what actually happened, and hope the machine will, you know, learn to do 00:13:09.360 --> 00:13:12.317 a better job at predicting next time around. There's a huge amount of 00:13:12.400 --> 00:13:17.597 information you give the machine because it's an entire video clip. It's a of uh 00:13:17.680 --> 00:13:21.277 you know, the of the the future after the video clip you fed it um in the 00:13:21.360 --> 00:13:24.717 first place. >> So, both for language and for vision, 00:13:24.800 --> 00:13:29.677 there's a subtle, seemingly trivial construction, but maybe that's 00:13:29.760 --> 00:13:33.717 representative of what is required to create intelligence, which is 00:13:33.800 --> 00:13:38.200 filling the gap. So 00:13:38.360 --> 00:13:41.637 >> Sounds dumb, but can you it's 00:13:41.720 --> 00:13:45.837 it is possible that you can solve all of intelligence in this way. Just for both 00:13:45.920 --> 00:13:48.237 language, just 00:13:48.320 --> 00:13:50.997 give a sentence and continue it or give 00:13:51.080 --> 00:13:56.077 a sentence and there's a gap in it, some words blanked out and you fill in 00:13:56.160 --> 00:13:59.837 what words go there. For vision, you give a sequence of 00:13:59.920 --> 00:14:04.157 images and predict what's going to happen next or you fill in what happened 00:14:04.240 --> 00:14:07.277 in between. Do you think it's possible that 00:14:07.360 --> 00:14:10.557 formulation alone 00:14:10.640 --> 00:14:14.957 as a signal for self-supervised learning can solve intelligence for vision and 00:14:15.040 --> 00:14:16.797 language? >> I think that's a bit short at the 00:14:16.880 --> 00:14:20.357 moment. Um so whether this will 00:14:20.440 --> 00:14:23.837 take us all the way to, you know, human-level intelligence or something or 00:14:23.920 --> 00:14:28.357 just cat-level intelligence uh is not clear, but among all the 00:14:28.440 --> 00:14:32.477 possible approaches that people have proposed, I think it's a bit short. So 00:14:32.560 --> 00:14:35.157 I think this idea of 00:14:35.240 --> 00:14:37.077 an intelligent system filling in the 00:14:37.160 --> 00:14:39.837 blanks, either, you know, predicting the 00:14:39.920 --> 00:14:44.477 future, inferring the past, filling in missing information, 00:14:44.560 --> 00:14:48.637 uh you know, I'm currently filling the blank of what is behind your head and 00:14:48.720 --> 00:14:52.757 what you what your head looks like and, you know, from from the back. Uh because 00:14:52.840 --> 00:14:55.677 I have, you know, basic knowledge about how humans are made. 00:14:55.760 --> 00:14:58.237 And I don't know if you're going to, you know, what where you're going to say, at 00:14:58.320 --> 00:14:59.997 which point you're going to speak, whether you're going to move your head 00:15:00.080 --> 00:15:02.917 this way or that way, which way you're going to look, but I know you're not 00:15:03.000 --> 00:15:06.157 going to just dematerialize and reappear 3 m 00:15:06.240 --> 00:15:10.237 down the hall. Uh you know, because I know what's possible and what's 00:15:10.320 --> 00:15:12.957 impossible uh according to intuitive physics. 00:15:13.040 --> 00:15:15.877 >> But you have a model of what's possible and impossible and then you'd be very 00:15:15.960 --> 00:15:19.757 surprised if it happens and then then you'll have to reconstruct your model. 00:15:19.840 --> 00:15:23.277 >> Right. So that that's the model of the world. It's what tells you you know what 00:15:23.360 --> 00:15:26.957 fills in the blanks. So given your partial information about the state of 00:15:27.040 --> 00:15:31.997 the world given by your perception your your model of the world fills in 00:15:32.080 --> 00:15:35.677 the missing information and that includes predicting the future, 00:15:35.760 --> 00:15:38.837 retrodicting the past, you know, and filling in things you 00:15:38.920 --> 00:15:42.597 don't immediately perceive. >> And that doesn't have to be purely 00:15:42.680 --> 00:15:46.717 generic vision or visual information or generic language. You can go to 00:15:46.800 --> 00:15:50.877 specifics like uh predicting 00:15:50.960 --> 00:15:53.757 what control decision you make when you're driving in a lane. You have a 00:15:53.840 --> 00:15:57.557 sequence of images from a a vehicle 00:15:57.640 --> 00:16:00.477 and then you can you have information if 00:16:00.560 --> 00:16:04.917 you record it on video where the car ended up going so you can go back in 00:16:05.000 --> 00:16:08.757 time and predict where the car went based on the visual information. That's 00:16:08.840 --> 00:16:11.317 very specific, domain specific. 00:16:11.400 --> 00:16:12.917 >> Right, but the question is whether we 00:16:13.000 --> 00:16:15.357 can come up with sort of a generic 00:16:15.440 --> 00:16:18.597 uh method for you know, training 00:16:18.680 --> 00:16:20.437 machines to do this kind of prediction 00:16:20.520 --> 00:16:23.437 or filling in the blanks. So right now 00:16:23.520 --> 00:16:28.117 uh this type of approach has been unbelievably successful in the context 00:16:28.200 --> 00:16:30.037 of natural language processing. 00:16:30.120 --> 00:16:34.197 Every modern natural language processing is pre-trained in self-supervised manner 00:16:34.280 --> 00:16:37.677 to fill in the blanks. So you you show it a sequence of words, you remove 10% 00:16:37.760 --> 00:16:40.677 of them and then you train some gigantic neural net to predict the words that are 00:16:40.760 --> 00:16:44.157 missing. That and once you pre-train that 00:16:44.240 --> 00:16:46.277 network, you can use the 00:16:46.360 --> 00:16:48.677 internal representation learned by it as 00:16:48.760 --> 00:16:54.077 input to you know, something that you train supervised or or whatever. Um 00:16:54.160 --> 00:16:57.877 that's been incredibly successful, not so successful in images although it's 00:16:57.960 --> 00:17:02.597 making progress. And uh and it's it's based on uh sort of 00:17:02.680 --> 00:17:06.797 manual data augmentation. Uh we can go into this later, but what has not been 00:17:06.880 --> 00:17:09.077 successful yet is training from video. 00:17:09.160 --> 00:17:13.437 So getting a machine to learn to represent the visual world for example 00:17:13.520 --> 00:17:16.757 by just watching video. Nobody has really succeeded in doing this. 00:17:16.840 --> 00:17:19.757 >> Okay, well, let's kind of give a high-level overview. What's the 00:17:19.840 --> 00:17:21.997 difference in 00:17:22.080 --> 00:17:24.917 kind and in difficulty between vision 00:17:25.000 --> 00:17:29.477 and language? So, you said people haven't been able to 00:17:29.560 --> 00:17:33.397 really kind of crack the problem of vision open in terms of self-supervised 00:17:33.480 --> 00:17:36.677 learning, but that may not be necessarily because it's fundamentally 00:17:36.760 --> 00:17:38.957 more difficult. Maybe like when we're 00:17:39.040 --> 00:17:42.197 talking about achieving like passing 00:17:42.280 --> 00:17:46.037 the Turing test in the full spirit of the Turing test in language might be 00:17:46.120 --> 00:17:49.917 harder than vision. That's That's not obvious. So, what In your view, what 00:17:50.000 --> 00:17:55.117 which is harder or perhaps are they just the same problem? When uh 00:17:55.200 --> 00:17:58.597 the farther we get to solving each, the more we realize it's all the same thing. 00:17:58.680 --> 00:18:02.317 It's all the same cake. >> I think what I'm looking for are methods that 00:18:02.400 --> 00:18:05.877 make make them look essentially like the same cake, but 00:18:05.960 --> 00:18:10.637 currently they're not. And the main issue with learning world models or learning 00:18:10.720 --> 00:18:12.077 predictive models 00:18:12.160 --> 00:18:15.917 is that uh the prediction is never a 00:18:16.000 --> 00:18:20.957 single thing cuz the world is not entirely predictable. 00:18:21.040 --> 00:18:22.597 >> Yeah. >> It may be deterministic or stochastic. 00:18:22.680 --> 00:18:25.997 We can get into the philosophical discussion about it, but uh 00:18:26.080 --> 00:18:29.357 but even if it's deterministic, it's not entirely predictable. 00:18:29.440 --> 00:18:34.597 And so if I play a short video clip and then I ask you to 00:18:34.680 --> 00:18:37.357 predict what's going to happen next, there's many, many plausible 00:18:37.440 --> 00:18:42.437 continuations for that uh video clip, and the number of continuation grows 00:18:42.520 --> 00:18:45.477 with the interval of time that you're asking the 00:18:45.560 --> 00:18:48.397 system to make a prediction for. 00:18:48.480 --> 00:18:52.757 And so, one big question with self-supervised learning is how you 00:18:52.840 --> 00:18:57.077 represent this uncertainty, how you represent multiple discrete outcomes, 00:18:57.160 --> 00:18:59.197 how you represent a sort of continuum of 00:18:59.280 --> 00:19:00.877 possible outcomes, 00:19:00.960 --> 00:19:04.757 um etc. And you know, if you are a sort 00:19:04.840 --> 00:19:08.357 of a classical machine learning person, you say, "Oh, you just represent a 00:19:08.440 --> 00:19:13.477 distribution, right?" And that we know how to do when we're 00:19:13.560 --> 00:19:17.037 predicting words, missing words in the text because 00:19:17.120 --> 00:19:20.517 uh you can have a neural net give a score for every word in the dictionary. 00:19:20.600 --> 00:19:24.357 It's a big, you know, it's a big list of numbers, you know, maybe 100,000 or so. 00:19:24.440 --> 00:19:26.157 You can turn them into a a probability 00:19:26.240 --> 00:19:28.117 distribution that gives that tells you 00:19:28.200 --> 00:19:33.477 when I say a sentence, you know, the you know, the cat is chasing the blank 00:19:33.560 --> 00:19:35.917 in the kitchen. You know, there are only a few words 00:19:36.000 --> 00:19:39.957 that make sense there. You know, it could be a mouse, it could be a lizard 00:19:40.040 --> 00:19:42.207 spot or you know, something like that, right? 00:19:42.290 --> 00:19:46.157 >> [snorts] >> Uh and and if I if I say the the blank is 00:19:46.240 --> 00:19:49.677 chasing the blank in the savanna, you also have a bunch of plausible options 00:19:49.760 --> 00:19:53.797 for those two words, right? Um that that that because you you have 00:19:53.880 --> 00:19:56.877 kind of a you know, underlying reality that you can refer to to sort of fill in 00:19:56.960 --> 00:20:01.357 those those blanks. Um so 00:20:01.440 --> 00:20:05.837 you cannot say for sure in the savanna if it's a you know, a lion or a cheetah 00:20:05.920 --> 00:20:08.237 or whatever. You cannot know if it's a 00:20:08.320 --> 00:20:11.437 zebra or gnu or you know, whatever. 00:20:11.520 --> 00:20:14.437 Wildebeest, same thing. Um 00:20:14.520 --> 00:20:17.237 but uh 00:20:17.320 --> 00:20:21.677 but you can represent the uncertainty by just a a long list of numbers. Now, 00:20:21.760 --> 00:20:23.637 if I if I do the same thing with video 00:20:23.720 --> 00:20:26.197 and I ask you to predict a video clip, 00:20:26.280 --> 00:20:30.717 it's not a discrete set of potential frames. You have to have 00:20:30.800 --> 00:20:34.037 somewhere representing a sort of infinite number of plausible 00:20:34.120 --> 00:20:38.437 continuations of multiple frames in a you know, high-dimensional continuous 00:20:38.520 --> 00:20:42.037 space. And we just have no idea how to do this properly. 00:20:42.120 --> 00:20:44.797 >> Uh finite high-dimensional. 00:20:44.880 --> 00:20:47.197 So, like you cuz >> It's finite high-dimensional, yes. 00:20:47.280 --> 00:20:50.597 >> Just like the words. Like try to get it to uh 00:20:50.680 --> 00:20:56.157 down to a a small finite set of like under a million, something like that. 00:20:56.240 --> 00:20:59.957 >> Something like that. >> I mean, it's kind of ridiculous that we're 00:21:00.040 --> 00:21:04.757 doing a distribution over every single possible word for language and it works. 00:21:04.840 --> 00:21:07.757 It It like that's a really dumb way to do it. 00:21:07.840 --> 00:21:10.557 Um like there seems to there there seems 00:21:10.640 --> 00:21:14.997 to be like there should be some more compressed representation of the 00:21:15.080 --> 00:21:18.037 distribution over the words. >> You're right about that. 00:21:18.120 --> 00:21:21.117 >> And so >> I agree. >> do you have any interesting ideas about 00:21:21.200 --> 00:21:24.637 how to represent all of reality in a compressed way such you can form a 00:21:24.720 --> 00:21:26.797 distribution over it? >> That's one of the big questions, you 00:21:26.880 --> 00:21:28.237 know, how do how do you do that? Right. 00:21:28.320 --> 00:21:32.557 I mean what's what's kind of you know, another thing that that really is 00:21:32.640 --> 00:21:36.917 stupid about um I shouldn't say stupid but like simplistic about current 00:21:37.000 --> 00:21:41.317 approaches to self-supervision in in NLP in text 00:21:41.400 --> 00:21:45.957 is that not only do you represent a joint distribution over words but for 00:21:46.040 --> 00:21:48.837 multiple words that are missing those distributions are essentially 00:21:48.920 --> 00:21:51.397 independent of each other. 00:21:51.480 --> 00:21:53.677 Um and you know, you don't pay too much 00:21:53.760 --> 00:21:58.597 of a price for this. So you So you can't So that you know, the system you know, 00:21:58.680 --> 00:22:01.517 in the the the sentence that I gave earlier 00:22:01.600 --> 00:22:05.837 uh if it gives a a certain probability for a lion and and cheetah and then a 00:22:05.920 --> 00:22:07.717 certain probability for 00:22:07.800 --> 00:22:12.917 uh you know, gazelle uh wildebeest and and and zebra 00:22:13.000 --> 00:22:17.477 uh those two probabilities are independent of each other. 00:22:17.560 --> 00:22:20.637 Uh and it's not the case that those things are independent. Lions actually 00:22:20.720 --> 00:22:25.197 attack like bigger animals than than cheetahs. So you know, there's a huge 00:22:25.280 --> 00:22:29.757 independence hypothesis in in this process which is not actually true. The 00:22:29.840 --> 00:22:33.637 reason for this is that we don't know how to represent uh 00:22:33.720 --> 00:22:38.077 properly distributions over combinatorial uh sequences of symbols 00:22:38.160 --> 00:22:41.037 essentially when they're because the number grows exponentially with the 00:22:41.120 --> 00:22:44.597 length of the of the symbols. And so we have to use tricks for this 00:22:44.680 --> 00:22:49.717 but um those techniques can you know, get around like don't even deal with it. 00:22:49.800 --> 00:22:52.837 So So the big question is like would there 00:22:52.920 --> 00:22:57.557 be some sort of abstract latent representation of text 00:22:57.640 --> 00:22:59.837 that would say that you know, when I 00:22:59.920 --> 00:23:02.997 when I switch lion for gazelle uh lion 00:23:03.080 --> 00:23:07.437 for cheetah, I also have to switch zebra for gazelle. 00:23:07.520 --> 00:23:10.877 >> Yeah, so this independence assumption let me 00:23:10.960 --> 00:23:14.877 throw some criticism at you that I often hear and see how you respond. 00:23:14.960 --> 00:23:18.957 So, this kind of filling in the blanks is just statistics. You're not learning 00:23:19.040 --> 00:23:20.837 anything 00:23:20.920 --> 00:23:23.517 like the deep underlying concepts. 00:23:23.600 --> 00:23:27.077 You're just mimicking stuff from 00:23:27.160 --> 00:23:30.597 the past. You're not learning anything new such that you can use it to 00:23:30.680 --> 00:23:33.357 generalize about the world. 00:23:33.440 --> 00:23:35.637 Uh or okay, let me just say the crude 00:23:35.720 --> 00:23:38.157 version, which is it's just statistics. 00:23:38.240 --> 00:23:42.197 It's not intelligence. Uh what do you have to say to that? What do you usually 00:23:42.280 --> 00:23:44.557 say to that if you kind of hear this kind of thing? 00:23:44.640 --> 00:23:47.197 >> I don't get into those discussions because they are they're kind of 00:23:47.280 --> 00:23:50.637 pointless. Um so, first of all, it's quite possible 00:23:50.720 --> 00:23:52.397 that intelligence is just statistics. 00:23:52.480 --> 00:23:54.637 It's just statistics of a particular kind. 00:23:54.720 --> 00:23:58.637 >> Yes. >> Uh where >> the philosophical question. Is Is kind of is 00:23:58.720 --> 00:24:02.197 is intel- Is it possible that intelligence is just statistics? 00:24:02.280 --> 00:24:06.797 >> Yeah. But, what kind of statistics? So, uh if 00:24:06.880 --> 00:24:10.917 you are asking the question are the model of the world the models of 00:24:11.000 --> 00:24:14.717 the world that we learn, um do they have some notion of causality? Yes. 00:24:14.800 --> 00:24:17.317 >> Mhm. >> So, if the criticism comes from people 00:24:17.400 --> 00:24:20.637 who say you know, uh current machine learning system don't care about 00:24:20.720 --> 00:24:25.637 causality, which by the way is wrong. Uh you know, I agree with them. Yeah, you 00:24:25.720 --> 00:24:29.877 should, you know, your model of the world should have your actions as one of your 00:24:29.960 --> 00:24:32.557 the inputs. And that will drive you to learn causal 00:24:32.640 --> 00:24:36.717 models of the world where you know what you know, what uh intervention in the 00:24:36.800 --> 00:24:41.237 world will cause what result. Or you can do this by observation of other agents 00:24:41.320 --> 00:24:44.437 uh acting in the world and and observing the effect. 00:24:44.520 --> 00:24:48.197 Uh other humans, for example. So, I think 00:24:48.280 --> 00:24:52.797 you know, at some level of description, uh intelligence is just statistics. 00:24:52.880 --> 00:24:57.077 Uh but that doesn't mean you don't you don't you know, you won't have models 00:24:57.160 --> 00:25:02.037 that have, you know, deep mechanistic explanation for what goes on. Uh, the 00:25:02.120 --> 00:25:03.597 question is how do you learn them? 00:25:03.680 --> 00:25:05.797 That's that's the question I'm interested in. 00:25:05.880 --> 00:25:09.197 Uh, because you know, a lot of people who actually 00:25:09.280 --> 00:25:13.277 voiced a criticism say that those mechanistic model has to 00:25:13.360 --> 00:25:15.997 have to come from someplace else. They have to come from human designers, they 00:25:16.080 --> 00:25:18.037 have to come from, I don't know what. 00:25:18.120 --> 00:25:22.597 And obviously, we learn them. Uh, or if we don't learn them as an 00:25:22.680 --> 00:25:27.557 individual, nature learn them for us using evolution. So, 00:25:27.640 --> 00:25:29.237 regardless of what you think, those 00:25:29.320 --> 00:25:32.077 processes have been learned somehow. 00:25:32.160 --> 00:25:34.877 >> So, if you look at the the human brain, 00:25:34.960 --> 00:25:39.477 just like when we humans introspect about how the brain works, it seems like 00:25:39.560 --> 00:25:43.437 when we think about what is intelligence, we think about the 00:25:43.520 --> 00:25:46.957 high-level stuff like the models we've constructed, concepts like cognitive 00:25:47.040 --> 00:25:51.717 science, like concepts of memory and reasoning module, almost like these 00:25:51.800 --> 00:25:55.317 high-level modules. Is there's Is this serve as a good 00:25:55.400 --> 00:25:57.237 analogy? 00:25:57.320 --> 00:26:01.677 Like, are we ignoring the uh 00:26:01.760 --> 00:26:06.477 the dark matter, the the basic low-level mechanisms, just like we ignore the way 00:26:06.560 --> 00:26:09.357 the operating system works, we're just using the 00:26:09.440 --> 00:26:11.637 uh the the high-level software, we're 00:26:11.720 --> 00:26:16.637 ignoring that at the low level, the neural network might be doing something 00:26:16.720 --> 00:26:20.957 like statistics. Like, I mean, sorry to use this word 00:26:21.040 --> 00:26:24.557 probably incorrectly and crudely, but doing this kind of filling the gap kind 00:26:24.640 --> 00:26:28.357 of learning and just kind of updating the model constantly in order to be able 00:26:28.440 --> 00:26:32.637 to support the raw sensory information mation to predict it and then adjust to 00:26:32.720 --> 00:26:36.197 the prediction when it's wrong. But like, high level when we look at our 00:26:36.280 --> 00:26:40.237 brain at the high level, it feels like we're doing like we're playing chess, 00:26:40.320 --> 00:26:44.837 like we're we're like playing with high-level concepts and we're stitching 00:26:44.920 --> 00:26:48.797 them together and we're putting them into long-term memory, but really what's 00:26:48.880 --> 00:26:52.837 going underneath is something we're not able to introspect, which is this kind 00:26:52.920 --> 00:26:57.957 of uh uh simple large neural network that's just filling in the gaps. 00:26:58.040 --> 00:27:01.757 >> Right. Well, okay, so there's a lot of questions and a lot of answers there. 00:27:01.840 --> 00:27:04.597 Okay, so first of all, there's a whole school of thought in neuroscience, 00:27:04.680 --> 00:27:06.837 computational neuroscience in particular, 00:27:06.920 --> 00:27:11.717 um that likes the idea of predictive coding, which is really related to the 00:27:11.800 --> 00:27:14.437 idea I was talking about in self-supervised learning. So, everything 00:27:14.520 --> 00:27:17.917 is about prediction. The essence of intelligence is the ability to predict. 00:27:18.000 --> 00:27:21.037 >> Yeah. >> And everything the brain does is trying to predict 00:27:21.120 --> 00:27:24.037 uh predict everything from everything else. 00:27:24.120 --> 00:27:28.077 Okay, and that that's really sort of the underlying principle, if you want, that 00:27:28.160 --> 00:27:32.077 uh self-supervised learning is trying to kind of reproduce this idea of 00:27:32.160 --> 00:27:34.997 prediction as kind of an essential mechanism 00:27:35.080 --> 00:27:38.237 of uh task-independent learning, if you want. 00:27:38.320 --> 00:27:41.437 The next step is what kind of intelligence are you 00:27:41.520 --> 00:27:45.437 interested in reproducing? And of course, you know, we all think about, 00:27:45.520 --> 00:27:49.197 you know, trying to reproduce sort of, you know, high-level cognitive processes 00:27:49.280 --> 00:27:53.477 in humans. But like with machines, we're not even at the level of 00:27:53.560 --> 00:27:58.237 even reproducing the learning processes in a in a cat brain. 00:27:58.320 --> 00:28:01.757 Um you know, the most intelligent of our intelligent systems don't don't have as 00:28:01.840 --> 00:28:04.597 much common sense as as a house cat. 00:28:04.680 --> 00:28:07.437 >> [snorts] >> So, um how is it that cats learn? And 00:28:07.520 --> 00:28:10.757 you know, cats don't do a whole lot of uh reasoning. They certainly have causal 00:28:10.840 --> 00:28:14.917 models. They certainly have uh because, you know, many cats can 00:28:15.000 --> 00:28:16.317 figure out like how they can act on the 00:28:16.400 --> 00:28:18.637 world to get what they want. Um they 00:28:18.720 --> 00:28:21.997 certainly have uh a fantastic model of 00:28:22.080 --> 00:28:25.757 intuitive physics, uh certainly of the the the dynamics of 00:28:25.840 --> 00:28:29.757 their own bodies, but but also of prey and things like that, right? So, 00:28:29.840 --> 00:28:31.797 um they they're they're pretty smart. 00:28:31.880 --> 00:28:35.037 They They only do this with about 800 million neurons. 00:28:35.120 --> 00:28:38.117 Um we are not anywhere close to 00:28:38.200 --> 00:28:40.117 reproducing this kind of uh thing. So, 00:28:40.200 --> 00:28:43.237 to some extent, I I could I could say 00:28:43.320 --> 00:28:47.557 let's not even worry about like the high-level cognition 00:28:47.640 --> 00:28:50.357 uh and kind of, you know, long-term planning and reasoning that humans can 00:28:50.440 --> 00:28:53.357 do until we figure out like, you know, can we even reproduce what Cassio is 00:28:53.440 --> 00:28:54.357 doing? 00:28:54.440 --> 00:28:57.917 Now, that said, this ability to learn 00:28:58.000 --> 00:29:01.237 world models, I think is the key to the 00:29:01.320 --> 00:29:05.677 possibility of learning machines that can also reason. So, whenever I give a 00:29:05.760 --> 00:29:08.597 talk, I say there are there are three challenges in the three main challenges 00:29:08.680 --> 00:29:11.917 in machine learning. The first one is uh you know, getting machines to learn to 00:29:12.000 --> 00:29:15.997 represent the world um and I'm proposing self-supervised learning. 00:29:16.080 --> 00:29:20.277 >> [snorts] >> Uh the second is uh getting machines to reason in ways 00:29:20.360 --> 00:29:23.517 that are compatible with essentially gradient-based learning 00:29:23.600 --> 00:29:26.437 because this is what deep learning is all about, really. 00:29:26.520 --> 00:29:28.717 Um [snorts] and the third one is something we have 00:29:28.800 --> 00:29:33.197 no idea how to solve, at least I have no idea how to solve, is uh 00:29:33.280 --> 00:29:37.117 can we get machines to learn hierarchical representations of action 00:29:37.200 --> 00:29:41.517 plans? You know, like, you know, we know how to train them to learn hierarchical 00:29:41.600 --> 00:29:43.507 representations of perception. 00:29:43.590 --> 00:29:45.237 >> [snorts] >> You know, we accomplish on that sensing 00:29:45.320 --> 00:29:49.357 like that and so on. But, what about action plans? Can we get them to 00:29:49.440 --> 00:29:52.397 spontaneously learn good hierarchical representations of actions? 00:29:52.480 --> 00:29:56.317 >> Also gradient-based. >> Yeah, all of that has, you know, needs 00:29:56.400 --> 00:29:59.517 to be somewhat differentiable so that you can apply sort of gradient-based 00:29:59.600 --> 00:30:03.147 learning, uh which is really what deep learning is about. 00:30:03.230 --> 00:30:03.877 >> [snorts] 00:30:03.960 --> 00:30:07.837 >> So, it's background knowledge, ability 00:30:07.920 --> 00:30:11.317 to reason in a way uh 00:30:11.400 --> 00:30:15.717 that's differentiable that is somehow connected deeply integrated 00:30:15.800 --> 00:30:19.637 with our background knowledge or builds on top of that background knowledge and 00:30:19.720 --> 00:30:23.157 then given that background knowledge, be able to make hierarchical plans. 00:30:23.240 --> 00:30:26.877 >> Right. >> So, if if you take classical optimal 00:30:26.960 --> 00:30:30.637 control, there's something in classical optimal control called uh 00:30:30.720 --> 00:30:34.277 model predictive control and it's, you know, it's been around since the the 00:30:34.360 --> 00:30:37.677 early '60s. NASA uses that to compute trajectories 00:30:37.760 --> 00:30:42.517 of rockets. And the basic idea is that you You a pre- a predictive model >> [snorts] 00:30:42.600 --> 00:30:47.037 >> of the rocket, let's say, or whatever system you are you intend to control, 00:30:47.120 --> 00:30:49.637 >> [snorts] >> which, given the state of the system at 00:30:49.720 --> 00:30:54.237 time t and given an action that you're taking on the system, so for 00:30:54.320 --> 00:30:58.157 a rocket it would be thrust and, you know, all the controls you can have. 00:30:58.240 --> 00:31:01.637 Uh it gives you the state of the system at time t plus delta t, right? So 00:31:01.720 --> 00:31:04.067 basically differential equation, something like that. 00:31:04.150 --> 00:31:08.157 >> [snorts] >> Um and if you have this model and you have 00:31:08.240 --> 00:31:12.997 this model in the form of some sort of neural net or some sort of uh set of 00:31:13.080 --> 00:31:16.037 formula that you can back propagate gradient through, you can do what's 00:31:16.120 --> 00:31:19.277 called model predictive control or gradient based uh 00:31:19.360 --> 00:31:22.077 model predictive control. So, you have 00:31:22.160 --> 00:31:25.077 uh you can unroll that uh 00:31:25.160 --> 00:31:27.917 that model in time, you you 00:31:28.000 --> 00:31:30.157 you you feed it a 00:31:30.240 --> 00:31:33.037 hypothesized sequence of actions, 00:31:33.120 --> 00:31:37.197 and then you have some objective function that measures how well at the 00:31:37.280 --> 00:31:40.397 end of the trajectory the system has uh succeeded or matched what you wanted to 00:31:40.480 --> 00:31:44.517 do. Um you know, is the robot harmed? Have you grasped the object you want to 00:31:44.600 --> 00:31:47.037 grasp? If it's a rocket, you know, are 00:31:47.120 --> 00:31:51.757 you at the right place near the space station? Things like that. Um and by 00:31:51.840 --> 00:31:54.957 back propagation through time, and again this was invented in the 1960s by 00:31:55.040 --> 00:31:58.837 optimal control theories, uh you can figure out 00:31:58.920 --> 00:32:02.357 uh what is the optimal sequence of actions that will 00:32:02.440 --> 00:32:07.077 uh you know, get my system to the the best final state. 00:32:07.160 --> 00:32:11.277 So, that's a form of reasoning. It's basically planning, and a lot of 00:32:11.360 --> 00:32:14.637 planning uh systems in robotics are actually based on this. 00:32:14.720 --> 00:32:17.597 Uh and uh uh and you can think of this as a form 00:32:17.680 --> 00:32:21.957 of reasoning. So, you know, to take the example of the teenager driving a car 00:32:22.040 --> 00:32:25.397 again, you have a pretty good dynamical model of the car. It doesn't need to be 00:32:25.480 --> 00:32:29.357 very accurate, but you know, again, that if you turn the wheel to the right and 00:32:29.440 --> 00:32:31.757 there is a cliff, you're going to run off the cliff, right? You don't need to 00:32:31.840 --> 00:32:34.077 have a very accurate model to predict that. 00:32:34.160 --> 00:32:38.037 And you can run this in your mind and decide not to do it for that reason. 00:32:38.120 --> 00:32:40.757 Because you can predict in advance that the result is going to be bad. So, you 00:32:40.840 --> 00:32:45.637 can sort of imagine different scenarios, and uh and then, you know, employ 00:32:45.720 --> 00:32:48.437 uh or take the first step in the scenario that is most favorable, and 00:32:48.520 --> 00:32:49.757 then repeat the process of planning. 00:32:49.840 --> 00:32:53.037 That's called receding horizon model predictive control. So, you know, you 00:32:53.120 --> 00:32:57.757 know, all those things have names, you know, uh going back, you know, decades. Um 00:32:57.840 --> 00:33:02.317 and so, if you're not not the you know, in classical optimal control, the model of 00:33:02.400 --> 00:33:04.717 the world is not generally learned. 00:33:04.800 --> 00:33:07.557 Uh this, you know, sometimes a few parameters you have to identify. That's 00:33:07.640 --> 00:33:11.477 called systems identification. But, uh but generally, 00:33:11.560 --> 00:33:14.877 the model is mostly deterministic and mostly built by 00:33:14.960 --> 00:33:19.237 hand. So, the big question of AI, uh I think the big challenge of AI for 00:33:19.320 --> 00:33:23.037 the next decade is how do we get machines to learn predictive models of 00:33:23.120 --> 00:33:27.077 the world that deal with uncertainty, and deal with the real world in all its 00:33:27.160 --> 00:33:31.317 complexity. So, it's not just trajectory of a rocket, which you can reduce to 00:33:31.400 --> 00:33:35.117 first principles. It's not It's not even just a trajectory of a robot arm, which 00:33:35.200 --> 00:33:38.517 again, you can model by, you know, careful mathematics. But, it's 00:33:38.600 --> 00:33:42.797 everything else. Everything we observe in the world, you know, people behavior, 00:33:42.880 --> 00:33:45.997 um you know, physical systems that involve 00:33:46.080 --> 00:33:50.957 collective phenomena, like water, or or, you know, 00:33:51.040 --> 00:33:56.277 trees and you know, branches in a in a tree or something, or or uh like complex 00:33:56.360 --> 00:33:58.637 things that, you know, humans have no trouble 00:33:58.720 --> 00:34:02.237 developing abstract representations and predictive model for, but we still don't 00:34:02.320 --> 00:34:05.757 know how to do with machines. >> Where do you put in in these three, 00:34:05.840 --> 00:34:09.117 maybe in the in the planning stages, 00:34:09.200 --> 00:34:11.997 the the game theoretic nature of this 00:34:12.080 --> 00:34:16.437 world, where your actions not only respond to the dynamic nature of the 00:34:16.520 --> 00:34:19.477 world, the environment, but also affect it? 00:34:19.560 --> 00:34:24.477 So, if there's other humans involved, is this is this point number four, or is it 00:34:24.560 --> 00:34:28.557 somehow integrated into the hierarchical representation of action, in your view? 00:34:28.640 --> 00:34:31.997 >> I think it's integrated. It's just um it's just that now your model of the 00:34:32.080 --> 00:34:35.037 world has to deal with, you know, it just makes it more complicated, right? 00:34:35.120 --> 00:34:38.037 The fact that uh humans are complicated and not easily 00:34:38.120 --> 00:34:41.997 predictable, that makes your model of the world much more complicated. That 00:34:42.080 --> 00:34:45.437 much more complicated. >> Well, there's a che- I mean, I I suppose 00:34:45.520 --> 00:34:50.077 chess is an analogy. So, Monte Carlo tree search. 00:34:50.160 --> 00:34:53.437 >> Right. >> It There is a I go, you go, I go, you 00:34:53.520 --> 00:34:56.117 go. Like um Andrej Karpathy recently 00:34:56.200 --> 00:34:59.877 gave a talk at MIT about car doors. 00:34:59.960 --> 00:35:03.797 I think there's some machine learning too, but mostly car doors. And there's a 00:35:03.880 --> 00:35:07.677 dynamic nature to the car like the person opening the door checking 00:35:07.760 --> 00:35:10.317 I mean, he wasn't talking about that. He was talking about the perception problem 00:35:10.400 --> 00:35:13.517 of what the the anthology of what defines a car door, this big 00:35:13.600 --> 00:35:17.757 philosophical question. But to me it was interesting cuz like it's obvious that 00:35:17.840 --> 00:35:21.517 the person opening the car doors they're trying to get out like here in New York 00:35:21.600 --> 00:35:25.117 trying to get out of the car. You slowing down is going to signal 00:35:25.200 --> 00:35:28.397 something. You speeding up is going to signal something. And that's a dance. 00:35:28.480 --> 00:35:33.277 It's a asynchronous chess game. I don't know. 00:35:33.360 --> 00:35:37.677 >> [laughter] >> So, I It feels like um 00:35:37.760 --> 00:35:41.397 it's not just I mean, I guess you can integrate all of that into one giant 00:35:41.480 --> 00:35:46.477 model. Like the entirety of the the the these little interactions. Cuz it's 00:35:46.560 --> 00:35:49.677 not as complicated as chess. It's just like a little dance. We do like a little 00:35:49.760 --> 00:35:51.917 dance together. And then we figure it out. 00:35:52.000 --> 00:35:55.757 >> Well, in some ways it's way more complicated than chess because 00:35:55.840 --> 00:35:59.597 uh because it's continuous, it's uncertain in a continuous manner. 00:35:59.680 --> 00:36:01.797 Uh >> It doesn't feel more complicated. 00:36:01.880 --> 00:36:04.997 >> But it doesn't feel more complicated because that's what we are we've evolved 00:36:05.080 --> 00:36:07.757 to solve. This is the kind of problem we've evolved to solve. And so we're 00:36:07.840 --> 00:36:12.437 good at it because, you know, nature has made us good at it. 00:36:12.520 --> 00:36:16.037 Nature has not made us good at chess. We completely suck at chess. 00:36:16.120 --> 00:36:19.437 >> Yeah. >> Um in fact, that's why we designed it as a 00:36:19.520 --> 00:36:20.837 as a game is to be challenging. 00:36:20.920 --> 00:36:23.477 >> Mhm. >> [clears throat] >> Uh but and if there is something that, 00:36:23.560 --> 00:36:28.477 you know, recent progress in the chess and go has made us realize is that humans are 00:36:28.560 --> 00:36:32.197 really terrible at those things. Like really bad, you know? There was a story, 00:36:32.280 --> 00:36:37.037 right, before AlphaGo that uh uh you know, the best Go player thought 00:36:37.120 --> 00:36:40.477 they were maybe two or three stones behind, you know, an ideal player that 00:36:40.560 --> 00:36:44.757 they would call God. Uh in fact, no, they are like nine or 10 00:36:44.840 --> 00:36:47.277 stones behind. I mean, we're just bad. 00:36:47.360 --> 00:36:52.237 So, we're not good at and it's because we have limited working memory. We know 00:36:52.320 --> 00:36:55.717 we're not very good at like doing this tree exploration. You know, computers 00:36:55.800 --> 00:36:59.397 are much better um at doing than we are. But we are much 00:36:59.480 --> 00:37:02.477 better at learning differentiable models of the world. 00:37:02.560 --> 00:37:06.917 I mean, I say differentiable in a kind of you know, I should say 00:37:07.000 --> 00:37:09.317 not differentiable in the sense that, you know, we run backprop through it, 00:37:09.400 --> 00:37:14.077 but in the sense that uh our brain has some mechanism for estimating gradients 00:37:14.160 --> 00:37:15.957 of of some kind. >> Yeah. 00:37:16.040 --> 00:37:17.917 >> And that's what, you know, makes us 00:37:18.000 --> 00:37:21.117 efficient. So, if you have an agent that 00:37:21.200 --> 00:37:25.557 consists of a a model of the world, which, you know, 00:37:25.640 --> 00:37:28.557 in the human brain is basically the entire front half of your brain. 00:37:28.640 --> 00:37:33.357 >> [snorts] >> Um an objective function which 00:37:33.440 --> 00:37:37.477 uh in human in in humans is a combination of two things. There is your 00:37:37.560 --> 00:37:40.877 sort of intrinsic motivation module, which is in the basal ganglia, you know, 00:37:40.960 --> 00:37:44.517 at the base of your brain. That's the thing that measures pain and hunger and 00:37:44.600 --> 00:37:47.357 things like that, like immediate um 00:37:47.440 --> 00:37:51.917 uh feelings and emotions. Um and then there is 00:37:52.000 --> 00:37:54.437 you know, the equivalent of what people in reinforcement learning call a critic, 00:37:54.520 --> 00:37:57.517 which is a sort of module that predicts 00:37:57.600 --> 00:38:01.557 ahead what the outcome of a 00:38:01.640 --> 00:38:05.677 of a situation will be. And so, it's it's not a cost function, 00:38:05.760 --> 00:38:09.557 but it's sort of not an objective function, but it's sort of a 00:38:09.640 --> 00:38:13.117 you know, trained predictor of the ultimate objective function. And that 00:38:13.200 --> 00:38:16.917 also is differentiable. And so, if all of this is differentiable, your cost 00:38:17.000 --> 00:38:19.517 function, your your critic, your 00:38:19.600 --> 00:38:22.317 uh Uh, you know, your your world model. 00:38:22.400 --> 00:38:26.877 Then you can use gradient-based type methods to do planning, to do reasoning, 00:38:26.960 --> 00:38:32.677 to do learning, you know, to do all the things that we'd like an intelligent agent 00:38:32.760 --> 00:38:36.117 uh, to do. >> And the gradient-based learning 00:38:36.200 --> 00:38:39.677 like what's your intuition? That's probably at the core of what can solve 00:38:39.760 --> 00:38:41.997 intelligence. So, you don't need 00:38:42.080 --> 00:38:47.517 like um logic-based reasoning uh, in your view. 00:38:47.600 --> 00:38:50.717 >> I don't know how to make logic-based reasoning compatible with 00:38:50.800 --> 00:38:52.077 uh, efficient learning. >> Yeah. 00:38:52.160 --> 00:38:55.197 >> And Okay. I mean, there is a big question, perhaps a philosophical 00:38:55.280 --> 00:38:58.117 question. I mean, it's not that philosophical, but 00:38:58.200 --> 00:39:01.997 uh, that we can ask is is that, you know, all the learning algorithms we 00:39:02.080 --> 00:39:07.037 know from engineering and computer science proceed by optimizing some objective 00:39:07.120 --> 00:39:12.837 function. >> Yeah. >> Right? Uh, so one question we may ask is uh, 00:39:12.920 --> 00:39:16.637 is does learning in the brain minimize an objective function? 00:39:16.720 --> 00:39:19.637 I mean, it could be a, you know, a composite of multiple 00:39:19.720 --> 00:39:22.877 objective functions, but it's still an objective function. 00:39:22.960 --> 00:39:25.437 Uh, second, if it does optimize an 00:39:25.520 --> 00:39:30.157 objective function, does it do does it do it by uh, some sort of gradient 00:39:30.240 --> 00:39:33.837 estimation? You know, it doesn't need to be backprop, but, you know, some way of 00:39:33.920 --> 00:39:36.757 estimating the gradient in efficient manner. 00:39:36.840 --> 00:39:41.117 Whose complexity is on the same order of magnitude as you know, actually running 00:39:41.200 --> 00:39:44.717 the inference. >> [snorts] >> Uh, 00:39:44.800 --> 00:39:48.197 cuz you can't afford to do things like, you know, perturbing a weight in your 00:39:48.280 --> 00:39:50.037 brain to figure out what the effect is. 00:39:50.120 --> 00:39:52.637 And then sort of uh, you know, you you can't do sort of 00:39:52.720 --> 00:39:54.597 estimating gradient by perturbation. 00:39:54.680 --> 00:39:57.317 It's To me, it seems very implausible 00:39:57.400 --> 00:40:00.797 that the brain uses some sort of uh, 00:40:00.880 --> 00:40:04.917 you know, zeroth-order black box gradient-free optimization 00:40:05.000 --> 00:40:08.717 because it's so much less efficient than gradient optimization. So, it has to 00:40:08.800 --> 00:40:11.157 have a way of estimating gradients. 00:40:11.240 --> 00:40:13.077 >> Is it possible that some kind of 00:40:13.160 --> 00:40:16.397 logic-based reasoning emerges in pockets 00:40:16.480 --> 00:40:20.477 as a useful, like you said, if the brain has an objective function, maybe it's a 00:40:20.560 --> 00:40:23.277 mechanism for creating objective functions. 00:40:23.360 --> 00:40:25.917 It's It's It's a mechanism for um 00:40:26.000 --> 00:40:30.717 creating knowledge bases, for example, that can then be queried. Like maybe 00:40:30.800 --> 00:40:33.517 it's like an efficient representation of knowledge that's learned in a 00:40:33.600 --> 00:40:35.717 gradient-based way or something like that. 00:40:35.800 --> 00:40:39.277 >> Well, so I think there is a lot of different types of uh intelligence. So, 00:40:39.360 --> 00:40:43.157 first of all, I think the type of logical reasoning that we think about um 00:40:43.240 --> 00:40:46.997 that we are >> [snorts] >> you know, maybe stemming from you know, 00:40:47.080 --> 00:40:49.997 sort of classical AI of the 1970s and '80s. 00:40:50.080 --> 00:40:54.917 Um I think humans use that relatively rarely 00:40:55.000 --> 00:40:56.637 and are not particularly good at it. 00:40:56.720 --> 00:40:58.877 >> But we judge each other based on our 00:40:58.960 --> 00:41:02.437 ability to uh solve those rare problems. 00:41:02.520 --> 00:41:07.117 It's called an IQ test. >> so. Like I'm I'm not very good at chess. 00:41:07.200 --> 00:41:10.437 >> Yes, I'm I'm judging you this whole time because of it. 00:41:10.520 --> 00:41:13.317 Well, we we actually >> With With your with your uh you know, 00:41:13.400 --> 00:41:15.187 heritage, I'm sure you're good at chess. 00:41:15.270 --> 00:41:18.117 >> [laughter] >> No, stereotypes. Not all stereotypes are 00:41:18.200 --> 00:41:19.917 true. 00:41:20.000 --> 00:41:22.317 >> Well, I'm terrible at chess. So, um 00:41:22.400 --> 00:41:28.077 you know, but I think perhaps uh another type of intelligence that I have is this 00:41:28.160 --> 00:41:32.597 uh uh you know, ability of sort of building models of the world from 00:41:32.680 --> 00:41:37.557 uh you know, reasoning obvious obviously, but also also data. 00:41:37.640 --> 00:41:41.117 Uh and those those models generally are more kind of analogical, right? So, it's 00:41:41.200 --> 00:41:44.237 it's uh it's reasoning by simulation 00:41:44.320 --> 00:41:48.957 and but by analogy, where you use one model to apply to a new situation. Even 00:41:49.040 --> 00:41:52.997 though you've never seen that situation, you can sort of um connect it to a 00:41:53.080 --> 00:41:57.957 situation you've encountered before. Uh and and your reasoning is more, 00:41:58.040 --> 00:42:01.877 you know, akin to some sort of internal simulation. So, you you are kind of 00:42:01.960 --> 00:42:04.677 simulating what's happening when you're building, I don't know, a box out of 00:42:04.760 --> 00:42:08.317 wood or something, right? You kind of imagine in in advance 00:42:08.400 --> 00:42:10.437 like what would be the result of, you know, cutting the wood in this 00:42:10.520 --> 00:42:14.037 particular way. Are you going to use, you know, screws or nails or whatever. 00:42:14.120 --> 00:42:17.117 Um when you are interacting with someone, you also have a model of that 00:42:17.200 --> 00:42:21.237 person and and sort of interact with that person, 00:42:21.320 --> 00:42:24.917 you know, having this model in mind uh to kind of uh 00:42:25.000 --> 00:42:29.277 tell the person what what you think is useful to them. So, I think this 00:42:29.360 --> 00:42:34.157 this ability to construct models of the world is basically the essence the 00:42:34.240 --> 00:42:35.797 essence of intelligence. 00:42:35.880 --> 00:42:37.917 And the ability to use it then to 00:42:38.000 --> 00:42:41.357 uh plan uh actions that will 00:42:41.440 --> 00:42:44.957 uh fulfill a particular uh criterion um 00:42:45.040 --> 00:42:47.357 of course is is is necessary as well. 00:42:47.440 --> 00:42:51.037 >> So, I'm going to ask you a series of impossible questions as we keep ask as 00:42:51.120 --> 00:42:55.397 I've been doing. So, so if that's the fundamental sort of dark matter of 00:42:55.480 --> 00:42:59.517 intelligence, this ability to form a background model, what's your intuition 00:42:59.600 --> 00:43:03.397 about how much knowledge is required? 00:43:03.480 --> 00:43:07.197 You you know, I think dark matter you you put a percentage of 00:43:07.280 --> 00:43:12.077 uh on it of of uh the composition of the universe and 00:43:12.160 --> 00:43:13.637 how much of it is dark matter, how much 00:43:13.720 --> 00:43:17.117 of it is dark energy. How much 00:43:17.200 --> 00:43:21.877 uh information do you think is required to to to be a house cat? 00:43:21.960 --> 00:43:25.397 So, you have to be able to uh when you see a box, go in it. When you see a 00:43:25.480 --> 00:43:29.957 human, compute the most evil action. If there's a thing that's near an edge, you 00:43:30.040 --> 00:43:34.277 knock it off. All of that plus the extra stuff you 00:43:34.360 --> 00:43:38.317 mentioned, which is a uh great self-awareness of the physics of your of 00:43:38.400 --> 00:43:40.677 your own body and the and the world. 00:43:40.760 --> 00:43:43.997 How much knowledge is required do you think to solve it? 00:43:44.080 --> 00:43:47.597 >> Um >> I don't even know how to measure an answer to that question. 00:43:47.680 --> 00:43:51.077 >> I'm not sure how to measure it, but whatever it is, it fits in about 00:43:51.160 --> 00:43:54.317 about 800,000 neurons. Uh 00:43:54.400 --> 00:43:58.157 800 million neurons, sorry. >> The representation does. 00:43:58.240 --> 00:44:00.917 >> Everything. All knowledge, everything, right? 00:44:01.000 --> 00:44:05.597 Um you know, it's less than a billion. A dog is 2 billion, but a cat is less than 00:44:05.680 --> 00:44:10.277 1 billion. >> Right. And so multiply that by 1,000 and you get 00:44:10.360 --> 00:44:14.237 the number of synapses. And I think almost all of it is is 00:44:14.320 --> 00:44:18.597 learned through this you know a sort of self-supervised learning although you 00:44:18.680 --> 00:44:21.877 know I I think a tiny sliver is learned through reinforcement learning and 00:44:21.960 --> 00:44:25.157 certainly very little through you know classical supervised learning 00:44:25.240 --> 00:44:28.557 although it's not even clear how supervised learning actually works in 00:44:28.640 --> 00:44:33.277 in biological world. So I think almost all of it is 00:44:33.360 --> 00:44:37.557 is self-supervised learning. >> It's driven by 00:44:37.640 --> 00:44:42.477 the the sort of ingrained objective functions that a cat or human have at 00:44:42.560 --> 00:44:45.357 the base of their brain which kind of drives their 00:44:45.440 --> 00:44:49.677 their behavior. So you know nature tells us 00:44:49.760 --> 00:44:53.117 you're hungry. It doesn't tell us how to feed a feed 00:44:53.200 --> 00:44:56.237 ourselves. That's that's something that the rest of our brain has to figure out 00:44:56.320 --> 00:44:58.837 right? >> Well it's interesting cuz there might be 00:44:58.920 --> 00:45:01.517 more like deeper objective functions 00:45:01.600 --> 00:45:06.117 that are lying the whole thing. So hun- hunger may be some kind of 00:45:06.200 --> 00:45:10.277 Now you go to like neurobiology might be just the brain 00:45:10.360 --> 00:45:14.397 um trying to maintain homeostasis. 00:45:14.480 --> 00:45:18.877 So hunger is just one of the human perceivable 00:45:18.960 --> 00:45:23.437 symptoms of the brain being unhappy with the way things are currently. So it 00:45:23.520 --> 00:45:26.877 could be just like one really dumb objective function at the core. 00:45:26.960 --> 00:45:30.357 >> But that's how that's how a behavior is is is driven. 00:45:30.440 --> 00:45:33.957 The fact that you know the orbital ganglia 00:45:34.040 --> 00:45:38.397 drive us to do things that are that are different from say an orangutan or 00:45:38.480 --> 00:45:41.917 certainly a cat is what makes you know human nature 00:45:42.000 --> 00:45:44.877 versus orangutan nature versus cat nature. 00:45:44.960 --> 00:45:46.757 So for example 00:45:46.840 --> 00:45:50.437 you know our basal ganglia drives us to 00:45:50.520 --> 00:45:54.157 seek the company of other humans. 00:45:54.240 --> 00:45:58.037 And that's because nature has figured out that we need to be social animals 00:45:58.120 --> 00:46:03.277 for our species to survive and it's true of of uh uh 00:46:03.360 --> 00:46:07.557 It's not true of orangutans. Orangutans are solitary animals. Um they don't seek 00:46:07.640 --> 00:46:10.677 the company of others. In fact, they avoid them. 00:46:10.760 --> 00:46:13.317 Uh in fact, they scream at them when they come too close because they're 00:46:13.400 --> 00:46:17.477 territorial. Uh cuz for for their survival, you know, 00:46:17.560 --> 00:46:20.197 uh evolution has figured out that's the best thing. 00:46:20.280 --> 00:46:24.157 I mean, they're occasionally social, of course, for, you know, um 00:46:24.240 --> 00:46:28.757 reproduction and stuff like that. But um but but they're mostly solitary. So, 00:46:28.840 --> 00:46:31.837 uh so all of those behaviors are not part of intelligence. You know, people 00:46:31.920 --> 00:46:34.197 say, "Oh, you're never going to have intelligent machines because, you know, 00:46:34.280 --> 00:46:37.757 human intelligence is social." But then you look at orangutans, you look at 00:46:37.840 --> 00:46:40.717 octopus. Octopus never know their parents. 00:46:40.800 --> 00:46:44.517 They barely interact with any other. And they And they get to be really smart in 00:46:44.600 --> 00:46:47.357 less than a less than a year, in like half a year. 00:46:47.440 --> 00:46:49.517 Um you know, in a year or they're adults. 00:46:49.600 --> 00:46:52.717 In 2 years, they're dead. So, um 00:46:52.800 --> 00:46:56.677 there are things that we think as humans are intimately linked with 00:46:56.760 --> 00:46:59.677 intelligence, like social interaction, 00:46:59.760 --> 00:47:01.677 like language. 00:47:01.760 --> 00:47:06.757 We think I think we give way too much importance to language as a substrate of 00:47:06.840 --> 00:47:10.357 intelligence as humans. Because we think our reasoning is so 00:47:10.440 --> 00:47:14.197 linked with language. >> So, for for to solve the house cat 00:47:14.280 --> 00:47:18.317 intelligence problem, you think you could do it on a desert island. You could have 00:47:18.400 --> 00:47:22.677 >> Pretty much. >> You could just have a cat sitting there um 00:47:22.760 --> 00:47:27.637 looking at the waves at the ocean waves and figure a lot of it out. 00:47:27.720 --> 00:47:33.157 >> It needs to have sort of, you know, the right set of drives uh to kind of, 00:47:33.240 --> 00:47:37.637 you know, get it to do the thing and learn the appropriate things, right? But uh 00:47:37.720 --> 00:47:42.557 um like for example, you know, uh baby humans are are driven to 00:47:42.640 --> 00:47:46.517 learn to stand up and walk. Okay, you know, it's not that's kind of 00:47:46.600 --> 00:47:50.037 this desire is hardwired. How to do it precisely is not. That's learned. 00:47:50.120 --> 00:47:52.357 >> Mhm. >> But the desire to >> To walk? 00:47:52.440 --> 00:47:56.837 >> move around and stand up. Um that's sort of 00:47:56.920 --> 00:47:58.957 hardwired. It's very simple to hardwire 00:47:59.040 --> 00:48:03.877 this kind of stuff. >> What Oh, like the desire to Well, 00:48:03.960 --> 00:48:07.557 that's interesting. You're hardwired to want to walk. 00:48:07.640 --> 00:48:12.397 That's not a There's got to be a deeper need for walking. 00:48:12.480 --> 00:48:15.557 I I think it was probably socially imposed by society that you need to 00:48:15.640 --> 00:48:19.797 walk. All the other bipedal >> There are a lot of simple animals that, 00:48:19.880 --> 00:48:24.117 you know, would probably walk without ever watching any 00:48:24.200 --> 00:48:27.877 other members of the species. >> It seems like a scary thing to have to 00:48:27.960 --> 00:48:30.397 do cuz you suck at bipedal walking at 00:48:30.480 --> 00:48:33.877 first. It seems crawling is much safer. 00:48:33.960 --> 00:48:37.317 Much more like why are you in a hurry? 00:48:37.400 --> 00:48:39.197 >> [laughter] >> Well, because because you have this 00:48:39.280 --> 00:48:42.277 thing that drives you to do it, you know? Um 00:48:42.360 --> 00:48:46.997 which is sort of part of uh the sort of human development. 00:48:47.080 --> 00:48:49.717 >> Is that understood actually what >> Not entirely. No. 00:48:49.800 --> 00:48:53.037 >> What is the What's the reason to get on two feet? It's really hard. Like most 00:48:53.120 --> 00:48:54.717 animals don't get on two feet. 00:48:54.800 --> 00:48:57.717 >> They get on four feet. You know, many mammals get on four feet. 00:48:57.800 --> 00:49:00.477 >> Yeah, they feel >> Very quickly. Some of them extremely quickly. 00:49:00.560 --> 00:49:04.517 >> But I don't you know, like from the last time I've interacted with a table, 00:49:04.600 --> 00:49:08.397 that's much more stable than a thing on two legs. It's just a really hard problem. 00:49:08.480 --> 00:49:11.317 >> Yeah, I mean birds have figured it out with two feet. 00:49:11.400 --> 00:49:15.477 >> It Well, technically we can go into anthology. They have four. I guess they 00:49:15.560 --> 00:49:17.077 have two feet. >> They have two feet. 00:49:17.160 --> 00:49:20.757 >> Chickens. >> You know, dinosaurs had two feet. Many of them. 00:49:20.840 --> 00:49:25.237 >> Allegedly. >> [laughter] >> I'm just now learning that T-Rex was 00:49:25.320 --> 00:49:29.077 eating grass, not other animals. T-Rex might have been a friendly 00:49:29.160 --> 00:49:32.157 friendly pet. What do you think about uh 00:49:32.240 --> 00:49:36.677 I don't know if you looked at the test for general intelligence that 00:49:36.760 --> 00:49:40.117 François Chollet put together. I don't know if you got a chance to look at that 00:49:40.200 --> 00:49:43.717 kind of thing. Like what What's your intuition about how to solve 00:49:43.800 --> 00:49:47.237 like an IQ type of test? >> I don't know. I think it's so outside of 00:49:47.320 --> 00:49:50.157 my radar screen that it's not really 00:49:50.240 --> 00:49:52.517 relevant, I think, in the short term. 00:49:52.600 --> 00:49:55.837 >> Well, I guess one way to ask another way 00:49:55.920 --> 00:49:58.517 perhaps more closer to what 00:49:58.600 --> 00:50:00.637 to your work is like how do you solve 00:50:00.720 --> 00:50:04.597 MNIST uh with very little example data? 00:50:04.680 --> 00:50:07.637 >> That's right. And that's the answer to this probably is self-supervised 00:50:07.720 --> 00:50:11.717 learning. Just learn to represent images and then learning uh you know, to 00:50:11.800 --> 00:50:15.477 recognize handwritten digits on top of this will only require a few samples. 00:50:15.560 --> 00:50:17.397 And we observe this in in humans, right? 00:50:17.480 --> 00:50:22.157 You you show a a young child a a picture book with a couple picture of an 00:50:22.240 --> 00:50:26.037 elephant and that's it. The The child knows what an elephant is. And 00:50:26.120 --> 00:50:29.977 we we we see this today with practical systems that we you know, we train 00:50:30.060 --> 00:50:32.397 [snorts] image recognition systems with 00:50:32.480 --> 00:50:36.157 uh enormous amounts of of images either 00:50:36.240 --> 00:50:40.237 either completely self-supervised or very weakly supervised. For example, you 00:50:40.320 --> 00:50:44.957 can uh train a neural net to predict uh whatever hashtag people type on 00:50:45.040 --> 00:50:47.917 Instagram, right? And you can do this with billions of images cuz there's 00:50:48.000 --> 00:50:52.357 billions per day that are showing up. Um so, the the amount of training data 00:50:52.440 --> 00:50:55.517 there is essentially unlimited. And [snorts] then you take the output 00:50:55.600 --> 00:50:59.997 representation, you know, a couple layers down from the output uh of what 00:51:00.080 --> 00:51:04.997 the system learned and feed this as input to a classifier for any object in 00:51:05.080 --> 00:51:07.637 the world that you want and it works pretty well. So, that's transfer 00:51:07.720 --> 00:51:09.637 learning, okay? [clears throat] Uh or 00:51:09.720 --> 00:51:12.197 weakly supervised uh transfer learning. 00:51:12.280 --> 00:51:17.037 Uh people are making very very fast progress using self-supervised learning 00:51:17.120 --> 00:51:19.757 uh for for the with this kind of scenario as well. 00:51:19.840 --> 00:51:22.557 >> [snorts] >> Um and you know, my guess is that that's 00:51:22.640 --> 00:51:24.317 that's going to be the the future. 00:51:24.400 --> 00:51:26.757 >> For self-supervised learning, how much 00:51:26.840 --> 00:51:29.597 cleaning do you think is needed for 00:51:29.680 --> 00:51:34.637 filtering um uh malicious signal or what's a better 00:51:34.720 --> 00:51:38.557 term? But like a lot of people use hashtags on Instagram 00:51:38.640 --> 00:51:41.957 to uh get like good SEO >> Mhm. 00:51:42.040 --> 00:51:44.917 >> that doesn't fully represent the contents of the image. 00:51:45.000 --> 00:51:49.437 Like they'll put a picture of a cat and hashtag it with like science, awesome, 00:51:49.520 --> 00:51:54.357 fun, I don't know. All kind of Why would you put science? That's not very good SEO. 00:51:54.440 --> 00:51:58.157 >> The way the way my my colleagues who worked on this project at 00:51:58.240 --> 00:52:02.477 at Facebook now Meta Meta AI a few years ago 00:52:02.560 --> 00:52:05.757 dealt with this is that they only selected something like 17,000 tags that 00:52:05.840 --> 00:52:08.637 correspond to kind of physical things or 00:52:08.720 --> 00:52:12.917 or situations. Like you know, that has some visual content. 00:52:13.000 --> 00:52:17.797 Um so you know you wouldn't have like hash TBT or 00:52:17.880 --> 00:52:21.757 anything like that. >> Oh, so they they keep a very select set 00:52:21.840 --> 00:52:23.197 of hashtags is what you're saying. 00:52:23.280 --> 00:52:26.557 >> Yeah. >> Okay. >> But it's still it's still on the order of 00:52:26.640 --> 00:52:29.757 you know, 10 to 20,000. So it's fairly large. 00:52:29.840 --> 00:52:33.197 >> Okay. Can you tell me about data augmentation? 00:52:33.280 --> 00:52:37.717 What the heck is data augmentation and how is it used maybe contrastive 00:52:37.800 --> 00:52:42.717 learning for for video? What are some cool ideas here? 00:52:42.800 --> 00:52:45.917 >> Right. So data augmentation, I mean first data augmentation, you know, is 00:52:46.000 --> 00:52:49.637 the idea of artificially increasing the size of your training set by distorting 00:52:49.720 --> 00:52:52.677 the images that you have in ways that don't 00:52:52.760 --> 00:52:54.237 change the nature of the image, right? 00:52:54.320 --> 00:52:57.797 So you take you do MNIST, you can do data augmentation on MNIST. And people have 00:52:57.880 --> 00:53:00.917 done this since 1990s, right? You take a 00:53:01.000 --> 00:53:05.437 MNIST digit and you shift it a little bit or you change the size or 00:53:05.520 --> 00:53:08.877 rotate it, skew it you know, etc. 00:53:08.960 --> 00:53:12.717 >> Add noise. >> Add noise, etc. And it it works better. 00:53:12.800 --> 00:53:15.837 If you train a supervised classifier with augmented data, you're going to get 00:53:15.920 --> 00:53:17.437 better results. 00:53:17.520 --> 00:53:20.717 Now, it's become really interesting over 00:53:20.800 --> 00:53:23.717 the last couple years because 00:53:23.800 --> 00:53:25.597 a lot of self-supervised learning 00:53:25.680 --> 00:53:28.037 techniques to pre-train vision systems 00:53:28.120 --> 00:53:31.517 are based on data augmentation. And the 00:53:31.600 --> 00:53:35.237 the basic techniques is originally inspired by 00:53:35.320 --> 00:53:38.917 techniques that I worked on in the early '90s and Jeff Hinton worked on also in 00:53:39.000 --> 00:53:42.757 the early '90s. They were sort of parallel work. I used to call this 00:53:42.840 --> 00:53:47.237 Siamese network. So, basically you take uh two identical copies of the same 00:53:47.320 --> 00:53:49.717 network. They share the same weights. 00:53:49.800 --> 00:53:54.517 And you show two uh different views of the same object. Either those two 00:53:54.600 --> 00:53:57.917 different views may have been obtained by data augmentation, or maybe it's two 00:53:58.000 --> 00:54:01.837 different views of the same scene from a camera that you moved, or at different 00:54:01.920 --> 00:54:05.397 times, or something like that, right? Or two pictures of the same person, things 00:54:05.480 --> 00:54:09.997 like that. And then you train this neural net, those two identical copies of this 00:54:10.080 --> 00:54:13.957 neural net, to produce an output representation, a vector, 00:54:14.040 --> 00:54:17.357 uh in such a way that the representation for those two 00:54:17.440 --> 00:54:21.997 images are as close to each other as possible, as identical to each other as 00:54:22.080 --> 00:54:26.597 possible, right? Because you want the system to be basically learn a function 00:54:26.680 --> 00:54:29.837 that will that will be invariant, that will not change, whose output will not 00:54:29.920 --> 00:54:34.397 change when you transform those inputs uh in in those in those particular ways, 00:54:34.480 --> 00:54:37.637 right? >> [snorts] >> Um so, that's easy to do. 00:54:37.720 --> 00:54:40.917 What's complicated is how do you make sure that when you show two images that 00:54:41.000 --> 00:54:42.917 are different, the system will produce different things? 00:54:43.000 --> 00:54:47.917 >> Mhm. >> Because if you don't have a specific provision for this, 00:54:48.000 --> 00:54:51.437 uh the system will just ignore the inputs uh when you train it. It will end 00:54:51.520 --> 00:54:54.477 up ignoring the input and just produce a constant vector that is the same for 00:54:54.560 --> 00:54:56.597 every input, right? >> Yes. >> That's called a collapse. 00:54:56.680 --> 00:55:00.357 >> Uh-huh. >> Now, how do you avoid collapse? So, there's two ideas. 00:55:00.440 --> 00:55:04.877 Uh one idea that I proposed in the early '90s uh with my colleagues at Bell Labs, 00:55:04.960 --> 00:55:09.837 uh Jan Bromley and a couple other people, uh which we now call contrastive 00:55:09.920 --> 00:55:13.397 learning, which is to have negative examples, right? So, you have pairs 00:55:13.480 --> 00:55:16.397 of images that you know are different. 00:55:16.480 --> 00:55:20.597 And you show them to the network, and uh those two copies, and then you you push 00:55:20.680 --> 00:55:22.517 the two output vectors away from each other. 00:55:22.600 --> 00:55:24.957 >> Mhm. >> And that would eventually guarantee that 00:55:25.040 --> 00:55:28.557 things that are semantically similar produce similar representations, and 00:55:28.640 --> 00:55:30.677 things that are different produce different representations. 00:55:30.760 --> 00:55:33.357 >> Mhm. >> Uh we actually came up with this idea 00:55:33.440 --> 00:55:35.717 for a project of doing uh signature 00:55:35.800 --> 00:55:40.797 verification. So, we would collect signature signatures from like multiple 00:55:40.880 --> 00:55:43.957 signatures on the same person and then train a neural net to produce the same 00:55:44.040 --> 00:55:47.677 representation. And then you know 00:55:47.760 --> 00:55:52.437 force the system to produce different representation for different signatures. 00:55:52.520 --> 00:55:56.677 This was actually the the problem was proposed by people from a what was a 00:55:56.760 --> 00:56:00.237 subsidiary of AT&T at the time called NCR. 00:56:00.320 --> 00:56:02.317 And they were interested in storing 00:56:02.400 --> 00:56:07.477 representation of the signature on the 80 bytes of the magnetic strip of a 00:56:07.560 --> 00:56:10.037 credit card. >> [snorts] >> So, we came up with this idea of having 00:56:10.120 --> 00:56:13.397 a neural net with 80 outputs. You know, that we would quantize on 00:56:13.480 --> 00:56:16.957 bytes to so that we could encode the the >> And then encoding was then used to 00:56:17.040 --> 00:56:19.037 compare whether signature matches or not. 00:56:19.120 --> 00:56:22.157 >> That's right. So, then you would you know sign you would run through the 00:56:22.240 --> 00:56:25.037 neural net and then you would compare the output vector to whatever is stored 00:56:25.120 --> 00:56:26.637 on your card. >> Did it actually work? 00:56:26.720 --> 00:56:28.927 >> It worked but they ended up not using it. 00:56:29.010 --> 00:56:33.397 >> [laughter] >> Um because nobody cares actually. I mean the 00:56:33.480 --> 00:56:38.997 American you know finish payment system is incredibly lax in that respect compared 00:56:39.080 --> 00:56:41.597 to Europe for example. >> the signatures. What's the purpose of 00:56:41.680 --> 00:56:45.037 signatures anyway? This is very >> Nobody looks at them. Nobody cares, you know. 00:56:45.120 --> 00:56:49.517 >> And and it's uh yeah. >> Yeah, no. So, uh so that that's contrastive running, 00:56:49.600 --> 00:56:53.517 right? So, you need positive and negative pairs and the problem with that is that 00:56:53.600 --> 00:56:56.717 you know, even though I had the original paper on this 00:56:56.800 --> 00:56:59.997 I'm actually not very positive about it because it doesn't work in high 00:57:00.080 --> 00:57:02.997 dimension. If your representation is high dimensional 00:57:03.080 --> 00:57:06.237 there's just too many ways for two things to be different. 00:57:06.320 --> 00:57:09.957 And and so you would need lots and lots and lots of negative pairs. 00:57:10.040 --> 00:57:12.957 Uh so, there is a a particular implementation of this which is 00:57:13.040 --> 00:57:17.197 relatively recent from actually the Google Toronto group 00:57:17.280 --> 00:57:21.077 uh where you know, Jeff Hinton is the the senior member there. It's called 00:57:21.160 --> 00:57:23.917 SimCLR, S I M C L C L R. >> Mhm. 00:57:24.000 --> 00:57:27.037 >> Uh and it you know, basically a particular way of implementing this idea 00:57:27.120 --> 00:57:30.557 of contrastive learning. The particular objective function. 00:57:30.640 --> 00:57:34.397 Now what I'm much more enthusiastic about 00:57:34.480 --> 00:57:36.557 these days is non-contrasting methods. 00:57:36.640 --> 00:57:39.077 So, other ways to guarantee that >> [snorts] 00:57:39.160 --> 00:57:42.197 >> uh uh the representations will be 00:57:42.280 --> 00:57:46.197 different for different uh different inputs. 00:57:46.280 --> 00:57:49.757 And it's actually based on an idea that um 00:57:49.840 --> 00:57:53.877 Jeff Hinton proposed in the early '90s with his student at the time, Subbarao. 00:57:53.960 --> 00:57:56.517 And it's based on the idea of maximizing the mutual information between the 00:57:56.600 --> 00:58:00.197 outputs of the two systems. You only show positive pairs. You only show pairs 00:58:00.280 --> 00:58:03.247 of images that you know are somewhat similar. 00:58:03.330 --> 00:58:06.997 >> [snorts] >> And you train the two networks to be informative, 00:58:07.080 --> 00:58:11.757 but also to be as informative uh of each other as possible. So, 00:58:11.840 --> 00:58:16.717 basically, one representation has to be predictable from the other, essentially. 00:58:16.800 --> 00:58:20.717 Uh and, you know, he proposed that idea had, you know, 00:58:20.800 --> 00:58:24.077 a couple of papers in the early '90s, and then nothing was done about it for 00:58:24.160 --> 00:58:29.397 decades. And I kind of revived this idea together with my postdocs at at FAIR. 00:58:29.480 --> 00:58:32.717 Uh particularly a postdoc called Stéphane Deny, who's now a 00:58:32.800 --> 00:58:35.997 junior professor in in Finland at University of Alto. 00:58:36.080 --> 00:58:41.197 Um we came [snorts] up with something called uh that we called Barlow Twins. 00:58:41.280 --> 00:58:42.837 And it's a a particular way of 00:58:42.920 --> 00:58:46.357 maximizing the information content of uh 00:58:46.440 --> 00:58:50.837 of a vector, you know, using some uh hypotheses. 00:58:50.920 --> 00:58:54.557 And we have uh kind of a another version of it that's more recent 00:58:54.640 --> 00:58:58.957 now called VICReg, VICREG. That that means variance invariance covariance 00:58:59.040 --> 00:59:02.477 regularization. it's the thing I'm the most excited about in in machine 00:59:02.560 --> 00:59:05.517 learning in the last 15 years. I mean, I'm I'm not I'm really really excited 00:59:05.600 --> 00:59:09.557 about this. >> What uh kind of data augmentation is useful for 00:59:09.640 --> 00:59:12.117 that non-contrastive learning method? 00:59:12.200 --> 00:59:15.437 Are we talking about Does that not matter that much? Or it seems like a 00:59:15.520 --> 00:59:17.837 very important part of the step. 00:59:17.920 --> 00:59:19.597 >> Yeah. >> How you generate the images that are 00:59:19.680 --> 00:59:21.517 similar but sufficiently different? 00:59:21.600 --> 00:59:24.317 >> Yeah, that's right. It's a important step, and it's also an annoying step 00:59:24.400 --> 00:59:28.197 because you need to have that knowledge of what the augmentation uh 00:59:28.280 --> 00:59:32.117 you can do that do not change the nature of the of the object. 00:59:32.200 --> 00:59:35.277 >> Yeah. >> And so um the standard scenario which 00:59:35.360 --> 00:59:39.557 you know a lot of people working in this area are using is you use uh 00:59:39.640 --> 00:59:43.477 the the type of distortion So so basically you do uh geometric 00:59:43.560 --> 00:59:46.397 distortion. So one basically just shift the image a little bit, it's called 00:59:46.480 --> 00:59:50.317 cropping. Another one kind of changes the scale a little bit. Another one kind 00:59:50.400 --> 00:59:53.797 of rotates it. Another one changes the colors, you know, you kind of do a shift 00:59:53.880 --> 00:59:56.197 in color balance or something like that. 00:59:56.280 --> 01:00:00.757 Um saturation another one sort of blurs it. Another one adds noise. So you have 01:00:00.840 --> 01:00:04.277 like a a catalog of kind of standard things and people try to use the same 01:00:04.360 --> 01:00:07.437 ones for different algorithms so that they can compare. 01:00:07.520 --> 01:00:10.717 Um but some algorithms uh some self-supervised algorithm actually can 01:00:10.800 --> 01:00:14.397 deal with much bigger like more aggressive data augmentation 01:00:14.480 --> 01:00:18.957 and some can't don't. So that kind of makes the whole thing difficult. But but 01:00:19.040 --> 01:00:21.797 that's the kind of distortions we're talking about and 01:00:21.880 --> 01:00:24.237 and so you you you you train with those 01:00:24.320 --> 01:00:26.717 distortions. And then uh 01:00:26.800 --> 01:00:29.237 you you chop off the last layer or 01:00:29.320 --> 01:00:33.077 couple layers of the of the the network. 01:00:33.160 --> 01:00:36.717 And you use the representation as input to a classifier. You train the classifier. 01:00:36.800 --> 01:00:40.637 >> Mhm. >> [clears throat] >> Um on ImageNet let's say or whatever. 01:00:40.720 --> 01:00:45.717 And measure the performance. And interestingly enough the methods that 01:00:45.800 --> 01:00:49.037 are really good at eliminating the information that is irrelevant which is 01:00:49.120 --> 01:00:53.547 the distortions between those images do a good job at eliminating it. 01:00:53.630 --> 01:00:56.837 >> [laughter] >> And uh as a consequence you cannot use 01:00:56.920 --> 01:01:00.957 those the representations in those systems for things like object detection 01:01:01.040 --> 01:01:04.027 and localization because that information is gone. 01:01:04.110 --> 01:01:06.117 >> [snorts] >> So the type of data augmentation you 01:01:06.200 --> 01:01:09.157 need to do depends on the task you want eventually 01:01:09.240 --> 01:01:13.037 the system to to solve and the type of data augmentation standard data 01:01:13.120 --> 01:01:16.717 augmentation that we use today are only appropriate for object recognition or 01:01:16.800 --> 01:01:19.557 image classification. They're not appropriate for things like 01:01:19.640 --> 01:01:22.997 >> Okay, help me out and what uh localization is So, you're 01:01:23.080 --> 01:01:27.037 saying it's just not good at the negative uh like at cla- at classifying the 01:01:27.120 --> 01:01:29.877 negative, so that's why it can't be used for the localization. 01:01:29.960 --> 01:01:32.517 >> No, it's just that you train the system, 01:01:32.600 --> 01:01:35.517 you know, you you you give it an image and then you give it the same image 01:01:35.600 --> 01:01:38.677 shifted and scaled. You tell it that's the same image. 01:01:38.760 --> 01:01:41.117 >> Yeah. >> So, the system basically is trying to 01:01:41.200 --> 01:01:44.037 eliminate the information about position and size. 01:01:44.120 --> 01:01:47.957 So, now And now you want to use that >> Oh, like literally 01:01:48.040 --> 01:01:49.997 >> out where an object is and what size it is. 01:01:50.080 --> 01:01:51.957 >> bounding box. It did be able to actually 01:01:52.040 --> 01:01:56.397 Okay. It can still find It can still find the object in the image. It's just 01:01:56.480 --> 01:02:00.917 not very good at finding the exact boundaries of that object. Interesting. 01:02:01.000 --> 01:02:03.957 Interesting, which you know, there that's an interesting 01:02:04.040 --> 01:02:07.677 sort of philosophical question. How important How important is object 01:02:07.760 --> 01:02:11.837 localization anyway? We're We're like obsessed by measuring like image 01:02:11.920 --> 01:02:16.077 segmentation. Obsessed by measuring perfectly knowing the boundaries of 01:02:16.160 --> 01:02:17.677 objects when 01:02:17.760 --> 01:02:19.917 arguably 01:02:20.000 --> 01:02:24.717 that's not that essential to an understanding what are the contents of 01:02:24.800 --> 01:02:29.357 the scene. >> On the other hand, I think evolutionarily the first vision systems 01:02:29.440 --> 01:02:32.797 in animals were basically all about localization, very little about 01:02:32.880 --> 01:02:36.317 recognition. And in the human brain, you have two 01:02:36.400 --> 01:02:41.357 separate pathways for um recognizing the nature of of a scene 01:02:41.440 --> 01:02:45.597 or an object and localizing objects. So, you use the 01:02:45.680 --> 01:02:49.317 first pathway called the ventral pathway for, you know, telling what you're 01:02:49.400 --> 01:02:51.957 looking at. >> [snorts] >> Uh the other pathway, the dorsal 01:02:52.040 --> 01:02:56.637 pathway, is used for navigation, for grasping, for everything else. And, you 01:02:56.720 --> 01:03:00.237 know, basically a lot of the things you need for survival are >> [snorts] 01:03:00.320 --> 01:03:03.600 >> localization and detection. 01:03:03.880 --> 01:03:08.077 >> Is similarity learning or contrastive learning or these non-contrastive 01:03:08.160 --> 01:03:10.837 methods the same as understanding something? 01:03:10.920 --> 01:03:14.917 Just because you know a distorted cat is the same as a non-distorted cat, does 01:03:15.000 --> 01:03:18.677 Does mean you understand what it means to be a cat? 01:03:18.760 --> 01:03:22.037 >> To some extent. I mean, it's a superficial understanding, obviously. 01:03:22.120 --> 01:03:25.797 >> But like what is the ceiling of this method, do you think? Is this just one 01:03:25.880 --> 01:03:30.357 trick on the path to doing self-supervised learning? Can we go 01:03:30.440 --> 01:03:33.877 really, really far? >> I think we can go really far. So, if we 01:03:33.960 --> 01:03:38.557 figure out how to uh use techniques of that type, perhaps 01:03:38.640 --> 01:03:41.397 very different, but you know, of the same nature, 01:03:41.480 --> 01:03:46.637 to uh train a system from from video to do video prediction, essentially. 01:03:46.720 --> 01:03:51.957 Um I think we'll have a path, um you know, towards 01:03:52.040 --> 01:03:56.317 I you know, I wouldn't say unlimited, but but a path towards some level of 01:03:56.400 --> 01:04:02.187 uh you know, physical common sense in in machines. And I also think that 01:04:02.270 --> 01:04:04.757 >> [snorts] >> um that 01:04:04.840 --> 01:04:09.597 ability to learn how the world works from a sort of high-throughput channel 01:04:09.680 --> 01:04:11.757 like like vision >> Mhm. 01:04:11.840 --> 01:04:15.437 >> is a necessary step towards uh 01:04:15.520 --> 01:04:19.357 sort of real artificial intelligence. In other words, I believe in grounded 01:04:19.440 --> 01:04:22.877 intelligence. I don't think we can train a machine to be intelligent purely from 01:04:22.960 --> 01:04:25.757 text. Because I think the amount of 01:04:25.840 --> 01:04:26.917 information about the world that's 01:04:27.000 --> 01:04:32.877 contained in text is tiny compared to what we need to know. 01:04:32.960 --> 01:04:35.437 >> Mhm. >> So, for example, let let's uh and and 01:04:35.520 --> 01:04:38.677 you know, people have attempted to do this for for 30 years, right? The The 01:04:38.760 --> 01:04:40.317 Cyc project and things like that, right? 01:04:40.400 --> 01:04:43.917 Of basically kind of writing down all the facts that are known and >> Mhm. 01:04:44.000 --> 01:04:47.757 >> uh hoping that some some sort of common sense will emerge. Um I think it's 01:04:47.840 --> 01:04:51.917 basically hopeless. Uh but let me take an example. You take an object, I I 01:04:52.000 --> 01:04:55.637 describe a situation to you. I take an object, I put it on the table, and I 01:04:55.720 --> 01:04:59.677 push the table. It's completely obvious to you that the object will be pushed 01:04:59.760 --> 01:05:03.317 with the table, right? Because it's sitting on it. 01:05:03.400 --> 01:05:06.997 Um there's no text in the world, I believe, that explains this. 01:05:07.080 --> 01:05:11.197 And so, if you train a machine as powerful as it could be, you know, your 01:05:11.280 --> 01:05:13.717 GPT 5000 01:05:13.800 --> 01:05:16.517 um whatever it is, it's never going to 01:05:16.600 --> 01:05:21.197 learn about this. Um that information is just is not 01:05:21.280 --> 01:05:24.637 present in any text. >> Well, the the question like with the Cyc 01:05:24.720 --> 01:05:26.317 project, the dream I think is to have 01:05:26.400 --> 01:05:28.477 like 01:05:28.560 --> 01:05:32.357 like 10 million say facts like that 01:05:32.440 --> 01:05:37.397 that give you a head start. Like a parent guiding you. 01:05:37.480 --> 01:05:41.117 Now, we humans don't need a parent to tell us that the table will move uh 01:05:41.200 --> 01:05:44.797 sorry, the smartphone will move with the table. But 01:05:44.880 --> 01:05:47.797 we get a lot of guidance in other ways. 01:05:47.880 --> 01:05:50.357 So, it's possible that we can give it a quick shortcut. 01:05:50.440 --> 01:05:52.957 >> What about cat? A cat knows that. 01:05:53.040 --> 01:05:57.277 >> No, but they evolved. So >> No, they learn like us. 01:05:57.360 --> 01:05:59.277 >> The the sorry, the physics of stuff? 01:05:59.360 --> 01:06:04.037 >> Yeah. >> [sighs and gasps] >> Well, yeah, so you're saying it's 01:06:04.120 --> 01:06:08.997 uh so you're putting a lot of intelligence onto the nurture side, not the nature. 01:06:09.080 --> 01:06:13.837 >> Yes. >> Cuz we've seen to have you know, there's a very inefficient 01:06:13.920 --> 01:06:15.837 arguably process of evolution that got 01:06:15.920 --> 01:06:19.717 us from bacteria to who we are today. 01:06:19.800 --> 01:06:21.717 Started at the bottom, now we're here. 01:06:21.800 --> 01:06:26.077 So, uh the question is how Okay, so 01:06:26.160 --> 01:06:30.237 the question is how fundamental is that that the nature of the whole hardware 01:06:30.320 --> 01:06:33.797 uh and then is there any way to shortcut it if it's 01:06:33.880 --> 01:06:36.917 fundamental? If it's not, if it's most of intelligence, most of the cool stuff 01:06:37.000 --> 01:06:38.877 we've been talking about is mostly 01:06:38.960 --> 01:06:41.157 nurture mostly trained, we figured out 01:06:41.240 --> 01:06:43.997 by observing the world. We can form that 01:06:44.080 --> 01:06:48.277 uh big beautiful sexy background model that you're talking about just by 01:06:48.360 --> 01:06:50.837 sitting there. Um 01:06:50.920 --> 01:06:53.717 then okay, then you need to then like 01:06:53.800 --> 01:06:58.597 maybe uh it is all supervised learning all the 01:06:58.680 --> 01:07:00.917 way down. It's all supervised learning, say. 01:07:01.000 --> 01:07:03.997 >> Whatever it is that makes uh you know, human intelligence 01:07:04.080 --> 01:07:08.237 different from other animals, which you know, a lot of people think is language 01:07:08.320 --> 01:07:10.597 and logical reasoning and this kind of stuff. 01:07:10.680 --> 01:07:14.717 It cannot be that complicated because it only popped up in the last million years. 01:07:14.800 --> 01:07:19.037 >> Yeah. >> Uh and >> Yeah, it it >> you know, it it and it only involves, 01:07:19.120 --> 01:07:21.717 you know, less than 1% of our genome, right? Which 01:07:21.800 --> 01:07:25.277 is the difference between human genome and chimps or whatever. 01:07:25.360 --> 01:07:28.757 So, uh it can't be that complicated, you know, 01:07:28.840 --> 01:07:32.757 it can't be that fundamental. I mean, the most of the so complicated stuff 01:07:32.840 --> 01:07:37.037 already exists in cats and dogs and you know, certainly primates, non-human 01:07:37.120 --> 01:07:40.517 primates. >> Yeah, that little thing with humans 01:07:40.600 --> 01:07:43.597 might be just a something about social 01:07:43.680 --> 01:07:48.717 interaction and ability to maintain ideas across like a collective of 01:07:48.800 --> 01:07:54.117 people. It it's it sounds very dramatic and very impressive, but it probably isn't 01:07:54.200 --> 01:07:56.877 mechanistically speaking. >> It is, but we're not there yet. Like, we 01:07:56.960 --> 01:08:01.397 you know, we we have I mean, this is number 634, 01:08:01.480 --> 01:08:05.317 you know, in the list of problems we have to solve. 01:08:05.400 --> 01:08:10.277 >> So, basic physics of the world is is is is number one. What do you um 01:08:10.360 --> 01:08:14.237 Just a quick tangent on data augmentation. So, a lot 01:08:14.320 --> 01:08:15.837 of it is 01:08:15.920 --> 01:08:19.717 uh hard-coded versus learned. 01:08:19.800 --> 01:08:22.837 Do you have any intuition that maybe 01:08:22.920 --> 01:08:27.197 there could be some weird data augmentation like generative type of 01:08:27.280 --> 01:08:31.197 data augmentation, like doing something weird to images which then 01:08:31.280 --> 01:08:34.197 improves the the similarity learning 01:08:34.280 --> 01:08:36.917 process? So, not just kind of dumb 01:08:37.000 --> 01:08:41.557 simple distortions, but by you shaking your head just saying that even simple 01:08:41.640 --> 01:08:44.917 distortions are enough. >> I think no. I think data augmentation is 01:08:45.000 --> 01:08:48.037 a temporary necessary evil. >> Huh. 01:08:48.120 --> 01:08:51.997 >> Um so, what people are working on now is is two things. One is 01:08:52.080 --> 01:08:56.637 uh the type of self-supervised learning like trying to to translate the type of 01:08:56.720 --> 01:09:00.557 self-supervised learning people use in language, translating this to images, 01:09:00.640 --> 01:09:02.077 which is basically a denoising 01:09:02.160 --> 01:09:04.277 autoencoder method. Right? So, you you 01:09:04.360 --> 01:09:09.237 take an image, you you block you mask some parts of it. 01:09:09.320 --> 01:09:12.677 Uh and then you you train some giant neural net to reconstruct the parts that 01:09:12.760 --> 01:09:18.157 you that are that are missing. And until very recently 01:09:18.240 --> 01:09:20.997 there was no there was no working methods for that. 01:09:21.080 --> 01:09:24.557 All the auto encoder type methods for images weren't producing very good 01:09:24.640 --> 01:09:27.877 representation. But there's a paper now coming out of 01:09:27.960 --> 01:09:32.677 the fair group at Menlo Park that actually works very well. So, 01:09:32.760 --> 01:09:35.037 that doesn't require the data augmentation, that requires only 01:09:35.120 --> 01:09:40.077 masking. >> Okay. Only masking for images 01:09:40.160 --> 01:09:42.157 Okay. [snorts] >> Right. So, you mask part of the image 01:09:42.240 --> 01:09:46.917 and you train a system which you know in this case is a transformer because 01:09:47.000 --> 01:09:51.237 you can you can the transformer represents the image as 01:09:51.320 --> 01:09:55.237 uh non-overlapping patches, so it's easy to mask patches and things like that. 01:09:55.320 --> 01:09:59.077 >> Okay, then my question transfers to that problem then masking. Like why should 01:09:59.160 --> 01:10:01.797 the mask be a square or rectangle? 01:10:01.880 --> 01:10:04.997 >> Yeah, so it doesn't matter. Like you know, I think we're going to come up 01:10:05.080 --> 01:10:09.877 probably in the future with sort of uh you know, ways to mask that are it's you 01:10:09.960 --> 01:10:14.797 know, kind of random essentially. I mean, they are random already, but 01:10:14.880 --> 01:10:18.717 >> No, no, but like something that's challenging. 01:10:18.800 --> 01:10:22.437 Like optimally challenging. So, like I mean, 01:10:22.520 --> 01:10:23.917 maybe it's a metaphor that doesn't 01:10:24.000 --> 01:10:26.557 apply, but you're it seems like there's 01:10:26.640 --> 01:10:32.077 an data augmentation or masking is an interactive element with it. Like 01:10:32.160 --> 01:10:33.997 you're almost like playing with an image. 01:10:34.080 --> 01:10:37.597 >> Yeah. >> And like it's like the way we play with an image in our minds. 01:10:37.680 --> 01:10:41.597 >> No, but it's like dropout. It's like Boltzmann machine training. You 01:10:41.680 --> 01:10:46.317 you know, every every every time you see a percept you also 01:10:46.400 --> 01:10:49.957 you you can you can perturb it in some way and then 01:10:50.040 --> 01:10:53.437 the the principle of the training procedure 01:10:53.520 --> 01:10:57.917 is to minimize the difference of the output of the representation between the 01:10:58.000 --> 01:11:02.797 the clean version and the corrupted version essentially, right? And you you 01:11:02.880 --> 01:11:05.717 you can do this in real time, right? So, you know, Boltzmann machine works like 01:11:05.800 --> 01:11:09.997 this, right? You you you show a percept, uh you tell the machine that's a good 01:11:10.080 --> 01:11:12.157 combination of activities of your input 01:11:12.240 --> 01:11:17.397 neurons. Uh and then you either uh 01:11:17.480 --> 01:11:22.517 let them go their merry way without clamping them to values or you only do 01:11:22.600 --> 01:11:26.157 this with a subset. >> Yeah. >> [snorts] >> And what you do is you're training the 01:11:26.240 --> 01:11:30.397 system so that the the stable state of the entire network is the same 01:11:30.480 --> 01:11:35.077 regardless of whether it's his entire input or whether it's his only part of it. 01:11:35.160 --> 01:11:38.397 Um you know, the noise in autoencoder method is basically the same thing, 01:11:38.480 --> 01:11:41.557 right? You you're you're training a system to reproduce the input the 01:11:41.640 --> 01:11:45.637 complete input and filling the blanks regardless of which which parts are are 01:11:45.720 --> 01:11:49.437 missing and that's really the underlying principle. And you could imagine sort of 01:11:49.520 --> 01:11:53.917 even in the brain some sort of neural principle where, you know, neurons can uh 01:11:54.000 --> 01:11:56.717 oscillate, right? So, they they take their activity and then 01:11:56.800 --> 01:12:00.877 temporarily they kind of shut off to um you know, force the rest of the 01:12:00.960 --> 01:12:02.717 system to basically 01:12:02.800 --> 01:12:05.237 reconstruct the input without their 01:12:05.320 --> 01:12:10.437 help, you know? And and and I mean, you you could imagine, you know, 01:12:10.520 --> 01:12:13.117 you know, more or less biologically possible processes. 01:12:13.200 --> 01:12:14.317 >> Something like that. And I guess with 01:12:14.400 --> 01:12:18.317 this denoising autoencoder and masking 01:12:18.400 --> 01:12:21.637 and data augmentation, you don't have to worry about being 01:12:21.720 --> 01:12:25.517 super efficient. You can just do do as much as you want. 01:12:25.600 --> 01:12:28.077 >> Yeah. >> And and get better over time. 01:12:28.160 --> 01:12:31.637 Cuz I was thinking like you might want to be clever about the way you do all of 01:12:31.720 --> 01:12:36.077 these procedures, you know, but that's only if it's 01:12:36.160 --> 01:12:39.877 somehow costly to do every iteration, but it's not really. 01:12:39.960 --> 01:12:43.557 >> Not really. Maybe. And then there is, you know, data 01:12:43.640 --> 01:12:46.117 augmentation without explicit data augmentation. 01:12:46.200 --> 01:12:49.237 This data augmentation by waiting, which is, you know, the the sort of video 01:12:49.320 --> 01:12:55.157 prediction. >> Right. >> Um you're observing a video clip observing the 01:12:55.240 --> 01:12:58.317 you know the continuation of that video clip and try 01:12:58.400 --> 01:13:02.197 you try to learn a representation using the joint embedding architectures 01:13:02.280 --> 01:13:06.597 in such a way that the representation of the future clip is easily predictable 01:13:06.680 --> 01:13:08.517 from the representation of the of the 01:13:08.600 --> 01:13:10.397 observed clip. 01:13:10.480 --> 01:13:14.637 >> Do you think YouTube has enough raw data 01:13:14.720 --> 01:13:18.237 from which to learn how to be a cat? 01:13:18.320 --> 01:13:23.277 >> I think so. >> So it the amount of data is not the constraint. I 01:13:23.360 --> 01:13:25.957 mean >> it would require some selection I think. 01:13:26.040 --> 01:13:29.797 >> Some actually >> Some some selection of maybe the right 01:13:29.880 --> 01:13:33.077 type of data you need some >> the rabbit hole of just cat videos that 01:13:33.160 --> 01:13:36.397 might you might need to watch some lectures or something. 01:13:36.480 --> 01:13:39.397 >> No, you >> How matter would that be 01:13:39.480 --> 01:13:42.397 if it like watches lectures about 01:13:42.480 --> 01:13:46.457 intelligence and then learns watches your lectures and NYU and learns 01:13:46.540 --> 01:13:48.157 [clears throat] from that how to be intelligent. 01:13:48.240 --> 01:13:49.737 >> I don't think it would be enough. 01:13:49.820 --> 01:13:54.357 >> [laughter] [gasps] >> What's your do you find multimodal 01:13:54.440 --> 01:13:57.717 learning interesting we've been talking about visual language like if combining 01:13:57.800 --> 01:14:00.117 those together maybe audio all those kinds of things. 01:14:00.200 --> 01:14:05.117 >> lot of things that I find interesting in the short term but are not addressing 01:14:05.200 --> 01:14:08.877 the important problem that I think are really kind of the big challenges. So I 01:14:08.960 --> 01:14:10.397 think you know things like multitask 01:14:10.480 --> 01:14:15.237 learning continual learning you know 01:14:15.320 --> 01:14:19.077 adversarial issues. I mean those have you know great practical interests in 01:14:19.160 --> 01:14:22.797 the relative short term. Possibly but I don't think they're 01:14:22.880 --> 01:14:25.677 fundamental you know active learning even to some extent reinforcement 01:14:25.760 --> 01:14:28.677 learning. I think those things will 01:14:28.760 --> 01:14:32.157 become either obsolete or or useless or 01:14:32.240 --> 01:14:36.117 easy once we figured out how to do 01:14:36.200 --> 01:14:37.837 self-supervised representation learning 01:14:37.920 --> 01:14:41.157 or or learning predictive world models. 01:14:41.240 --> 01:14:46.317 And so I think that's what you know the entire community should be focusing on 01:14:46.400 --> 01:14:48.997 at least people who are interested in sort of fundamental questions or you 01:14:49.080 --> 01:14:52.637 know, really kind of pushing the envelope of AI towards the next the next 01:14:52.720 --> 01:14:55.477 stage. But of course there is like a huge amount of you know, very 01:14:55.560 --> 01:14:58.637 interesting work to do in sort of practical questions that that have you 01:14:58.720 --> 01:15:01.317 know, short-term impact. >> Well, you know, 01:15:01.400 --> 01:15:05.557 it's it's difficult to talk about that the temporal scale because all of human 01:15:05.640 --> 01:15:09.837 civilization will eventually be destroyed because the the the the sun 01:15:09.920 --> 01:15:14.077 will die out and even if Elon Musk is successful multi-planetary 01:15:14.160 --> 01:15:16.437 colonization across the galaxy, 01:15:16.520 --> 01:15:18.797 eventually the entirety of it will just 01:15:18.880 --> 01:15:22.317 become giant black holes and um 01:15:22.400 --> 01:15:24.077 >> That's going to take a while though. 01:15:24.160 --> 01:15:27.797 >> So but but what I'm saying is then that logic can be used to say it's all 01:15:27.880 --> 01:15:31.477 meaningless. I'm saying all that to say that 01:15:31.560 --> 01:15:33.797 multitask learning 01:15:33.880 --> 01:15:36.797 might be you're you're calling it 01:15:36.880 --> 01:15:39.437 practical or pragmatic or whatever. That 01:15:39.520 --> 01:15:40.837 might be the thing that achieves 01:15:40.920 --> 01:15:44.517 something very akin to intelligence 01:15:44.600 --> 01:15:47.357 while we're trying to solve the more 01:15:47.440 --> 01:15:51.757 uh general problem of self-supervised learning of background knowledge. So the 01:15:51.840 --> 01:15:55.477 reason I bring that up, maybe one way to ask that question. I've been very 01:15:55.560 --> 01:15:59.037 impressed by what Tesla autopilot team is doing. I don't know if you got any 01:15:59.120 --> 01:16:02.557 chance to glance at this particular one example of 01:16:02.640 --> 01:16:06.917 multitask learning where they're literally taking the problem 01:16:07.000 --> 01:16:09.317 like I don't know, Charles Darwin start 01:16:09.400 --> 01:16:14.157 studying animals. They're studying the problem of driving and asking, okay, 01:16:14.240 --> 01:16:16.917 what are all the things you have to perceive? 01:16:17.000 --> 01:16:21.157 And the way they're solving it is one, there's an ontology where you're 01:16:21.240 --> 01:16:24.117 bringing that to the table. So you're formulating a bunch of different tasks. 01:16:24.200 --> 01:16:26.957 It's like over 100 tasks or something like that that they're involved in 01:16:27.040 --> 01:16:31.197 driving. And then they're deploying it and then getting data back from people 01:16:31.280 --> 01:16:35.637 that run into trouble and then trying to figure out do we add task? Do we like we 01:16:35.720 --> 01:16:39.837 focus on each individual task separately? In fact, half So the I would 01:16:39.920 --> 01:16:44.757 say I'll classify Andrej Karpathy's talk into it. So, one was about doors and the 01:16:44.840 --> 01:16:46.597 other one about how much ImageNet sucks. 01:16:46.680 --> 01:16:50.557 He kept going back and forth on those two topics, 01:16:50.640 --> 01:16:53.197 which ImageNet sucks meaning you can't 01:16:53.280 --> 01:16:55.717 just use a single benchmark. There's so 01:16:55.800 --> 01:17:00.557 like you you have to have like a a giant suite of benchmarks to understand how 01:17:00.640 --> 01:17:02.117 well your your system actually works. 01:17:02.200 --> 01:17:06.997 >> with him. I mean, he's he's he's very sensible guy. Um Now, 01:17:07.080 --> 01:17:11.157 okay, it's it's very clear that if you're faced with a an engineering 01:17:11.240 --> 01:17:14.797 problem that you need to solve in a relatively short time, particularly if 01:17:14.880 --> 01:17:17.837 you have Elon Musk breathing down your neck, 01:17:17.920 --> 01:17:21.637 you're going to have to take shortcuts, right? You you you might think about the 01:17:21.720 --> 01:17:26.037 the fact that the the the right thing to do and the long-term solution involves, 01:17:26.120 --> 01:17:29.997 you know, some fancy self-supervisioning, but you have 01:17:30.080 --> 01:17:34.477 you know, Elon Musk breathing down your neck, uh and, you know, this involves 01:17:34.560 --> 01:17:38.397 uh you know, human lives, and so you you have to 01:17:38.480 --> 01:17:43.077 basically just do the systematic engineering and, you know, 01:17:43.160 --> 01:17:47.997 uh fine-tuning and refinements and trial and error and and and all that stuff. 01:17:48.080 --> 01:17:49.317 Um there's nothing wrong with that. 01:17:49.400 --> 01:17:52.637 That's that's called engineering. That's called, you know, uh 01:17:52.720 --> 01:17:56.077 putting technology out uh in the 01:17:56.160 --> 01:17:58.877 yeah, in the world, um and and you have 01:17:58.960 --> 01:18:03.517 to kind of iron clad it before before you do this, you know, um 01:18:03.600 --> 01:18:08.837 uh so much for, you know, grand grand ideas and principles. 01:18:08.920 --> 01:18:11.917 Um But, you know what? I'm placing myself 01:18:12.000 --> 01:18:16.357 sort of, you know, some you know, upstream of this, you know, 01:18:16.440 --> 01:18:19.357 quite a bit upstream of this. >> You're Plato, thinking about Platonic 01:18:19.440 --> 01:18:21.197 forms. You're You're 01:18:21.280 --> 01:18:23.757 >> No, Platonic because eventually I want 01:18:23.840 --> 01:18:28.957 us stuff to get used, but uh it's okay if it takes 5 or 10 years for the 01:18:29.040 --> 01:18:33.117 community to realize this is the right thing to do. I've I've done this before. 01:18:33.200 --> 01:18:36.997 It's been the case before that, you You I've made that that case. I mean, if you 01:18:37.080 --> 01:18:40.917 look back in the mid-2000s, for example, and you ask yourself the question, 01:18:41.000 --> 01:18:44.487 "Okay, I want to recognize cars or faces or whatever." 01:18:44.570 --> 01:18:47.477 >> [snorts] >> Um you know, I can use convolutional nets, 01:18:47.560 --> 01:18:51.637 or I can use some more conventional uh kind of computer vision techniques, you 01:18:51.720 --> 01:18:56.157 know, using interest point detectors or SIFT dense features, and you know, 01:18:56.240 --> 01:18:58.157 sticking an SVM on top. At that time, 01:18:58.240 --> 01:19:00.717 the data sets were so small that 01:19:00.800 --> 01:19:05.717 those methods that used more engineering worked better than ConvNets. There was 01:19:05.800 --> 01:19:09.317 just not enough data for ConvNets. And ConvNets were were a little a little 01:19:09.400 --> 01:19:12.757 slow with the kind of hardware that was available at the time. 01:19:12.840 --> 01:19:17.517 And there was a sea change when uh basically when, you know, 01:19:17.600 --> 01:19:21.877 data sets became bigger and and GPUs became available. That that's what, you 01:19:21.960 --> 01:19:25.757 know, the two of the main factors that basically 01:19:25.840 --> 01:19:28.557 made people change their change their mind. 01:19:28.640 --> 01:19:32.317 Um >> [snorts] >> and you can you can look at the history 01:19:32.400 --> 01:19:36.477 of like all sub-branches of AI or pattern 01:19:36.560 --> 01:19:40.557 recognition. And there's a similar trajectory 01:19:40.640 --> 01:19:44.877 followed by techniques where people start by, you know, engineering the hell 01:19:44.960 --> 01:19:49.277 out of it. Um you know, be it 01:19:49.360 --> 01:19:52.957 optical character recognition, speech recognition, 01:19:53.040 --> 01:19:56.037 computer vision, like image recognition in general, 01:19:56.120 --> 01:19:59.677 uh natural language understanding, like, you know, translation, things like that, 01:19:59.760 --> 01:20:02.397 right? You start to engineer the hell out of it. 01:20:02.480 --> 01:20:05.557 Um you start to acquire all the knowledge that prior knowledge you know 01:20:05.640 --> 01:20:09.277 about image formation, about, you know, the shape of characters, about, you 01:20:09.360 --> 01:20:13.037 know, morphological operations, about like feature extraction, Fourier 01:20:13.120 --> 01:20:17.037 transforms, you know, Zernike moments, you know, whatever, right? People have 01:20:17.120 --> 01:20:20.197 come up with thousands of ways of representing images so that you they 01:20:20.280 --> 01:20:24.077 could be easily classified afterwards. Same for speech 01:20:24.160 --> 01:20:26.957 recognition, right? There is, you know, two decades for people to 01:20:27.040 --> 01:20:30.837 figure out a good front end to preprocess uh, 01:20:30.920 --> 01:20:34.157 uh, speech signals so that, you know, the information about what is being said 01:20:34.240 --> 01:20:39.437 is preserved, but most of the information about the identity of the speaker is gone. 01:20:39.520 --> 01:20:41.277 Um, you know, [snorts] catch all 01:20:41.360 --> 01:20:43.837 coefficients or whatever, right? Um, 01:20:43.920 --> 01:20:47.157 and same for for text, right? Uh, you do 01:20:47.240 --> 01:20:49.397 named entity recognition and you parse 01:20:49.480 --> 01:20:53.517 and you you you do uh, tagging of uh, of 01:20:53.600 --> 01:20:57.517 of the parts of speech and, you know, you do this sort of tree representation 01:20:57.600 --> 01:21:02.157 of uh, clauses and all that stuff, right? Before you can do anything. 01:21:02.240 --> 01:21:07.197 Um, so, that's how it starts, right? Just 01:21:07.280 --> 01:21:10.997 engineer the hell out of it. And then, you start having data 01:21:11.080 --> 01:21:13.917 and maybe you have more powerful computers, maybe you know something 01:21:14.000 --> 01:21:16.877 about statistical learning. So, you start using machine learning and it's 01:21:16.960 --> 01:21:20.157 usually a small sliver on top of your kind of hand crafted system where, you 01:21:20.240 --> 01:21:22.517 know, you extract features by hand. 01:21:22.600 --> 01:21:26.157 Okay, and now, you know, nowadays, the standard way of doing this is that you 01:21:26.240 --> 01:21:28.957 train the entire thing end to end with a deep learning system and it learns its 01:21:29.040 --> 01:21:33.717 own features and and, you know, speech recognition systems nowadays, 01:21:33.800 --> 01:21:37.157 uh, or OCR systems are completely end to end. It's the uh, you know, it's some 01:21:37.240 --> 01:21:40.837 giant neural net that takes raw waveforms 01:21:40.920 --> 01:21:44.317 and produces a sequence of characters coming out. And it's just a huge neural 01:21:44.400 --> 01:21:47.317 net, right? There's no you know, Markov model, there's no 01:21:47.400 --> 01:21:51.037 language model that is explicit other than, you know, something that's 01:21:51.120 --> 01:21:54.197 ingrained in the in the sort of neural language model, if you want. Same for 01:21:54.280 --> 01:21:59.317 translation, same for all kinds of stuff. So, you see this continuous evolution 01:21:59.400 --> 01:22:03.277 from you know, less and less hand crafting 01:22:03.360 --> 01:22:07.477 and more and more learning. Um, and uh, 01:22:07.560 --> 01:22:09.877 I I think uh, 01:22:09.960 --> 01:22:12.597 I mean, it's true in biology as well. 01:22:12.680 --> 01:22:17.197 >> So, I mean, we might disagree about this, maybe not uh, in the this one 01:22:17.280 --> 01:22:21.437 little piece at the end. You mentioned active learning. 01:22:21.520 --> 01:22:25.357 It feels like active learning, which is the selection of data and also the 01:22:25.440 --> 01:22:29.597 interactivity, needs to be part of this giant neural network. You cannot just be 01:22:29.680 --> 01:22:33.757 an observer to do self-supervised learning. You have to Well, 01:22:33.840 --> 01:22:38.077 I don't self-supervised learning is just a word, but I would whatever this giant 01:22:38.160 --> 01:22:43.037 stack of a neural network that's automatically learning, it feels My 01:22:43.120 --> 01:22:46.077 intuition is that 01:22:46.160 --> 01:22:48.837 you have to have a system, whether it's 01:22:48.920 --> 01:22:52.117 a physical robot or a digital robot 01:22:52.200 --> 01:22:57.117 that's interacting with the world and doing so in a flawed way and improving 01:22:57.200 --> 01:23:02.317 over time. In order to to form the self-supervised 01:23:02.400 --> 01:23:06.917 learning, well, you can't just give it a giant sea of data. 01:23:07.000 --> 01:23:08.317 >> Okay, I agree and I disagree. 01:23:08.400 --> 01:23:12.237 >> Okay. >> I agree in the sense that I think uh 01:23:12.320 --> 01:23:17.037 I agree I agree in two ways. The first The first way I agree is that if you want uh 01:23:17.120 --> 01:23:20.397 and you certainly need a causal model of the world that allows you to predict the 01:23:20.480 --> 01:23:23.677 consequences of your actions, to train that model, you need to take 01:23:23.760 --> 01:23:28.077 actions, right? You need to be able to act in a world and see the effect 01:23:28.160 --> 01:23:30.357 for you to be to learn causal models of the world. 01:23:30.440 --> 01:23:33.477 >> Well, so that's not That's not obvious because you can observe others. 01:23:33.560 --> 01:23:35.997 >> You can observe others. >> And you can infer that they're similar 01:23:36.080 --> 01:23:37.997 to you, and then you can learn from that. 01:23:38.080 --> 01:23:41.037 >> Yeah, but then you have to kind of hardwire that part, right? And you know, 01:23:41.120 --> 01:23:44.077 mirror neurons and all that stuff, right? So, um 01:23:44.160 --> 01:23:47.237 and it's not clear to me how you would do this in a machine. So, 01:23:47.320 --> 01:23:51.157 um um so, I think the the action part would 01:23:51.240 --> 01:23:53.837 be necessary for having causal models of 01:23:53.920 --> 01:23:54.957 of the world. 01:23:55.040 --> 01:23:58.517 Uh the second reason it may be necessary 01:23:58.600 --> 01:24:03.597 or at least more efficient is that uh active learning basically, you know, 01:24:03.680 --> 01:24:07.397 goes for the jugular of what you're what you don't know, right? Mhm. It's It's 01:24:07.480 --> 01:24:10.397 you know, obvious areas of uncertainty 01:24:10.480 --> 01:24:12.837 uh about your your world and about the 01:24:12.920 --> 01:24:14.877 how the world behaves. 01:24:14.960 --> 01:24:19.117 And you can resolve this uncertainty by 01:24:19.200 --> 01:24:22.877 systematic exploration of that part that you don't you don't know. And if you 01:24:22.960 --> 01:24:25.477 know that you don't know, then, you know, it makes you curious. You kind of 01:24:25.560 --> 01:24:28.437 look into situations that And 01:24:28.520 --> 01:24:31.237 you know, across the animal world, 01:24:31.320 --> 01:24:35.677 different species have different levels of curiosity, right? [laughter] 01:24:35.760 --> 01:24:36.997 Depending on how they're built, right? 01:24:37.080 --> 01:24:41.597 So, you know, [snorts] cats and rats are incredibly curious. Dogs not so much. I 01:24:41.680 --> 01:24:45.837 mean, less. >> Yeah. So, it could be useful to have that kind of curiosity. 01:24:45.920 --> 01:24:48.997 >> So, it would be useful, but curiosity just makes the process faster. It 01:24:49.080 --> 01:24:51.437 doesn't make the process exist. 01:24:51.520 --> 01:24:55.877 The So, what process what learning process 01:24:55.960 --> 01:25:00.557 is it that active learning makes more efficient? 01:25:00.640 --> 01:25:05.117 And I'm asking that first question, you know, >> [snorts] 01:25:05.200 --> 01:25:07.717 >> you know, we haven't answered that question yet. So, you know, I'll worry 01:25:07.800 --> 01:25:09.997 about active learning once this question is uh 01:25:10.080 --> 01:25:11.597 >> So, that's the more fundamental question 01:25:11.680 --> 01:25:15.237 to ask. And if active learning or 01:25:15.320 --> 01:25:17.957 interaction increases the efficiency of 01:25:18.040 --> 01:25:20.437 the learning, see, sometimes 01:25:20.520 --> 01:25:25.597 it becomes very different if the increase is several orders of magnitude, 01:25:25.680 --> 01:25:29.637 right? Like >> That's true. >> But fundamentally it's still the same 01:25:29.720 --> 01:25:33.117 thing and building up the intuition about how to 01:25:33.200 --> 01:25:36.357 in a self-supervised way to construct background models, efficient or 01:25:36.440 --> 01:25:40.557 inefficient, is is the core problem. 01:25:40.640 --> 01:25:44.797 What do you think about Yoshua Bengio's talking about consciousness 01:25:44.880 --> 01:25:46.437 and all of these kinds of concepts? 01:25:46.520 --> 01:25:50.317 >> Okay. I don't know what consciousness is. 01:25:50.400 --> 01:25:53.877 But uh >> That's a good opener. 01:25:53.960 --> 01:25:57.597 >> And to some extent, a lot of the things that are said about consciousness remind 01:25:57.680 --> 01:26:00.117 me of the questions people were asking 01:26:00.200 --> 01:26:02.797 themselves in the 18th century or 17th 01:26:02.880 --> 01:26:07.197 century when they discovered that uh you know, how the eye works and the 01:26:07.280 --> 01:26:10.277 fact that the image at the back of the eye was 01:26:10.360 --> 01:26:14.397 upside down, right? Because you have a lens and And so, on your retina, the 01:26:14.480 --> 01:26:18.357 image that forms is an the of the world, but it's upside-down. How is it that you 01:26:18.440 --> 01:26:21.797 see right-side up? And you know, with what we know today in 01:26:21.880 --> 01:26:25.757 science, you know, we realize this question doesn't make any sense. 01:26:25.840 --> 01:26:29.317 Or or is kind of ridiculous in some way, right? So, I think a lot of what is said 01:26:29.400 --> 01:26:30.917 about consciousness is of that nature. 01:26:31.000 --> 01:26:34.077 Now, that said, there's a lot of really smart people that 01:26:34.160 --> 01:26:37.197 for for whom I have a lot of respect who are talking about this topic, people 01:26:37.280 --> 01:26:40.397 like David Chalmers, who is a colleague of mine at NYU. 01:26:40.480 --> 01:26:43.717 Um I have [snorts] kind of a 01:26:43.800 --> 01:26:47.677 unorthodox folk uh 01:26:47.760 --> 01:26:51.197 speculative hypothesis about consciousness. So, 01:26:51.280 --> 01:26:53.037 we're talking about this idea of world model. 01:26:53.120 --> 01:26:56.477 >> Mhm. >> [clears throat] >> And uh I think, you know, our entire 01:26:56.560 --> 01:26:58.877 prefrontal cortex basically is 01:26:58.960 --> 01:27:02.437 uh the engine for our world model. 01:27:02.520 --> 01:27:06.517 Uh but when we are attending at a particular situation, 01:27:06.600 --> 01:27:09.637 we're focused on that situation. We basically cannot attend to anything 01:27:09.720 --> 01:27:10.517 else. 01:27:10.600 --> 01:27:13.997 And that seems to suggest 01:27:14.080 --> 01:27:17.597 that we basically have only one 01:27:17.680 --> 01:27:21.397 world model engine in our prefrontal cortex. 01:27:21.480 --> 01:27:25.397 Uh that engine is configurable to the situation at hand. So, we're building a 01:27:25.480 --> 01:27:29.597 box out of wood, or we are, you know, driving uh 01:27:29.680 --> 01:27:34.477 down the highway playing chess, we we we basically have uh a single model of the 01:27:34.560 --> 01:27:37.277 world that we configure into the situation at hand. 01:27:37.360 --> 01:27:41.117 Which is why we can only attend to one task at a time. 01:27:41.200 --> 01:27:44.757 Now, if there is a task that we do repeatedly, um 01:27:44.840 --> 01:27:48.797 it it goes from the sort of deliberate reasoning using model of the world and 01:27:48.880 --> 01:27:51.517 prediction and perhaps something like model predictive control, which I was 01:27:51.600 --> 01:27:55.197 talking about earlier, to something that is more subconscious, 01:27:55.280 --> 01:27:58.917 that becomes automatic. So, I don't know if you've ever played against a uh chess 01:27:59.000 --> 01:28:03.837 grandmaster, uh you know, I get wiped out in, you know, 10 01:28:03.920 --> 01:28:07.557 10 plays, right? Uh and, you know, I have to think about my 01:28:07.640 --> 01:28:11.197 move for, you know, like 15 minutes. 01:28:11.280 --> 01:28:13.397 Uh and the person in front of me, the 01:28:13.480 --> 01:28:18.317 grandmaster, you know, would just like react within seconds, right? Uh, 01:28:18.400 --> 01:28:21.357 you know, he doesn't need to think about it. That's become part of his 01:28:21.440 --> 01:28:23.837 subconscious because, you know, it's basically just 01:28:23.920 --> 01:28:25.477 pattern recognition at this point. 01:28:25.560 --> 01:28:28.517 >> [snorts] >> Um, same, you know, you the first few 01:28:28.600 --> 01:28:31.517 hours you drive a car, you are really attentive, you can't do anything else, 01:28:31.600 --> 01:28:35.157 and then after 20, 30 hours of practice, 50 hours, 01:28:35.240 --> 01:28:37.437 you know, it's subconscious, you can talk to the person next to you, you 01:28:37.520 --> 01:28:38.997 know, things like that, right? 01:28:39.080 --> 01:28:41.757 Unless the situation becomes unpredictable, and then you have to stop 01:28:41.840 --> 01:28:46.637 talking. Um, so that suggests you only have one model in your head. 01:28:46.720 --> 01:28:50.837 And it might suggest the idea that consciousness basically is the module 01:28:50.920 --> 01:28:54.597 that configures this world model of yours. You know, you need to have some 01:28:54.680 --> 01:28:59.277 sort of uh, executive kind of overseer that 01:28:59.360 --> 01:29:02.477 configures your world model for the situation at hand. 01:29:02.560 --> 01:29:06.477 And that that leads to the kind of the really curious concept that uh, 01:29:06.560 --> 01:29:10.037 consciousness is not a consequence of the power of our minds, but of the 01:29:10.120 --> 01:29:13.637 limitation of our brains. That because we have only one world 01:29:13.720 --> 01:29:18.397 model, we have to be conscious. If we had as many world models as 01:29:18.480 --> 01:29:21.957 there are as situations we encounter, then we could do all of them 01:29:22.040 --> 01:29:24.997 simultaneously, and we wouldn't need this sort of executive control that we 01:29:25.080 --> 01:29:28.117 call consciousness. >> Yeah, interesting. And somehow maybe 01:29:28.200 --> 01:29:32.037 that executive controller, I mean, the the hard problem of 01:29:32.120 --> 01:29:35.597 consciousness, there's some kind of chemicals and biology that's creating a 01:29:35.680 --> 01:29:38.877 feeling like it feels to experience some of 01:29:38.960 --> 01:29:44.397 these things. That that's kind of like the hard question is 01:29:44.480 --> 01:29:47.237 what the heck [clears throat] is that, and why is that useful? Maybe the more 01:29:47.320 --> 01:29:51.837 pragmatic question, why is it useful to feel like this is really you 01:29:51.920 --> 01:29:55.317 experiencing this versus just like information 01:29:55.400 --> 01:29:59.837 being processed. Um, yeah, it could be just a very nice 01:29:59.920 --> 01:30:02.277 side effect of um, 01:30:02.360 --> 01:30:04.797 of the way we evolved that's just very 01:30:04.880 --> 01:30:06.957 um, useful to 01:30:07.040 --> 01:30:11.957 to uh, feel a sense of ownership to the decisions you make, to 01:30:12.040 --> 01:30:15.597 the perceptions you make, to the model you're trying to maintain. Like you own 01:30:15.680 --> 01:30:18.477 this thing and it's the only one you got and if you 01:30:18.560 --> 01:30:20.477 lose it, it's going to really suck. And 01:30:20.560 --> 01:30:23.117 so you should really send the brain some 01:30:23.200 --> 01:30:25.757 signals about it. 01:30:25.840 --> 01:30:28.717 What ideas do you believe might be true 01:30:28.800 --> 01:30:33.197 that most or at least many people disagree with you with? 01:30:33.280 --> 01:30:35.717 Let's say in the space of machine learning. 01:30:35.800 --> 01:30:38.957 >> Well, depends who you talk about, but I think 01:30:39.040 --> 01:30:42.797 So certainly there is uh a bunch of people who are nativist, 01:30:42.880 --> 01:30:45.517 right? Who think that a lot of the basic things about the world are kind of 01:30:45.600 --> 01:30:47.677 hardwired in our in our minds. 01:30:47.760 --> 01:30:51.037 Um things like you know, the world is three-dimensional for example. Is that 01:30:51.120 --> 01:30:55.197 hardwired? Things like uh you know, object permanence. Is it something that we 01:30:55.280 --> 01:30:58.677 learn uh you know, before the age of 3 months or 01:30:58.760 --> 01:31:03.437 so? Or or are we born with it? And there are you know, very you know, 01:31:03.520 --> 01:31:08.517 wide disagreement among the you know, the cognitive scientists uh for this. 01:31:08.600 --> 01:31:11.237 I think those things are actually very simple to learn. 01:31:11.320 --> 01:31:14.237 Um you know, is it the case that the 01:31:14.320 --> 01:31:18.317 oriented edge detectors in V1 are learned or are they hardwired? I think 01:31:18.400 --> 01:31:21.437 they are learned. They might be learned before birth because it's really easy to 01:31:21.520 --> 01:31:24.877 generate signals from the retina that are actually will train edge detectors. 01:31:24.960 --> 01:31:28.877 So Um and again, those are things that can be learned within 01:31:28.960 --> 01:31:32.997 minutes of opening your eyes, right? I mean uh >> [snorts] 01:31:33.080 --> 01:31:37.117 >> you know, since the 1990s we have uh algorithms that can learn oriented edge 01:31:37.200 --> 01:31:40.717 detectors completely unsupervised with the equivalent of a few minutes of real 01:31:40.800 --> 01:31:43.877 time. So uh so those things have to be learned. 01:31:43.960 --> 01:31:47.757 Um and there's also those you know, MIT experiments where you kind of 01:31:47.840 --> 01:31:50.957 plug the optical nerve on the auditory cortex of 01:31:51.040 --> 01:31:54.117 a baby ferret, right? And that auditory cortex become a visual cortex 01:31:54.200 --> 01:31:58.117 essentially. So you know, clearly um there's learning 01:31:58.200 --> 01:32:01.717 taking place there. Um so you know, I think a lot of what 01:32:01.800 --> 01:32:05.117 people think are so basic that they need to be hardwired, 01:32:05.200 --> 01:32:07.677 I think a lot of those things are learned because they're easy to learn. 01:32:07.760 --> 01:32:12.077 >> Do you So, you put a lot of value in the power of learning. What 01:32:12.160 --> 01:32:15.277 kind of things do you suspect might not be learned? 01:32:15.360 --> 01:32:17.957 Is there something that could not be learned? 01:32:18.040 --> 01:32:22.477 >> So, your intrinsic drives are not learned. They They They are the things 01:32:22.560 --> 01:32:24.597 that, you know, make humans 01:32:24.680 --> 01:32:29.517 uh human or make, you know, cats different from dogs, right? It's 01:32:29.600 --> 01:32:34.957 the the basic drives that are kind of hardwired in our basal ganglia. Um 01:32:35.040 --> 01:32:37.197 I mean, there are people who are working on on this kind of stuff. It's called 01:32:37.280 --> 01:32:41.037 intrinsic motivation in the context of reinforcement learning. Um so, these are 01:32:41.120 --> 01:32:44.957 objective functions where the reward doesn't come from the external world. 01:32:45.040 --> 01:32:48.357 It's computed by your own brain. Your own brain computes 01:32:48.440 --> 01:32:52.797 whether you're happy or not, right? It measures your degree of uh 01:32:52.880 --> 01:32:57.317 comfort or discomfort. Uh and and because it's your brain 01:32:57.400 --> 01:33:00.437 computing this, presumably it knows also how to estimate gradient and so that's 01:33:00.520 --> 01:33:02.597 right. So, um 01:33:02.680 --> 01:33:05.397 uh so, it's easier to to learn when your 01:33:05.480 --> 01:33:08.237 objective is is intrinsic. 01:33:08.320 --> 01:33:11.517 Um so, that has to be hardwired. 01:33:11.600 --> 01:33:15.677 Uh the critic that makes long-term prediction of the outcome, which is the 01:33:15.760 --> 01:33:19.317 eventual result of this, that's learned. 01:33:19.400 --> 01:33:23.277 Uh and perception is learned and your model of the world is learned. But, let 01:33:23.360 --> 01:33:26.917 me take take an example of, you know, why the critic uh I mean, example of why 01:33:27.000 --> 01:33:31.757 the critic may be learned, right? If I uh if I come to you um 01:33:31.840 --> 01:33:35.477 you know, I reach across the table and I I pinch your arm, right? Complete 01:33:35.560 --> 01:33:38.197 surprise for you. You would not have expected this from 01:33:38.280 --> 01:33:40.877 >> expecting that the whole time, but yes, right. [laughter] Let's say for the the 01:33:40.960 --> 01:33:42.997 sake of the story, yes. 01:33:43.080 --> 01:33:49.597 >> Um okay, your basal ganglia is going to light up cuz it's going to hurt, right? 01:33:49.680 --> 01:33:51.997 Uh >> [snorts] >> and now your model of the world includes 01:33:52.080 --> 01:33:55.917 the fact that I may pinch you if I approach my uh 01:33:56.000 --> 01:33:59.357 my uh >> Don't trust humans. Right. My hand to 01:33:59.440 --> 01:34:04.037 your arm. So, if I try again, you're going to recoil. And that's your critic, 01:34:04.120 --> 01:34:09.557 uh your predictive you know, your predictor of your uh ultimate pain 01:34:09.640 --> 01:34:13.717 uh uh system that predicts that something bad is 01:34:13.800 --> 01:34:15.757 going to happen and you recoil to avoid it. 01:34:15.840 --> 01:34:17.077 >> So, even that can be learned. 01:34:17.160 --> 01:34:20.797 >> That is learned, definitely. This is what allows you also to 01:34:20.880 --> 01:34:22.517 uh you know, define some goals, right? 01:34:22.600 --> 01:34:25.637 So, um the fact that you know, you're a 01:34:25.720 --> 01:34:29.877 school child, you wake up in the morning and you go to school and you know, 01:34:29.960 --> 01:34:34.957 um it's not because you necessarily like waking up early and going to school, but 01:34:35.040 --> 01:34:37.797 you know that there is a long-term objective you're trying to optimize. 01:34:37.880 --> 01:34:40.517 >> So, Ernest Becker, I'm not sure if you're familiar with him. He's the 01:34:40.600 --> 01:34:44.077 philosopher who wrote the book The Denial of Death. And his idea is that 01:34:44.160 --> 01:34:47.797 one of the core motivations of human beings is our terror of death, our fear 01:34:47.880 --> 01:34:50.397 of death. Uh that's what makes us unique from 01:34:50.480 --> 01:34:52.717 cats. Cats are just surviving. They do 01:34:52.800 --> 01:34:54.317 not have a 01:34:54.400 --> 01:34:58.637 deep under- like um cognizance, 01:34:58.720 --> 01:35:03.757 introspection that over the horizon is the end. 01:35:03.840 --> 01:35:06.837 And he says that I mean, there's a terror management theory that just all 01:35:06.920 --> 01:35:11.117 these psychological experiments that show the basically 01:35:11.200 --> 01:35:13.437 this idea that 01:35:13.520 --> 01:35:15.637 all of human civilization, everything we 01:35:15.720 --> 01:35:21.157 create is kind of uh trying to forget if even for a brief moment that we're going 01:35:21.240 --> 01:35:25.677 to die. When When do you think humans understand 01:35:25.760 --> 01:35:28.757 that they're going to die? Is it learned early on 01:35:28.840 --> 01:35:33.397 also, like >> I don't know to at what point I mean, 01:35:33.480 --> 01:35:36.797 it's a it's a question like, you know, at what point do you realize that, you 01:35:36.880 --> 01:35:40.197 know, what death really is. And I think most people don't actually realize what 01:35:40.280 --> 01:35:42.877 death is, right? I mean, most people believe that you go to heaven or 01:35:42.960 --> 01:35:46.437 something, right? >> Well, so the the pushback on that what 01:35:46.520 --> 01:35:51.237 Ernest Becker says and um Sheldon Solomon, all of those folks, 01:35:51.320 --> 01:35:54.877 and I find those ideas a little bit compelling is that there is moments in 01:35:54.960 --> 01:35:58.517 life early in life. A lot of this fun happens early in life 01:35:58.600 --> 01:36:00.837 when you are 01:36:00.920 --> 01:36:05.957 uh when you do deeply experience the terror of this realization. And all the 01:36:06.040 --> 01:36:09.517 things you think about about religion, all those kinds of things that we kind 01:36:09.600 --> 01:36:14.037 of think about more like teenage years and later, we're talking about way earlier. 01:36:14.120 --> 01:36:15.957 >> No, it's like 7 or 8 years, something like that. Yeah. 01:36:16.040 --> 01:36:20.717 >> You realize holy crap, this is 01:36:20.800 --> 01:36:25.077 uh like the mystery, the terror. Like it's almost like you're a little prey, a 01:36:25.160 --> 01:36:29.397 little baby deer sitting in the darkness of the jungle, of the woods, looking all 01:36:29.480 --> 01:36:33.117 around you. There's darkness full of terror. I mean, that's that realization 01:36:33.200 --> 01:36:37.437 says, "Okay, I'm going to go go back in the comfort of my mind where there is a 01:36:37.520 --> 01:36:42.397 where there is a deep meaning, where there is a maybe like pretend I'm immortal in 01:36:42.480 --> 01:36:47.117 however way however kind of idea I can construct to 01:36:47.200 --> 01:36:49.037 help me understand that I'm immortal." 01:36:49.120 --> 01:36:52.677 Religion helps with that. You can you can delude yourself in all kinds of 01:36:52.760 --> 01:36:57.317 ways. Like lose yourself in the busyness of each day, have little goals in mind, 01:36:57.400 --> 01:37:00.397 all those kinds of things to think that it's going to go on forever. And you 01:37:00.480 --> 01:37:04.317 kind of know you're going to die, yeah, and it's going to be sad, but you don't 01:37:04.400 --> 01:37:08.317 really understand that you're going to die. And so that's that's their idea. 01:37:08.400 --> 01:37:10.517 And if I find that compelling 01:37:10.600 --> 01:37:13.437 because it does seem to be a core unique 01:37:13.520 --> 01:37:17.717 aspect of human nature that we we're able to think that we're go we're able 01:37:17.800 --> 01:37:22.597 to really understand that this life is finite. I find that important. 01:37:22.680 --> 01:37:23.997 >> This There's a bunch of different things 01:37:24.080 --> 01:37:26.557 there. So, first of all, I don't think there is a qualitative difference 01:37:26.640 --> 01:37:29.397 between between us and cats in the term. 01:37:29.480 --> 01:37:33.477 I think the difference is that we just have a better long-term 01:37:33.560 --> 01:37:37.957 ability to predict, you know, in the long term. And so, we have better 01:37:38.040 --> 01:37:39.237 understanding of how the world works. 01:37:39.320 --> 01:37:42.517 So, we have better understanding of, you know, finiteness of life and things like 01:37:42.600 --> 01:37:45.437 that. >> So, we have a better planning engine than cats? 01:37:45.520 --> 01:37:49.877 >> Yeah. >> Okay. >> Um you know, perhaps >> what's the motivation for 01:37:49.960 --> 01:37:52.077 >> for planning that far >> Well, I think it's just a side effect of 01:37:52.160 --> 01:37:55.917 the fact that we have just a better planning engine because it makes us 01:37:56.000 --> 01:37:59.277 as I said, you know, the essence of intelligence is the ability to predict 01:37:59.360 --> 01:38:03.637 and so the because we're smarter as a side effect we also have this 01:38:03.720 --> 01:38:09.437 ability to kind of make predictions about our own future existence or lack thereof. 01:38:09.520 --> 01:38:14.477 >> Okay. >> Um you say religion helps with that. I think religion hurts actually. 01:38:14.560 --> 01:38:17.917 Um It makes people worry about like, you know, what's going to happen after their 01:38:18.000 --> 01:38:22.877 death, etc. If you believe that, you know, you just don't exist after death. Like, 01:38:22.960 --> 01:38:24.837 you know, it solves completely the problem. At least 01:38:24.920 --> 01:38:28.317 >> you're saying if you don't believe in God, you don't worry about what happens 01:38:28.400 --> 01:38:31.077 after death? >> Yeah. >> I don't know. 01:38:31.160 --> 01:38:36.157 >> worry about the about you know, this life because that's the only one you have. 01:38:36.240 --> 01:38:39.117 >> I think it's well, I don't know I don't know. If I were to say what Ernest 01:38:39.200 --> 01:38:42.837 Becker says then I'll say I agree with him more 01:38:42.920 --> 01:38:47.837 than not is um you do deeply worry. 01:38:47.920 --> 01:38:52.357 Uh if you if you believe there's no God, there's still a deep worry like of the 01:38:52.440 --> 01:38:54.357 mystery of it all. Like, how does that 01:38:54.440 --> 01:38:57.637 make any sense that it just ends? I 01:38:57.720 --> 01:39:00.837 don't think we can truly understand that 01:39:00.920 --> 01:39:02.997 this right I mean, so much of our life, 01:39:03.080 --> 01:39:06.317 the consciousness, the ego is invested 01:39:06.400 --> 01:39:09.157 in this in this being and then 01:39:09.240 --> 01:39:13.477 >> Science keeps bringing humanity down from its pedestal. 01:39:13.560 --> 01:39:16.717 And that's that's just another another example of it. 01:39:16.800 --> 01:39:20.957 >> That's wonderful, but for us individual humans, we don't like to be brought down 01:39:21.040 --> 01:39:25.437 from a pedestal. You're saying like what but see you're fine with it because 01:39:25.520 --> 01:39:28.957 well, so what Ernest Becker would say is you're fine with it because that's just 01:39:29.040 --> 01:39:32.277 a more peaceful existence for you, but you're not really fine. You're hiding 01:39:32.360 --> 01:39:34.037 from In fact, some of the people that 01:39:34.120 --> 01:39:37.237 experienced the deepest trauma that it 01:39:37.320 --> 01:39:39.917 earlier in life, they often before they 01:39:40.000 --> 01:39:44.317 to extensive therapy will say that I'm fine. It's like when you talk to people 01:39:44.400 --> 01:39:46.037 who are truly angry, how are you doing? 01:39:46.120 --> 01:39:49.757 I'm fine. The The question is, what's going on? 01:39:49.840 --> 01:39:53.717 >> Now, I had a near-death experience. I had a very bad 01:39:53.800 --> 01:39:56.517 motorbike accident when I was 17, so. 01:39:56.600 --> 01:40:01.197 >> And >> Uh but that didn't have any impact on my 01:40:01.280 --> 01:40:04.197 reflection on that topic. >> So, I'm I'm basically just playing a bit 01:40:04.280 --> 01:40:07.837 of devil's advocate pushing back and wondering 01:40:07.920 --> 01:40:09.517 is it truly possible to accept death? 01:40:09.600 --> 01:40:13.397 And the flip side that's more interesting, I think, for AI and 01:40:13.480 --> 01:40:16.477 robotics, is how important is it to have 01:40:16.560 --> 01:40:19.077 this as one of the suite of motivations 01:40:19.160 --> 01:40:24.117 is to uh not just avoid 01:40:24.200 --> 01:40:27.357 falling off the roof or something like that, but 01:40:27.440 --> 01:40:28.797 ponder 01:40:28.880 --> 01:40:31.797 the the end of the ride. 01:40:31.880 --> 01:40:34.557 Uh if you listen to the Stoics, 01:40:34.640 --> 01:40:37.357 it's uh it's a great motivator. It adds 01:40:37.440 --> 01:40:40.477 a sense of urgency. So, maybe a to truly 01:40:40.560 --> 01:40:43.917 fear death or be cognizant of it might 01:40:44.000 --> 01:40:49.597 give uh a deeper meaning and urgency to the moment to live 01:40:49.680 --> 01:40:54.317 fully. >> I may I may Maybe I don't disagree with that. 01:40:54.400 --> 01:40:58.797 I mean, I think what motivates me here is uh 01:40:58.880 --> 01:41:02.117 you know, knowing more about about human nature. I mean, I think 01:41:02.200 --> 01:41:05.837 human nature and human intelligence is a big mystery. It's a scientific mystery. 01:41:05.920 --> 01:41:08.437 >> Mhm. [snorts] >> Uh in addition to, you know, 01:41:08.520 --> 01:41:13.957 philosophical and etc., but, you know, I'm a true believer in science, so 01:41:14.040 --> 01:41:18.077 um I and and and I do have kind of a belief 01:41:18.160 --> 01:41:21.917 that for complex systems like like the brain and the mind, 01:41:22.000 --> 01:41:24.477 the the way to 01:41:24.560 --> 01:41:28.997 understand it is to try to reproduce it with, you know, artifacts that you build 01:41:29.080 --> 01:41:32.157 because you know what's essential to it when you try to build it. You know, the 01:41:32.240 --> 01:41:36.317 same way um I've used this analogy before with you, I believe. Um 01:41:36.400 --> 01:41:39.437 the same way [clears throat] we we only started to understand uh 01:41:39.520 --> 01:41:42.357 aerodynamics when we started building airplanes and that helped us understand 01:41:42.440 --> 01:41:45.557 how birds fly. Uh, you know, so I think there's kind of 01:41:45.640 --> 01:41:49.997 a a similar process here where we don't have a theory of a full theory 01:41:50.080 --> 01:41:53.077 of intelligence but building, you know, intelligent 01:41:53.160 --> 01:41:56.397 artifacts will help us perhaps develop some 01:41:56.480 --> 01:42:00.677 you know, underlying theory that encompasses not just artificial 01:42:00.760 --> 01:42:05.797 implements but also uh human and biological intelligence in general. 01:42:05.880 --> 01:42:07.117 >> So, you're an interesting person to ask 01:42:07.200 --> 01:42:09.997 this question about sort of all kinds of 01:42:10.080 --> 01:42:13.317 different other intelligent entities 01:42:13.400 --> 01:42:17.397 or intelligences. What are your thoughts about kind of 01:42:17.480 --> 01:42:19.117 like the Turing or the Chinese room 01:42:19.200 --> 01:42:21.157 question? 01:42:21.240 --> 01:42:24.837 If we create an AI system that exhibits 01:42:24.920 --> 01:42:29.477 a lot of properties of intelligence and consciousness, 01:42:29.560 --> 01:42:34.277 how comfortable are you thinking of that entity as intelligent or conscious? 01:42:34.360 --> 01:42:37.597 So, so you're trying to build novel systems that have intelligence and 01:42:37.680 --> 01:42:42.237 there's metrics about their performance but that metric is 01:42:42.320 --> 01:42:46.957 external. >> Okay. >> So, how are you are you okay calling a 01:42:47.040 --> 01:42:50.957 thing intelligent or you going to be like most humans 01:42:51.040 --> 01:42:56.277 and be once again unhappy to be brought down from a pedestal of consciousness / 01:42:56.360 --> 01:43:02.157 intelligence? >> No, I'm I'll be very happy to understand 01:43:02.240 --> 01:43:05.677 uh more about uh human nature, human mind, 01:43:05.760 --> 01:43:10.117 and human intelligence through the construction of machines that 01:43:10.200 --> 01:43:14.157 uh have similar abilities. And if a 01:43:14.240 --> 01:43:18.157 consequence of this is to bring down humanity one notch down from 01:43:18.240 --> 01:43:21.437 its already low pedestal, I'm just [laughter] fine with it. That's just 01:43:21.520 --> 01:43:25.917 reality of life. Um, so I'm fine with that. Now, you're asking me about things 01:43:26.000 --> 01:43:29.797 that uh opinions I have that a lot of people may disagree with. 01:43:29.880 --> 01:43:34.797 I think uh if we think about the design of 01:43:34.880 --> 01:43:38.717 autonomous intelligence systems, so assuming that we are somewhat successful 01:43:38.800 --> 01:43:41.957 at some at some level of getting machines to learn models of 01:43:42.040 --> 01:43:46.117 the world, predictive models of the world, uh we have we build intrinsic 01:43:46.200 --> 01:43:50.317 motivation objective functions to drive the behavior of that system. Uh the 01:43:50.400 --> 01:43:53.957 system also has perception modules that allows it to estimate the state of the 01:43:54.040 --> 01:43:57.757 world and then have some way of figuring out a sequence of actions that, you 01:43:57.840 --> 01:44:01.197 know, to optimize a particular objective. 01:44:01.280 --> 01:44:05.357 If it has a critic of the type that I was describing before, the thing that 01:44:05.440 --> 01:44:10.477 makes you recoil your arm the second time I tried to pinch you, um 01:44:10.560 --> 01:44:13.597 intelligent autonomous machine will have emotions. 01:44:13.680 --> 01:44:16.237 I think emotions are an integral part of 01:44:16.320 --> 01:44:18.317 autonomous intelligence. 01:44:18.400 --> 01:44:20.917 If you have an intelligent system 01:44:21.000 --> 01:44:25.917 that is driven by uh intrinsic motivation, by objectives, 01:44:26.000 --> 01:44:28.877 if it has a critic that allows it to predict in 01:44:28.960 --> 01:44:32.517 advance whether the outcome of a of a situation is going to be good or 01:44:32.600 --> 01:44:36.117 bad, it's going to have emotions. It's going to have fear >> Yes. 01:44:36.200 --> 01:44:40.037 >> uh when it predicts that the outcome is going to is going to be bad 01:44:40.120 --> 01:44:43.117 and and something to avoid, it's going to have elation when it predicts that 01:44:43.200 --> 01:44:44.837 it's going to be good. >> [snorts] 01:44:44.920 --> 01:44:48.677 >> Um uh if it has drives to relate with 01:44:48.760 --> 01:44:53.117 humans, um you know, in some ways, the way humans have, 01:44:53.200 --> 01:44:57.797 um you know, it's it's going to be social, right? And so it's going to have 01:44:57.880 --> 01:45:01.437 emotions about attachment and and things of that type. So, 01:45:01.520 --> 01:45:06.197 um so I think uh you know, the the sort of sci-fi 01:45:06.280 --> 01:45:09.957 thing where, you know, you see Commander Data like having an emotion chip that he 01:45:10.040 --> 01:45:13.637 can turn off, right? Um I think that's ridiculous. 01:45:13.720 --> 01:45:18.357 >> So, I mean, here's the difficult philosophical social 01:45:18.440 --> 01:45:21.997 question. Do you think there will be a time 01:45:22.080 --> 01:45:25.517 like a civil rights movement for robots where 01:45:25.600 --> 01:45:27.517 um okay, forget the movement, but a 01:45:27.600 --> 01:45:31.677 discussion at the Supreme Court 01:45:31.760 --> 01:45:34.797 that particular kinds of robots, 01:45:34.880 --> 01:45:37.557 you know, particular kinds of systems, 01:45:37.640 --> 01:45:42.957 um, deserve the same rights as humans because they can suffer just as humans 01:45:43.040 --> 01:45:46.677 can. Um, all those kinds of things. 01:45:46.760 --> 01:45:48.077 >> Well, 01:45:48.160 --> 01:45:49.917 perhaps, perhaps not. Like imagine that 01:45:50.000 --> 01:45:53.797 humans were that that you could, uh, 01:45:53.880 --> 01:45:57.677 you know, die and be restored. Like, you know, you could be sort of, you know, be 01:45:57.760 --> 01:46:01.757 3D reprinted and, you know, your brain could be reconstructed in its finest 01:46:01.840 --> 01:46:04.157 details. >> Our ideas of rights will change in that 01:46:04.240 --> 01:46:07.717 case. If you can always just >> [laughter] 01:46:07.800 --> 01:46:11.637 >> there's always a backup, you can always restore, maybe like the the importance 01:46:11.720 --> 01:46:13.877 of murder will go down one notch. 01:46:13.960 --> 01:46:17.717 >> That's right. But also the, uh, your your 01:46:17.800 --> 01:46:19.437 you know, desire to do dangerous things 01:46:19.520 --> 01:46:20.877 like, you know, 01:46:20.960 --> 01:46:25.197 you know, skydiving or you know, uh, 01:46:25.280 --> 01:46:28.917 or or or, you know, race car driving, you know, car racing or that kind of 01:46:29.000 --> 01:46:31.357 stuff, you know, would probably increase. 01:46:31.440 --> 01:46:34.757 Or or, you know, airplane aerobatics or that kind of stuff, right? You would be 01:46:34.840 --> 01:46:38.757 fine to do a lot of those things or explore, you know, dangerous areas and 01:46:38.840 --> 01:46:42.677 things like that. It would kind of change your relationship. So, now, it's 01:46:42.760 --> 01:46:47.197 very likely that robots would be like that because, you know, they'll be 01:46:47.280 --> 01:46:51.077 based on perhaps technology that is somewhat similar to 01:46:51.160 --> 01:46:54.197 uh, today's technology and you can you can always have a backup. 01:46:54.280 --> 01:46:57.597 >> So, it's possible I don't know if you like video games, 01:46:57.680 --> 01:47:01.237 but there's a there's a game called Diablo and, um, 01:47:01.320 --> 01:47:03.797 >> Oh, my my sons are huge fans of this. 01:47:03.880 --> 01:47:07.517 >> Yes. Uh, and >> In fact, they made a game that's 01:47:07.600 --> 01:47:10.917 inspired by it. >> Awesome. Like built a game? 01:47:11.000 --> 01:47:14.557 >> My my three sons have a a game design studio between them, yeah. 01:47:14.640 --> 01:47:17.077 >> That's awesome. >> They came out with a game >> Like just came out with a game? 01:47:17.160 --> 01:47:20.077 >> Last year. No, this was last year, early this year, about a year ago. 01:47:20.160 --> 01:47:23.957 >> That's awesome. But so, in Diablo, there's a something called hardcore mode 01:47:24.040 --> 01:47:27.797 which if you die, there's no you're gone. 01:47:27.880 --> 01:47:30.637 >> Right. >> That's it. And so it's possible with AI 01:47:30.720 --> 01:47:34.477 systems for them to be able to operate 01:47:34.560 --> 01:47:38.237 successfully and for us to treat them in a certain way for cuz if they have to be 01:47:38.320 --> 01:47:40.197 integrated in human society, 01:47:40.280 --> 01:47:43.077 they have to be able to die, no copies 01:47:43.160 --> 01:47:45.757 allowed. In fact, copying is illegal. 01:47:45.840 --> 01:47:49.397 It's possible with humans as well, like cloning will be illegal even when 01:47:49.480 --> 01:47:51.917 possible. >> but cloning is not copying, right? I 01:47:52.000 --> 01:47:56.357 mean, you don't reproduce the the the mind of the person and experience. 01:47:56.440 --> 01:47:58.237 >> It's just a delayed twin, so. 01:47:58.320 --> 01:48:01.397 >> But then it's well but we were talking about with computers that you will be 01:48:01.480 --> 01:48:03.357 able to copy. You You will be able to 01:48:03.440 --> 01:48:07.637 perfectly save, pickle the the the mind 01:48:07.720 --> 01:48:10.357 state. And it's possible that that will 01:48:10.440 --> 01:48:15.157 be illegal because that goes against um 01:48:15.240 --> 01:48:17.797 that will destroy the motivations of the system. 01:48:17.880 --> 01:48:22.117 >> Okay, so let's say you you have a domestic robot, okay, 01:48:22.200 --> 01:48:24.277 sometime in the future. >> Yes. 01:48:24.360 --> 01:48:27.197 >> And uh the domestic robot, you know, comes to 01:48:27.280 --> 01:48:30.957 you kind of somewhat pre-trained, you know, it can do a bunch of things, but 01:48:31.040 --> 01:48:33.597 it has a particular personality that makes it slightly different from the 01:48:33.680 --> 01:48:37.237 other robots because that makes them more interesting. And then because it's, 01:48:37.320 --> 01:48:40.877 you know, it's lived with you for 5 years, you've you've grown some 01:48:40.960 --> 01:48:44.197 attachment to it and vice versa, and 01:48:44.280 --> 01:48:48.877 it's learned a lot about you. Or maybe it's not a household robot, maybe it's 01:48:48.960 --> 01:48:52.037 maybe it's a virtual assistant that lives in your, you know, augmented 01:48:52.120 --> 01:48:53.757 reality glasses or whatever, right? 01:48:53.840 --> 01:48:56.037 >> Yeah. >> Uh you know, the the Her movie type 01:48:56.120 --> 01:48:58.517 thing, right? Um 01:48:58.600 --> 01:49:02.557 and that system to some extent the the 01:49:02.640 --> 01:49:06.877 intelligence in that system is a bit like your child or maybe your PhD 01:49:06.960 --> 01:49:11.797 student in the sense that there's a lot of you in that in that machine now, right? 01:49:11.880 --> 01:49:15.757 >> Yeah. >> And so if it were a living thing, you 01:49:15.840 --> 01:49:19.677 would do this for free if you want, right? If it's your child, your child 01:49:19.760 --> 01:49:23.997 can, you know, then uh live his or her own life and, you 01:49:24.080 --> 01:49:26.997 know, the fact that they learned stuff from you doesn't mean that you have any 01:49:27.080 --> 01:49:29.877 ownership of it, right? But if it's a robot 01:49:29.960 --> 01:49:33.677 uh that you've trained, perhaps you have some 01:49:33.760 --> 01:49:37.597 intellectual property claim about >> Intellectual property? Oh, I thought you 01:49:37.680 --> 01:49:42.077 meant like uh permanent value in the sense that it's part of you is in 01:49:42.160 --> 01:49:43.597 >> Well, there is permanent value, right? 01:49:43.680 --> 01:49:47.237 So, you would lose a lot if that robot were to be destroyed and you you had no 01:49:47.320 --> 01:49:49.877 backup, you would lose a lot. You know, you would lose a lot of investment, you 01:49:49.960 --> 01:49:53.517 know, kind of like a uh you know, a person dying, you know? 01:49:53.600 --> 01:49:57.917 Um that that uh a friend of a friend of yours dying or or co-worker or something 01:49:58.000 --> 01:50:00.397 like that. Um [snorts] 01:50:00.480 --> 01:50:03.237 >> But also uh you have like intellectual 01:50:03.320 --> 01:50:08.157 property rights in the sense that that that system is fine-tuned to your 01:50:08.240 --> 01:50:12.597 particular existence, so that's now a very unique instantiation of that 01:50:12.680 --> 01:50:16.197 original background model, whatever it was, that arrived. 01:50:16.280 --> 01:50:19.957 >> And then there are issues of privacy, right? Because now imagine that that 01:50:20.040 --> 01:50:24.197 robot has its own kind of volition and decides to work for someone else. 01:50:24.280 --> 01:50:29.237 >> Yes. >> Or kind of, you know, thinks uh life with you is sort of untenable or 01:50:29.320 --> 01:50:33.437 whatever. >> Right. >> Um >> [snorts] >> Now, all the all the things that that 01:50:33.520 --> 01:50:37.957 system learned from you, uh you know, how 01:50:38.040 --> 01:50:40.557 can you like, you know, delete all the personal information that that system 01:50:40.640 --> 01:50:43.717 knows about you? >> Yeah. >> Uh I mean, that would be kind of an ethical 01:50:43.800 --> 01:50:45.317 question, like, you know, can you erase 01:50:45.400 --> 01:50:51.717 the the mind of a of a intelligent robot uh to protect your your privacy? 01:50:51.800 --> 01:50:53.477 >> Yeah. >> You can't do this with humans. 01:50:53.560 --> 01:50:57.557 You can ask them to shut up, but that you don't have complete power over them. 01:50:57.640 --> 01:51:00.637 >> Can't erase humans. Yeah, that's the problem with the relationships, you 01:51:00.720 --> 01:51:04.597 know? If they you break up, you can't you can't erase the other human. 01:51:04.680 --> 01:51:08.397 With robots, I think it'll have to be the same thing with robots. That that 01:51:08.480 --> 01:51:11.957 risk that that has to be um, 01:51:12.040 --> 01:51:15.797 some risk to our interactions to truly 01:51:15.880 --> 01:51:18.077 experience them deeply, it feels like. 01:51:18.160 --> 01:51:21.517 So, you have to be able to lose your robot friend. 01:51:21.600 --> 01:51:25.597 And that robot friend to go tweeting about how much of an you were. 01:51:25.680 --> 01:51:30.077 >> But then, are you allowed to, you know, murder the robot to protect your private 01:51:30.160 --> 01:51:31.877 information if the robot decides to leave? 01:51:31.960 --> 01:51:36.437 >> I have this intuition that for robots with with certain 01:51:36.520 --> 01:51:40.357 like it's it's almost like, uh, regulation. If you declare your robot to 01:51:40.440 --> 01:51:44.477 be, let's call it sentient or something like that. Like, this this robot is 01:51:44.560 --> 01:51:47.437 designed for human interaction, then you're not allowed to murder these 01:51:47.520 --> 01:51:50.077 robots, it's the same as murdering other humans. 01:51:50.160 --> 01:51:53.957 >> Well, but what about you do a backup of the robot you do preserve on the on a on 01:51:54.040 --> 01:51:55.757 a hard drive or the equivalent in the future? 01:51:55.840 --> 01:51:59.957 >> That might be illegal. Just like it's like piracy piracy is illegal. 01:52:00.040 --> 01:52:01.717 >> No, but it's your own it's your own robot, right? 01:52:01.800 --> 01:52:06.477 >> But you can't you don't >> But then but then you can wipe out, >> [laughter] 01:52:06.560 --> 01:52:09.757 >> its brain so the this robot doesn't know anything about you anymore, but you 01:52:09.840 --> 01:52:13.557 still have technically it's still in existence because you backed it up. 01:52:13.640 --> 01:52:17.317 >> And then there'll be these great speeches at the Supreme Court by saying, 01:52:17.400 --> 01:52:20.997 "Oh, sure you can erase the mind of the robot just like you can erase the mind 01:52:21.080 --> 01:52:22.917 of a human. We both can suffer." 01:52:23.000 --> 01:52:24.757 There'll be some epic like Obama-type 01:52:24.840 --> 01:52:27.837 character with a speech that we we like 01:52:27.920 --> 01:52:30.357 the robots and the humans are the same. 01:52:30.440 --> 01:52:34.717 Uh, we can both suffer, we can both hope. We can both 01:52:34.800 --> 01:52:36.757 all of those all those kinds of things. 01:52:36.840 --> 01:52:39.117 Raise families, all that kind of stuff. 01:52:39.200 --> 01:52:41.157 It's it's, uh, interesting for these 01:52:41.240 --> 01:52:42.877 just like you said, emotion seems to be 01:52:42.960 --> 01:52:46.797 a fascinatingly powerful aspect of human 01:52:46.880 --> 01:52:50.237 human interaction, human robot interaction. And if they're able to 01:52:50.320 --> 01:52:54.197 exhibit emotions at the end of the day, that's probably 01:52:54.280 --> 01:52:57.797 going to have us deeply consider 01:52:57.880 --> 01:53:02.237 human rights. Like what we value in humans, what we value in other animals. 01:53:02.320 --> 01:53:06.317 That's why robots and AI is great. It makes us ask, uh, really good questions. 01:53:06.400 --> 01:53:10.357 >> questions, yeah. But you I mean you ask about you asked about the Chinese room 01:53:10.440 --> 01:53:12.877 type argument, you know, is it real if it looks real? 01:53:12.960 --> 01:53:16.997 >> Yeah. >> I think the Chinese room argument is a ridiculous one. So 01:53:17.080 --> 01:53:19.637 >> So so for people don't know, Chinese room is 01:53:19.720 --> 01:53:21.917 uh you you can I don't even know how to 01:53:22.000 --> 01:53:26.837 formulate it well, but basically you can mimic the behavior of an 01:53:26.920 --> 01:53:28.717 intelligent system by just following a 01:53:28.800 --> 01:53:32.037 giant algorithm codebook that tells you 01:53:32.120 --> 01:53:36.597 exactly how to respond in exactly each case, but is that really intelligent? 01:53:36.680 --> 01:53:40.437 It's like a giant lookup table. When this person says this, you answer this. 01:53:40.520 --> 01:53:42.917 When this person says this, you answer this. 01:53:43.000 --> 01:53:46.277 And um if you understand how that works, you 01:53:46.360 --> 01:53:50.677 have this giant nearly infinite lookup table, is that really intelligence? Cuz 01:53:50.760 --> 01:53:55.277 intelligence seems to be a mechanism that's much more interesting and complex 01:53:55.360 --> 01:53:56.517 than this lookup table. 01:53:56.600 --> 01:53:58.997 >> I don't think so. So the I mean the real 01:53:59.080 --> 01:54:03.077 question comes down to do you think uh 01:54:03.160 --> 01:54:06.957 you know, you can you can make an AI's uh intelligence in some way, even if if 01:54:07.040 --> 01:54:11.317 that involves uh learning. And the answer is of course yes. There's 01:54:11.400 --> 01:54:15.197 no question. Uh there's a second question then, which is 01:54:15.280 --> 01:54:19.397 uh assuming you can uh reproduce intelligence in sort of different 01:54:19.480 --> 01:54:24.317 hardware than biological hardware, you know, what I computers, um 01:54:24.400 --> 01:54:26.517 can you uh you know 01:54:26.600 --> 01:54:30.237 match uh human intelligence in 01:54:30.320 --> 01:54:34.757 all the domains in which humans are intelligent? 01:54:34.840 --> 01:54:39.037 Is it possible, right? So this is the quite the hypothesis of strong AI. The 01:54:39.120 --> 01:54:42.837 answer to this in my opinion is a unqualified yes. This 01:54:42.920 --> 01:54:44.517 will this will happen at some point. 01:54:44.600 --> 01:54:47.997 There's no question that machines at some point will become more intelligent 01:54:48.080 --> 01:54:51.517 than humans in all domains where humans are intelligent. This is not for 01:54:51.600 --> 01:54:56.637 tomorrow. It's going to take a long time uh [snorts] regardless of what, you know, 01:54:56.720 --> 01:55:01.597 Elon and others have claimed or believed. Um this is a lot a lot harder 01:55:01.680 --> 01:55:04.477 than many of many of those guys think it is. 01:55:04.560 --> 01:55:08.757 Um and many of those guys who thought it was simpler than that years you know, 5 01:55:08.840 --> 01:55:13.197 years ago now think it's hard because it's been 5 years and they realize it's 01:55:13.280 --> 01:55:15.277 it's going to take a lot longer. 01:55:15.360 --> 01:55:17.997 That includes [snorts] a bunch of people at DeepMind for example, but um 01:55:18.080 --> 01:55:21.197 >> Oh, interesting. I haven't actually uh touched base with the DeepMind folks. 01:55:21.280 --> 01:55:23.757 But some of it Elon or 01:55:23.840 --> 01:55:26.637 uh Demis Hassabis, I mean sometimes in 01:55:26.720 --> 01:55:30.037 your role you have to kind of create 01:55:30.120 --> 01:55:35.077 deadlines that are nearer than farther away to to kind of create an urgency cuz 01:55:35.160 --> 01:55:37.437 you know, you have to believe the impossible is possible in order to 01:55:37.520 --> 01:55:41.037 accomplish it. And there's of course a flip side to that coin, but it's it's a 01:55:41.120 --> 01:55:44.357 weird You can't be too cynical if you want to get something done. 01:55:44.440 --> 01:55:48.597 >> Absolutely. I I I agree with that, but um I mean you have to inspire people, 01:55:48.680 --> 01:55:51.197 right? To work on certain ambitious things. 01:55:51.280 --> 01:55:55.437 Um Uh so, you know, it's it's uh it's 01:55:55.520 --> 01:55:58.237 certainly uh a lot harder than we believe, but there's no question in my 01:55:58.320 --> 01:56:00.037 mind that this will this will happen. 01:56:00.120 --> 01:56:03.637 And now, you know, people are kind of worried about what does that mean for 01:56:03.720 --> 01:56:07.077 humans? They they are going to be brought down from their pedestal, you 01:56:07.160 --> 01:56:11.757 know, a bunch of notches. With that and uh 01:56:11.840 --> 01:56:14.397 you know, is that going to be good or bad? I mean, it's just going to give 01:56:14.480 --> 01:56:18.037 more power, right? To sort of amplify your for human intelligence, really. 01:56:18.120 --> 01:56:20.357 >> So, speaking of doing cool ambitious 01:56:20.440 --> 01:56:24.437 things, FAIR, the Facebook AI research 01:56:24.520 --> 01:56:28.397 group, has recently celebrated its eighth birthday. Or 01:56:28.480 --> 01:56:32.557 maybe you can correct me on that. Uh looking back, what has been the 01:56:32.640 --> 01:56:36.357 successes, the failures, the lessons learned from the 8 years of FAIR? And 01:56:36.440 --> 01:56:39.477 maybe you can also give context of 01:56:39.560 --> 01:56:44.477 the newly minted Meta AI fit into uh how does it relate to FAIR? 01:56:44.560 --> 01:56:48.077 >> Right. So, let me tell you a little bit about the organization of all this. Uh 01:56:48.160 --> 01:56:52.597 uh Yeah, FAIR was created almost exactly 8 years ago. Uh it wasn't called FAIR 01:56:52.680 --> 01:56:55.877 yet. It took that name a few a few months later. 01:56:55.960 --> 01:56:58.557 Um and at the time uh I joined uh 01:56:58.640 --> 01:57:03.567 Facebook, um there was a a group called the AI group that had about 12 engineers 01:57:03.650 --> 01:57:06.717 [snorts] and a few sci- a few scientists, like, you know, 10 engineers 01:57:06.800 --> 01:57:08.157 and two scientists or something like that. 01:57:08.240 --> 01:57:12.597 >> Mhm. >> I run it for 3 and 1/2 years as a director. 01:57:12.680 --> 01:57:16.237 You know, hired the first few scientists and kind of set up the culture and 01:57:16.320 --> 01:57:20.837 organized it, you know, explained to the Facebook leadership what what 01:57:20.920 --> 01:57:24.677 fundamental research was about and how it can work within uh 01:57:24.760 --> 01:57:27.917 uh industry and how it needs to be open and everything. 01:57:28.000 --> 01:57:32.877 Um and uh I think it's been an an qualified 01:57:32.960 --> 01:57:35.797 success in the sense that 01:57:35.880 --> 01:57:39.757 FAIR has simultaneously produced um you 01:57:39.840 --> 01:57:43.517 know, top-level research and advanced the the science and the technology, 01:57:43.600 --> 01:57:45.477 provided tools, open-source tools like 01:57:45.560 --> 01:57:47.837 PyTorch and many others. 01:57:47.920 --> 01:57:51.237 Uh but at the same time has had a 01:57:51.320 --> 01:57:57.397 a direct or mostly indirect impact on uh Facebook at the time, now Meta. 01:57:57.480 --> 01:58:00.477 >> Mhm. >> Uh in the sense that a lot of systems 01:58:00.560 --> 01:58:05.037 that are that Meta is built around um now are 01:58:05.120 --> 01:58:10.237 are based on research projects that uh started at uh at FAIR. 01:58:10.320 --> 01:58:13.557 So, if [snorts] you were to take out, you know, deep learning out of uh 01:58:13.640 --> 01:58:18.717 Facebook services now and and Meta more generally, um I mean, the company would 01:58:18.800 --> 01:58:21.557 would literally crumble. I mean, it's completely built around 01:58:21.640 --> 01:58:25.877 uh around AI these days. It's really essential to the operations. 01:58:25.960 --> 01:58:30.117 So, what happened uh after 3 and 1/2 years is that uh 01:58:30.200 --> 01:58:33.277 I changed role. I became chief scientist, so I'm I'm not doing 01:58:33.360 --> 01:58:36.757 day-to-day management of uh of FAIR anymore. 01:58:36.840 --> 01:58:40.117 I'm more of a kind of you know, think about strategy and 01:58:40.200 --> 01:58:43.797 things like that and and I carry my I I conduct my own research. I have, you 01:58:43.880 --> 01:58:46.437 know, my own kind of research group working on self-supervised learning and 01:58:46.520 --> 01:58:49.677 things like this, which I didn't have time to do when I was director. 01:58:49.760 --> 01:58:53.117 >> [snorts] >> Um so, now uh FAIR is run by, uh, Joelle 01:58:53.200 --> 01:58:57.437 Pineau and, uh, Antoine Bord >> [snorts] >> uh, together because FAIR is kind of 01:58:57.520 --> 01:59:01.277 split in two now. There's something called FAIR Labs, which is sort of 01:59:01.360 --> 01:59:04.917 bottom-up scientist-driven research, and FAIR Excel, which is slightly more 01:59:05.000 --> 01:59:09.957 organized for bigger projects that require a little more kind of focus and more 01:59:10.040 --> 01:59:13.677 engineering support and things like that. So, Joelle leads FAIR Lab and 01:59:13.760 --> 01:59:14.837 Antoine Bord leads FAIR Excel. 01:59:14.920 --> 01:59:19.957 >> Where are they located? >> Uh, it's decentralized all over. Um, 01:59:20.040 --> 01:59:23.277 so there's there's there's no question that the leadership, uh, 01:59:23.360 --> 01:59:25.997 of the company believes that 01:59:26.080 --> 01:59:30.357 this was a very worthwhile investment, and what that means is that 01:59:30.440 --> 01:59:34.757 uh, it's it's there for the long run. 01:59:34.840 --> 01:59:38.437 Right? So, there is, uh, if you if you want to talk in these 01:59:38.520 --> 01:59:41.437 terms, which I don't like, this is a this is a business model, if you want, 01:59:41.520 --> 01:59:46.157 where where uh, FAIR, despite being a very fundamental research lab, brings a lot 01:59:46.240 --> 01:59:49.797 of value to the company, either mostly indirectly through other groups. 01:59:49.880 --> 01:59:53.197 >> Mhm. >> Um, now, what happened 3 and 1/2 years 01:59:53.280 --> 01:59:55.037 ago when I stepped down is, uh, was also 01:59:55.120 --> 02:00:00.317 the creation of Facebook AI, which was basically a larger organization that, 02:00:00.400 --> 02:00:05.877 uh, covers FAIR, so FAIR is included in it, but also has other organizations that 02:00:05.960 --> 02:00:10.357 are uh, focused on applied research or advanced 02:00:10.440 --> 02:00:15.757 development uh, of AI technology that is more, you know, focused on the >> Right. 02:00:15.840 --> 02:00:18.597 >> products of the company. >> So, less emphasis on fundamental research. 02:00:18.680 --> 02:00:21.037 >> Less fundamental, but it's still a research. I mean, there's a lot of 02:00:21.120 --> 02:00:26.077 papers coming out of those organizations and uh, the people are some awesome and, 02:00:26.160 --> 02:00:30.917 you know, wonderful to interact with, and but it it serves, uh, as kind of, 02:00:31.000 --> 02:00:34.957 uh, uh, a way to kind of scale up, if you 02:00:35.040 --> 02:00:38.237 want, um, sort of AI technology which, you know, 02:00:38.320 --> 02:00:41.917 may be very experimental and and sort of lab prototypes into things that are 02:00:42.000 --> 02:00:44.157 usable. >> So, FAIR is [snorts] a subset of Meta 02:00:44.240 --> 02:00:47.237 AI. Is FAIR become like KFC? It it'll 02:00:47.320 --> 02:00:49.717 just keep the F. Nobody cares what the F 02:00:49.800 --> 02:00:51.317 stands for. 02:00:51.400 --> 02:00:55.157 >> We'll know soon enough by probably 02:00:55.240 --> 02:00:57.597 probably by the end of the of 2021. 02:00:57.680 --> 02:01:00.277 >> not a giant change, mayor fair. 02:01:00.360 --> 02:01:04.077 >> Well, mayor doesn't sound too good, but you know, the the brand people are kind 02:01:04.160 --> 02:01:08.397 of deciding on this. And they've been hesitating for for a while now, and they 02:01:08.480 --> 02:01:10.877 you know, they tell us they're going to come up with an answer as to whether 02:01:10.960 --> 02:01:13.797 fair is going to change name or whether we're going to change just the meaning 02:01:13.880 --> 02:01:16.557 of the F. >> Oh, that's a good call. I would keep 02:01:16.640 --> 02:01:18.037 fair and change the meaning of the F. 02:01:18.120 --> 02:01:20.197 That's a good one. >> be my preference, you know, I would turn 02:01:20.280 --> 02:01:22.757 I would turn the F into fundamental. 02:01:22.840 --> 02:01:24.237 >> Oh, that's good. >> Fundamentally I research. 02:01:24.320 --> 02:01:25.157 >> Oh, that's really good. Yeah, yeah. 02:01:25.240 --> 02:01:28.357 >> Within meta AI. So, this would be sort of meta fair. 02:01:28.440 --> 02:01:30.157 >> Yeah. But, you know, people will call it fair, right? 02:01:30.240 --> 02:01:34.197 >> Yeah, exactly. I like it. >> meta AI 02:01:34.280 --> 02:01:38.997 uh is part of the reality lab. So, you 02:01:39.080 --> 02:01:43.157 know, uh uh meta now the new Facebook right is 02:01:43.240 --> 02:01:45.317 called meta and it's kind of 02:01:45.400 --> 02:01:51.107 divided into you know, Facebook, Instagram, WhatsApp. 02:01:51.190 --> 02:01:54.157 >> [snorts] >> And 02:01:54.240 --> 02:01:58.797 reality lab. And reality lab is about, you know, AR, VR, um 02:01:58.880 --> 02:02:02.197 you know, telepresence, communication part of uh technology and stuff like 02:02:02.280 --> 02:02:06.117 that. It's kind of the you can think of it as the sort of a 02:02:06.200 --> 02:02:10.957 combination of um sort of new products and and technology 02:02:11.040 --> 02:02:15.557 part of uh of uh meta. >> Is that where the touch sensing for 02:02:15.640 --> 02:02:18.197 robots? I saw that you were posting about. That's that's 02:02:18.280 --> 02:02:21.317 >> sensing for robot is part of fair, actually. That's that's a fair part. 02:02:21.400 --> 02:02:25.877 >> Okay, cool. >> Yeah. This is also the No, but there is the the other way the 02:02:25.960 --> 02:02:29.517 the haptic glove, right? That that has like >> more reality lab. 02:02:29.600 --> 02:02:32.557 >> That's that's reality lab research. 02:02:32.640 --> 02:02:35.997 >> Reality lab research. But, by the way, the touch sensors are super interesting. 02:02:36.080 --> 02:02:39.237 Uh like integrating that modality into the whole 02:02:39.320 --> 02:02:42.037 uh sensing uh suite is very interesting. 02:02:42.120 --> 02:02:45.597 So, uh what do you think about the metaverse? 02:02:45.680 --> 02:02:50.317 What do you think about this whole this whole kind of expansion of the view 02:02:50.400 --> 02:02:52.837 of the role of Facebook and Meta in the world? 02:02:52.920 --> 02:02:56.197 >> Well, a metaverse really should be thought of as the next step in the 02:02:56.280 --> 02:03:00.237 internet, right? Sort of trying to 02:03:00.320 --> 02:03:03.157 uh kind of, you know, make the 02:03:03.240 --> 02:03:07.597 experience more compelling um of, you know, being 02:03:07.680 --> 02:03:10.517 connected either with other people or with content. 02:03:10.600 --> 02:03:15.437 Um and you know, we are evolved and trying to 02:03:15.520 --> 02:03:20.677 evolve in, you know, 3D environments where uh you know, we can see other people, we 02:03:20.760 --> 02:03:24.157 can talk to them when you're when when near them or 02:03:24.240 --> 02:03:26.397 you know, and other people who are far away can't hear us, you know, things 02:03:26.480 --> 02:03:30.517 like that, right? So, it it's uh there's a lot of social conventions that exist 02:03:30.600 --> 02:03:33.997 in the real world that we can try to transpose. Now, what is going to be 02:03:34.080 --> 02:03:38.117 eventually the the uh how compelling is it going to be? 02:03:38.200 --> 02:03:40.797 Like are, you know, uh is it going to be the case that 02:03:40.880 --> 02:03:44.557 people are going to be willing to do this if they have to wear, you know, 02:03:44.640 --> 02:03:49.477 a huge pair of goggles all day? Maybe not. Um >> But then again, if the experience is 02:03:49.560 --> 02:03:52.237 sufficiently compelling, maybe so. 02:03:52.320 --> 02:03:55.877 >> Or if the device that you have to wear is just basically a pair of glasses, you 02:03:55.960 --> 02:03:58.877 know, technology makes sufficient progress for that. 02:03:58.960 --> 02:04:04.357 Um um you know, AR is a much easier concept to grasp that you're going to 02:04:04.440 --> 02:04:09.517 have, you know, augmented reality glasses that basically contain some sort of, you 02:04:09.600 --> 02:04:12.157 know, virtual assistant that can help you in your daily lives. 02:04:12.240 --> 02:04:14.637 >> But at the same at the same time with the AR, you have to contend with 02:04:14.720 --> 02:04:17.917 reality. With VR, you can completely detach yourself from reality, so it 02:04:18.000 --> 02:04:22.237 gives you freedom. It might be easier to design worlds in in VR. 02:04:22.320 --> 02:04:26.597 >> Yeah, but you you you can imagine how, you know, the metaverse being uh 02:04:26.680 --> 02:04:29.637 >> a mix. >> A mix, right? Or or like you can have 02:04:29.720 --> 02:04:32.597 objects that exist in the metaverse that, you know, pop up on top of the 02:04:32.680 --> 02:04:36.277 real world or only exist in virtual reality. 02:04:36.360 --> 02:04:38.917 >> Okay, let me ask the hard question. 02:04:39.000 --> 02:04:40.437 >> Oh, because all of this was easy so far. 02:04:40.520 --> 02:04:42.517 >> This was easy. Uh 02:04:42.600 --> 02:04:45.237 the Facebook, now Meta, the social 02:04:45.320 --> 02:04:48.477 network has been painted by the media as 02:04:48.560 --> 02:04:52.957 a net negative for society, even destructive and evil at times. You've 02:04:53.040 --> 02:04:54.877 pushed back against this, defending 02:04:54.960 --> 02:04:58.477 Facebook. Can you explain your defense? 02:04:58.560 --> 02:05:02.917 >> Yeah, so the the description, the company that is being described in the 02:05:03.000 --> 02:05:07.797 in some media uh is not the company we know when we 02:05:07.880 --> 02:05:11.357 work inside. And you know 02:05:11.440 --> 02:05:14.997 it could be claimed that a lot of employees are uninformed about 02:05:15.080 --> 02:05:18.747 what really goes on in the company, but you know, I'm a vice president. I mean, 02:05:18.830 --> 02:05:21.197 [snorts] I have a pretty good vision of what goes on, you know, I don't know 02:05:21.280 --> 02:05:25.077 everything obviously. I'm not involved in in in everything, but certainly not 02:05:25.160 --> 02:05:28.837 in decision about like, you know, content moderation or anything like this, but 02:05:28.920 --> 02:05:32.077 but I have, you know, some decent vision of what what goes on. 02:05:32.160 --> 02:05:35.557 And this evil that is being described, I just don't see it. 02:05:35.640 --> 02:05:40.077 And then, you know, I think there is an easy story to buy 02:05:40.160 --> 02:05:44.157 which is that, you know, all the bad things in the in the world and you know, 02:05:44.240 --> 02:05:47.757 the the reason your friend believe crazy stuff. 02:05:47.840 --> 02:05:51.557 Um you know, there's an easy scapegoat, right? In the 02:05:51.640 --> 02:05:56.397 uh in in in in social media in general, Facebook in particular. 02:05:56.480 --> 02:05:59.157 But you have to look at the data. Like, is it the case 02:05:59.240 --> 02:06:04.197 that uh Facebook, for example, uh polarizes people politically? 02:06:04.280 --> 02:06:07.157 Um are there academic studies that show this? 02:06:07.240 --> 02:06:12.517 Is it the case that uh you know, teenagers uh think of themselves less if they use 02:06:12.600 --> 02:06:16.517 Instagram more? Is it the case that uh 02:06:16.600 --> 02:06:20.717 you know, people get more riled up against uh 02:06:20.800 --> 02:06:25.117 you know, opposite sides in a in a debate or political opinion if they if 02:06:25.200 --> 02:06:27.597 they are more on Facebook or if they are less? 02:06:27.680 --> 02:06:32.797 And uh study after study show that none of this is true. 02:06:32.880 --> 02:06:36.957 This is independent studies by academic that are not funded by Facebook or Meta. 02:06:37.040 --> 02:06:40.557 Um, you know, studied by Stanford, by some of my colleagues at NYU actually 02:06:40.640 --> 02:06:42.317 with whom I have no connection. 02:06:42.400 --> 02:06:44.477 Um, you know, there's a study recently 02:06:44.560 --> 02:06:47.237 they they they paid people, I think it 02:06:47.320 --> 02:06:50.077 was in um 02:06:50.160 --> 02:06:54.357 in in in the former Yugoslavia, I'm not exactly sure in what what part, but they 02:06:54.440 --> 02:06:57.757 paid people to not use Facebook for a while in the 02:06:57.840 --> 02:07:04.197 period um uh before the anniversary of the Srebrenica 02:07:04.280 --> 02:07:07.477 >> Mhm. >> massacres, right? So, you know, people get riled up like should you know, 02:07:07.560 --> 02:07:11.437 should we have a celebration, I mean a memorial kind of 02:07:11.520 --> 02:07:14.437 celebration for it or not. So, they paid [snorts] a bunch of people 02:07:14.520 --> 02:07:18.157 to not use Facebook for a few weeks. 02:07:18.240 --> 02:07:22.277 It turns out that those people ended up being more 02:07:22.360 --> 02:07:25.357 polarized than they were at the beginning and the people who were more 02:07:25.440 --> 02:07:27.757 on Facebook were less polarized. 02:07:27.840 --> 02:07:31.157 Um, there's a study, you know, from Stanford of 02:07:31.240 --> 02:07:35.877 uh economists at Stanford that tried to identify the causes of increasing 02:07:35.960 --> 02:07:39.717 polarization in the US. And it's been going on for 40 years 02:07:39.800 --> 02:07:42.437 before, you know, Mark Zuckerberg was born. 02:07:42.520 --> 02:07:46.837 >> Yeah. >> Uh continuously. And um and uh so the if 02:07:46.920 --> 02:07:50.077 there is a cause, it's not Facebook or social media. So, you could say if 02:07:50.160 --> 02:07:54.277 social media just accelerated but no, I mean it's basically a continuous 02:07:54.360 --> 02:07:57.797 uh evolution by some measure of uh polarization in the US. 02:07:57.880 --> 02:08:01.517 And then you compare this with other countries like uh >> [snorts] 02:08:01.600 --> 02:08:05.757 >> uh the the the west half of Germany because you can go 40 years in the east 02:08:05.840 --> 02:08:09.237 east side or Denmark or or other countries. 02:08:09.320 --> 02:08:12.557 Uh and they use Facebook just as much and they're not getting more polarized, 02:08:12.640 --> 02:08:16.397 they're getting less polarized. So, if you want to look for, you know, uh a 02:08:16.480 --> 02:08:21.077 causal relationship there um you can find a scapegoat, but you can't 02:08:21.160 --> 02:08:24.397 find a cause. Now, if you want to fix the problem, you have to find the right 02:08:24.480 --> 02:08:26.837 cause. And what riled me up is that 02:08:26.920 --> 02:08:31.477 people now are accusing Facebook of bad deeds that are done by others and those 02:08:31.560 --> 02:08:33.557 others are we're not doing anything about them. 02:08:33.640 --> 02:08:36.477 >> Mhm. >> And by the way, those others include the 02:08:36.560 --> 02:08:39.557 owner of the Wall Street Journal in which all of those papers were published. 02:08:39.640 --> 02:08:44.077 >> So, I should mention that I'm talking to Sheryl Mike Sheryl for on this podcast 02:08:44.160 --> 02:08:46.837 and also Mark Zuckerberg and probably these are conversations you can have 02:08:46.920 --> 02:08:49.837 with them. Uh cuz it's very interesting to me even 02:08:49.920 --> 02:08:53.957 if Facebook has an has some measurable negative effect. You 02:08:54.040 --> 02:08:55.677 can't just consider that in isolation. 02:08:55.760 --> 02:08:58.797 You have to consider about all the positive ways that it connects us. So, 02:08:58.880 --> 02:09:00.757 like every technology 02:09:00.840 --> 02:09:03.237 >> It's a question. You can't just say like 02:09:03.320 --> 02:09:08.117 uh there's an increase in division. Yes, probably Google search engine has 02:09:08.200 --> 02:09:11.797 created increase in division. We have to consider about how much information are 02:09:11.880 --> 02:09:15.757 brought to the world. Like I'm sure Wikipedia created more division. If you 02:09:15.840 --> 02:09:19.517 just look at the division, we have to look at the full context of the world 02:09:19.600 --> 02:09:22.837 and did it make a better world? You have to >> the printing press has created more 02:09:22.920 --> 02:09:26.677 division, right? >> Exactly. >> So, you know, when the 02:09:26.760 --> 02:09:30.757 printing press was invented, the first books that were that were 02:09:30.840 --> 02:09:34.677 printed were things like the Bible and that allowed people to read the Bible by 02:09:34.760 --> 02:09:37.117 themselves, not get the message uniquely 02:09:37.200 --> 02:09:41.757 from priests in Europe. And that created, you know, the Protestant 02:09:41.840 --> 02:09:46.637 movement and 200 years of religious persecution and wars. So, that's a bad 02:09:46.720 --> 02:09:49.677 side effect of the printing press. You know, social networks aren't being 02:09:49.760 --> 02:09:52.597 nearly as bad as the printing press, but nobody would say the printing press was 02:09:52.680 --> 02:09:57.197 a bad idea. >> [sighs and gasps] >> Yeah, a lot of it is perception and 02:09:57.280 --> 02:10:00.197 there's a lot of different incentives operating here. 02:10:00.280 --> 02:10:02.757 Um maybe a quick comment since you're 02:10:02.840 --> 02:10:07.477 one of the top leaders at Facebook and and at meta, sorry, that's in the tech 02:10:07.560 --> 02:10:10.997 space. I'm sure Facebook involves a lot of 02:10:11.080 --> 02:10:15.077 incredible technological challenges that need to be solved. A lot 02:10:15.160 --> 02:10:18.117 of it probably is on the computer infrastructure, the hardware, the I 02:10:18.200 --> 02:10:20.437 mean, it's just a huge amount. 02:10:20.520 --> 02:10:23.317 Uh maybe can you give me context about 02:10:23.400 --> 02:10:28.277 how much of Sherp's life is AI and how much of it is low-level compute, how 02:10:28.360 --> 02:10:31.477 much of it is flying all around doing business stuff 02:10:31.560 --> 02:10:33.917 in the same way as Zuckerberg, Mark Zuckerberg. 02:10:34.000 --> 02:10:37.237 >> They really focus on AI. I mean, certainly, uh, 02:10:37.320 --> 02:10:41.397 in the, uh, in the run-up of the creation of Fair 02:10:41.480 --> 02:10:45.557 and for, you know, at least a year after that, if not more, 02:10:45.640 --> 02:10:47.717 uh, Mark was was very, very much focused 02:10:47.800 --> 02:10:52.677 on on AI and was spending quite a lot of effort, um, on it. And that's his style. 02:10:52.760 --> 02:10:55.797 When he gets interested in something, he reads everything about it. 02:10:55.880 --> 02:10:57.237 You know, he read some of my papers, for 02:10:57.320 --> 02:11:03.397 example, before he joined. Um, um, and, uh, so he he learned a lot 02:11:03.480 --> 02:11:08.357 about it. >> notes. >> [laughter] >> Right? Um, >> [snorts] 02:11:08.440 --> 02:11:13.357 >> and, uh, uh, you know, Sherp was really tweeting also. I mean, Sherp 02:11:13.440 --> 02:11:18.237 is really kind of, um, you know, has 02:11:18.320 --> 02:11:22.237 something I've tried to preserve also, despite my, uh, not so young age, which 02:11:22.320 --> 02:11:25.997 is a a sense of wonder about, uh, science and technology. And he certainly 02:11:26.080 --> 02:11:29.357 he certainly has that. Um, he's also a wonderful person. I 02:11:29.440 --> 02:11:33.237 mean, in terms of, uh, like, uh, as a manager, like, dealing with people 02:11:33.320 --> 02:11:35.517 and everything. Mark also, actually. 02:11:35.600 --> 02:11:39.677 Um, so I mean, they're very like, you know, very human people. 02:11:39.760 --> 02:11:44.077 For in the case of Mark, it's, uh, shockingly human, you know, given his, uh, 02:11:44.160 --> 02:11:49.077 his trajectory. Uh, I mean, the personality of of him that 02:11:49.160 --> 02:11:51.477 is painted in the press is just completely wrong. 02:11:51.560 --> 02:11:55.197 >> Yeah. But you have to know how to play the press. So that's 02:11:55.280 --> 02:11:56.997 I I put some of that responsibility on 02:11:57.080 --> 02:12:00.197 him, too. You have to 02:12:00.280 --> 02:12:04.717 It's like, um, you know, like, the director, the 02:12:04.800 --> 02:12:08.877 conductor of an orchestra, you have to play the press and the public 02:12:08.960 --> 02:12:11.997 in a certain kind of way where you convey your true self to them. If there 02:12:12.080 --> 02:12:13.597 is a depth and kindness to you. 02:12:13.680 --> 02:12:17.957 >> And he's probably And he's probably not the best at it. So, yeah. 02:12:18.040 --> 02:12:20.557 >> [laughter] >> You have to learn. And it's it's sad to 02:12:20.640 --> 02:12:25.957 see I'll talk to him about it, but the strap is slowly stepping down. 02:12:26.040 --> 02:12:31.317 It's always sad to see folks sort of be there for a long time and slowly 02:12:31.400 --> 02:12:34.357 I guess time is up. >> You know, I think he's I think he's done 02:12:34.440 --> 02:12:38.477 the thing he set out to do and, you know, he's he's 02:12:38.560 --> 02:12:44.637 got, you know, uh family priorities and stuff like that and 02:12:44.720 --> 02:12:49.797 um I I understand, you know, after 13 years or something. 02:12:49.880 --> 02:12:53.337 >> It's been a good run. >> Which in Silicon Valley is basically a 02:12:53.420 --> 02:12:56.917 [laughter] lifetime. >> Yeah. >> You know, cuz you know, it's dog years. 02:12:57.000 --> 02:13:00.317 >> So, NeurIPS the conference just wrapped up. 02:13:00.400 --> 02:13:03.717 Let [clears throat] me just go back to something else. You posted a paper you 02:13:03.800 --> 02:13:06.357 co-authored was rejected from NeurIPS. 02:13:06.440 --> 02:13:09.697 As you said, proudly in quotes, rejected. 02:13:09.780 --> 02:13:12.837 >> [laughter] >> No joke. >> Yeah, I know. 02:13:12.920 --> 02:13:15.837 Uh can you describe this paper and like 02:13:15.920 --> 02:13:20.397 what what what was the idea in it? And also, maybe this is a good opportunity 02:13:20.480 --> 02:13:24.277 to ask what are the pros and cons, what works and what doesn't about the review 02:13:24.360 --> 02:13:26.837 process. >> Yeah, let me talk about the paper first. 02:13:26.920 --> 02:13:30.557 I'll talk about the review talk about the review process afterwards. 02:13:30.640 --> 02:13:34.477 Um the paper is called VICReg. So, this is I mentioned that before, 02:13:34.560 --> 02:13:38.277 variance-invariance covariance regularization. And it's a technique in 02:13:38.360 --> 02:13:42.557 non-contrastive learning technique for what I call joint embedding 02:13:42.640 --> 02:13:44.797 architecture. So, Siamese nets are an 02:13:44.880 --> 02:13:46.757 example of joint embedding architecture. 02:13:46.840 --> 02:13:51.157 So, joint embedding architecture is uh uh 02:13:51.240 --> 02:13:54.837 Let me back up a little bit, right? So, if you want to do self-supervised learning, 02:13:54.920 --> 02:13:57.917 uh you can you can do it by prediction. 02:13:58.000 --> 02:14:00.917 Um so, let's say you want to train a system to predict video, right? You show 02:14:01.000 --> 02:14:05.557 it a video clip and uh and you train the system to predict the next the 02:14:05.640 --> 02:14:09.677 continuation of that video clip. Now, because you need to handle uncertainty, 02:14:09.760 --> 02:14:13.557 cuz there There of you, you know, many continuations that are plausible. 02:14:13.640 --> 02:14:15.397 You need to have You need to handle this 02:14:15.480 --> 02:14:17.917 in some way. You need to have a way for 02:14:18.000 --> 02:14:22.407 the system to be able to produce multiple predictions. 02:14:22.490 --> 02:14:25.437 >> [snorts] >> And the way The only way I know to do this 02:14:25.520 --> 02:14:29.597 is through what's called a latent variable. So, you have some sort of 02:14:29.680 --> 02:14:33.677 hidden vector of a variable that you can vary over a set or draw from a 02:14:33.760 --> 02:14:37.037 distribution. And as you vary this vector over a set, the output, the 02:14:37.120 --> 02:14:39.997 prediction varies over a set of plausible predictions. 02:14:40.080 --> 02:14:42.677 >> Mhm. >> Okay. So, [clears throat] that's called I call this a 02:14:42.760 --> 02:14:46.037 a generative latent variable model. 02:14:46.120 --> 02:14:48.477 >> Got it. >> Okay? Now, there is an alternative to 02:14:48.560 --> 02:14:50.597 this to handle uncertainty. 02:14:50.680 --> 02:14:54.077 And instead of directly predicting the 02:14:54.160 --> 02:14:57.757 the next frames of the of the of the clip, 02:14:57.840 --> 02:15:02.637 you also run those through another neural net. 02:15:02.720 --> 02:15:07.437 Uh so, that you now have two neural nets. One that looks at the the 02:15:07.520 --> 02:15:11.437 you know, the initial segment of the video clip. And another one that looks 02:15:11.520 --> 02:15:14.807 at the the continuation during training, right? 02:15:14.890 --> 02:15:19.557 >> [snorts] >> And what what you're trying to do is learn a representation 02:15:19.640 --> 02:15:23.757 of those two video clips that is maximally informative about the video 02:15:23.840 --> 02:15:28.117 clips themselves, but is such that you can predict the 02:15:28.200 --> 02:15:32.357 representation of the second video clip from the representation of the first one 02:15:32.440 --> 02:15:35.517 easily. >> Okay. >> And you can sort of formalize this in 02:15:35.600 --> 02:15:38.357 terms of maximizing mutual information and some stuff like that, but it doesn't 02:15:38.440 --> 02:15:41.517 matter. Uh what you want is 02:15:41.600 --> 02:15:43.837 uh informative representative represent 02:15:43.920 --> 02:15:48.637 you know, informative representations of the two video clips that are mutually 02:15:48.720 --> 02:15:51.997 predictable. >> Mhm. >> [clears throat] >> What that means is that there's a lot of 02:15:52.080 --> 02:15:55.397 details in the second video clip that are irrelevant. 02:15:55.480 --> 02:15:57.757 Uh you know, um 02:15:57.840 --> 02:16:00.757 I let's say video clip consists in, you 02:16:00.840 --> 02:16:03.597 know, a camera panning the scene. 02:16:03.680 --> 02:16:06.797 Uh there's going to be a piece of that room that is going to be revealed, and I 02:16:06.880 --> 02:16:10.277 can somewhat predict what the what that room is going to look like, but I may 02:16:10.360 --> 02:16:14.637 not be able to predict the details of the texture of the ground and where the 02:16:14.720 --> 02:16:16.877 tiles are ending and stuff and stuff like that, right? So, those are 02:16:16.960 --> 02:16:21.637 irrelevant details that perhaps my representation will eliminate. 02:16:21.720 --> 02:16:24.797 And so, what I need is to train this second 02:16:24.880 --> 02:16:27.157 neural net in such a way that 02:16:27.240 --> 02:16:33.117 uh whenever the the the the continuation video clip varies over all the possible 02:16:33.200 --> 02:16:35.517 continuations, >> Mhm. 02:16:35.600 --> 02:16:37.557 >> the representation doesn't change. 02:16:37.640 --> 02:16:40.437 >> Got it. So, it's the Yeah, yeah, got it. 02:16:40.520 --> 02:16:44.197 Over the space of representations, doing the same kind of thing as you're 02:16:44.280 --> 02:16:46.237 doing with similarity learning. 02:16:46.320 --> 02:16:49.957 >> Right. >> Yeah. >> So, so these are two ways to handle 02:16:50.040 --> 02:16:53.797 multimodality in a prediction, right? In the first way, you parametrize the 02:16:53.880 --> 02:16:57.717 prediction with a latent variable, but you predict pixels essentially, right? 02:16:57.800 --> 02:17:01.237 In the second one, you don't predict pixels, you predict an abstract 02:17:01.320 --> 02:17:03.877 representation of pixels, and you guarantee that this abstract 02:17:03.960 --> 02:17:06.957 representation has as much information as possible 02:17:07.040 --> 02:17:10.397 about the input, but sort of, you know, drops all the stuff that you really 02:17:10.480 --> 02:17:15.157 can't predict, essentially. Uh I used to be a big fan of the first 02:17:15.240 --> 02:17:19.037 approach, and in fact, in this paper with Ishan Mishra, this this blog post, >> Mhm. 02:17:19.120 --> 02:17:21.877 >> uh the dark matter intelligence, I was kind of advocating for this. And in the 02:17:21.960 --> 02:17:24.237 last [clears throat] year and a half, I've completely changed 02:17:24.320 --> 02:17:26.277 my mind. I'm now a big fan of the second one. 02:17:26.360 --> 02:17:30.117 >> Mhm. >> Uh and it's because of uh 02:17:30.200 --> 02:17:34.597 uh a small collection of algorithms that have been proposed over the last uh 02:17:34.680 --> 02:17:39.717 year and a half or so, 2 years, uh to do this, uh including VICReg, uh 02:17:39.800 --> 02:17:43.957 its predecessor called Barlow Twins, which I mentioned, uh 02:17:44.040 --> 02:17:47.077 a method from our friends at DeepMind called BYOL, 02:17:47.160 --> 02:17:51.637 um and and and there's a bunch of others now that kind of work similarly. So, 02:17:51.720 --> 02:17:55.157 they're they're all based on this idea of joint embedding. Some of them have an 02:17:55.240 --> 02:17:58.397 explicit criterion that is an approximation of mutual information. 02:17:58.480 --> 02:18:02.037 Some others like BYOL work, but we don't really know why. And there's been like 02:18:02.120 --> 02:18:05.117 lots of theoretical papers about why Barlow Twins works. No, it's not that 02:18:05.200 --> 02:18:08.277 because we take it out and it still works and you know, blah blah blah. I mean, so 02:18:08.360 --> 02:18:12.557 there's like a big a big debate, but um uh but the important point is that we 02:18:12.640 --> 02:18:16.317 now have a collection of non-contrastive joint embedding methods, which I think 02:18:16.400 --> 02:18:18.357 is the best thing since sliced bread. 02:18:18.440 --> 02:18:23.117 So, I'm super excited about this because I think it's our best shot 02:18:23.200 --> 02:18:26.757 for techniques that would allow us to kind of build uh 02:18:26.840 --> 02:18:30.157 predictive world models and at the same time learn how I pick up 02:18:30.240 --> 02:18:33.877 representations of the world where what matters about the world is preserved and 02:18:33.960 --> 02:18:36.317 what is irrelevant is eliminated. 02:18:36.400 --> 02:18:39.877 >> Uh by the way, the representations before and after is 02:18:39.960 --> 02:18:43.957 in the space in a sequence of images or is it for single images? 02:18:44.040 --> 02:18:46.637 >> Uh it would be either for a single image, for a sequence. It could It 02:18:46.720 --> 02:18:49.277 doesn't have to be images. This could be applied to text. This could be applied 02:18:49.360 --> 02:18:53.197 to just about any signal. I'm looking at you know, I'm looking for 02:18:53.280 --> 02:18:56.837 methods that are generally applicable, that are not specific to you know, one 02:18:56.920 --> 02:18:59.517 particular modality. You know, it could be audio or whatever. 02:18:59.600 --> 02:19:03.517 >> Got it. So, what's the story behind this paper? This This paper is what? Is Is 02:19:03.600 --> 02:19:05.437 describing one of the one such method? 02:19:05.520 --> 02:19:09.157 >> This this VicReg method. So, this is co-authored uh the first author is uh a 02:19:09.240 --> 02:19:11.677 student uh called Adrien Bard, who is a 02:19:11.760 --> 02:19:14.597 resident PhD student at Fair Paris. 02:19:14.680 --> 02:19:19.637 He's co-advised by me and Jean Ponce, um who's a professor at École Normale 02:19:19.720 --> 02:19:23.037 Supérieure, also uh research director at Inria. 02:19:23.120 --> 02:19:27.317 Um so, this is wonderful program in France where PhD students can basically 02:19:27.400 --> 02:19:29.117 do their PhD in industry, and that's 02:19:29.200 --> 02:19:31.477 kind of what what's happening here. 02:19:31.560 --> 02:19:35.717 Um and this paper is a follow-up on uh 02:19:35.800 --> 02:19:37.797 the the this Barlow Twins paper by yeah, 02:19:37.880 --> 02:19:41.437 my former postdoc now Stefano De uh with 02:19:41.520 --> 02:19:43.917 Li Jing and and Yurish Montal and bunch 02:19:44.000 --> 02:19:46.557 of other people from from from Fair. 02:19:46.640 --> 02:19:49.157 And one of the main criticism from 02:19:49.240 --> 02:19:50.997 reviewers is that VicReg is not 02:19:51.080 --> 02:19:54.877 different enough from Barlow Twins. But, 02:19:54.960 --> 02:19:58.797 you know, my impression is that it's, you know, 02:19:58.880 --> 02:20:02.797 bottom twins with a few bugs fixed essentially. And 02:20:02.880 --> 02:20:05.037 in the end, this is what people will use. 02:20:05.120 --> 02:20:08.957 >> Right. So, >> But, you know, I'm used to stuff >> Yeah. 02:20:09.040 --> 02:20:10.997 >> that I submit being rejected for a while. 02:20:11.080 --> 02:20:14.157 >> might be rejected and actually accepted and well-cited cuz people use it. 02:20:14.240 --> 02:20:16.317 >> Well, it's already decided like a bunch of times. 02:20:16.400 --> 02:20:18.517 >> So, I mean, the the the question is then 02:20:18.600 --> 02:20:22.637 to the deeper question about peer review and conferences. I mean, computer 02:20:22.720 --> 02:20:26.877 science as a field is kind of unique that the conference is highly prized. 02:20:26.960 --> 02:20:29.997 That's one. >> Right. >> And it's interesting because the peer 02:20:30.080 --> 02:20:33.517 review process there is similar, I suppose, to the journals, but it's 02:20:33.600 --> 02:20:38.437 accelerated significantly. Well, not significantly, but it goes fast. 02:20:38.520 --> 02:20:43.437 It's a nice way to get stuff out quickly, to peer review quickly, go to present it 02:20:43.520 --> 02:20:46.997 quickly to the community. So, not quickly, but quicker. 02:20:47.080 --> 02:20:48.957 >> Yeah. >> But, nevertheless, it has many of the 02:20:49.040 --> 02:20:53.357 same flaws of peer review cuz it's a limited number of people look at it. 02:20:53.440 --> 02:20:57.397 There's bias in the following like if you if you want to do new ideas, 02:20:57.480 --> 02:21:01.637 you're going to get pushback. Um there's self-interested people that 02:21:01.720 --> 02:21:05.517 kind of can infer who submitted it 02:21:05.600 --> 02:21:09.637 and kind of you know, be cranky about it, all that kind of stuff. 02:21:09.720 --> 02:21:12.397 >> Yeah, I mean, there's a lot of, you know, social [snorts] phenomena there. 02:21:12.480 --> 02:21:16.397 Um there's one social phenomenon which is that because the field has been 02:21:16.480 --> 02:21:20.437 growing exponentially, the vast majority of people in the field 02:21:20.520 --> 02:21:22.477 are extremely junior. >> Yeah. 02:21:22.560 --> 02:21:26.757 >> So, as a consequence, and that's just a consequence of the field growing, right? 02:21:26.840 --> 02:21:31.437 So, as the number of as the size of the field kind of start saturating, you you 02:21:31.520 --> 02:21:35.877 have less of that problem of reviewers being very uh 02:21:35.960 --> 02:21:42.077 inexperienced. A consequence of this is that, you know, young reviewers 02:21:42.160 --> 02:21:46.517 I mean, there's a phenomenon which is that reviewers try to make their life easy. 02:21:46.600 --> 02:21:49.397 And to make their life easy when reviewing a paper 02:21:49.480 --> 02:21:53.157 is very simple. You just have to find a flaw in the paper, right? So, basically 02:21:53.240 --> 02:21:57.917 they see that task as finding flaws in papers and most papers have flaws, 02:21:58.000 --> 02:21:59.557 even the good ones. 02:21:59.640 --> 02:22:02.637 So it's it's easy to you know 02:22:02.720 --> 02:22:06.717 to do that. You you your job is easier as a reviewer 02:22:06.800 --> 02:22:10.877 if you just focus on this. But what's important is 02:22:10.960 --> 02:22:16.077 like is there a new idea in that paper that is likely to influence 02:22:16.160 --> 02:22:19.317 It doesn't matter if the experiments are not that great, if the protocol is, you 02:22:19.400 --> 02:22:23.757 know, so-so, you know, things like that. As long as 02:22:23.840 --> 02:22:28.677 there is a worthy idea in it that will influence the way people think 02:22:28.760 --> 02:22:32.517 about the problem >> [snorts] >> even if they make it better, you know, 02:22:32.600 --> 02:22:36.437 eventually. I think that's that's really what what makes a paper 02:22:36.520 --> 02:22:37.437 useful. 02:22:37.520 --> 02:22:40.157 And so this combination of 02:22:40.240 --> 02:22:43.677 social phenomena creates 02:22:43.760 --> 02:22:47.477 a disease that has plagued, you know, other fields in the past like speech 02:22:47.560 --> 02:22:51.877 recognition where basically, you know, people chase these numbers on uh 02:22:51.960 --> 02:22:55.917 on benchmarks and and it's much easier to get a paper 02:22:56.000 --> 02:23:00.117 accepted if it brings an incremental improvement on a 02:23:00.200 --> 02:23:04.317 sort of mainstream well-accepted uh method 02:23:04.400 --> 02:23:07.997 or problem. And those are to me boring papers. I 02:23:08.080 --> 02:23:12.957 mean they're not useless, right? Because industry, you know, strives on on those 02:23:13.040 --> 02:23:15.157 kind of progress. But they're not the one that I'm 02:23:15.240 --> 02:23:18.397 interested in in terms of like new concepts and new ideas. So 02:23:18.480 --> 02:23:23.077 papers that are really trying to strike kind of new advances 02:23:23.160 --> 02:23:26.117 generally don't make it. Now thankfully we have archive. 02:23:26.200 --> 02:23:30.757 >> Archive, exactly. And then there's open review type of situations where you 02:23:30.840 --> 02:23:34.917 And then I mean Twitter is a kind of open review. I I'm a huge believer that 02:23:35.000 --> 02:23:38.277 review should be done by thousands of people, not two people. 02:23:38.360 --> 02:23:41.997 >> And so archive Like do you see a future where a lot of 02:23:42.080 --> 02:23:45.797 really strong papers, it's already the present, but a growing future where 02:23:45.880 --> 02:23:49.877 it'll just be archive and 02:23:49.960 --> 02:23:52.437 you're presenting an ongoing continuous 02:23:52.520 --> 02:23:55.717 conference called Twitter / the internet 02:23:55.800 --> 02:23:59.557 / archive sanity. Andre just released a new version. 02:23:59.640 --> 02:24:02.197 So just not you know not being so 02:24:02.280 --> 02:24:05.437 elitist about this particular gating. 02:24:05.520 --> 02:24:06.837 >> not a question of being elitist or not. 02:24:06.920 --> 02:24:11.157 It's a question of being basically 02:24:11.240 --> 02:24:15.557 recommendation and sort of approvals for people who don't see themselves as 02:24:15.640 --> 02:24:18.997 having the ability to do so by themselves, right? And so it saves time, 02:24:19.080 --> 02:24:20.837 right? If you rely on other people's 02:24:20.920 --> 02:24:23.997 opinion and you trust those people or 02:24:24.080 --> 02:24:29.637 those groups to evaluate a paper for you. 02:24:29.720 --> 02:24:34.477 Um that saves you time cuz you know you don't have to like scrutinize 02:24:34.560 --> 02:24:37.357 the paper as much, you know, it is brought to your attention. I mean it's 02:24:37.440 --> 02:24:40.637 the whole idea of sort of you know collective recommendation system, right? 02:24:40.720 --> 02:24:43.957 So I actually thought about this a lot um 02:24:44.040 --> 02:24:46.157 you know, about 10 15 years ago. Of 02:24:46.240 --> 02:24:48.597 course the what discussions at 02:24:48.680 --> 02:24:53.597 NIPS and you know and we're about to create ICLR with Yoshua Bengio and so I 02:24:53.680 --> 02:24:58.157 wrote a document um kind of describing a reviewing system 02:24:58.240 --> 02:25:01.837 which basically was you know, you post your paper on some repository, let's say 02:25:01.920 --> 02:25:06.997 archive or now could be open review. Um and then you can form 02:25:07.080 --> 02:25:11.037 a reviewing entity which is equivalent to a reviewing board, you know, of a 02:25:11.120 --> 02:25:15.837 journal or um program committee of a conference. Um 02:25:15.920 --> 02:25:19.917 you have to list the members and then um that 02:25:20.000 --> 02:25:24.437 group reviewing entity can choose to review a particular paper 02:25:24.520 --> 02:25:28.277 spontaneously or not. There is no exclusive relationship anymore between a 02:25:28.360 --> 02:25:31.037 paper and a venue or reviewing entity. 02:25:31.120 --> 02:25:33.597 Any reviewing entity can review any paper. 02:25:33.680 --> 02:25:38.517 Um or may choose not to. Um and then you know, given evaluation 02:25:38.600 --> 02:25:42.437 it's not published not published, it's just an evaluation and a comment which 02:25:42.520 --> 02:25:45.597 would be public, signed by the reviewing entity. 02:25:45.680 --> 02:25:47.757 And if it's signed by the reviewing entity, 02:25:47.840 --> 02:25:50.517 you know it's one of the members of reviewing entity. So, if the reviewing 02:25:50.600 --> 02:25:54.877 entity is, you know, Lex Fridman's, you know, preferred 02:25:54.960 --> 02:25:57.517 papers, right? You know it's Lex Fridman writing the review. 02:25:57.600 --> 02:26:02.077 >> Yes. What So, for me one that's a beautiful 02:26:02.160 --> 02:26:06.437 system, I think, but what in addition to that it feels like there should be a 02:26:06.520 --> 02:26:08.877 reputation system for the reviewers. 02:26:08.960 --> 02:26:10.957 >> Absolutely. For the reviewing entities. 02:26:11.040 --> 02:26:12.397 Not the reviewers individually. 02:26:12.480 --> 02:26:13.677 >> The reviewing entities, sure. 02:26:13.760 --> 02:26:17.157 >> But even that within that the reviewers too, because uh 02:26:17.240 --> 02:26:19.397 I There's another thing here. It's not 02:26:19.480 --> 02:26:22.197 just the reputation, it's an incentive 02:26:22.280 --> 02:26:24.677 for an individual person to do great. 02:26:24.760 --> 02:26:29.117 Right now in in the academic setting, the incentive is kind of uh 02:26:29.200 --> 02:26:32.677 internal, just wanting to do a good job, but honestly that's not a strong enough 02:26:32.760 --> 02:26:36.237 incentive to do a really good job in reading a paper, in finding the 02:26:36.320 --> 02:26:39.997 beautiful amidst the mistakes and the flaws and all that kind of stuff. Like 02:26:40.080 --> 02:26:44.397 if you're the person that first discovered a powerful paper, 02:26:44.480 --> 02:26:49.517 and you get to be proud of that discovery, then that gives a huge incentive to you. 02:26:49.600 --> 02:26:50.837 >> That's That's a big part of my proposal, 02:26:50.920 --> 02:26:52.877 actually. You know, I describe that as 02:26:52.960 --> 02:26:58.117 you know, if if your evaluation of papers is predictive of future success, 02:26:58.200 --> 02:26:59.437 >> Yes. Yes. That's a good one. 02:26:59.520 --> 02:27:03.477 >> then your reputation should go up as a reviewing entity. 02:27:03.560 --> 02:27:06.957 Um So, yeah, exactly. I mean, that that um 02:27:07.040 --> 02:27:11.397 I even had a master's student who was um a master's student in uh library science 02:27:11.480 --> 02:27:14.797 and computer science. Actually, kind of worked out exactly how should that 02:27:14.880 --> 02:27:16.997 should work with formulas and everything, but 02:27:17.080 --> 02:27:20.517 >> So, in terms of implementation, do you think that's something that's doable? 02:27:20.600 --> 02:27:23.477 >> I mean, I've been sort of, you know, talking about this to sort of various 02:27:23.560 --> 02:27:27.757 people like, you know, Andrew McCallum who started OpenReview. 02:27:27.840 --> 02:27:29.637 >> [snorts] >> And the reason why we picked OpenReview 02:27:29.720 --> 02:27:31.517 for ICLR initially, even though it was 02:27:31.600 --> 02:27:33.957 very early for them, um is because my 02:27:34.040 --> 02:27:38.557 hope was that ICLR was it was eventually going to kind of 02:27:38.640 --> 02:27:40.917 inaugurate this type of uh system. So, I 02:27:41.000 --> 02:27:43.797 clear kept the idea of open reviews. 02:27:43.880 --> 02:27:47.517 Uh, so whether reviews are you know, published uh with the paper, which I 02:27:47.600 --> 02:27:51.677 think is very useful. But, in many ways has kind of reverted 02:27:51.760 --> 02:27:56.157 to kind of more of a conventional type conferences for everything else. And 02:27:56.240 --> 02:28:00.677 that I mean, I I don't run I clear. I'm just the 02:28:00.760 --> 02:28:06.357 president of the uh of the foundation, but um you know, people who run it should make 02:28:06.440 --> 02:28:09.597 decisions about how to run it and I'm not going to tell them um because 02:28:09.680 --> 02:28:13.957 they're volunteers and I'm really thankful that they do that. So, but I'm 02:28:14.040 --> 02:28:15.997 saddened by the fact that uh we're not 02:28:16.080 --> 02:28:18.877 being um innovative enough. 02:28:18.960 --> 02:28:23.197 >> Yeah, me too. I hope that changes. Uh, yeah. Cuz the communication [snorts] 02:28:23.280 --> 02:28:27.237 science broadly, but communication computer science ideas 02:28:27.320 --> 02:28:30.317 is is is how you make those ideas have impact, I think. 02:28:30.400 --> 02:28:31.677 >> Yeah, and I think you know, a lot of 02:28:31.760 --> 02:28:35.077 this is um because people have in their 02:28:35.160 --> 02:28:38.877 mind kind of an objective which is you 02:28:38.960 --> 02:28:43.677 know, fairness for authors and the ability to count points 02:28:43.760 --> 02:28:46.037 basically and and give credits accurately. 02:28:46.120 --> 02:28:50.357 >> Mhm. >> But, that comes at the expense of the progress of science. 02:28:50.440 --> 02:28:53.917 >> Yeah. >> So, to some extent we're slowing down the progress of science. 02:28:54.000 --> 02:28:56.197 >> And are we actually achieving fairness? 02:28:56.280 --> 02:29:00.077 >> And we're not achieving fairness, you know, we have biases, you know, we're 02:29:00.160 --> 02:29:05.797 doing you know, a double blind review, but uh you know, the the the biases are are 02:29:05.880 --> 02:29:08.477 still there. They're different different kinds of biases. 02:29:08.560 --> 02:29:12.677 >> You write that the phenomenon of emergence, collective behavior exhibited 02:29:12.760 --> 02:29:17.197 by large collection of simple elements in interaction is one of the things that 02:29:17.280 --> 02:29:19.197 got you into neural nets in the first 02:29:19.280 --> 02:29:24.517 place. I love cellular automata. I love simple interacting elements and uh the 02:29:24.600 --> 02:29:27.197 things that emerge from them. Do you think 02:29:27.280 --> 02:29:31.677 we understand how complex systems can emerge from such simple components that 02:29:31.760 --> 02:29:34.197 interact simply? >> No, we don't. 02:29:34.280 --> 02:29:36.997 It's a big mystery. It also is a mystery for physicists. It's a mystery for 02:29:37.080 --> 02:29:42.037 biologists. Um you know, how is it that uh >> [snorts] 02:29:42.120 --> 02:29:45.757 >> the uh universe around us seems to be increasing in complexity and not 02:29:45.840 --> 02:29:49.197 decreasing? I mean, that that that is a kind of 02:29:49.280 --> 02:29:52.717 curious property of uh physics that despite the second law of 02:29:52.800 --> 02:29:54.277 thermodynamics, 02:29:54.360 --> 02:29:57.037 we seem to be, you know, evolution and 02:29:57.120 --> 02:30:00.437 and learning and uh etc. seems to be 02:30:00.520 --> 02:30:04.757 kind of at least locally uh to increase uh 02:30:04.840 --> 02:30:07.917 complexity and not decrease it. So, perhaps the ultimate purpose of the 02:30:08.000 --> 02:30:10.957 universe is uh to just get more complex. 02:30:11.040 --> 02:30:16.637 >> Have these uh I mean, uh small pockets of beautiful complexity. 02:30:16.720 --> 02:30:19.317 Uh does that the cellular automata, these 02:30:19.400 --> 02:30:21.677 kinds of emergence and complex systems 02:30:21.760 --> 02:30:26.997 give you some intuition or guide your understanding of machine learning 02:30:27.080 --> 02:30:28.597 systems and neural networks and so on? 02:30:28.680 --> 02:30:31.397 Or are these for you right now disparate concepts? 02:30:31.480 --> 02:30:36.477 >> Well, you got you got me into it. You know, I I discovered the existence of uh 02:30:36.560 --> 02:30:40.357 the perceptron when I was a college student. Um 02:30:40.440 --> 02:30:44.797 you know, by reading a book on it was a debate between Chomsky and Piaget and 02:30:44.880 --> 02:30:48.637 Seymour Papert from MIT was kind of singing the praise of the perceptron in 02:30:48.720 --> 02:30:52.037 that in the book and I I first time I heard about Turing machine. Right? So, I 02:30:52.120 --> 02:30:55.757 started digging the literature and I found those paper those books which were 02:30:55.840 --> 02:31:00.477 basically transcri- transcription of, you know, workshops or conferences 02:31:00.560 --> 02:31:03.877 uh from the '50s and '60s about self-organizing systems. 02:31:03.960 --> 02:31:07.277 Mhm. So, there were there was a series of conferences on self-organizing 02:31:07.360 --> 02:31:11.917 systems and the these books on this um some of them are you can actually get 02:31:12.000 --> 02:31:16.077 them at the internet archive, you know, the digital version. 02:31:16.160 --> 02:31:21.077 Uh and there are like uh fascinating articles in there by this guy whose name 02:31:21.160 --> 02:31:24.677 has been largely forgotten, Heinz von Foerster. 02:31:24.760 --> 02:31:29.357 So, it's a German physicist who immigrated to the US and uh worked on 02:31:29.440 --> 02:31:33.277 self-organizing systems in the in the 50s 02:31:33.360 --> 02:31:36.917 and in the 60s he created at University of Illinois Urbana-Champaign he created 02:31:37.000 --> 02:31:41.877 the Biological Computer Laboratory BCL which was you know all about neural 02:31:41.960 --> 02:31:44.757 nets. Unfortunately that was kind of towards 02:31:44.840 --> 02:31:49.117 the end of the popularity of neural nets so that that lab never kind of strive 02:31:49.200 --> 02:31:52.557 very much but but he wrote a bunch of papers about self-organization and about 02:31:52.640 --> 02:31:55.477 the the mystery of self-organization. An 02:31:55.560 --> 02:32:00.077 example he has is you take imagine you are in space there's no gravity you have 02:32:00.160 --> 02:32:04.037 a big box with magnets in it okay 02:32:04.120 --> 02:32:07.477 you know kind of rectangular magnets with north pole on one end 02:32:07.560 --> 02:32:11.197 south pole on the other end. You shake the box gently and the magnets will kind 02:32:11.280 --> 02:32:17.117 of stick to themselves and probably form a complex structure you know spontaneously 02:32:17.200 --> 02:32:19.597 you know that could be an example of self-organization but you know you have 02:32:19.680 --> 02:32:23.157 lots of examples neural nets are an example of self-organization too 02:32:23.240 --> 02:32:27.597 you know in many respect. And it's it's a bit of a mystery 02:32:27.680 --> 02:32:31.117 you know how like what what is possible with this 02:32:31.200 --> 02:32:33.877 you know pattern formation in physical systems 02:32:33.960 --> 02:32:37.997 in chaotic system and things like that you know you know the emergence of life 02:32:38.080 --> 02:32:41.437 you know things like that so you know how does does that happen? 02:32:41.520 --> 02:32:44.477 It's a it's big puzzle for for physicists as well. 02:32:44.560 --> 02:32:47.117 >> It feels like understanding this the the 02:32:47.200 --> 02:32:52.317 mathematics of emergence in some constrained situations might help us 02:32:52.400 --> 02:32:57.837 create intelligence. Like help us add a little spice to the systems 02:32:57.920 --> 02:33:01.437 because you seem to be able to 02:33:01.520 --> 02:33:04.077 in complex systems with emergence to be 02:33:04.160 --> 02:33:06.517 able to get a lot from little. 02:33:06.600 --> 02:33:11.637 And so that seems like a shortcut to get big leaps in performance. 02:33:11.720 --> 02:33:14.317 But um >> But there's there's a missing concept 02:33:14.400 --> 02:33:16.877 theoretical concept that we are we don't have. 02:33:16.960 --> 02:33:19.437 >> Yeah. >> And it's it's something also I've been 02:33:19.520 --> 02:33:22.677 fascinated by since my undergrad days 02:33:22.760 --> 02:33:25.557 And it's how you measure complexity. 02:33:25.640 --> 02:33:29.717 Right. So, we don't actually have good ways of measuring or at least we don't 02:33:29.800 --> 02:33:33.197 have good ways of interpreting the measures that we we have at our 02:33:33.280 --> 02:33:36.477 disposal. Like, how do you measure the complexity of something, right? 02:33:36.560 --> 02:33:38.317 So, there's all those things, you know, like, you know, 02:33:38.400 --> 02:33:42.317 Kolmogorov-Chaitin-Solomonoff complexity of, you know, the length of the shortest 02:33:42.400 --> 02:33:46.357 program that would generate a bit string can be thought of as the complexity of 02:33:46.440 --> 02:33:47.837 that bit string, right? 02:33:47.920 --> 02:33:49.957 Um I've been fascinated by that concept. 02:33:50.040 --> 02:33:53.437 But, the problem with that is that 02:33:53.520 --> 02:33:57.837 that complexity is defined up to a constant, which can be very large. 02:33:57.920 --> 02:34:00.957 Right? Uh there's there are similar concepts that are derived from, you 02:34:01.040 --> 02:34:05.317 know, you know, Bayesian probability theory, 02:34:05.400 --> 02:34:09.597 where, you know, the complexity of something is uh negative log of its 02:34:09.680 --> 02:34:12.717 probability, essentially, right? And you have a complete equivalence between the 02:34:12.800 --> 02:34:15.157 two things. And there you would think, you know, the 02:34:15.240 --> 02:34:18.677 probability is something that's well defined mathematically, which means 02:34:18.760 --> 02:34:22.517 complexity is well defined. But, it's not true. You need to have a model of uh 02:34:22.600 --> 02:34:25.797 of of the distribution. Uh you may need to have a prior if 02:34:25.880 --> 02:34:29.317 you're doing Bayesian inference. And the prior plays the same role as the choice 02:34:29.400 --> 02:34:32.477 of the computer with which you measure your Kolmogorov complexity. 02:34:32.560 --> 02:34:37.197 And so, every measure of complexity we have has some arbitrariness in it. 02:34:37.280 --> 02:34:40.757 Um you know, an additive constant which 02:34:40.840 --> 02:34:45.717 is can be arbitrarily large. And so, you know, how can we come up with a good 02:34:45.800 --> 02:34:48.317 theory of how things become more complex if we don't have a good measure of 02:34:48.400 --> 02:34:52.317 complexity? >> Yeah. Which we need for this one way 02:34:52.400 --> 02:34:56.397 that people study this in space biology, the people that study the origin of life 02:34:56.480 --> 02:34:59.757 or try to recreate the life in in laboratory. 02:34:59.840 --> 02:35:03.957 And the the more interesting one is the alien one is when we go to other planets, 02:35:04.040 --> 02:35:06.637 how do we recognize that it's life? 02:35:06.720 --> 02:35:10.997 Cuz you know, complexity we associate complexity maybe some level of mobility 02:35:11.080 --> 02:35:14.277 with life. You know, we have to be able to like 02:35:14.360 --> 02:35:16.677 have concrete 02:35:16.760 --> 02:35:21.877 algorithms for like um measuring the level of complexity we 02:35:21.960 --> 02:35:25.277 see to in order to know the difference between life and non-life. 02:35:25.360 --> 02:35:26.717 >> And the problem is that complexity is in 02:35:26.800 --> 02:35:29.077 the eye of the beholder. So, let me give 02:35:29.160 --> 02:35:31.357 you an example. If I um 02:35:31.440 --> 02:35:34.997 if I give you uh an image of the MNIST 02:35:35.080 --> 02:35:39.877 digits, right? And I I flip through MNIST digits. There is some obviously some 02:35:39.960 --> 02:35:43.237 structure to it because local structure, you know, neighboring 02:35:43.320 --> 02:35:45.797 pixels are correlated uh 02:35:45.880 --> 02:35:48.037 across the entire data set. 02:35:48.120 --> 02:35:53.837 Now, imagine that um I I apply a random permutation to all the pixels. 02:35:53.920 --> 02:35:56.477 >> Mhm. >> A fixed random permutation. 02:35:56.560 --> 02:36:01.397 Now, I show you those images. They will look you know, really disorganized to you. 02:36:01.480 --> 02:36:04.717 More complex. In fact, they're not more complex in 02:36:04.800 --> 02:36:08.397 absolute terms. They're exactly the same as originally, right? And if you knew 02:36:08.480 --> 02:36:11.957 what the permutation was, you know, you could undo the permutation. 02:36:12.040 --> 02:36:16.597 Now, imagine I give you special glasses that undo that permutation. 02:36:16.680 --> 02:36:19.477 Now, all of a sudden what looked complicated becomes simple. 02:36:19.560 --> 02:36:22.637 >> Right. >> So, if you have two if you have, you 02:36:22.720 --> 02:36:26.597 know, humans on one end and then another race of aliens that sees the universe 02:36:26.680 --> 02:36:29.397 with permutation glasses, >> Yeah, with the permutation glasses. 02:36:29.480 --> 02:36:32.357 >> Okay. >> [laughter] >> What we perceive as simple to them is 02:36:32.440 --> 02:36:34.237 hardly complicated. It's probably heat. 02:36:34.320 --> 02:36:39.477 >> Yeah. Heat, yeah. >> Okay. And what they perceive as simple to us is is random 02:36:39.560 --> 02:36:42.157 uh fluctuation. It's heat. >> Yeah. 02:36:42.240 --> 02:36:44.717 >> Uh so >> in the eye of the beholder. 02:36:44.800 --> 02:36:46.877 >> Depends what kind of glasses you're wearing. 02:36:46.960 --> 02:36:50.317 >> Depends what kind of algorithm you're running in your perception system. 02:36:50.400 --> 02:36:54.117 >> So, I don't think we'll have a theory of intelligence, self-organization, 02:36:54.200 --> 02:36:57.077 evolution, things like this, until we 02:36:57.160 --> 02:37:01.397 have a good handle on a notion of complexity, which we know is in the eye 02:37:01.480 --> 02:37:05.837 of the eye of the beholder. >> [sighs] >> Yeah, it's sad to think that we might 02:37:05.920 --> 02:37:10.357 not be able to detect or interact with alien species because we're wearing 02:37:10.440 --> 02:37:13.517 different glasses. >> Um because their notion of locality may 02:37:13.600 --> 02:37:14.837 be different from ours. >> Yeah. 02:37:14.920 --> 02:37:17.757 >> This actually connects with uh fascinating questions in physics at the 02:37:17.840 --> 02:37:21.597 moment, like modern physics. Uh quantum physics, like you know, 02:37:21.680 --> 02:37:25.037 questions about like you know, can we recover the information that's lost in a 02:37:25.120 --> 02:37:26.477 black hole and things like that, right? 02:37:26.560 --> 02:37:30.957 And uh and that relies on notions of complexity. Um 02:37:31.040 --> 02:37:33.597 Yeah, which you know, I find I find it fascinating. 02:37:33.680 --> 02:37:35.397 >> Can you describe your personal quest to 02:37:35.480 --> 02:37:39.237 build an expressive electronic wind 02:37:39.320 --> 02:37:44.077 instrument, EWI? What is it? What does it take 02:37:44.160 --> 02:37:47.557 to uh to build it? >> Well, I'm a tinkerer. I like building 02:37:47.640 --> 02:37:51.517 things. Uh I like building things with combinations of electronics and, you 02:37:51.600 --> 02:37:55.597 know, mechanical stuff. Um you know, I have a bunch of different 02:37:55.680 --> 02:37:59.517 hobbies, but um you know, probably my first one was 02:37:59.600 --> 02:38:02.397 little was building model airplanes and stuff like that, and I still do that to 02:38:02.480 --> 02:38:05.157 some extent. Uh but also electronics. I taught myself 02:38:05.240 --> 02:38:09.517 electronics before I studied it. Um and the reason I taught myself 02:38:09.600 --> 02:38:13.477 electronics is because of music. Um my cousin 02:38:13.560 --> 02:38:17.397 was an inspiring electronic musician and he had an analog synthesizer, and I was, 02:38:17.480 --> 02:38:19.277 you know, basically modifying it for him. 02:38:19.360 --> 02:38:21.237 >> Mhm. [snorts] >> Uh and building sequencers and stuff 02:38:21.320 --> 02:38:24.557 like that, right, for him. I was I was in high school when I was doing this. 02:38:24.640 --> 02:38:28.557 >> And the interest in like progressive rock, like '80s, like what's what's the 02:38:28.640 --> 02:38:31.397 greatest band of all time according to Yann LeCun? 02:38:31.480 --> 02:38:35.357 >> Um there's too there's too many of them, but you know, it's a combination of 02:38:35.440 --> 02:38:38.757 uh uh you [clears throat] know, Mahavishnu 02:38:38.840 --> 02:38:44.437 Orchestra, uh Weather Report, Yes, Genesis, uh you 02:38:44.520 --> 02:38:47.877 know, >> Genesis. >> pre-Peter Gabriel, >> Uh-huh. 02:38:47.960 --> 02:38:51.037 >> um >> [laughter] >> Gentle Giant, you know, things like that. 02:38:51.120 --> 02:38:55.197 >> Great. Okay, so this this love of electronics and this love of music 02:38:55.280 --> 02:38:59.277 combine together. >> Right. So, I was actually trying to play uh 02:38:59.360 --> 02:39:03.437 uh Baroque and Renaissance uh music, and I I played in a uh 02:39:03.520 --> 02:39:07.677 orchestra when I was in high school and uh first year of college. 02:39:07.760 --> 02:39:11.517 And I played the recorder, crumhorn, uh uh a little bit of oboe, you know, 02:39:11.600 --> 02:39:14.997 things like that. So, I'm a I'm a wind instrument player. But, I always wanted 02:39:15.080 --> 02:39:18.677 to play improvised music, even though I don't know anything about it. Uh and the 02:39:18.760 --> 02:39:23.197 only way I figured, you know, short of like learning to play the saxophone was 02:39:23.280 --> 02:39:27.277 to play electronic wind instrument. So, they behave you know, the fingering is 02:39:27.360 --> 02:39:31.157 similar to a saxophone, but you know, you have a wide variety of sound because 02:39:31.240 --> 02:39:34.997 you control the synthesizer with it. So, I I had a a bunch of those, you know, 02:39:35.080 --> 02:39:37.077 going back to the late '80s >> [snorts] 02:39:37.160 --> 02:39:41.157 >> uh from either Yamaha or Akai. They 02:39:41.240 --> 02:39:44.757 They're both kind of the main manufacturers of those. So, they were 02:39:44.840 --> 02:39:47.357 classically, you know, going back several decades. 02:39:47.440 --> 02:39:50.277 Uh but, I've never been completely satisfied with them because of lack of 02:39:50.360 --> 02:39:54.357 expressivity. And you know, those things, you know, 02:39:54.440 --> 02:39:56.757 are somewhat expressive. I mean, they measure the breath pressure, they 02:39:56.840 --> 02:40:00.437 measure the lip pressure, and you know, uh 02:40:00.520 --> 02:40:04.317 you have like various parameters you can you can vary with with fingers, but they 02:40:04.400 --> 02:40:08.997 they they're not really as expressive as a acoustic instrument, right? You 02:40:09.080 --> 02:40:12.237 you hear John Coltrane play two notes and you hear you know it's John 02:40:12.320 --> 02:40:14.157 Coltrane, you know it's got a unique sound. 02:40:14.240 --> 02:40:18.117 >> Mhm. >> Uh or or Miles Davis, right? You can hear it's Miles Davis 02:40:18.200 --> 02:40:22.357 uh playing the trumpet because the the sound 02:40:22.440 --> 02:40:27.077 reflects their you know, physiognomy, basically, the 02:40:27.160 --> 02:40:32.077 shape of the vocal tract um kind of shapes the the sound. So, how 02:40:32.160 --> 02:40:35.917 how do you do this with uh electronic wind instrument? And it was um many 02:40:36.000 --> 02:40:39.837 years ago I met a guy called David Wessel. He he was a professor at 02:40:39.920 --> 02:40:44.077 Berkeley and created the the center for like, you know, music 02:40:44.160 --> 02:40:46.917 technology there. And he was interested in that question. 02:40:47.000 --> 02:40:49.277 >> Mhm. >> [clears throat] >> And so, I kept kind of thinking about 02:40:49.360 --> 02:40:53.717 this for many years and finally, because of COVID, you know, I was at home. I was 02:40:53.800 --> 02:40:58.597 in my workshop. My workshop serves also as my kind of zoom uh room and and and 02:40:58.680 --> 02:41:00.677 home office. And uh >> This is in New Jersey? 02:41:00.760 --> 02:41:03.917 >> In New Jersey. And um I started uh 02:41:04.000 --> 02:41:07.757 really being serious about you know building my own EWI instrument. 02:41:07.840 --> 02:41:09.557 >> What else is going on in that New Jersey 02:41:09.640 --> 02:41:11.397 workshop? Is there some Is there some 02:41:11.480 --> 02:41:16.597 crazy stuff you built that just or or like left on the workshop floor? Left 02:41:16.680 --> 02:41:19.477 behind. >> Uh a lot of crazy stuff is uh you know 02:41:19.560 --> 02:41:21.397 electronics we built with 02:41:21.480 --> 02:41:23.997 microcontrollers of various kinds and 02:41:24.080 --> 02:41:28.137 you know weird flying contraptions. 02:41:28.220 --> 02:41:30.637 >> [laughter] >> So you still love flying? 02:41:30.720 --> 02:41:35.397 >> It's a family disease. My My dad got me into it when I was a kid. 02:41:35.480 --> 02:41:39.637 And he was building model airplanes when he was a kid and 02:41:39.720 --> 02:41:43.717 and he was a mechanical engineer. He taught himself electronics also so he he 02:41:43.800 --> 02:41:46.917 built his early radio control systems in the 02:41:47.000 --> 02:41:51.117 late '60s, early '70s. Um and so that that that's what got me 02:41:51.200 --> 02:41:54.517 into I mean he got me into kind of you know engineering and science and 02:41:54.600 --> 02:41:56.597 technology. >> Do you also have an interest and 02:41:56.680 --> 02:42:00.637 appreciation of flight in other forms like with drones, quadrocopters or do 02:42:00.720 --> 02:42:04.277 you Is it Is it Is it model airplane the the thing that 02:42:04.360 --> 02:42:08.357 >> I you know before drones were you know kind of 02:42:08.440 --> 02:42:10.037 consumer products 02:42:10.120 --> 02:42:12.517 um you know I built my own you know with 02:42:12.600 --> 02:42:17.357 also building a microcontroller with gyroscopes and accelerometers for 02:42:17.440 --> 02:42:20.557 stabilization writing the firmware for it you know and then when it became kind 02:42:20.640 --> 02:42:23.037 of a standard thing you could buy it was boring you know I stopped doing it it 02:42:23.120 --> 02:42:26.597 was not fun anymore. Um >> Yeah. 02:42:26.680 --> 02:42:28.157 You were doing it before it was cool. 02:42:28.240 --> 02:42:31.477 >> Yeah. >> What advice would you give to a young person 02:42:31.560 --> 02:42:34.197 today in high school in college that 02:42:34.280 --> 02:42:37.077 dreams of doing something big like Yann 02:42:37.160 --> 02:42:42.317 LeCun. Like let's talk in the space of intelligence. Dreams of having a chance 02:42:42.400 --> 02:42:45.917 to solve some fundamental problem in space of intelligence 02:42:46.000 --> 02:42:50.637 both for their career and just in life being somebody who was a part of 02:42:50.720 --> 02:42:55.637 creating something special. >> So try to get 02:42:55.720 --> 02:43:00.797 interested by big questions. Things like, you know, what is intelligence? Uh 02:43:00.880 --> 02:43:05.397 what is the universe made of? What's life all about? Things like that. 02:43:05.480 --> 02:43:09.557 Um Like even like crazy big questions like 02:43:09.640 --> 02:43:13.277 um what's time? Like nobody knows what time is. 02:43:13.360 --> 02:43:18.077 Um And uh and then 02:43:18.160 --> 02:43:23.357 learn basic things like basic methods either from math from physics or from 02:43:23.440 --> 02:43:27.157 engineering. Uh things that have a long shelf life. 02:43:27.240 --> 02:43:30.597 Um like if you have a choice between like, you know, 02:43:30.680 --> 02:43:34.517 learning uh you know, mobile programming on iPhone 02:43:34.600 --> 02:43:38.360 or quantum mechanics, take quantum mechanics. 02:43:38.520 --> 02:43:40.357 Um because you're going to learn things 02:43:40.440 --> 02:43:45.477 that you have no idea exist. It and it you may not you you you may never be a 02:43:45.560 --> 02:43:48.677 quantum physicist but you will learn about path integrals 02:43:48.760 --> 02:43:51.837 and path integrals are used um everywhere. It's the same formula that 02:43:51.920 --> 02:43:55.277 you use for, you know, Bayesian integration and stuff like that. 02:43:55.360 --> 02:43:58.037 >> So the ideas the little ideas 02:43:58.120 --> 02:44:03.397 within quantum mechanics or within some of these kind of more solidified fields 02:44:03.480 --> 02:44:08.277 will have a longer shelf life. They will you somehow use indirectly in in your 02:44:08.360 --> 02:44:11.917 work. >> Learn classical mechanics like you learn about Lagrangians for example. 02:44:12.000 --> 02:44:16.357 >> Mhm. >> Um which is like a a huge hugely useful 02:44:16.440 --> 02:44:19.237 concept, you know, for all kinds of different things. Uh 02:44:19.320 --> 02:44:22.517 learn uh statistical physics because uh 02:44:22.600 --> 02:44:26.877 all the math that comes out of you know, for machine learning 02:44:26.960 --> 02:44:30.997 uh basically comes out of was figured out by statistical physicist in the you 02:44:31.080 --> 02:44:35.637 know, late 19th early 20th century, right? So um and for some of them actually more 02:44:35.720 --> 02:44:39.997 recently for by people like like Giorgio Parisi who just got the Nobel Prize for 02:44:40.080 --> 02:44:42.077 the replica method among other things. 02:44:42.160 --> 02:44:45.957 Um it's used for a lot of different things. Uh you know, variational 02:44:46.040 --> 02:44:48.917 inference uh that math comes from statistical 02:44:49.000 --> 02:44:53.837 physics. Um so um so a lot of those kind of, you 02:44:53.920 --> 02:44:56.957 know, basic courses, you know, you'll you'll 02:44:57.040 --> 02:44:59.717 If you do electrical engineering, you take signal processing, you'll you'll 02:44:59.800 --> 02:45:01.757 learn about Fourier transforms. 02:45:01.840 --> 02:45:04.037 Again, something super useful. It's at 02:45:04.120 --> 02:45:06.037 the basis of things like graph neural 02:45:06.120 --> 02:45:09.517 nets, which is a an entirely new 02:45:09.600 --> 02:45:13.397 sub-area of, you know, AI machine learning deep learning, which I think is 02:45:13.480 --> 02:45:15.797 super promising for all kinds of applications. 02:45:15.880 --> 02:45:18.797 Um, something very promising if you're more interested in applications is the 02:45:18.880 --> 02:45:23.397 applications of, uh, AI machine learning and deep learning to science. 02:45:23.480 --> 02:45:26.997 Or to science that can help solve big problems 02:45:27.080 --> 02:45:29.437 in the world. I've I've colleagues at, 02:45:29.520 --> 02:45:34.117 uh, at Meta at FAIR who started this project called Open Catalyst, and it's, 02:45:34.200 --> 02:45:36.437 um, it's an open project collaborative. 02:45:36.520 --> 02:45:39.757 And the idea is to use, uh, deep learning to 02:45:39.840 --> 02:45:43.117 help design, uh, new chemical compounds 02:45:43.200 --> 02:45:47.677 or materials that would facilitate the separation of hydrogen from oxygen. 02:45:47.760 --> 02:45:49.797 If you can efficiently separate oxygen 02:45:49.880 --> 02:45:52.837 from hydrogen with electricity, 02:45:52.920 --> 02:45:55.397 you you solve climate change. 02:45:55.480 --> 02:45:59.957 It's as simple as that, because, um, you you cover, you know, some random 02:46:00.040 --> 02:46:03.957 desert with, uh, solar panels, uh, and you have it have them work all 02:46:04.040 --> 02:46:07.397 day, produce hydrogen, and then you ship the hydrogen wherever it's needed. You 02:46:07.480 --> 02:46:09.357 don't need anything else. 02:46:09.440 --> 02:46:13.037 Um, you know, you you you have 02:46:13.120 --> 02:46:15.917 controllable power that's, you know, 02:46:16.000 --> 02:46:17.557 what can be transported, uh, anywhere. 02:46:17.640 --> 02:46:20.917 So, if we if we have a large-scale 02:46:21.000 --> 02:46:25.437 efficient, uh, energy storage technology like producing 02:46:25.520 --> 02:46:28.717 hydrogen, uh, we we solve climate change. Here's 02:46:28.800 --> 02:46:31.837 another way to solve climate change is, uh, if you can figure out how to make 02:46:31.920 --> 02:46:34.677 fusion work. Now, the problem with fusion is that you make a super hot 02:46:34.760 --> 02:46:38.637 plasma, and the plasma is unstable, and you can't control it. Maybe with deep 02:46:38.720 --> 02:46:41.517 learning you can find controllers that will stabilize plasma and make, you 02:46:41.600 --> 02:46:45.477 know, practical fusion reactors. I mean, that's very speculative, but, you know, 02:46:45.560 --> 02:46:50.197 it's worth trying because, um, you know, it uh, the payoff is huge. 02:46:50.280 --> 02:46:53.117 There's a group at Google working on this led by John Platt. 02:46:53.200 --> 02:46:55.997 >> So, control convert as many problems in 02:46:56.080 --> 02:47:00.437 science and physics and biology and chemistry into a 02:47:00.520 --> 02:47:03.477 into a learnable problem and see if a machine can learn it. 02:47:03.560 --> 02:47:07.677 >> Right. I mean, there's properties of you know, complex materials that we don't 02:47:07.760 --> 02:47:10.277 understand from first principle, for example. 02:47:10.360 --> 02:47:14.677 Right? So, you know, if we could design new 02:47:14.760 --> 02:47:18.317 you know, new materials, we could make more efficient batteries. 02:47:18.400 --> 02:47:21.477 You know, we could make maybe faster electronics. We could I mean, there's a 02:47:21.560 --> 02:47:25.797 lot of things we can imagine doing or you know, lighter 02:47:25.880 --> 02:47:28.997 materials for for cars or airplanes or things like that. Maybe better fuel 02:47:29.080 --> 02:47:32.397 cells. There's all kinds of stuff we can imagine. If we had good fuel cells, 02:47:32.480 --> 02:47:36.677 hydrogen fuel cells, we could use them to power airplanes and you know, 02:47:36.760 --> 02:47:39.237 transportation wouldn't be or or cars. 02:47:39.320 --> 02:47:43.237 We wouldn't have emission problem CO2 emission problems 02:47:43.320 --> 02:47:47.397 for for air transportation anymore. So, this 02:47:47.480 --> 02:47:52.037 there's a lot of those things I think where AI you know, can be used it 02:47:52.120 --> 02:47:55.877 and and this is not even talking about all the sort of medicine, biology and 02:47:55.960 --> 02:47:57.597 and everything like that, right? 02:47:57.680 --> 02:48:01.197 You know, like protein folding, you know, figuring out 02:48:01.280 --> 02:48:03.877 like how can you design a protein that it sticks to another protein that a 02:48:03.960 --> 02:48:07.277 particular sites because that's how you design drugs in the end. 02:48:07.360 --> 02:48:10.637 So, you know, deep learning would be useful and those are and those are kind 02:48:10.720 --> 02:48:14.117 of you know, would be sort of enormous progress if we could 02:48:14.200 --> 02:48:17.477 use it for that. Here's an example. If you take 02:48:17.560 --> 02:48:19.397 this is like from recent material 02:48:19.480 --> 02:48:23.197 physics. You take a monoatomic layer of 02:48:23.280 --> 02:48:25.637 graphene, right? So, it's just carbon on 02:48:25.720 --> 02:48:31.037 a hexagonal mesh and then make this single single atom thick. 02:48:31.120 --> 02:48:34.237 You put another one on top. You twist them by some magic 02:48:34.320 --> 02:48:36.717 number of degrees, 3 degrees or something. 02:48:36.800 --> 02:48:40.437 It becomes superconductor. Nobody has any idea why. 02:48:40.520 --> 02:48:44.357 >> [laughter] >> I want to know how that was discovered, 02:48:44.440 --> 02:48:46.637 but that's the kind of thing that machine learning can actually discover 02:48:46.720 --> 02:48:50.037 these kinds of things. >> maybe not, but but there is 02:48:50.120 --> 02:48:51.997 a hint perhaps that with machine 02:48:52.080 --> 02:48:54.797 learning we could train a system to 02:48:54.880 --> 02:48:59.477 basically be a phenomenological model of some complex emergent phenomenon which, 02:48:59.560 --> 02:49:03.037 you know, superconductivity is one of those. 02:49:03.120 --> 02:49:07.757 Uh where, you know, think this collective phenomenon is too difficult to describe 02:49:07.840 --> 02:49:11.957 from first principles with the current, you know, the usual sort of, you know, 02:49:12.040 --> 02:49:15.757 reductionist type method. But we could have 02:49:15.840 --> 02:49:19.037 uh deep learning systems that predict the properties 02:49:19.120 --> 02:49:23.317 of a system from a description of it uh after being trained with sufficiently uh 02:49:23.400 --> 02:49:26.117 many samples. Um 02:49:26.200 --> 02:49:28.517 There's um this guy Pascal Fua at EPFL. 02:49:28.600 --> 02:49:32.517 He has a startup company um that >> [snorts] >> where he he 02:49:32.600 --> 02:49:36.557 basically trained a um a convolutional net essentially to 02:49:36.640 --> 02:49:39.877 predict the aerodynamic properties of solids. 02:49:39.960 --> 02:49:42.957 And you can generate as much data as you want by just running uh computational 02:49:43.040 --> 02:49:47.597 fluid dynamics, right? So, you give a like a 02:49:47.680 --> 02:49:51.677 a wing airfoil or something shape of some kind, 02:49:51.760 --> 02:49:55.957 and you run computational fluid dynamics, you get as a result the drag 02:49:56.040 --> 02:49:59.677 and, you know, uh lift and all that stuff, right? And 02:49:59.760 --> 02:50:03.077 you can you can generate lots of data, train a neural net to make those 02:50:03.160 --> 02:50:05.957 predictions, and now what you have is a differentiable model 02:50:06.040 --> 02:50:10.677 of, let's say, drag and and lift as a function of the shape of that solid, and 02:50:10.760 --> 02:50:13.837 so you can do back gradient descent, you can optimize the shape so you get the 02:50:13.920 --> 02:50:18.117 properties you want. >> [snorts] >> Um >> Yeah, that's incredible. That's 02:50:18.200 --> 02:50:21.597 incredible. And then on top of all that, probably 02:50:21.680 --> 02:50:22.597 you should read a little bit of 02:50:22.680 --> 02:50:25.517 literature and a little bit of history 02:50:25.600 --> 02:50:30.717 for inspiration and for wisdom, cuz after all all of these technologies will 02:50:30.800 --> 02:50:32.237 have to work in a human world. 02:50:32.320 --> 02:50:34.477 >> Yes. >> And the human world is complicated. 02:50:34.560 --> 02:50:39.117 >> It is. So, yeah. >> Yeah, and this is uh 02:50:39.200 --> 02:50:41.957 an amazing conversation. I'm really honored that you would talk with me 02:50:42.040 --> 02:50:46.077 today. Thank you for all the amazing work you're doing at FAIR at Meta. And 02:50:46.160 --> 02:50:50.117 thank you for being so passionate after all these years about everything that's 02:50:50.200 --> 02:50:53.877 going on. You're you're a beacon of hope for the machine learning community. And 02:50:53.960 --> 02:50:56.557 thank you so much for spending your valuable time with me today. That was 02:50:56.640 --> 02:51:00.717 awesome. >> Thanks for having me on. That was It was a pleasure. 02:51:00.800 --> 02:51:03.757 >> Thanks for listening to this conversation with Yann LeCun. To support 02:51:03.840 --> 02:51:07.637 this podcast, please check out our sponsors in the description. 02:51:07.720 --> 02:51:12.517 And now, let me leave you with some words from Isaac Asimov. 02:51:12.600 --> 02:51:15.677 Your assumptions are your windows on the world. 02:51:15.760 --> 02:51:20.677 Scrub them off every once in a while, or the light won't come in. 02:51:20.760 --> 02:51:25.520 Thank you for listening, and hope to see you next time.