WEBVTT 00:00:00.000 --> 00:00:01.517 Our mission is to put intelligence on a 00:00:01.600 --> 00:00:04.197 billion machines. And we think that can 00:00:04.280 --> 00:00:07.077 have a profound impact on society. 00:00:07.160 --> 00:00:10.477 Applied Intuition is a physical AI company. We put intelligence on 00:00:10.560 --> 00:00:15.397 machines. Cars, trucks, tanks, drones, it's a physical moving thing. We make it 00:00:15.480 --> 00:00:17.317 intelligent. >> Digital AI of course is building 00:00:17.400 --> 00:00:21.477 software and optimizing ads and creating videos. That's all interesting and good, 00:00:21.560 --> 00:00:25.517 but really we talk about the global economy, that's physical AI. 00:00:25.600 --> 00:00:28.957 >> In this intelligence revolution, the companies that impact the physical world 00:00:29.040 --> 00:00:31.357 might actually be bigger >> [music] >> than the companies that impact the 00:00:31.440 --> 00:00:34.077 digital world. >> How many things are there where the idea 00:00:34.160 --> 00:00:36.477 of physical AI, physical intelligence are going to matter? 00:00:36.560 --> 00:00:38.037 >> There's no reason autonomy should be 00:00:38.120 --> 00:00:41.317 this obscure, difficult technology. Our 00:00:41.400 --> 00:00:45.677 vision for that is a high school kid that can make iPhone apps should be able 00:00:45.760 --> 00:00:49.277 to make autonomous systems. That platform for designing and developing is 00:00:49.360 --> 00:00:51.077 what we're launching. It's called Dana. 00:00:51.160 --> 00:00:54.957 >> Everything that we've built and developed over the past nearly a decade, 00:00:55.040 --> 00:00:57.957 that's available in Dana. >> Which will we get first? A perfectly 00:00:58.040 --> 00:01:00.997 simulated real world environment for training autonomous devices or Grand 00:01:01.080 --> 00:01:04.900 Theft Auto 6? >> [laughter] 00:01:07.560 --> 00:01:09.997 >> Kash, Peter, welcome to the a16z podcast. 00:01:10.080 --> 00:01:12.617 >> Well, thanks for having us. Your name is? 00:01:12.700 --> 00:01:15.757 >> [laughter] >> Yeah, which is one of many. 00:01:15.840 --> 00:01:19.517 >> We've I think we've all each known each other for 00:01:19.600 --> 00:01:21.797 too long, more than I'd like to admit. 00:01:21.880 --> 00:01:23.597 >> [laughter] >> Long time. We're lucky to both be the 00:01:23.680 --> 00:01:26.357 first investor or among the first investor in the first round. Of course, 00:01:26.440 --> 00:01:29.957 different check sizes, but >> And I was an investor in you even before then. 00:01:30.040 --> 00:01:33.477 >> Exactly. Yeah. Um so, let's do that as a segue. We have a lot to talk about 00:01:33.560 --> 00:01:35.677 today. We've, you know, the the biggest launch in company history to talk about 00:01:35.760 --> 00:01:39.037 today. But first, why don't we just give an update or a status [clears throat] 00:01:39.120 --> 00:01:40.397 What does Applied Intuition do for those who don't 00:01:40.480 --> 00:01:43.117 >> Yeah, for the for the people who don't know, Applied Intuition is a physical AI 00:01:43.200 --> 00:01:47.517 company. We put intelligence on machines. That's the the simple way of 00:01:47.600 --> 00:01:51.437 describing it. Um and all types of machines. So, cars, 00:01:51.520 --> 00:01:56.277 trucks, tanks, drones, you name it. It's a physical moving thing. We we make it 00:01:56.360 --> 00:01:59.517 intelligent. And uh the history of the company is we originally started by 00:01:59.600 --> 00:02:02.397 making the tools that would make the intel and then we got into the actual 00:02:02.480 --> 00:02:05.077 intelligence itself. Um in a very like 00:02:05.160 --> 00:02:07.197 in some ways like a very uh boring AI 00:02:07.280 --> 00:02:12.197 company in sense of, you know, 83% of the company is engineering. We win by 00:02:12.280 --> 00:02:15.037 making really great products. It's not like a 00:02:15.120 --> 00:02:18.357 good sales or something like that. I don't think we're good enough for for uh 00:02:18.440 --> 00:02:20.477 for a sales enabled company, but uh 00:02:20.560 --> 00:02:23.357 yeah, over a thousand engineers um and 00:02:23.440 --> 00:02:25.437 based in Silicon Valley. Uh but we have 00:02:25.520 --> 00:02:28.997 offices globally, 18 offices. And uh you 00:02:29.080 --> 00:02:32.997 know, our mission is to to put intelligence on a billion machines. And 00:02:33.080 --> 00:02:38.397 uh we think that can have a profound impact on society both in the kind of 00:02:38.480 --> 00:02:41.237 busy things everyone talks about safety, you know, if you really talk to somebody 00:02:41.320 --> 00:02:45.757 who's been in a car accident or in a mining accident or, you know, uh in a 00:02:45.840 --> 00:02:50.837 farming accident. You Those Those are real gnarly situations. Beyond just fixing 00:02:50.920 --> 00:02:54.077 that, you just if you can unlock productivity, 00:02:54.160 --> 00:02:57.997 I think, you know, we've seen the unlock in the digital world. And everyone's 00:02:58.080 --> 00:03:01.637 super excited about it and you have trillion dollar companies emerging. I'm 00:03:01.720 --> 00:03:04.877 a pretty strong believer that I think when we look back 25 years If you look 00:03:04.960 --> 00:03:07.317 back at the internet now, you know, people actually If you look at the 00:03:07.400 --> 00:03:10.837 original internet companies that are doing, you know, serving or they're 00:03:10.920 --> 00:03:13.677 they're doing some analytics and those are interesting, but really when you 00:03:13.760 --> 00:03:17.797 look back 25 years from now, the big monolithic companies are Amazon that 00:03:17.880 --> 00:03:21.157 delivers you stuff, you know, Apple. These are the true kind of 00:03:21.240 --> 00:03:24.437 companies that come of age. And I think when we look back 25 years in this 00:03:24.520 --> 00:03:28.637 intelligent revolution, the companies that impact the physical world, 00:03:28.720 --> 00:03:31.477 you know, might actually be bigger than the companies that impact the digital 00:03:31.560 --> 00:03:33.277 world. >> I would love for you to talk about the 00:03:33.360 --> 00:03:35.437 following, which is when you first started the company, you know, the knock 00:03:35.520 --> 00:03:37.877 on the company I think was, oh, well, it's it's like, you know, car it's 00:03:37.960 --> 00:03:39.557 making making cars autonomous, right? 00:03:39.640 --> 00:03:42.437 Self-driving cars, but it's kind of like, okay, there's like whatever, 00:03:42.520 --> 00:03:44.837 you know, there's there's Tesla and Waymo building their own self-driving 00:03:44.920 --> 00:03:47.397 cars and then and then there's like six or eight other car companies that 00:03:47.480 --> 00:03:49.877 matter. And then the company just could never get that big cuz there just aren't 00:03:49.960 --> 00:03:53.757 that many customers. >> Yeah. >> So, what's the like how should people 00:03:53.840 --> 00:03:56.557 think about like how many things are there that are 00:03:56.640 --> 00:03:59.717 things that move where the idea of physical AI physical intelligence are 00:03:59.800 --> 00:04:02.637 going to matter? >> Yeah, I mean even today even if you do 00:04:02.720 --> 00:04:07.117 put that you know let's say view on us the automotive is like 30% of 00:04:07.200 --> 00:04:11.117 our business so 70% already is non-automotive and I think if you fast 00:04:11.200 --> 00:04:15.357 forward another 10 20 years even the manufacturers themselves as a 00:04:15.440 --> 00:04:19.197 customer base will be a small amount. I think our that mission just keep 00:04:19.280 --> 00:04:23.717 thinking a thought you know a billion machines becoming intelligent. 00:04:23.800 --> 00:04:26.397 And you think about all the types of machines that exist. Automotive is just 00:04:26.480 --> 00:04:29.397 an easy one. I think it sticks in people's heads because we all drive cars 00:04:29.480 --> 00:04:33.277 and it's a big market but I think it's it'll be a minority of 00:04:33.360 --> 00:04:35.677 of the business minority is a minority business. I think it'll be increasingly 00:04:35.760 --> 00:04:38.837 minority business but that doesn't necessarily mean it'll be small. 00:04:38.920 --> 00:04:42.597 Automotive is still huge just as a part of the globe's GDP automotive is 00:04:42.680 --> 00:04:46.237 something like 3% of all GDP. 00:04:46.320 --> 00:04:50.317 I think the way we always think about it like as you try to get to your mission 00:04:50.400 --> 00:04:54.877 initially the manufacturers were the distribution to that intelligence to 00:04:54.960 --> 00:04:58.117 consumers but then you start working in defense and you start working in 00:04:58.200 --> 00:05:01.677 construction and mining and agriculture and suddenly the manufacturers are 00:05:01.760 --> 00:05:04.597 important but maybe the mining operators actually really important or the 00:05:04.680 --> 00:05:07.677 Department of War is really important and suddenly they become customers and 00:05:07.760 --> 00:05:10.037 all of those are customers of ours as well. 00:05:10.120 --> 00:05:14.157 >> Yeah, I think if you split AI into digital AI and physical AI right digital 00:05:14.240 --> 00:05:18.437 AI of course is building software and optimizing ads and creating videos that 00:05:18.520 --> 00:05:21.637 sort of thing. That's all interesting and good but 00:05:21.720 --> 00:05:25.317 really when you talk about the global economy that's physical AI. Then we're 00:05:25.400 --> 00:05:30.597 talking about manufacturing and and mining and logistics and transportation 00:05:30.680 --> 00:05:32.157 all of these things that supply chains. 00:05:32.240 --> 00:05:33.037 >> Yeah, supply chains exactly. 00:05:33.120 --> 00:05:36.597 >> Let me let's build on that for a second which is like so things that move today 00:05:36.680 --> 00:05:39.077 or you know historically things that move are things that have human beings 00:05:39.160 --> 00:05:43.637 at the wheel or at the controls in some form right? And so and you know 00:05:43.720 --> 00:05:46.317 airplanes [snorts] have had to get designed around a human in the cockpit 00:05:46.400 --> 00:05:48.917 boats have had to you know, get designed around a human, you know, steering 00:05:49.000 --> 00:05:52.717 things. Um like in a world where in a world of 00:05:52.800 --> 00:05:56.517 autonomy like Do we already know what the things are 00:05:56.600 --> 00:05:58.877 that move? Or are we going to discover that there are a lot of new things that 00:05:58.960 --> 00:06:01.997 are going to get built when you don't need a a human in the in the driver's seat? 00:06:02.080 --> 00:06:06.397 >> I think both. Well, the the the the thing that you have to remember is like 00:06:06.480 --> 00:06:08.557 you take like a a haulage system that's 00:06:08.640 --> 00:06:13.557 on a in a in a port. Um like a caterpillar Komatsu, the dirt 00:06:13.640 --> 00:06:17.757 mover in a mine. Those are made for 20 25 years. 00:06:17.840 --> 00:06:20.957 >> Mhm. >> So, the buyers of that of those products, they're they might not have 00:06:21.040 --> 00:06:25.917 gotten their full cycle, you know, ROI on them. So, they're not immediately 00:06:26.000 --> 00:06:29.757 going to buy something new no matter how much better it is. So, one part of our 00:06:29.840 --> 00:06:32.397 strategy is you got to make those things intelligent cuz they're not going 00:06:32.480 --> 00:06:36.717 anywhere. The second is what you're talking about, which is well that 00:06:36.800 --> 00:06:40.237 depends on a human in a cab. If you don't have a human in a cab, you can run 00:06:40.320 --> 00:06:43.157 the the machine can be smaller. It can be shaped in very different ways. You 00:06:43.240 --> 00:06:45.917 need to talk about mining underground. 00:06:46.000 --> 00:06:47.877 >> Right. >> The constraint actually is the human 00:06:47.960 --> 00:06:50.557 because the human needs to breathe and it needs to it's very dangerous. And so, 00:06:50.640 --> 00:06:52.677 you can build a very very different machine. 00:06:52.760 --> 00:06:56.637 We're doing both of those things. And and then the thing that you were not 00:06:56.720 --> 00:07:00.357 talking about is we're all talking about intelligence almost like within a 00:07:00.440 --> 00:07:05.157 system, but the system level intelligence is where the unlock is. And 00:07:05.240 --> 00:07:07.917 we're already doing work like that where you say, "Hey, let's take an entire 00:07:08.000 --> 00:07:11.557 port. Let's take an entire mine. Let's take an entire quarry." 00:07:11.640 --> 00:07:14.077 And this heterogeneous mix of machines 00:07:14.160 --> 00:07:16.197 they're all can talk to each other. 00:07:16.280 --> 00:07:20.157 And they can optimize and be efficient when one machine goes down or one 00:07:20.240 --> 00:07:23.317 machine has an issue. The rest of the mine doesn't have to stop. When it's 00:07:23.400 --> 00:07:26.157 human driven, we don't even know the machine's going to go down because 00:07:26.240 --> 00:07:29.837 there's no announce. The human is not plugged into the core 00:07:29.920 --> 00:07:31.317 systems of the machine. >> Right. 00:07:31.400 --> 00:07:34.237 >> So, like a simple thing like knowing 00:07:34.320 --> 00:07:39.157 when a brake system is going to break is actually huge cuz you can start 00:07:39.240 --> 00:07:41.677 preparing for it in advance. You're like, "Oh, this wear and tear is higher 00:07:41.760 --> 00:07:45.637 than in other mines. He was using example. But like the the other macro 00:07:45.720 --> 00:07:48.677 point is if you look at agriculture an example, you know, average American 00:07:48.760 --> 00:07:52.157 farmer is 58 years old. The there's that the number something 00:07:52.240 --> 00:07:55.077 like under 35 it's like less than 10% of 00:07:55.160 --> 00:07:57.477 farmers are that young. So, what's going 00:07:57.560 --> 00:07:59.997 to happen? The need for food growth is 00:08:00.080 --> 00:08:04.717 continuing to grow. The need for rare earth materials is coming so these 00:08:04.800 --> 00:08:08.397 these demands are only growing, but the humans who are the bottleneck are 00:08:08.480 --> 00:08:11.437 decreasing. Trucking is the same way. 00:08:11.520 --> 00:08:16.677 Um and so you can you can really just unlock a lot more 00:08:16.760 --> 00:08:20.397 efficiency. So, I mean one way to think maybe think about this is like imagine 00:08:20.480 --> 00:08:24.917 if the cost for food decreases because it's way way more efficient. What what's 00:08:25.000 --> 00:08:28.957 the downstream impact? Then imagine for goods being transported. Let's say 00:08:29.040 --> 00:08:33.437 you know, instead of a few dollars a mile it's 20 cents a mile. And suddenly 00:08:33.520 --> 00:08:37.677 it's it's I think that the unlock is very very very big. Um and that isn't 00:08:37.760 --> 00:08:40.877 that I think doesn't necessarily need for all the machines to be redesigned 00:08:40.960 --> 00:08:42.917 from the ground up. >> Right. Right. Got it. Makes sense. 00:08:43.000 --> 00:08:45.077 >> And then maybe just one one more question would be just give give a give 00:08:45.160 --> 00:08:48.317 us a sense of parameterize like the scope and scale of the company today. 00:08:48.400 --> 00:08:51.117 >> Yeah, uh north of a thousand engineers. And 00:08:51.200 --> 00:08:54.077 those engineers are, you know, obviously 00:08:54.160 --> 00:08:58.917 the classic, you know, software and AI engineering teams. But 00:08:59.000 --> 00:09:02.037 we also have engineers who really know safety systems. We also have engineers 00:09:02.120 --> 00:09:05.757 who really know hardware. Because the important thing that we we're kind of 00:09:05.840 --> 00:09:09.917 just tipping around stepping around is all this stuff is hard because it 00:09:10.000 --> 00:09:11.517 ultimately has to meet the real world. 00:09:11.600 --> 00:09:15.717 And the real world is has way more complexity and and and has a lot more 00:09:15.800 --> 00:09:19.157 issues. And we have engineering teams that can I mean we've deployed our our 00:09:19.240 --> 00:09:21.677 our models onto like 50 some platforms. 00:09:21.760 --> 00:09:25.197 Like even that sounds trivial because when you think about mostly when you 00:09:25.280 --> 00:09:28.077 think about models you think about deploying them through a browser or on a 00:09:28.160 --> 00:09:31.997 phone and everything's abstracted away because you have iOS and you have 00:09:32.080 --> 00:09:34.237 Android and you have Windows and you have Linux and you have all these 00:09:34.320 --> 00:09:37.477 systems that have already taken care. In the real world you don't have that. 00:09:37.560 --> 00:09:39.197 >> Right. >> And so we we have engineering teams that 00:09:39.280 --> 00:09:44.477 can do that as well. Our kind of a you know, claim to fame is we've raised over 00:09:44.560 --> 00:09:47.637 about a billion dollars in the company's history. All that is sitting in the bank 00:09:47.720 --> 00:09:51.477 and I think that's I always say that with an asterisk because it doesn't mean 00:09:51.560 --> 00:09:54.077 we're not going to spend it, you know, next next month. 00:09:54.160 --> 00:09:55.957 >> good news, bad Good news, bad news. 00:09:56.040 --> 00:09:58.917 >> Yeah, good news, bad news. It's but it's 00:09:59.000 --> 00:10:02.957 the the and I think like we're at that phase, you know, you talk about scale. 00:10:03.040 --> 00:10:04.757 We're at that phase where 00:10:04.840 --> 00:10:08.517 these giant markets are around us and we 00:10:08.600 --> 00:10:12.197 can make the decision how aggressive do we want to pursue those because 00:10:12.280 --> 00:10:16.557 decade of frankly execution and deployment into production. I think the 00:10:16.640 --> 00:10:19.877 the hallmark of our engineering team is is putting 00:10:19.960 --> 00:10:23.077 you know, products into production. That that that that that really is is a big I 00:10:23.160 --> 00:10:24.917 don't know how do you think about scale? 00:10:25.000 --> 00:10:27.397 >> Yeah, I think that that's roughly it. I mean the the the mission of bringing 00:10:27.480 --> 00:10:30.357 intelligence to a billion machines. That is how we think about it. And then 00:10:30.440 --> 00:10:32.877 thinking about well, what are the types of machines that will have the most 00:10:32.960 --> 00:10:37.037 impact on and focusing on those areas first, but we'll we'll get there. 00:10:37.120 --> 00:10:41.117 >> I let's go deeper into the differences between digital and physical AI and more 00:10:41.200 --> 00:10:44.757 so into where where are we today? What progress 00:10:44.840 --> 00:10:47.517 has has been made? What are some of the main major bottlenecks in in physical 00:10:47.600 --> 00:10:49.877 AI? When you unpack some of that. 00:10:49.960 --> 00:10:53.437 >> Yeah, I mean I think a lot of times people think about the progress in 00:10:53.520 --> 00:10:57.597 physical AI is limited to basically two case use cases in there just cuz they're 00:10:57.680 --> 00:11:00.877 obvious and interesting, which is robo-taxis and humanoids. 00:11:00.960 --> 00:11:03.837 They're very visceral. They're they're they excite you and they're kind of 00:11:03.920 --> 00:11:08.837 sci-fi. I think they're those are very interesting 00:11:08.920 --> 00:11:12.077 and they there is real work being done by us and other people in in those 00:11:12.160 --> 00:11:16.477 domains. I think all the other uh all the other domains I think are going 00:11:16.560 --> 00:11:20.237 to be just as important. I mean you just you just think about what happens on a 00:11:20.320 --> 00:11:24.677 port. That's there's a huge unlock there 00:11:24.760 --> 00:11:27.557 and that's I think that's the that's the area where where where really focused 00:11:27.640 --> 00:11:30.037 on. It's like all the other nooks and crannies. 00:11:30.120 --> 00:11:34.157 If you look at like we we've talked before about uh the rise of Cisco and 00:11:34.240 --> 00:11:37.917 how, you know, networking kind of went from, you know, first individual 00:11:38.000 --> 00:11:40.597 machines and companies would get network and then entire countries were getting 00:11:40.680 --> 00:11:42.357 network. There's a similar thing 00:11:42.440 --> 00:11:45.597 happening with AI. AI is getting to that 00:11:45.680 --> 00:11:47.757 level of kind of sovereign AI is now a 00:11:47.840 --> 00:11:50.637 discussion. Sovereign AI really is about 00:11:50.720 --> 00:11:55.077 physical AI. Because that's where you're talking about AI in defense, you're 00:11:55.160 --> 00:11:58.157 talking about AI in the physical machines that are moving around. If you 00:11:58.240 --> 00:12:01.397 look just at the example of of Waymo 00:12:01.480 --> 00:12:05.637 from America and Pony from China, they're trying to deploy in, let's say, 00:12:05.720 --> 00:12:10.757 the other country, so not America, not Europe, not China, every one of those uh 00:12:10.840 --> 00:12:15.757 spaces, they're way more uh they're way more hesitant of saying, 00:12:15.840 --> 00:12:20.637 "Yeah, thumbs up, your robot taxis can run unfettered on our in our country." 00:12:20.720 --> 00:12:24.397 And so, if you look back just at kind of this arc of the internet, 00:12:24.480 --> 00:12:27.197 you know, when the first internet companies come, nobody's really thinking 00:12:27.280 --> 00:12:31.037 about sovereignty at all. It's like, the browser goes everywhere, the internet 00:12:31.120 --> 00:12:33.317 goes everywhere, that's almost the power 00:12:33.400 --> 00:12:38.277 of it. Then when social media emerges, there's a bit more of, "Hey, actually 00:12:38.360 --> 00:12:42.397 not every social media." And then you have China not allowing Facebook to come 00:12:42.480 --> 00:12:46.237 in and then you have some Then you get into the next level of like the online 00:12:46.320 --> 00:12:49.717 offline stuff. There's more resistance to Ubers, to DoorDashes. Suddenly, 00:12:49.800 --> 00:12:53.037 there's local players who are being favored very aggressively. 00:12:53.120 --> 00:12:56.357 When we get to physical AI, I think there's going to be huge and also 00:12:56.440 --> 00:13:00.997 there's like a larger geo you know, geopolitical theme of of kind of 00:13:01.080 --> 00:13:02.677 more fracturing than globalization. 00:13:02.760 --> 00:13:04.677 You're going to have this demand for 00:13:04.760 --> 00:13:08.677 this AI should somehow be localized. And 00:13:08.760 --> 00:13:11.917 I think that has to play into our strategy as well. We're technology 00:13:12.000 --> 00:13:15.677 provider. So, we can provide that technology across the globe. 00:13:15.760 --> 00:13:18.637 And I think that's that's that's that's that's that's something that's 00:13:18.720 --> 00:13:21.357 understated in this in this conversation. 00:13:21.440 --> 00:13:24.917 >> Yeah, a few other things on on digital versus physical AI. So, in digital AI, 00:13:25.000 --> 00:13:28.517 the state of the art is you you can train models effectively on the entirety 00:13:28.600 --> 00:13:32.717 of the internet and then maybe augment that with additional data that's been 00:13:32.800 --> 00:13:36.037 collected and refined with some hired experts, right? This is sort of a hot 00:13:36.120 --> 00:13:39.717 field right now. Uh but generally you're talking about a foundation model that's 00:13:39.800 --> 00:13:42.557 built on internet data. In physical AI, 00:13:42.640 --> 00:13:45.597 the internet data is is useful, too. 00:13:45.680 --> 00:13:49.677 However, to actually build a foundation model in physical AI, there's also a lot 00:13:49.760 --> 00:13:53.677 of private data collection. You uh when we're talking about 00:13:53.760 --> 00:13:58.037 mines or or logistics or any of these other uh other fields, the the data that's 00:13:58.120 --> 00:14:02.037 useful for training models there is not necessarily available. So, we have to do 00:14:02.120 --> 00:14:05.877 a lot of work ourselves actually going going out and collecting that data. And 00:14:05.960 --> 00:14:09.957 then the other key factor is safety, right? Uh if you're talking about uh 00:14:10.040 --> 00:14:13.037 building a smartphone app, you don't necessarily care about uh 00:14:13.120 --> 00:14:16.877 is is a safety-critical application, but when you're talking about moving a a 00:14:16.960 --> 00:14:20.757 machine that weighs many tons uh or think of a humanoid which could fall 00:14:20.840 --> 00:14:25.157 over on your children, uh you care a lot about safety and and the evaluation of 00:14:25.240 --> 00:14:28.517 that safety and and that is really sort of getting to the state of the art of 00:14:28.600 --> 00:14:32.637 physical AI and and really proving out uh the safety case around some of these 00:14:32.720 --> 00:14:34.677 state-of-the-art models. >> Yeah, and I think like you know, you you 00:14:34.760 --> 00:14:39.157 talk about like humanoid data collection has been its own, you know, uh little uh 00:14:39.240 --> 00:14:43.237 area of interest. But when you talk about collecting data like in places 00:14:43.320 --> 00:14:46.637 like Korea where they have North South Korea where you have North Korea, 00:14:46.720 --> 00:14:51.437 they don't allow mapping companies, let alone allowing a you know, a an American 00:14:51.520 --> 00:14:54.637 company to come in and data collect. So, we've figured out over the years whether 00:14:54.720 --> 00:14:59.037 it's the Middle East, whether it's uh LatAm, how to get into these countries, 00:14:59.120 --> 00:15:03.597 work with the governments, and get the thumbs up to collect proprietary data. 00:15:03.680 --> 00:15:06.997 And so, in in the way that it's it is similar to very you know, other uh 00:15:07.080 --> 00:15:11.717 digital AI systems, your proprietary data sets, scaling laws, all 00:15:11.800 --> 00:15:15.437 that stuff is the same. It's just applied in a very very different way. 00:15:15.520 --> 00:15:17.957 And uh it's almost like the way to think about 00:15:18.040 --> 00:15:19.997 it is like the diffusion of these models 00:15:20.080 --> 00:15:24.037 is very different because you can't it's 00:15:24.120 --> 00:15:27.237 not everyone can just access them through a phone. You're you're going to 00:15:27.320 --> 00:15:31.997 And so so that ironically is actually plays in our 00:15:32.080 --> 00:15:35.277 favor. Because once we have a massive proprietary data set which we've been 00:15:35.360 --> 00:15:37.997 building, we already have hundreds of petabytes of data. 00:15:38.080 --> 00:15:40.837 Um and then we have our own tools which are like, you know, synthetic data 00:15:40.920 --> 00:15:44.797 tools, uh neural sim. We can use our own tools and our own proprietary data and 00:15:44.880 --> 00:15:47.997 that allows us to build some of the best systems in the business. 00:15:48.080 --> 00:15:50.997 >> Cuz is it There's kind of a chicken and egg thing 00:15:51.080 --> 00:15:55.157 which is like in order to build an autonomous physical thing, you need a 00:15:55.240 --> 00:15:57.917 lot of data. To gather that data, you need a lot of 00:15:58.000 --> 00:16:00.877 physical autonomous things running around collecting the data. So it's like 00:16:00.960 --> 00:16:03.197 once you have a giant network of physical things running around, you have 00:16:03.280 --> 00:16:06.877 the data that makes them all work. Like is there a Is there a flywheel aspect to 00:16:06.960 --> 00:16:09.637 that? And is there Is there Is there like what what what's the level of 00:16:09.720 --> 00:16:11.397 difficulty involved in kind of booting up that flywheel? 00:16:11.480 --> 00:16:15.357 >> Uh it's it's it's difficult but it's also not difficult. I mean, I think we 00:16:15.440 --> 00:16:18.117 we have one of the largest data collection fleets on the planet, frankly 00:16:18.200 --> 00:16:23.037 speaking. Um so that's how you bootstrap your way into it. That's just money and 00:16:23.120 --> 00:16:24.837 resource and and technical knowledge. 00:16:24.920 --> 00:16:27.637 It's but it's not like there's probably more than five 00:16:27.720 --> 00:16:29.917 companies that have that technical knowledge. So it's not extremely 00:16:30.000 --> 00:16:33.757 obscure. I think what is more difficult is then 00:16:33.840 --> 00:16:35.677 how do you actually have that model 00:16:35.760 --> 00:16:37.797 which is going to work on lots of 00:16:37.880 --> 00:16:41.037 different hardware and is, you know, is 00:16:41.120 --> 00:16:45.517 is tested appropriately? Because the You saw it, you know, in Cruise, right? 00:16:45.600 --> 00:16:50.037 Cruise was this company that did amazing self-driving work and then one accident. 00:16:50.120 --> 00:16:53.117 General Motors owns them super scared and they pull back. So it's like just 00:16:53.200 --> 00:16:56.557 getting these things into production is actually more difficult than than the 00:16:56.640 --> 00:17:00.837 than it seems. Um I think like we believed synthetic data 00:17:00.920 --> 00:17:04.477 was going to be important. So we started our synthetic data team like 5 years ago 00:17:04.560 --> 00:17:08.037 now plus. >> than that, yeah. >> Yeah, more than that at this point. Uh 00:17:08.120 --> 00:17:12.917 and like when we were a strong believer that synthetic data can accelerate 00:17:13.000 --> 00:17:16.717 autonomy development. We've just seen that. And then there are like lots of 00:17:16.800 --> 00:17:20.677 other secondary and tertiary like um technical innovations that have Obviously, 00:17:20.760 --> 00:17:23.677 the transformer revolution hitting self-driving, massive. Basically, 00:17:23.760 --> 00:17:27.077 everything done in self-driving pre-21, 00:17:27.160 --> 00:17:30.757 '22, relevant, but you're almost like 00:17:30.840 --> 00:17:34.397 that's kind of the starting point. But it's also different than today being the 00:17:34.480 --> 00:17:37.357 starting point. Like there the those four, five years are actually there has 00:17:37.440 --> 00:17:41.157 been a lot of work done. You can see it most clearly with Tesla, but there's 00:17:41.240 --> 00:17:45.077 other folks. In that process, the actual techniques 00:17:45.160 --> 00:17:48.717 historically and let me just simplifying here. Imitation learning was the way the 00:17:48.800 --> 00:17:52.637 way the game, which was collect a bunch of data and then the models would 00:17:52.720 --> 00:17:55.717 basically imitate what human drivers do. 00:17:55.800 --> 00:17:57.637 The real state of the art right now is 00:17:57.720 --> 00:18:00.357 end-to-end reinforcement learning in in 00:18:00.440 --> 00:18:03.437 a closed loop in in your tools. And so, 00:18:03.520 --> 00:18:07.677 it's a little simplified to say the system learns itself. It identifies 00:18:07.760 --> 00:18:12.197 where the issues in the self-driving system are and essentially you then find 00:18:12.280 --> 00:18:15.997 data like that or you synthetically create data like that. And then you you 00:18:16.080 --> 00:18:19.037 know, you close that loop and you see are you performing the same scenarios 00:18:19.120 --> 00:18:22.837 better and better. I think if you fast forward some years, 00:18:22.920 --> 00:18:26.517 that will be a completely closed loop like with no humans intervening. Right 00:18:26.600 --> 00:18:30.877 now you still have like like what's the the fog error that we saw uh we still 00:18:30.960 --> 00:18:33.397 see errors in the real world that impact self-driving. 00:18:33.480 --> 00:18:36.157 >> Oh yeah, so so it's like what what what what are the bottlenecks, right? And and 00:18:36.240 --> 00:18:39.557 the the bottlenecks there's plenty of them, but whenever 00:18:39.640 --> 00:18:42.957 you're dealing with physical systems, you inevitably you hit a lot of 00:18:43.040 --> 00:18:45.197 gnarly hardware problems and it could be 00:18:45.280 --> 00:18:48.397 anything from overheating to 00:18:48.480 --> 00:18:51.957 sensor being slightly miscalibrated or 00:18:52.040 --> 00:18:55.277 funny funny issues I saw yesterday was 00:18:55.360 --> 00:18:59.557 basically a fogging sensor like fog impacting a sensor. And these are the 00:18:59.640 --> 00:19:02.437 things that you actually have to solve for this stuff to work very reliably in 00:19:02.520 --> 00:19:04.597 the real world. >> Yeah. So I'm I'm going to ask you a 00:19:04.680 --> 00:19:07.317 three-quarter a question and you can we can decide whether you guys want to 00:19:07.400 --> 00:19:09.837 engage on it or not. It might be an opportunity or might hit the question 00:19:09.920 --> 00:19:14.637 which is Were surprised So Cruise was a super high-flying Silicon Valley autonomy 00:19:14.720 --> 00:19:17.437 startup that was kind of running neck and neck with Tesla early on and so 00:19:17.520 --> 00:19:20.197 forth and you know, very top-end team and then they famously got bought by 00:19:20.280 --> 00:19:22.237 General Motors and they >> One of my first distributions 00:19:22.320 --> 00:19:23.877 personally, so I had enjoyed it. 00:19:23.960 --> 00:19:27.997 >> we go. Y Combinator Y Combinator Y Combinator Y Combinator company. Um 00:19:28.080 --> 00:19:31.757 and um and you know, top-end team and they they were you know, by all accounts 00:19:31.840 --> 00:19:34.077 making excellent progress. They got bought by General General Motors. They 00:19:34.160 --> 00:19:37.837 became the GM autonomy program. GM got a lot of praise at least you know, at 00:19:37.920 --> 00:19:40.557 least in the in the in the in tech circles for being like Okay, being like 00:19:40.640 --> 00:19:43.517 the the legacy automaker with the biggest investment 00:19:43.600 --> 00:19:46.597 >> I I called Peter when uh before before it was announced on that and I said, 00:19:46.680 --> 00:19:49.117 "Hey, Cruise just got bought." You know, he's also GM family. We're both GM 00:19:49.200 --> 00:19:52.597 families. And uh Peter guessed it was said, "Nvidia?" I 00:19:52.680 --> 00:19:54.277 said, "No." He said, I said, "Go fish." 00:19:54.360 --> 00:19:58.577 He said, "Apple?" I said, "No." I said, "General explicit of Motors." 00:19:58.660 --> 00:20:02.197 >> [laughter] >> So, that's surprising to people who are from GM. 00:20:02.280 --> 00:20:03.157 >> That they were willing to buy the thing. 00:20:03.240 --> 00:20:05.357 >> Yeah, that they did it. >> Okay, that they did it. And then by all 00:20:05.440 --> 00:20:08.237 accounts they were I mean, as far as I as far as I ever heard like they were 00:20:08.320 --> 00:20:09.637 making excellent progress. >> Yeah. 00:20:09.720 --> 00:20:10.917 >> And then they had this There was a There 00:20:11.000 --> 00:20:14.677 was an accident. There was a Was there Was there a injury or fatality or 00:20:14.760 --> 00:20:16.917 >> Uh it wasn't a fatality, but it was a serious injury. Somebody was dragged 00:20:17.000 --> 00:20:19.757 [clears throat] for 20 ft. >> Yeah, serious injury, bad press and then 00:20:19.840 --> 00:20:24.357 and then they put a bullet the the GM CEO and board put a bullet in the Cruise 00:20:24.440 --> 00:20:27.677 project and I know the at least some of the senior Cruise people were extremely 00:20:27.760 --> 00:20:30.837 upset you know, >> [clears throat] >> by the aftermath of that. Was it 00:20:30.920 --> 00:20:33.637 surprising that they reacted the way that they did? 00:20:33.720 --> 00:20:35.397 >> Uh so, you know, full disclosure General 00:20:35.480 --> 00:20:37.997 Motors is a customer and I went to the General Motors Institute. So, we have a 00:20:38.080 --> 00:20:42.997 lot of love for the company. Uh but incidentally and ironically, I'm reading 00:20:43.080 --> 00:20:45.997 uh or coincidentally I should say, I'm reading this very famous book which I 00:20:46.080 --> 00:20:49.667 had actually never read before called On a Clear Day You Can See General Motors. 00:20:49.750 --> 00:20:51.717 >> [laughter] >> And DeLorean's book. Have you read it? 00:20:51.800 --> 00:20:53.677 >> does. >> Have you read that book? 00:20:53.760 --> 00:20:56.517 >> So, I have years ago I have. It's one of the great all-time book titles and we 00:20:56.600 --> 00:20:59.437 should just pause to say John DeLorean was like what he was like the super 00:20:59.520 --> 00:21:01.237 genius of the car industry. >> Yeah, he was going to be the next 00:21:01.320 --> 00:21:03.717 president of General Motors. >> of General Motors and then later on he 00:21:03.800 --> 00:21:07.637 started his own car company which Which in Back to the Future and then, that 00:21:07.720 --> 00:21:11.397 whole thing collapsed for a variety of reasons. Uh but yeah, he he was like a 00:21:11.480 --> 00:21:14.757 legend. He was like one of the main principal drivers of innovation in the 00:21:14.840 --> 00:21:17.997 car industry. >> Exactly. Lee Iacocca bought lots of these categories. 00:21:18.080 --> 00:21:20.157 >> Yeah. And you got to remember this is in this rate 00:21:20.240 --> 00:21:21.757 >> Sorry, repeat the ti- repeat the title of the book. 00:21:21.840 --> 00:21:23.557 >> On a Clear Day You Can See General Motors. 00:21:23.640 --> 00:21:24.557 >> why was that the title of the book? 00:21:24.640 --> 00:21:29.277 >> Because uh there's a lot of >> [laughter] 00:21:29.360 --> 00:21:31.837 >> It's a very large complex. >> Yeah, complex. Yeah, yeah. 00:21:31.920 --> 00:21:33.197 >> It's like a It's like a nation state. 00:21:33.280 --> 00:21:37.917 >> Yeah, it I mean, really. I mean, it It really I think we say that like sometimes 00:21:38.000 --> 00:21:41.917 almost like flippantly. But these companies are like extension like 00:21:42.000 --> 00:21:43.237 Hyundai is an extension of the state. 00:21:43.320 --> 00:21:44.397 Kia is an extension of the state. 00:21:44.480 --> 00:21:48.077 Volkswagen is literally Volkswagen board members are members of the government. 00:21:48.160 --> 00:21:51.357 So these are extensions of the state in in in almost every uh And there used to 00:21:51.440 --> 00:21:54.077 be an old saying, "What's good for General Motors is good for America." 00:21:54.160 --> 00:21:59.077 >> Right. >> And you cannot understate how important General Motors is to the 00:21:59.160 --> 00:22:00.397 history of the American corporation. 00:22:00.480 --> 00:22:04.197 Sloan's uh My Years at General Motors and Adventures of a White-Collar Man. If 00:22:04.280 --> 00:22:07.077 you run a large engineering organization, you should read that. That 00:22:07.160 --> 00:22:11.277 is the this like you this com- this thing that that we talk as a 00:22:11.360 --> 00:22:13.077 modern corporation didn't just emerge. 00:22:13.160 --> 00:22:17.397 Sloan and Kettering create Kettering is a head of engineering created this with 00:22:17.480 --> 00:22:21.157 this arc you know, with levels and vice presidents and how do you do functional 00:22:21.240 --> 00:22:25.517 and matrix organizations? It's There really is like the source code. 00:22:25.600 --> 00:22:30.397 Comes along John uh you know, comes on DeLorean and he says 00:22:30.480 --> 00:22:32.757 um he writes his he's going to be president 00:22:32.840 --> 00:22:34.957 and he's so fed up with the company. 00:22:35.040 --> 00:22:39.277 But what was controversial was GM was doing really well at the time. GM 00:22:39.360 --> 00:22:43.677 was like a When we say like GM was number one in the Fortune 100, it was 00:22:43.760 --> 00:22:48.277 like number one, two, and three. It was everything and it was seen as the best 00:22:48.360 --> 00:22:51.957 company in America. So somebody to openly criticize the company. And so he 00:22:52.040 --> 00:22:56.117 has a whole he had writes this book as he quits out of out of uh 00:22:56.200 --> 00:22:59.797 how annoyed he was at General Motors for being read uh led. Uh he writes this 00:22:59.880 --> 00:23:02.997 book and then after he like sobers up, he's like, "I don't want that book 00:23:03.080 --> 00:23:05.757 published." [laughter] And so he fights for a years for his 00:23:05.840 --> 00:23:09.597 co-author not to publish the book. The co-author still publishes. So it's a 00:23:09.680 --> 00:23:12.637 real true insight into a large corporation. 00:23:12.720 --> 00:23:15.437 Um incidentally just reading it out even though I've you know worked at GM 20 00:23:15.520 --> 00:23:18.957 some years ago and had you know know a lot about the company. 00:23:19.040 --> 00:23:22.037 And what's shocking is it's not only about GM, most of the major 00:23:22.120 --> 00:23:26.717 manufacturers actually still operate that way on the inside. And so 00:23:26.800 --> 00:23:30.477 the question isn't the the the point I think for everyone to take away isn't 00:23:30.560 --> 00:23:34.637 these people who run these companies are stupid. They're not stupid. They're It's 00:23:34.720 --> 00:23:37.437 kind of like you know when you're selling to the Department of War and 00:23:37.520 --> 00:23:39.477 people say, "Well, why are you doing that?" It's like, "Well, the 00:23:39.560 --> 00:23:41.797 distribution defines the business." 00:23:41.880 --> 00:23:46.237 That like so like the distribution is this is a consumer product. Uh 00:23:46.320 --> 00:23:49.437 this this stat might be out of date, but when I worked in safety systems 20 years 00:23:49.520 --> 00:23:52.837 ago I remember GM used to pound in your head of the top five consumer lawsuits 00:23:52.920 --> 00:23:56.877 in in American history, three are automotive. We got the majority, right? 00:23:56.960 --> 00:23:59.637 So it's like you have to be extremely careful. We had these like weird things 00:23:59.720 --> 00:24:04.237 like inside the company you can It wasn't red, yellow, green. It was like 00:24:04.320 --> 00:24:07.877 purple or like you'd always have those decoder because you know why? Because 00:24:07.960 --> 00:24:11.317 when they go to lawsuit they're like, "You let a safety system that was marked 00:24:11.400 --> 00:24:14.687 red go to production." I was like, "No, it was marked magenta." 00:24:14.770 --> 00:24:18.917 >> [laughter] >> Like so like can you imagine how 00:24:19.000 --> 00:24:23.237 infuriating that is? Every time you're like "What does orange mean? Does 00:24:23.320 --> 00:24:27.677 this mean I have to like So fast forward to you're meeting that system." 00:24:27.760 --> 00:24:30.917 >> Well, for the Ford slogan for a very long time was it 00:24:31.000 --> 00:24:32.157 was quality is job one, right? 00:24:32.240 --> 00:24:33.957 >> Yeah, yeah, yeah. Which figures with safety as well. 00:24:34.040 --> 00:24:37.637 >> Yeah, yeah, exactly. And and that's the one-two punch of automotive. It's 00:24:37.720 --> 00:24:38.917 quality and safety. Quality and safety. 00:24:39.000 --> 00:24:42.997 And quality really becomes because the Japanese really reset that that stage in 00:24:43.080 --> 00:24:45.757 in because the that's a whole separate automotive 00:24:45.840 --> 00:24:48.357 history. We could talk about automotive history for for an hour, but the 00:24:48.440 --> 00:24:53.037 punchline is you have the Silicon Valley company meeting this immovable object. 00:24:53.120 --> 00:24:56.317 There is a parallel universe that cruises out there right now 00:24:56.400 --> 00:24:59.997 even as a part of General Motors. So, I think you always have to take it into 00:25:00.080 --> 00:25:03.877 the context of where the company is, where union negotiations are happening 00:25:03.960 --> 00:25:07.037 literally that year. And if you're the union, you're like, you can't make a 00:25:07.120 --> 00:25:10.157 million dollars for us, but you're funding this thing >> Right. 00:25:10.240 --> 00:25:12.197 >> that's killing people and it's sloppy. 00:25:12.280 --> 00:25:15.557 And so, I'm not saying precisely that's what happened, to be very clear. But, 00:25:15.640 --> 00:25:19.397 that it's a multi-varied problem. My other hot take is, you know, I worked at 00:25:19.480 --> 00:25:22.797 both companies, right? Google Google and General Motors. Those companies are way 00:25:22.880 --> 00:25:24.477 more similar than they are different. 00:25:24.560 --> 00:25:26.037 Way way more similar than they are. 00:25:26.120 --> 00:25:29.757 Literally, people don't need you know this, the Google leveling system is the 00:25:29.840 --> 00:25:33.317 same as the General Motors [laughter] leveling system. And I used to say, I 00:25:33.400 --> 00:25:35.757 used to say you know, this inside of Google meetings. It was like, "Hey, 00:25:35.840 --> 00:25:37.717 actually some of the engineers I knew at General Motors are better than the 00:25:37.800 --> 00:25:41.157 engineers here." And people would look at me like I'm saying there's no God in 00:25:41.240 --> 00:25:44.917 church. It's like, they're like, "How dare you, you metal butt bending 00:25:45.000 --> 00:25:47.157 monkey from Detroit?" >> [laughter] 00:25:47.240 --> 00:25:49.517 >> I was like, "No, actually like making a 00:25:49.600 --> 00:25:52.477 modern combustion engine is extremely 00:25:52.560 --> 00:25:54.477 complex. It's not just like, you know, 00:25:54.560 --> 00:25:59.637 it's it's not simple stuff." And so, the macro point I think is 00:25:59.720 --> 00:26:03.357 it's a bunch of things. I think safety is always at the top top of their top of 00:26:03.440 --> 00:26:07.797 their list. I do think, you know, we've hired lots of Cruise people. I 00:26:07.880 --> 00:26:11.597 think the way they dealt with that specific issue with the government, you 00:26:11.680 --> 00:26:14.477 got to you got to dance a particular way when that happens. And they just didn't 00:26:14.560 --> 00:26:17.237 dance exactly right. And that just gives 00:26:17.320 --> 00:26:21.517 government bureaucrats more ammo to go after. And you're big 00:26:21.600 --> 00:26:24.397 target like General Motors, you got to you know, 00:26:24.480 --> 00:26:29.797 it it reminded you guys ever see that movie like Goodfellas? You know, 00:26:29.880 --> 00:26:32.757 one of the last scenes of the House of the Rising Sun, you know, all the old 00:26:32.840 --> 00:26:35.797 bosses go in the back of the courtroom and they're like, 00:26:35.880 --> 00:26:37.677 And you know, that's what happened. They were like, 00:26:37.760 --> 00:26:42.067 the board was like, "What do we do about Cruise?" They're like, "What can we do?" 00:26:42.150 --> 00:26:47.037 >> [laughter] >> It's like, "Kyle's a GOOD GUY, BUT >> [laughter] 00:26:47.120 --> 00:26:50.117 >> AND THEN IT'S LIKE, cue House of the Rising Sun. 00:26:50.200 --> 00:26:54.157 People are running through the San Francisco just kidding. [laughter] 00:26:54.240 --> 00:26:56.717 Don't don't make that an AI video. 00:26:56.800 --> 00:27:00.277 It's going to get a >> [laughter] >> going to get a mean text from Kyle. 00:27:00.360 --> 00:27:03.077 So, I think I think there is a universe where it would it would survive, but 00:27:03.160 --> 00:27:05.517 it's it's it's it's tough. >> So, then a lot of a lot of what Applied 00:27:05.600 --> 00:27:08.237 Intuition does is kind of as you said, like that dance. It's like how to how to 00:27:08.320 --> 00:27:09.837 be a great partner to these companies. 00:27:09.920 --> 00:27:14.077 >> Exactly. >> Bearing in mind their own real very real issues and constraints. 00:27:14.160 --> 00:27:18.317 >> I I I think GM was also had the topic of business model, right? So, you have 00:27:18.400 --> 00:27:22.277 Cruise was going after the robot taxi concept, but GM makes its profits from 00:27:22.360 --> 00:27:25.357 personal car ownership. And those things can be a bit odd. So, I think that was 00:27:25.440 --> 00:27:27.317 also a bit of the >> Oh, right. 00:27:27.400 --> 00:27:31.557 >> Yeah, and I think it wasn't clear. I mean, the by the way, you know, um 00:27:31.640 --> 00:27:35.797 you actually, of all people, you spoke at YC at 20 in 2013. I was in the 00:27:35.880 --> 00:27:37.917 audience. I was a partner at the time. 00:27:38.000 --> 00:27:41.997 And um you said something which I think is it's it's very like recursive here. 00:27:42.080 --> 00:27:47.717 We're we're feeding each other your own advice. It's uh the key thing in in new 00:27:47.800 --> 00:27:50.597 technology in the new technology business, actually everyone kind of 00:27:50.680 --> 00:27:52.877 figures out the technology. Though, that's still hard. It's still hard 00:27:52.960 --> 00:27:55.917 sometimes to build really complex things. It's when and how you deploy 00:27:56.000 --> 00:27:59.717 them into the market. The when becomes really important. You're 2 years early 00:27:59.800 --> 00:28:02.197 and you're doomed. You're 2 years late, there's too many competitors. You have 00:28:02.280 --> 00:28:06.357 to like hit it at the right spot. And I think it's like like I mean, a 00:28:06.440 --> 00:28:09.517 controversial thing to say is like I actually think Cruise, you know, they 00:28:09.600 --> 00:28:13.477 were certainly moving at a much faster pace than Waymo. They started way behind 00:28:13.560 --> 00:28:16.477 and you're talking about neck and neck when, you know, ultimately the the the 00:28:16.560 --> 00:28:18.637 plug was pulled. So, who knows what 00:28:18.720 --> 00:28:20.917 happens in the long term. Our hypothesis 00:28:21.000 --> 00:28:26.117 in that same equation is actually the distribution, you let the manufacturers 00:28:26.200 --> 00:28:29.517 do that. Like we run self-driving trucks right now in Japan. They carry 00:28:29.600 --> 00:28:32.197 commercial loads. They're they're, you know, they're safety drivers there, but 00:28:32.280 --> 00:28:36.517 they're autonomously running. And but you won't know that because the brand is 00:28:36.600 --> 00:28:39.557 Isuzu. That's the customer. And and why it's so 00:28:39.640 --> 00:28:44.197 good for us to partner with Isuzu in that case is that company's been around 00:28:44.280 --> 00:28:48.197 almost 100 years, right? If I find if I'm not mistaken, uh, pre-pre-World War 00:28:48.280 --> 00:28:52.557 II company, and they are, uh, you know, they know the government. They have test 00:28:52.640 --> 00:28:55.957 tracks. They know safety. They know their own trucks very well. So, when we 00:28:56.040 --> 00:28:58.997 go and provide them with the intelligence and the integration into 00:28:59.080 --> 00:29:00.597 their physical machinery, that's a 00:29:00.680 --> 00:29:04.077 fantastic one-two punch. I think today 00:29:04.160 --> 00:29:08.117 the world is ready to consume AI in the real world. And that's a lot because of 00:29:08.200 --> 00:29:11.397 ChatGPT and Anthropic and all these, you know, everything that's happened. So, 00:29:11.480 --> 00:29:14.477 people are no longer like, "What's a self-driving car?" And this is because 00:29:14.560 --> 00:29:18.717 of Waymo and Tesla. So, the market is ready to consume. And 00:29:18.800 --> 00:29:21.957 I think you just have to meet the market in the way that the best way possible. 00:29:22.040 --> 00:29:25.477 In our view, that has always been you go through some of the people who run the 00:29:25.560 --> 00:29:28.237 economy right now. Whether it's a mining operator, whether it's a Department of 00:29:28.320 --> 00:29:31.757 War, whether it's, uh, the manufacturers. And we work with, you 00:29:31.840 --> 00:29:35.557 know, within each vertical, uh, with with the right partner. But that's a 00:29:35.640 --> 00:29:38.837 fundamentally different view than a Tesla or a Waymo, which are going to be 00:29:38.920 --> 00:29:42.677 vertical. Where we're really playing the horizontal. And I think the way we can 00:29:42.760 --> 00:29:45.517 always think of we think about that our company. We're kind of like a like a 00:29:45.600 --> 00:29:50.117 chip maker. You know, we we we actually look and talk and walk a lot like a 00:29:50.200 --> 00:29:52.797 silicon company, except we don't obviously we don't make chips. But, you 00:29:52.880 --> 00:29:56.757 know, we're our we have design wins and then we have really large long-term 00:29:56.840 --> 00:30:00.557 relationships. And once we're in, we're in. It's really hard to, you know, to 00:30:00.640 --> 00:30:02.437 take us out. So, you need deep trust. 00:30:02.520 --> 00:30:05.797 Our partners have really a lot a lot of deep trust. And we know their markets 00:30:05.880 --> 00:30:09.277 really really well. I think the things that Jensen knows is he knows his 00:30:09.360 --> 00:30:12.477 customers. That's why Nvidia does well beyond the 00:30:12.560 --> 00:30:15.757 fact obviously they make a very complex technology. 00:30:15.840 --> 00:30:20.957 >> So, how are these legacy car companies preparing for the future? Are they 00:30:21.040 --> 00:30:22.677 making or more acquisitions or going to? 00:30:22.760 --> 00:30:24.877 Are they building and, you know, partnering with you? Or they how are 00:30:24.960 --> 00:30:26.957 they going to compete with, you know, tech tech native companies? 00:30:27.040 --> 00:30:29.317 >> It's like saying like how are governments dealing with AI? It's such a 00:30:29.400 --> 00:30:33.237 it's the such a broad topic. Um, and each manufacturer, like even like you 00:30:33.320 --> 00:30:37.237 take Honda, Nissan, Toyota, three Japanese manufacturers with, you know, 00:30:37.320 --> 00:30:40.317 long legacies. They all approach it very differently. They're roughly in a 00:30:40.400 --> 00:30:43.717 spectrum of we're going to build to we're going to buy Um, 00:30:43.800 --> 00:30:48.517 and other both extremes more than ever we're going to buy is the common answer 00:30:48.600 --> 00:30:52.317 because they've been trying. And we've been there the whole time. 00:30:52.400 --> 00:30:55.557 For the folks that are going to build we provide them tools 00:30:55.640 --> 00:30:59.677 and we talk a little bit about our our our new product that we're 00:30:59.760 --> 00:31:03.677 announcing here. And then on the ones that just want to buy we we sell them 00:31:03.760 --> 00:31:07.357 the actual intelligence that goes on the machines. And so we meet the customer 00:31:07.440 --> 00:31:09.117 wherever they're ready in their in their 00:31:09.200 --> 00:31:11.917 journey. The more nuanced version of 00:31:12.000 --> 00:31:15.157 that is you know the reality is like 00:31:15.240 --> 00:31:20.117 every every product is a different product and so the amount of silicon and 00:31:20.200 --> 00:31:24.117 amount of dollars you can put towards it towards sensors what the customer's 00:31:24.200 --> 00:31:28.317 willing to pay all that depends on what actually 00:31:28.400 --> 00:31:32.077 gets in in the long horizon all these things will be fully autonomous. 00:31:32.160 --> 00:31:34.877 But the intermittent steps are very much 00:31:34.960 --> 00:31:39.277 what we saw in the PC. Where you you have the slow step up to 00:31:39.360 --> 00:31:42.317 one day that'll be like now nobody really looks at laptop specs and even 00:31:42.400 --> 00:31:43.997 maybe frankly your phone specs. But 00:31:44.080 --> 00:31:46.917 that's not the case from basically 85 to 00:31:47.000 --> 00:31:51.717 2002 2005 where finally people stop actually specking at all and and and 00:31:51.800 --> 00:31:53.877 then they're really moving to laptops. 00:31:53.960 --> 00:31:57.757 But there's a similar kind of 20 year I think horizon there. 00:31:57.840 --> 00:32:00.717 >> And Brody when you talk about machines and machines becoming intelligent right 00:32:00.800 --> 00:32:04.397 fundamentally a machine is it's a collection of these different components 00:32:04.480 --> 00:32:07.717 that are integrated right and and whoever does that final integration is 00:32:07.800 --> 00:32:11.837 often times the company that puts their their badge on it the brand name but 00:32:11.920 --> 00:32:14.757 many many companies are building technology that goes into that machines. 00:32:14.840 --> 00:32:19.237 And so we now have a bunch of technology components and platforms that can go 00:32:19.320 --> 00:32:22.637 into these machines. And but we also sell the core technology that can be 00:32:22.720 --> 00:32:25.277 used to develop them as well. >> And if you look by the way under the 00:32:25.360 --> 00:32:29.557 hood of a of a dirt mover or like combine or diesel truck they'll have 00:32:29.640 --> 00:32:33.597 Cummins engines in them. But nobody says well because all these guys buy Cummins 00:32:33.680 --> 00:32:37.517 this this means that they're you know whatever Caterpillar's is a good 00:32:37.600 --> 00:32:41.037 company. It's like no, that's just a component that they buy. Uh 00:32:41.120 --> 00:32:44.357 they they had they have a different role. So, if you when you look in the 00:32:44.440 --> 00:32:48.037 any of these verticals, there's it's just a complex web of folks. That's why 00:32:48.120 --> 00:32:52.197 I always say like the the chip kind of analogy actually works quite effectively 00:32:52.280 --> 00:32:56.757 because some none of those companies make chips, but they all buy chips. 00:32:56.840 --> 00:32:59.677 Uh and so I think that that's it's a good way to think about it. 00:32:59.760 --> 00:33:01.637 >> So, self-driving cars, so you know, we've 00:33:01.720 --> 00:33:04.357 all been talking about self-driving cars for like I think the whole thing started 00:33:04.440 --> 00:33:07.717 like around 2012 2005 or something with the DARPA Grand Challenge originally. 00:33:07.800 --> 00:33:10.877 And so and then Google engaged on the program shortly after that. 00:33:10.960 --> 00:33:11.837 >> Yeah, late double O's, yeah. 00:33:11.920 --> 00:33:15.637 >> Late double O's. Um uh so almost 20 you basically around 20 less than 20 years, 00:33:15.720 --> 00:33:18.637 maybe. Um and there've been lots of predictions over the last 20 years of like 00:33:18.720 --> 00:33:22.797 self-driving cars are imminent at any moment. Um so I guess the bad news is 00:33:22.880 --> 00:33:25.277 we're sitting here today and most cars are not self-driving. The good news is 00:33:25.360 --> 00:33:26.637 there are now self-driving cars. 00:33:26.720 --> 00:33:28.517 >> Yeah. >> And so the Waymo cars are driving all 00:33:28.600 --> 00:33:30.837 over, you know, in the places they're deployed. It's become a you know, I 00:33:30.920 --> 00:33:33.717 think people in San Francisco are I think treated now as routine that they 00:33:33.800 --> 00:33:35.917 get into a car. >> think you can call I think Tesla It's 00:33:36.000 --> 00:33:39.557 kind of like the AGI thing. It's like, you know, if we're talking 20 years ago, 00:33:39.640 --> 00:33:43.837 everything we're seeing right now is like mind-blowingly AGI. The the host 00:33:43.920 --> 00:33:45.437 keeps moving. The Tesla stuff's amazing. 00:33:45.520 --> 00:33:49.077 You can look at a bunch of manufacturers. Blue Cruise, Supercruise, 00:33:49.160 --> 00:33:51.797 BMW, Volvo's Pilot. They're all [clears throat] quite impressive 00:33:51.880 --> 00:33:53.597 systems. They're not full self-driving. 00:33:53.680 --> 00:33:57.647 >> Right. >> But yeah. >> Well, it's it's full self-driving X whatever remote 00:33:57.730 --> 00:33:59.477 >> [snorts] >> monitoring is happening. [laughter] Um 00:33:59.560 --> 00:34:02.477 uh the the Tesla uh we have we have a home in in in Los Angeles and if you 00:34:02.560 --> 00:34:05.037 guys may recall, there was a there was a large fire in Los Angeles last 00:34:05.120 --> 00:34:08.117 >> Yeah, yeah, yeah. >> And the power and then the California power grid was 00:34:08.200 --> 00:34:11.117 buckling even before that. So, it actually turns out among the things 00:34:11.200 --> 00:34:13.957 Cybertrucks are good at is they're they're very good batteries >> Yeah. 00:34:14.040 --> 00:34:15.397 >> uh for powering your house. 00:34:15.480 --> 00:34:17.157 >> Yeah. >> Um and so literally we have a Cybertruck 00:34:17.240 --> 00:34:20.157 as our backup battery for the house. And as of last year, whatever the FSD 00:34:20.240 --> 00:34:22.917 release, I forget the exact one, but there was one where it like at least a 00:34:23.000 --> 00:34:24.557 lot of people thought it like really turned the corner. 00:34:24.640 --> 00:34:26.197 >> 14, yeah. >> And like that thing drives people I 00:34:26.280 --> 00:34:28.957 talked to somebody yesterday uh talked to somebody yesterday who has a Model Y 00:34:29.040 --> 00:34:31.677 who let the let the thing let the thing do the full the full route all the way 00:34:31.760 --> 00:34:32.877 up highway one through Big Sur. 00:34:32.960 --> 00:34:35.797 >> Yeah, I think mean disengagement like uh the the 00:34:35.880 --> 00:34:39.957 mean time and like miles per disengagement are are 00:34:40.040 --> 00:34:43.677 really high and I think miles is like in the thousands. Yeah, which is like very 00:34:43.760 --> 00:34:45.197 impressive. >> Yeah, I know the Big Sur for people who 00:34:45.280 --> 00:34:49.197 have driven the Big Sur highway one Big Sur like that's a that's a that's a 00:34:49.280 --> 00:34:52.957 stressful drive. He said it was great the the whole anyway. So and I wouldn't 00:34:53.040 --> 00:34:56.157 have been talking to him had it not been >> [laughter] 00:34:56.240 --> 00:34:59.117 >> would have gone right off right off >> Cuz he unbolted the steering wheel so 00:34:59.200 --> 00:35:01.197 >> right off right off the >> cliff. 00:35:01.280 --> 00:35:05.517 >> So and then you know Tesla's rolling out their robo-taxi, you know, that it's 00:35:05.600 --> 00:35:08.797 it's starting starting to show up in in the wild. So so on the one hand those 00:35:08.880 --> 00:35:12.117 exist on the other hand, you know, 99.999999% 00:35:12.200 --> 00:35:13.597 of cars are are still not self-driving. 00:35:13.680 --> 00:35:15.997 And then I would say maybe just one other would be the self-driving trucks. 00:35:16.080 --> 00:35:19.077 There's been this recurring kind of panic in the press of like this trucks 00:35:19.160 --> 00:35:21.637 become self-driving and the employment, you know, all these truck drivers will 00:35:21.720 --> 00:35:25.197 be out of a job. And sitting here today, I don't think I don't know is there 00:35:25.280 --> 00:35:27.597 are there any trucks on the road that are self-driving that don't have at 00:35:27.680 --> 00:35:31.037 least a safety driver in the truck? And I think the answer is probably still 00:35:31.120 --> 00:35:33.037 >> Yeah, Silver if you So let's let's split 00:35:33.120 --> 00:35:36.317 let's split the there's multiple points that were brought up here. One is on the 00:35:36.400 --> 00:35:38.837 let's say personally owned vehicles and 00:35:38.920 --> 00:35:40.957 why are they not more ubiquitous? 00:35:41.040 --> 00:35:45.357 The part of that is the manufacturers are not good at deploying technology. 00:35:45.440 --> 00:35:48.957 Part of that is they're they're they want to be safety conscious but most of 00:35:49.040 --> 00:35:51.117 it is cost cost cost. 00:35:51.200 --> 00:35:53.797 The the what you're seeing in China, 00:35:53.880 --> 00:35:59.077 which is a China's kind of a different EV ecosystem mainly because they don't 00:35:59.160 --> 00:36:01.797 care about profits. When you're talking about business that doesn't care about 00:36:01.880 --> 00:36:05.637 profits, it changes the entire calculus of the entire industry that that doesn't 00:36:05.720 --> 00:36:08.757 care. But what you're seeing is you're seeing 00:36:08.840 --> 00:36:12.557 L2++ systems. So we can simplify the entire self-driving conversation to is 00:36:12.640 --> 00:36:15.797 there a driver behind the steering wheel? Right. So this is a driver behind 00:36:15.880 --> 00:36:19.357 the steering wheel still there but generally like Tesla drives everywhere. 00:36:19.440 --> 00:36:21.557 They're like sub $1,000. There's an 00:36:21.640 --> 00:36:26.477 aggressive That's that's chip, sensors, the package, the software, everything. 00:36:26.560 --> 00:36:29.917 We, you know, anticipate that there there's a very aggressive Once you get 00:36:30.000 --> 00:36:33.117 to like 500, the automotive OEMs will actually subsidize it for free. They'll 00:36:33.200 --> 00:36:35.877 just give it to you. This happened in nav systems. If you guys remember nav 00:36:35.960 --> 00:36:39.117 systems, it used to be a big thing where you'd pay four grand, 3,500, to get a 00:36:39.200 --> 00:36:42.157 nav system, and then suddenly it became free and it just became default. I think 00:36:42.240 --> 00:36:46.517 that'll happen. The There's a weird thing which is like actually getting 00:36:46.600 --> 00:36:49.157 into a subset of your cars costs X 00:36:49.240 --> 00:36:53.597 dollars, and to get into all the cars costs X plus just a small incremental 00:36:53.680 --> 00:36:57.637 amount. Because there's a fixed cost and the way that how many vehicles and the 00:36:57.720 --> 00:37:00.957 way the assembly line comes in the way you have homologation, all these testing 00:37:01.040 --> 00:37:04.037 regimes, all this stuff. So, I think you'll have wait, wait, wait, and then a 00:37:04.120 --> 00:37:08.277 lot. >> Yeah. >> Every single OEM without exception, even 00:37:08.360 --> 00:37:10.117 the lowest dollar OEMs are working on an 00:37:10.200 --> 00:37:15.917 FSD competitor. So, it'll come. But it's just like, you know, with 00:37:16.000 --> 00:37:18.677 a good analogy to think about self-driving in the personally owned 00:37:18.760 --> 00:37:23.797 ecosystem is mobile phones. We had We had the the satellite phones, 00:37:23.880 --> 00:37:26.877 then we had the Qualcomm, you know, brick phones, then we had the Motorola 00:37:26.960 --> 00:37:31.677 Razrs, and from, you know, the late '90s to the late to double zeros, there was 00:37:31.760 --> 00:37:33.397 like, "When's mobile going to come?" 00:37:33.480 --> 00:37:36.397 There was a huge like, and then it it comes and we 00:37:36.480 --> 00:37:39.917 by by '07 from the iPhone launch, like 00:37:40.000 --> 00:37:44.877 it's like four years when you get Uber, Instagram, WhatsApp, Snapchat. Those are 00:37:44.960 --> 00:37:48.757 the killer applications. So, I think there's a very, very similar kind of 00:37:48.840 --> 00:37:52.677 wait, wait, wait, and then it's just basically ubiquitous in every vehicle. 00:37:52.760 --> 00:37:54.997 If you had to ask me for what that number is, 00:37:55.080 --> 00:37:58.837 28 SOP, 29 start of production, 29, 30, 00:37:58.920 --> 00:38:02.717 and then by the early '30s, then it'll start becoming very cheap to free. 00:38:02.800 --> 00:38:06.037 >> I think routinely by by the early '30s, you would just you buy a car and you just 00:38:06.120 --> 00:38:09.717 assume it's self-driving. >> Exactly. Or it has the driver in seat 00:38:09.800 --> 00:38:11.677 L2++ system, being very specific. 00:38:11.760 --> 00:38:14.997 >> Like like Cybertruck or Tesla the Tesla equivalent of what Tesla people have 00:38:15.080 --> 00:38:18.077 what Tesla owners have today. >> Today, exactly. Yeah, it'll be 00:38:18.160 --> 00:38:20.517 default. So, then the question then the 00:38:20.600 --> 00:38:23.077 the the you know, the the other side of 00:38:23.160 --> 00:38:26.437 this is why don't we have a bunch of 00:38:26.520 --> 00:38:30.437 Waymos everywhere. Specifically, Waymo has a different 00:38:30.520 --> 00:38:32.157 technology. Without getting into the 00:38:32.240 --> 00:38:36.077 nuances here, but Tesla and many of the 00:38:36.160 --> 00:38:39.677 Chinese and applied were very much in this 00:38:39.760 --> 00:38:41.557 end-to-end model architecture. 00:38:41.640 --> 00:38:44.397 This is a new way of doing self-driving. 00:38:44.480 --> 00:38:48.317 Waymo, for the lack of a better word, is not that. It's not doesn't mean they're 00:38:48.400 --> 00:38:51.757 not learned. It doesn't It just It's not one end-to-end system. It's not one 00:38:51.840 --> 00:38:55.677 monolithic model. One of the proclivities of that of their 00:38:55.760 --> 00:38:57.717 approach is it does depend on HD maps. 00:38:57.800 --> 00:39:02.077 Therefore, there is a geofencing concept. I think Waymo's trying hard to 00:39:02.160 --> 00:39:05.357 remove that bottleneck so they can expand geographically faster, but the 00:39:05.440 --> 00:39:09.037 reality of today isn't there. The other thing is when you have researchers, 00:39:09.120 --> 00:39:12.437 which they which Waymo really was coming out of 00:39:12.520 --> 00:39:14.757 an Alphabet research organization, they 00:39:14.840 --> 00:39:19.637 didn't put commercial constraints. So, the sensors are bespoke and expensive. 00:39:19.720 --> 00:39:23.477 The cars and the then the compute that are in there, they're they're just not 00:39:23.560 --> 00:39:27.557 economically feasible. And they've tried a lot to get that down, but it's kind of 00:39:27.640 --> 00:39:30.317 like it's a lot easier to go from something that's really cheap and make 00:39:30.400 --> 00:39:35.317 it more uh you know, more featureful than something that's overbuilt and then 00:39:35.400 --> 00:39:39.197 trying to trim and make it really really cheap. And that's the big debate. Who's 00:39:39.280 --> 00:39:43.877 going to get there first? Tesla with full self-driving or Waymo with cost and 00:39:43.960 --> 00:39:47.157 geographic ubiquity? Uh but you know what we're not debating 00:39:47.240 --> 00:39:48.797 about? Is it going to happen? >> Right. 00:39:48.880 --> 00:39:49.717 >> You know what we're not debating about? 00:39:49.800 --> 00:39:52.837 Like is is is uh is there a big technical breakthrough that needs to 00:39:52.920 --> 00:39:56.677 happen? None of those things. So, now we're clearly in the engineering side of 00:39:56.760 --> 00:40:01.797 self-driving, which is just this grind down to uh like dollar per mile 00:40:01.880 --> 00:40:04.877 efficiency. And the moment that it's cheap, guess what? All the OEMs are 00:40:04.960 --> 00:40:06.317 smart, they'll just they just adopt it. 00:40:06.400 --> 00:40:10.437 It's not that OEMs are resistant because they don't think consumers want it or 00:40:10.520 --> 00:40:14.477 they don't understand the technology, it's because they want a 00:40:14.560 --> 00:40:16.277 price envelope, which allows them to 00:40:16.360 --> 00:40:18.917 keep their thin razor-thin margins and 00:40:19.000 --> 00:40:22.197 at a scale, which is deployed across 100 00:40:22.280 --> 00:40:25.797 plus countries in V1. And so, if you're 00:40:25.880 --> 00:40:29.917 just doing a small deployment, it's very different. And I think And the last 00:40:30.000 --> 00:40:32.397 thing I would say is the buyer of a 00:40:32.480 --> 00:40:36.037 Subaru or a buyer of a Suzuki have very 00:40:36.120 --> 00:40:40.597 different brand expectations than a buyer of a Tesla. And so, including the 00:40:40.680 --> 00:40:46.237 age of the consumer and what they think will happen or won't happen. Um so, that 00:40:46.320 --> 00:40:50.157 that also reason. So, if you're Suzuki, you're like, "Well, my buyer is like not 00:40:50.240 --> 00:40:53.477 doesn't want this stuff, so I'm not going to jam it into the car." It's not 00:40:53.560 --> 00:40:56.157 because they're not like technically competent. This is a different area. 00:40:56.240 --> 00:40:58.877 >> When do you think When do you think it'll be routine? Let's say the the 200 00:40:58.960 --> 00:41:02.917 biggest American cities, like um when will it be routine to walk [snorts] 00:41:03.000 --> 00:41:05.557 outside and you just you just take it for granted that a robot taxi can come 00:41:05.640 --> 00:41:11.757 pick you up? >> Uh it's 26 now. Um I mean, certainly by 30. 00:41:11.840 --> 00:41:14.797 >> Sorry. Okay. >> Yeah, certainly by 30. And I would say 00:41:14.880 --> 00:41:18.997 the the big like variable there really is like 00:41:19.080 --> 00:41:23.757 cuz you What Waymo will say is that the dollars and cents per city already work. 00:41:23.840 --> 00:41:26.397 And it's like, "Well, a company that has basically unlimited capital, why are 00:41:26.480 --> 00:41:28.917 they not already in 200 cities?" But then you see their launch schedule's 00:41:29.000 --> 00:41:33.477 pretty aggressive. You're like, "That can that can get there." So, 00:41:33.560 --> 00:41:35.397 maybe it has been aggressive. I'd say 28. 00:41:35.480 --> 00:41:36.517 >> Yeah, okay. >> Like 2 years. 00:41:36.600 --> 00:41:38.597 >> I I would say I would say available in 00:41:38.680 --> 00:41:42.317 30 but routine in maybe like 32 or 33. 00:41:42.400 --> 00:41:44.277 >> Cuz there's a scale up, there's a volume. 00:41:44.360 --> 00:41:49.117 >> And then also if if you you got you know, you're living in LA, so like 00:41:49.200 --> 00:41:52.557 5 years ago, I'd go to LA, people would be like, "What's Applied Intuition? I 00:41:52.640 --> 00:41:54.037 don't know what self-driving cars are." 00:41:54.120 --> 00:41:57.997 Uh in the last couple years, now they all know self-driving and some of them 00:41:58.080 --> 00:42:01.117 even know Applied Intuition because they know from the other manufacturers. 00:42:01.200 --> 00:42:05.437 Uh I think you fast forward another 2 to 4 years, everybody knows it. Now, does 00:42:05.520 --> 00:42:08.197 that mean everyone's taking Waymos exclusively? 00:42:08.280 --> 00:42:12.397 The answer is no, actually. Now, there is a a huge If you look at the numbers, if 00:42:12.480 --> 00:42:15.797 you're Uber, you got to be scared. I mean, they're just eating into into ride 00:42:15.880 --> 00:42:19.957 sharing. Um Yeah, but to get 100% equity, I mean, 00:42:20.040 --> 00:42:22.357 that's that's that's another Yes, to be extremely cheap. 00:42:22.440 --> 00:42:24.397 >> And what about long-haul trucking? 00:42:24.480 --> 00:42:27.397 >> So long So that's So that's what we That's So that's the pa- passenger side. 00:42:27.480 --> 00:42:30.037 The long-haul trucking, completely different economics, completely 00:42:30.120 --> 00:42:34.157 different um uh business model. Um there 00:42:34.240 --> 00:42:37.317 are many companies right now. I would say probably north of five that are 00:42:37.400 --> 00:42:41.597 running long-haul trucks with drivers carrying loads between America and 00:42:41.680 --> 00:42:44.957 China. I think China's probably getting into double digits. So it's there, but 00:42:45.040 --> 00:42:46.637 the reason uh you don't know it and the 00:42:46.720 --> 00:42:51.997 reason it's not top of mind is it's not a consumer product. And unlike uh on the 00:42:52.080 --> 00:42:53.637 Waymo and Tesla side, where investors 00:42:53.720 --> 00:42:56.557 are willing to essentially give you uh 00:42:56.640 --> 00:42:59.877 you know, some um market cap uh you 00:42:59.960 --> 00:43:03.877 know, uh adjustment for the potential of 00:43:03.960 --> 00:43:08.077 the the They say the the trucking business is like, you know, made What's the word? 00:43:08.160 --> 00:43:10.037 >> You buy a car with your heartstrings. 00:43:10.120 --> 00:43:11.317 You buy a truck with a calculator. 00:43:11.400 --> 00:43:12.437 >> Yeah. It's a calculator [laughter] 00:43:12.520 --> 00:43:15.317 business. And they And so it's like pure 00:43:15.400 --> 00:43:19.517 dollars and cents. And so I think um 00:43:19.600 --> 00:43:23.077 you as the provider of self-driving trucks, if you're doing the whole thing, 00:43:23.160 --> 00:43:25.237 like some of the companies are, which we're not, 00:43:25.320 --> 00:43:28.397 you have to show every mile I'm going to I'm going to save you this many dollars. 00:43:28.480 --> 00:43:32.277 And it's like, for sure, for sure, for sure, cuz the buyer's unsophisticated. 00:43:32.360 --> 00:43:35.317 And they're just like, well, I already got a staff that can drive. And it's 00:43:35.400 --> 00:43:38.637 like, and they're like they're just not inclined. 00:43:38.720 --> 00:43:41.797 Now, where we're playing in Japan's not random that we're doing trucking in 00:43:41.880 --> 00:43:45.477 Japan. There's a massive labor shortage today, and there's an imploding 00:43:45.560 --> 00:43:49.917 demographic uh uh you know, a situation. And so there's 00:43:50.000 --> 00:43:53.237 a demand from almost every sector, and that's why we've we've picked that 00:43:53.320 --> 00:43:57.077 market to to to to really grow. But I think like 00:43:57.160 --> 00:44:01.197 you can take like even more uh obscure like when we'll all 00:44:01.280 --> 00:44:04.757 uh quarries, you know, literally like where you would you know, you're you're 00:44:04.840 --> 00:44:07.837 you're moving cement, you're moving you're moving dirt. Uh 00:44:07.920 --> 00:44:11.117 not quarries, q u e a r s q u a r r y. 00:44:11.200 --> 00:44:13.597 >> Yeah. Quarries. Quarries. >> Those are rock rock stone. 00:44:13.680 --> 00:44:16.877 >> Rock stone cement. When when are those? 00:44:16.960 --> 00:44:20.277 I can tell you the people who own those things and run those things wanted it 00:44:20.360 --> 00:44:22.797 today. >> Right. >> So it's literally then you don't you 00:44:22.880 --> 00:44:25.637 don't have a We're literally we can't make this stuff fast enough. 00:44:25.720 --> 00:44:28.557 >> Right. >> Um the macro point though that people 00:44:28.640 --> 00:44:30.717 don't talk about I think all this stuff's going to happen. 00:44:30.800 --> 00:44:32.917 >> Right. >> It'll happen pretty soon. 00:44:33.000 --> 00:44:34.917 >> It'll happen fairly soon. But the macro 00:44:35.000 --> 00:44:37.677 point that in in legislation and and and 00:44:37.760 --> 00:44:40.677 kind of in in the in the kind of 00:44:40.760 --> 00:44:42.877 economics the political economy of of 00:44:42.960 --> 00:44:47.757 this conversation is AI is really you see you have this big pushback in 00:44:47.840 --> 00:44:49.717 digital AI because we're like counselor like I don't know what's going to happen 00:44:49.800 --> 00:44:54.317 to my job and VCs I'm sure all of your associates are very scared. 00:44:54.400 --> 00:44:56.797 But like in our universe >> They're debating whether they need us. 00:44:56.880 --> 00:44:59.517 >> Yeah yeah yeah. In our in our universe it's the other 00:44:59.600 --> 00:45:03.157 way around. It's like you They're literally I'll I'll I'll meet these you 00:45:03.240 --> 00:45:06.997 know operators and they're like We'll give you everything. Like if you 00:45:07.080 --> 00:45:08.237 can do this we'll give you everything. 00:45:08.320 --> 00:45:11.197 So then it's just up to us to like get there as as you know aggressively and 00:45:11.280 --> 00:45:14.077 physically as possible. >> Well you know the the fear for a long time has been for 00:45:14.160 --> 00:45:16.957 trucking for some reason triggers the at least the press's imagination on like 00:45:17.040 --> 00:45:19.837 you know sort of apocalyptic levels of job loss. Like will there 00:45:19.920 --> 00:45:21.357 >> But that's it's so wrong. >> Go ahead. 00:45:21.440 --> 00:45:25.277 >> There's not There's not enough truck drivers and guess what? Nobody wants to 00:45:25.360 --> 00:45:27.557 freaking be a truck driver. >> Why is that? Let me explain. 00:45:27.640 --> 00:45:31.277 >> Because it's a terrible job. It's like You're like you're like you're like you're 00:45:31.360 --> 00:45:33.717 >> I I by the way I grew up my the main feature of the town where I grew up was 00:45:33.800 --> 00:45:37.197 a truck stop so I >> It's like >> But yeah yeah yeah. Why is truck Why is 00:45:37.280 --> 00:45:38.357 truck driving not attractive to you? 00:45:38.440 --> 00:45:39.957 >> asking me it's you know what this is? 00:45:40.040 --> 00:45:43.717 This is like you know talking to my kid who's like well why can't I put my hand 00:45:43.800 --> 00:45:47.677 on the stove? It's like because it's going to burn your hand. 00:45:47.760 --> 00:45:49.637 >> [laughter] >> It's like but why? It's like after the 00:45:49.720 --> 00:45:53.567 third why it's like Come on buddy let's do this. 00:45:53.650 --> 00:45:55.637 >> [laughter] >> Yeah let's [clears throat] 00:45:55.720 --> 00:45:58.317 >> So what So what what's hard about >> This is make sure everybody knows I did 00:45:58.400 --> 00:45:59.437 not do that with [laughter] my son. 00:45:59.520 --> 00:46:01.397 >> Yes yes. So what's hard Why why is being 00:46:01.480 --> 00:46:03.757 a truck driver a difficult job or why would kids not want to do it when they 00:46:03.840 --> 00:46:06.277 grow up? >> So, what the let me use a parallel 00:46:06.360 --> 00:46:08.677 analogy which is very clear and then you can 00:46:08.760 --> 00:46:12.837 what what you know people will say like nobody wants to work anymore and they 00:46:12.920 --> 00:46:16.197 say well, you know, McDonald's has all these job openings. No, no, actually 00:46:16.280 --> 00:46:19.797 what it is is those people that used to work at McDonald's now DoorDash and Uber. 00:46:19.880 --> 00:46:21.797 >> Right. >> Because it's better for them because 00:46:21.880 --> 00:46:24.637 they can open they can still start their hours and end their hours and they don't 00:46:24.720 --> 00:46:26.877 have to there's no boss and they don't have to 00:46:26.960 --> 00:46:30.437 like stand on their feet and they can surf their phone in between, you know, 00:46:30.520 --> 00:46:32.357 orders and they don't like that's the 00:46:32.440 --> 00:46:35.157 reason. It's not random. The market is 00:46:35.240 --> 00:46:37.957 efficient and so in the truck driving 00:46:38.040 --> 00:46:40.317 example, why does somebody not want to 00:46:40.400 --> 00:46:43.077 be away from their family for 00:46:43.160 --> 00:46:45.837 4 to 8 days in a row doing long haul 00:46:45.920 --> 00:46:48.357 trucking? The more sharp example is in 00:46:48.440 --> 00:46:50.517 Australia, why don't people want to 00:46:50.600 --> 00:46:53.557 go literally by a plane to go to a mine 00:46:53.640 --> 00:46:58.397 and work on or you go off offshore oil rigs. Those jobs exist. If 00:46:58.480 --> 00:47:03.117 you want a job that pays six figures, they exist. Even with such lucrative pay 00:47:03.200 --> 00:47:06.557 packages, it's not enough cuz people are like, you know what? 00:47:06.640 --> 00:47:10.317 I like kind of being around my family and I'm willing to take an incremental 00:47:10.400 --> 00:47:12.637 decrease in cost and and how much money 00:47:12.720 --> 00:47:15.197 I make and and then also like I think 00:47:15.280 --> 00:47:18.197 today more than ever things like 00:47:18.280 --> 00:47:23.317 back pain and like being exposed to the sun and cancer and people that care 00:47:23.400 --> 00:47:26.317 about that's now a part of the >> This is the thing so tell me if I I have 00:47:26.400 --> 00:47:28.437 this right but I I believe it's because I think commercial long haul truck 00:47:28.520 --> 00:47:33.157 drivers die at life expectancy 10 years less than their peers and I think it's a 00:47:33.240 --> 00:47:37.157 consequence of several things. So, one is some combination of nutrition and 00:47:37.240 --> 00:47:40.077 sleep. It's you know, it's a basically, you know, >> It's yeah, it's 00:47:40.160 --> 00:47:43.157 >> It's very difficult It's very difficult to eat well and exercise and 00:47:43.240 --> 00:47:44.437 >> What's your sleep score if you're a long 00:47:44.520 --> 00:47:47.397 haul trucker? Let me guess, they don't eat sleep on that on that screen. 00:47:47.480 --> 00:47:50.237 >> Exactly and so like obesity and then heart disease, hypertension and so forth 00:47:50.320 --> 00:47:52.717 are all very high one and then two is I think the 00:47:52.800 --> 00:47:56.117 vibration is very difficult stress in the the and then the third is you 00:47:56.200 --> 00:47:59.317 mentioned cancer but I think it's the I think they have truck drivers have like 00:47:59.400 --> 00:48:01.037 a much higher rate of melanoma on their left arm. 00:48:01.120 --> 00:48:03.917 >> Exactly. Yeah, there's photos of like a truck driver who's been driving for 30 00:48:04.000 --> 00:48:05.997 years one half of their face the other half of their face cuz it's exposed to 00:48:06.080 --> 00:48:10.397 the sun. A more interesting even more stark stat, 00:48:10.480 --> 00:48:15.557 mining is 1% of the labor pool globally 8% of work related fatalities. Do you 00:48:15.640 --> 00:48:18.957 think people are rushing to work in mines when they hear stats like this? 00:48:19.040 --> 00:48:23.677 Most major mines have a fatality regularly which means once twice a year 00:48:23.760 --> 00:48:27.117 three times a year. And if you ever visit a mine you'll see that everything 00:48:27.200 --> 00:48:31.397 is based around safety because once you experience one of your co-workers dying 00:48:31.480 --> 00:48:33.397 then you're like what am I doing here? 00:48:33.480 --> 00:48:36.997 >> Yeah. >> Like there's other jobs I can take and so 00:48:37.080 --> 00:48:41.397 it's I I understand you're trying to enumerate for the audience like but 00:48:41.480 --> 00:48:45.437 these are not good jobs. And the best evidence is this this is not a 00:48:45.520 --> 00:48:49.277 mining podcast. This is not a podcast about hey long haul trucking is so 00:48:49.360 --> 00:48:51.717 great. It they're just not attractive jobs. 00:48:51.800 --> 00:48:54.317 >> even trucker even truckers don't want their kids to become truckers like it's 00:48:54.400 --> 00:48:57.437 it's a very for that reason they would you know they want their kids to be in a 00:48:57.520 --> 00:49:01.757 at the very least like safe safer safer line of work. But do the do the 00:49:01.840 --> 00:49:04.317 notwithstanding all that do they how long will there be do you think there'll 00:49:04.400 --> 00:49:07.557 be safety drivers in long haul trucks that are that are self-driving? Or or or 00:49:07.640 --> 00:49:10.357 let's say other even just somebody in the cab to deal with what happens when 00:49:10.440 --> 00:49:12.277 they arrive? >> We we know multiple companies that have 00:49:12.360 --> 00:49:16.157 driver out goals right now. So like they're they're working to get drivers 00:49:16.240 --> 00:49:20.547 out right now. You know, without going into our own details. 00:49:20.630 --> 00:49:22.837 >> [laughter] >> To be honest with you it's not long. 00:49:22.920 --> 00:49:26.477 We're talking we're talking a few years and >> I think on the long end. 00:49:26.560 --> 00:49:29.477 >> Yeah, on the long end. And the thing is it's it's there there's a software 00:49:29.560 --> 00:49:33.197 technology thing which is one part of the problem, but the other part is it's 00:49:33.280 --> 00:49:36.037 the redundancies that you need in hardware and the validation necessary 00:49:36.120 --> 00:49:40.037 for those redundancies. And in many cases that can actually be a long pull. 00:49:40.120 --> 00:49:43.197 It's like oh there productionizing a fully redundant 00:49:43.280 --> 00:49:46.957 steering system fully redundant braking system that's that's not in high that's 00:49:47.040 --> 00:49:48.877 not in high volume production yet and once you get that in high volume 00:49:48.960 --> 00:49:51.557 production now you get the quality up and then that's validated and now you 00:49:51.640 --> 00:49:53.237 can actually do these. Need the >> price down. 00:49:53.320 --> 00:49:55.517 >> Exactly. >> Do do you guys do do you like it's the 00:49:55.600 --> 00:49:56.877 you know these little delivery robots? 00:49:56.960 --> 00:49:59.477 Like is that you do you see a world where there's a billion of those running around? 00:49:59.560 --> 00:50:03.637 >> Yeah, I think so. I mean the uh the the the product that we're announcing uh I 00:50:03.720 --> 00:50:06.837 think it's probably come out around with this time is called uh Dana. So there's 00:50:06.920 --> 00:50:10.197 you can just simplify everything that Applied Intuition does into two buckets, 00:50:10.280 --> 00:50:14.637 which is the uh we've been talking mostly about the models that go on the 00:50:14.720 --> 00:50:18.717 machines. Then there's we say onboard software or onboard AI. Then there's 00:50:18.800 --> 00:50:23.677 offboard AI. This is the tools to design and develop these same systems. The the 00:50:23.760 --> 00:50:27.237 the models that actually go on the machines. Our uh you know vision for 00:50:27.320 --> 00:50:28.997 that is in the the delivery robot is a 00:50:29.080 --> 00:50:32.357 great example is like a high school kid 00:50:32.440 --> 00:50:36.077 or a middle schooler they can make iPhone apps they should be able to make 00:50:36.160 --> 00:50:38.157 autonomous systems. So why can't they? 00:50:38.240 --> 00:50:41.597 Just ask that very simple question. Why can't a 00:50:41.680 --> 00:50:45.197 ninth grader make a delivery robot in their in their home? Well, they don't 00:50:45.280 --> 00:50:50.597 have the the actual environment that they would first develop the scenarios in. They 00:50:50.680 --> 00:50:54.477 would define the requirements. Hey, I want this robot to go on my high school 00:50:54.560 --> 00:50:56.997 campus around these let's say four buildings. 00:50:57.080 --> 00:51:00.957 Uh then how Okay, now that you do define the the requirements then you have the 00:51:01.040 --> 00:51:05.037 scenarios get made. What are all the scenarios that I can that can uh that 00:51:05.120 --> 00:51:07.517 can be made by using let's say satellite image of 00:51:07.600 --> 00:51:10.237 the high school. Uh then now you have to 00:51:10.320 --> 00:51:13.157 train the robot. So you need some data. 00:51:13.240 --> 00:51:16.397 Where do you get that data? There's maybe enough publicly available data 00:51:16.480 --> 00:51:20.237 that can actually train a fairly uh rudimentary robot. You know, you got 00:51:20.320 --> 00:51:23.957 that data from online, maybe YouTube videos, couple of other places. Suddenly 00:51:24.040 --> 00:51:27.277 this robot's not doing Now you need to deploy it onto the actual machine. So 00:51:27.360 --> 00:51:30.797 then you deploy it onto the machine and then the robot runs into the wall. Okay, 00:51:30.880 --> 00:51:33.997 what happened there? The loop closes. That platform for 00:51:34.080 --> 00:51:37.797 designing and developing is what we're launching. It's called Dana, 00:51:37.880 --> 00:51:40.517 uh which is the street that Applied Intuition is >> [laughter] 00:51:40.600 --> 00:51:43.957 >> uh headquartered on. Uh and uh and 00:51:44.040 --> 00:51:45.917 our our our you know, this comes from 00:51:46.000 --> 00:51:47.917 our tooling background. And if you look 00:51:48.000 --> 00:51:51.397 at like how tooling has changed in the 00:51:51.480 --> 00:51:55.197 digital AI world, if you look at like what Claude did to all you you all 00:51:55.280 --> 00:52:00.317 remember like, you know, from Mixpanel to, you know, GitLab, GitHub, all these 00:52:00.400 --> 00:52:04.397 Now everything has moved into a very different almost IDE, frankly speaking. 00:52:04.480 --> 00:52:07.397 We think the same thing's going to happen in the physical world. And so, 00:52:07.480 --> 00:52:09.477 uh that that that's yeah, that's what we're that's what we're building. That's 00:52:09.560 --> 00:52:11.357 what we're that's what we built and that's what we're launching in. We 00:52:11.440 --> 00:52:15.557 already use it in-house to develop our autonomy system, which is you know, and 00:52:15.640 --> 00:52:20.237 we're working on the most kind of scaled complex systems on the planet in all 00:52:20.320 --> 00:52:22.437 these different verticals. So, we're pretty confident that it's actually 00:52:22.520 --> 00:52:25.957 quite useful. And we've seen massive productivity gains. 00:52:26.040 --> 00:52:30.637 Uh but also, you know, we think like other companies will use this to to 00:52:30.720 --> 00:52:33.637 build our own systems cuz it gets to that mission that 00:52:33.720 --> 00:52:35.357 a billion intelligent machines. 00:52:35.440 --> 00:52:39.237 >> Yeah, fundamentally our Dana is our agentic platform for physical AI. And 00:52:39.320 --> 00:52:43.357 and everything that we've built and developed over the past nearly a decade, 00:52:43.440 --> 00:52:44.917 every tool, every 00:52:45.000 --> 00:52:47.477 >> technique that's available in Dana. And 00:52:47.560 --> 00:52:52.517 it's very actually easy to use with the agentic interface. And so, workflows 00:52:52.600 --> 00:52:54.357 that used to maybe take 00:52:54.440 --> 00:52:59.677 days or weeks to run, you can now run those in in minutes in many cases. And 00:52:59.760 --> 00:53:02.237 this just lowers the barrier to entry to building these systems. And so, 00:53:02.320 --> 00:53:04.797 >> And we're just lowering the bar of like, 00:53:04.880 --> 00:53:07.877 you know, what it means to develop an autonomous system. Autonomy is still 00:53:07.960 --> 00:53:13.317 actually quite in the scope of software is quite exotic. It's not because of the 00:53:13.400 --> 00:53:16.717 things that we've talked about. And we've just brought that down very very 00:53:16.800 --> 00:53:20.477 aggressively. And it's kind of like, you know, the old adage of like how do you 00:53:20.560 --> 00:53:25.037 make a great product in software? It's like you either increase safety, 00:53:25.120 --> 00:53:28.957 convenience, or cost. And we want to try to do all three of those things with 00:53:29.040 --> 00:53:31.837 with Dana. And our hope is just like you said, 00:53:31.920 --> 00:53:36.317 like, you know, uh kids can develop robots for their for 00:53:36.400 --> 00:53:39.637 their own use. And and that extends to humanoids. So, we're not just talking 00:53:39.720 --> 00:53:44.357 about like land-based systems or or, you know, the ones that are so you can 00:53:44.440 --> 00:53:49.277 humanoids you can do drones. The fact that right now writing drone software 00:53:49.360 --> 00:53:51.237 and deploying it 00:53:51.320 --> 00:53:54.797 quite obscure and almost hobbyist. 00:53:54.880 --> 00:53:58.677 We want to just make that absolutely like that, you know, 00:53:58.760 --> 00:54:01.037 maybe not child's play but like teenager play. 00:54:01.120 --> 00:54:03.637 >> So this one still world like just fly like a lot more experimentation and 00:54:03.720 --> 00:54:06.597 entrepreneurship and like agriculture >> everything 00:54:06.680 --> 00:54:09.757 >> bots and like you basically in every domain construction >> Yeah. 00:54:09.840 --> 00:54:12.357 >> defense just all of a sudden have a much larger 00:54:12.440 --> 00:54:15.197 number of people who are applying creativity and coming up with ideas and 00:54:15.280 --> 00:54:17.917 making things that move. >> Yeah, and have you seen like with Claude 00:54:18.000 --> 00:54:22.197 it's like it's one thing just to make the engineer more efficient or bring more 00:54:22.280 --> 00:54:25.957 people into engineering but then when these agents really run you're getting 00:54:26.040 --> 00:54:29.037 into just like the iPhone example of you 00:54:29.120 --> 00:54:33.837 couldn't imagine Instagram before like 00:54:33.920 --> 00:54:37.397 iPhone. It's like imagine 2005 on laptops you're like you did 00:54:37.480 --> 00:54:39.677 10 years is going to be this this app. 00:54:39.760 --> 00:54:42.917 And you can put photos in like well the phones don't have cameras like yeah, but 00:54:43.000 --> 00:54:46.117 it's going to be like social like what the hell like so like Facebook it's like 00:54:46.200 --> 00:54:48.917 see it's just hard to so we think 00:54:49.000 --> 00:54:53.757 by lowering that barrier you're going to get way way more creative autonomy 00:54:53.840 --> 00:54:56.397 products. >> Yeah. Okay, so I want to say that you 00:54:56.480 --> 00:55:01.397 can decide whether to include this or not. So my kid is building autonomous bots in 00:55:01.480 --> 00:55:04.717 Factorio. >> Oh, nice. >> That's one of his projects and so and 00:55:04.800 --> 00:55:07.197 then but he's he's you know, he's had a role in cuz the tool kit's not available 00:55:07.280 --> 00:55:10.037 yet. So he's actually training and he's actually he's actually training models. 00:55:10.120 --> 00:55:12.597 >> Yeah. >> He's gathering data in the game and 00:55:12.680 --> 00:55:15.317 actually he has like a whole army of like bots that he's developed to go. 00:55:15.400 --> 00:55:18.517 >> Yeah. So like >> And then his mother is like why are you 00:55:18.600 --> 00:55:22.077 playing that game so much and he explains of course it's a purely educational 00:55:22.160 --> 00:55:26.597 process and experience but but it's you know, it's the kind of thing 00:55:26.680 --> 00:55:28.117 >> Like there's no reason autonomy should 00:55:28.200 --> 00:55:32.117 be this like you know, 00:55:32.200 --> 00:55:36.837 obscure difficult you know, alchemistic you know, 00:55:36.920 --> 00:55:39.077 technology And 00:55:39.160 --> 00:55:42.117 I think not only does that have a huge 00:55:42.200 --> 00:55:46.957 impact on on society, it also allows people to understand that these systems 00:55:47.040 --> 00:55:50.557 are not like you know, 00:55:50.640 --> 00:55:53.877 magic. It's like if I can develop a 00:55:53.960 --> 00:55:58.637 a Roomba for myself in my house on a weekend using Dana, 00:55:58.720 --> 00:56:02.237 then a why then it's not suddenly so scary. And I think that's like that's 00:56:02.320 --> 00:56:04.277 that's that's important. >> It can it can it can support it can 00:56:04.360 --> 00:56:06.837 support people in all kinds of ways that we haven't even imagined yet cuz they 00:56:06.920 --> 00:56:09.917 yeah. >> Absolutely. I mean you think about like you know, 00:56:10.000 --> 00:56:14.677 folks with disabilities. You know, we always think about humanoids as like this 00:56:14.760 --> 00:56:19.240 very important task of folding laundry which seems to be >> [laughter] 00:56:20.160 --> 00:56:24.157 >> So we focus on you know, the important task but the when you allow these tools 00:56:24.240 --> 00:56:27.757 to exist, I mean I I you know, we started a tooling company. I mean I feel 00:56:27.840 --> 00:56:31.757 so importantly that tools are like what separates actually advanced 00:56:31.840 --> 00:56:36.597 civilizations from you know, less advanced civil civilizations. And our our first 00:56:36.680 --> 00:56:39.757 mark for the company was a monkey's head and then we got a designer who said 00:56:39.840 --> 00:56:42.397 we're good. >> [laughter] 00:56:42.480 --> 00:56:47.200 >> Like this is stupid. I was like I thought it was pretty good. 00:56:47.720 --> 00:56:52.317 >> So you were talking earlier about how when you know, um 00:56:52.400 --> 00:56:55.077 the technology got got so good in mobile that there was a wave of these 00:56:55.160 --> 00:56:56.837 companies, you know, Uber, WhatsApp, 00:56:56.920 --> 00:56:59.717 Snap, you know, Airbnb etc. that emerged 00:56:59.800 --> 00:57:04.477 in quick succession. And so now the technology is getting there for the 00:57:04.560 --> 00:57:07.677 infrastructure for physical AI, what are some use cases or companies that you can 00:57:07.760 --> 00:57:10.317 obviously it's hard to predict the future but what were you most excited 00:57:10.400 --> 00:57:13.037 for? Like what what could we be talking about the equivalent here of in quick 00:57:13.120 --> 00:57:15.597 succession? >> I mean I think uh 00:57:15.680 --> 00:57:18.037 you know, mid-term we want Dana if not 00:57:18.120 --> 00:57:22.837 the short-term to really you know, make humanoids way more real. 00:57:22.920 --> 00:57:27.077 Uh there's I mean how many it's like a thousand core tasks in a home from from 00:57:27.160 --> 00:57:31.677 humanoids. And these companies it's like such I mean I'm sure if talk to people who 00:57:31.760 --> 00:57:35.157 work in these companies, it's everything is difficult. Every step of the way is 00:57:35.240 --> 00:57:36.957 difficult. Collecting data is difficult. 00:57:37.040 --> 00:57:38.877 You know, cleaning that data is difficult. Training those models or 00:57:38.960 --> 00:57:40.917 deploying the model is difficult. 00:57:41.000 --> 00:57:44.357 And the bar being I want a high school kid to make a humanoid. 00:57:44.440 --> 00:57:47.757 So, that that's our our our path. And we think there there could be a lot there. 00:57:47.840 --> 00:57:51.597 But, that's like these the obvious stuff. I think the true non-obvious 00:57:51.680 --> 00:57:55.357 stuff is going to be we'll look back will be will be way way more interesting. 00:57:55.440 --> 00:57:58.157 >> And there there's there's some core ingredients that we're bringing together 00:57:58.240 --> 00:58:01.917 in data, right? We're we're making it way easier to to actually get imitation 00:58:02.000 --> 00:58:05.077 learning to work. Way easier to make reinforcement learning work in 00:58:05.160 --> 00:58:08.717 combination with that. Um we're we we have pre-trained models 00:58:08.800 --> 00:58:12.437 that can be used as a baseline for a lot of things. Um world models, advanced 00:58:12.520 --> 00:58:15.517 simulation technique, all of these things come together and then you're 00:58:15.600 --> 00:58:18.037 sort of limited by your crea- creativity. Like, well, what what do I 00:58:18.120 --> 00:58:22.677 want to do? And if you think about any kind of physically AI task as it's it's 00:58:22.760 --> 00:58:27.517 a you are understanding the world and you're manipulating something and and we 00:58:27.600 --> 00:58:30.757 can build that. That can be built now much more easily in this in this tool. 00:58:30.840 --> 00:58:34.037 >> And I think sometimes people ask like us being a tooling company and like you 00:58:34.120 --> 00:58:37.437 take self-driving trucks. We deploy self-driving trucks and many of the 00:58:37.520 --> 00:58:41.157 self-driving truck companies use our tools. I I I think sometimes people ask, 00:58:41.240 --> 00:58:44.837 well, look, you know, with Dana, are you going to like enable 00:58:44.920 --> 00:58:46.197 all these competitors? That's great. 00:58:46.280 --> 00:58:48.277 >> Right. >> That's absolutely completely fine. If 00:58:48.360 --> 00:58:51.757 you look at Google and what Google did to web applications, there was a massive 00:58:51.840 --> 00:58:55.517 internet. Google still succeeded through, you know, search and YouTube 00:58:55.600 --> 00:59:00.317 and and and other web apps. And other folks learned and used open source 00:59:00.400 --> 00:59:02.677 products and then ultimately closed source products and ultimately venture 00:59:02.760 --> 00:59:06.517 back products. So, we think we think the the same thing can happen here. 00:59:06.600 --> 00:59:10.957 >> I was at a robotic startup a while back that you you guys know well. And they 00:59:11.040 --> 00:59:13.197 had they were training, you know, they were doing their training process, 00:59:13.280 --> 00:59:16.597 training one of their arms to do the particularly killer app that I thought 00:59:16.680 --> 00:59:19.887 was very appealing, which was picking up dog poop. 00:59:19.970 --> 00:59:21.757 >> [laughter] >> Literally, you know, training over and 00:59:21.840 --> 00:59:25.237 over again with different tools for the And so, you know, I don't know, why Why 00:59:25.320 --> 00:59:27.437 right? >> Yeah. >> Why not have the little Why not have the 00:59:27.520 --> 00:59:29.637 little robot follow you around when you walk the dog? 00:59:29.720 --> 00:59:32.317 And you'll pick up the poop. >> Yeah. And I think like like you 00:59:32.400 --> 00:59:35.317 >> I I know somebody who built a I forget who it was, but somebody built a a 00:59:35.400 --> 00:59:38.437 little lawn robot that would go around and individual pick up individual leaves. 00:59:38.520 --> 00:59:41.797 >> Yeah. >> Cuz you got that problem, right? You you rake you you rake you 00:59:41.880 --> 00:59:44.357 you rake your yard, it's completely clean, and then like 2 hours later 00:59:44.440 --> 00:59:46.837 there's like 14 leaves, and you're like >> Yeah, yeah. 00:59:46.920 --> 00:59:48.397 >> I'm just going to send out the little bot to pick up the leaves. 00:59:48.480 --> 00:59:52.157 >> It's like if development costs are zero, then people will do that. I mean you 00:59:52.240 --> 00:59:56.277 you guys remember like the early iPhone apps that hit were like the beer one or 00:59:56.360 --> 00:59:59.677 the fart app. If you >> [laughter] >> imagine that in like yeah, if you 00:59:59.760 --> 01:00:04.477 imagine that in '98 with you know, with the Symbian mobile, you know, whatever 01:00:04.560 --> 01:00:07.357 it was the OS from I think it was Ericsson or somebody. That'd be 01:00:07.440 --> 01:00:11.077 impossible. You didn't need a team of like 50 people to develop a like the 01:00:11.160 --> 01:00:14.597 beer thing for the Blackberry. So I think there's a similar type of thing 01:00:14.680 --> 01:00:17.317 that's happening where you know, we really want to be a 01:00:17.400 --> 01:00:18.837 part of that and we want to enable that. 01:00:18.920 --> 01:00:22.317 And if it like makes making like I think 01:00:22.400 --> 01:00:26.117 it'd still be a while before like making a robo taxi is like super super easy. 01:00:26.200 --> 01:00:27.437 >> Yeah. >> But that'll happen. 01:00:27.520 --> 01:00:29.997 >> But there's I mean the number of bots that could be the number of kinds of 01:00:30.080 --> 01:00:32.477 bots that could be deployed in health care is is almost endless. Health care 01:00:32.560 --> 01:00:34.997 alone is is endless. Home care home care. 01:00:35.080 --> 01:00:39.637 >> Yeah. >> Um and then in construction, you know, in all the physical trades. 01:00:39.720 --> 01:00:43.877 >> us sitting in 2007 and saying, "Let's we we should have an app store. What 01:00:43.960 --> 01:00:46.917 type of apps?" And we would come up with like a list of eight. 01:00:47.000 --> 01:00:50.397 And then they're like, "There'll be a messaging one and then there'll be a 01:00:50.480 --> 01:00:53.357 camera one." It's like and now you look at the app store and it's like, you 01:00:53.440 --> 01:00:57.917 know, there's an app for like the hotel you go to and it's like, you know, to to 01:00:58.000 --> 01:00:59.677 order, you know, food off the menu. 01:00:59.760 --> 01:01:00.957 >> Right. >> Yeah. >> Yeah, makes sense. 01:01:01.040 --> 01:01:03.717 >> Yeah. D, we're talking, you know, earlier about the the differences 01:01:03.800 --> 01:01:04.997 between digital AI and physical AI. 01:01:05.080 --> 01:01:09.197 We're sort of hinting at LLMs, but world models are you know, in vogue right now. 01:01:09.280 --> 01:01:12.277 Want to talk about sort of the the state of them in as it relates to physical AI 01:01:12.360 --> 01:01:13.837 and how should we think about them? 01:01:13.920 --> 01:01:17.757 >> So so first off, world models means about a hundred different things and 01:01:17.840 --> 01:01:21.957 we had a a team at CVPR recently and and I was joking with them about just how 01:01:22.040 --> 01:01:25.757 many different ways you can define what a what a world model is. But when we're 01:01:25.840 --> 01:01:28.237 thinking about a world model, right, we're we're typically thinking about it 01:01:28.320 --> 01:01:31.957 in the context of of a simulation, right? So something that is effectively 01:01:32.040 --> 01:01:34.757 >> We started a sim company. >> Yeah. Yeah, so so something that is like 01:01:34.840 --> 01:01:38.877 sufficiently able to represent the real world and and is reactive in a sense 01:01:38.960 --> 01:01:40.957 where you can actually have, let's say, 01:01:41.040 --> 01:01:46.317 a an autonomous agent that's acting in this world and the world model is is 01:01:46.400 --> 01:01:48.597 behaving appropriately in response to that autonomous agent. 01:01:48.680 --> 01:01:51.717 >> Maybe uh Peter, I think it's worth ex- being super explicit here. Or we just go 01:01:51.800 --> 01:01:55.597 just one level lower into Uh this you know 01:01:55.680 --> 01:01:59.837 determinism in simulators, kind of the sim-to-real gap, physics-based, you 01:01:59.920 --> 01:02:02.837 know, rendering all the way to like this generated world. 01:02:02.920 --> 01:02:05.997 >> Yeah, yeah. >> Where do we fit on it or or where, you 01:02:06.080 --> 01:02:08.997 know, yeah, just describe the landscape, I think, maybe. 01:02:09.080 --> 01:02:12.117 >> Yeah, yeah, so so this is like let's say simulation broadly, right? There's 01:02:12.200 --> 01:02:15.397 there's so many different ways of doing simulation. And so the the more 01:02:15.480 --> 01:02:20.037 classical approaches of simulation very physics-based and and you can 01:02:20.120 --> 01:02:23.477 decompose physics in all different ways and all different levels of abstraction, 01:02:23.560 --> 01:02:27.157 and you can simulate with sensors or without sensors, and and is it is just a 01:02:27.240 --> 01:02:30.277 body simulation or are we actually simulating, for example, the light in in 01:02:30.360 --> 01:02:32.917 the environment or the >> Almost think about like the way CGI is 01:02:33.000 --> 01:02:36.277 done. If we literally had technical artists, and we have technical artists 01:02:36.360 --> 01:02:40.597 who would create assets which would go in the simulator, which would mimic real 01:02:40.680 --> 01:02:44.437 like road signs and have you know, reflectivity and material properties 01:02:44.520 --> 01:02:46.037 that you would see in the real world. 01:02:46.120 --> 01:02:48.477 But as you guys know Hollywood [clears throat] is going 01:02:48.560 --> 01:02:52.437 through its own fundamental change. And yet now you've generated uh the same 01:02:52.520 --> 01:02:54.597 thing is happening in our universe as well. 01:02:54.680 --> 01:02:58.117 >> Yeah, so so that's sort of on the that's at the far end of physics-based 01:02:58.200 --> 01:03:03.237 simulation. And and then the opposite end is is purely neural simulation, but 01:03:03.320 --> 01:03:06.117 within that spectrum, there's there's many different things you can do that 01:03:06.200 --> 01:03:09.717 are each useful in their own right. And so one of those things is a 01:03:09.800 --> 01:03:14.037 Gaussian-based simulation, right? Where you have uh a a 01:03:14.120 --> 01:03:19.477 effectively a representation of the real world that that has a a a 3D 01:03:19.560 --> 01:03:22.197 representation. And that 3D representation is 01:03:22.280 --> 01:03:26.397 consistent, meaning that if if you let's say have some reference point, let's say 01:03:26.480 --> 01:03:31.237 a camera, and that camera moves within that 3D world, because the Gaussian is 01:03:31.320 --> 01:03:34.917 actually representing the 3D geometry of that world, you'll actually get very 01:03:35.000 --> 01:03:38.557 high-quality output from that. There's a lot of value in that, and and that's 01:03:38.640 --> 01:03:40.717 what I'd say one type of of world model. 01:03:40.800 --> 01:03:42.957 But when you go further on that on that 01:03:43.040 --> 01:03:47.837 spectrum into the really into neural simulation, then you get into these 01:03:47.920 --> 01:03:50.637 where you're actually generating the video feeds. You can think of a neural 01:03:50.720 --> 01:03:54.037 network that's actually outputting a video. 01:03:54.120 --> 01:03:58.037 As what's actually coming out of the neurons of that. And and 01:03:58.120 --> 01:04:00.917 that can be reactive, which gives you some very interesting properties. 01:04:01.000 --> 01:04:04.397 Reactive as in the ego >> does something in the environment and 01:04:04.480 --> 01:04:06.877 the other agents respond to the ego. 01:04:06.960 --> 01:04:10.237 >> Exactly. However, you're not guaranteed in that 01:04:10.320 --> 01:04:13.557 reactivity that it's it's accurate, right? And and and now it's a question 01:04:13.640 --> 01:04:18.317 of well, how can I align this this simulation, this world model, with the 01:04:18.400 --> 01:04:21.397 real world and the way that the real world would actually react? And if you 01:04:21.480 --> 01:04:25.557 have perfect alignment between the real world and and the world model, I mean I 01:04:25.640 --> 01:04:28.477 think you've just sort of solved the universe roughly, right? It's that's an 01:04:28.560 --> 01:04:33.197 impossibly difficult problem. But but as we make progress towards that, 01:04:33.280 --> 01:04:36.517 it makes training physically accurate models much 01:04:36.600 --> 01:04:40.637 easier because you can do more of that in simulation. But the hardest part 01:04:40.720 --> 01:04:42.557 though is we're always thinking about 01:04:42.640 --> 01:04:46.237 performance, right? So I like to say 01:04:46.320 --> 01:04:50.117 that the the labs, they they have it easy because they can they can make 01:04:50.200 --> 01:04:54.477 models that are trillions of parameters, and those models can be super slow, and 01:04:54.560 --> 01:04:57.637 that's fine. But we don't have that luxury in physically AI, right? We deal 01:04:57.720 --> 01:05:02.197 We we deal in real time. Like the actual actual clock real time. And so we we 01:05:02.280 --> 01:05:06.077 have so many milliseconds before we have to do something. And and those 01:05:06.160 --> 01:05:09.877 performance constraints, they actually constrain the problem in a lot of ways. 01:05:09.960 --> 01:05:13.317 So we can have very large models, and we do have very large models that are used 01:05:13.400 --> 01:05:16.717 in the off-world environment. But when once you go on board, all of those 01:05:16.800 --> 01:05:19.757 constraints are very real. And now we need to train a much smaller model that 01:05:19.840 --> 01:05:23.477 has these safety constraints, these determinism constraints. And 01:05:23.560 --> 01:05:26.677 and that's the hard part about physical That's also the moat, right? It's It's 01:05:26.760 --> 01:05:28.997 It is what makes our tooling and and our 01:05:29.080 --> 01:05:33.797 our competencies valuable because it's just really hard to to meet 01:05:33.880 --> 01:05:35.877 all of these constraints in in a physical system. 01:05:35.960 --> 01:05:39.557 >> When will you Which will we get first, a perfect perfectly simulated real-world 01:05:39.640 --> 01:05:43.717 environment for training autonomy autonomous devices or Grand Theft Auto 6? 01:05:43.800 --> 01:05:47.117 >> [laughter] >> You know, I as long as they keep putting 01:05:47.200 --> 01:05:50.997 out great trailers, I mean, I've I was I feel like I'm getting entertained 01:05:51.080 --> 01:05:54.847 without paying a dollar. I am re-introduced to Tom Petty because of 01:05:54.930 --> 01:05:56.950 >> [laughter] 01:05:57.920 --> 01:05:59.197 >> Will you give us some timelines? 01:05:59.280 --> 01:06:01.917 >> Soon as >> Well, well, so so this is >> Let's run with that for a second. 01:06:02.000 --> 01:06:03.837 >> Yeah, let's go for it. Go for [laughter] it. 01:06:03.920 --> 01:06:06.157 Well, no, look, I mean, so the whole thing with the whole thing with Grand 01:06:06.240 --> 01:06:09.197 Theft Auto is the big innovation was open open world open world sandbox 01:06:09.280 --> 01:06:12.997 gaming. So it's a It's a simi- it's a simulated city, and at least in theory. 01:06:13.080 --> 01:06:15.557 >> On that spectrum, I mean, we hire so many people out of the video game world. 01:06:15.640 --> 01:06:18.677 On that spectrum, it's absolutely real >> Well, then tell tell us about that. 01:06:18.760 --> 01:06:21.357 Yeah, what's the spectrum? >> So I here I I This is This is 01:06:21.440 --> 01:06:23.557 speculation, but I think Grand Theft 01:06:23.640 --> 01:06:25.797 Grand Theft Auto 6 will be perhaps the 01:06:25.880 --> 01:06:29.637 last major real-world video game that's 01:06:29.720 --> 01:06:34.077 still really developed, let's say, in that legacy era of traditional computer 01:06:34.160 --> 01:06:36.757 graphics tooling. >> Technical artists and yeah. 01:06:36.840 --> 01:06:40.677 >> Like I I think that that that Grand Theft Auto 7 will much more likely be 01:06:40.760 --> 01:06:42.317 like a world model-based video game. 01:06:42.400 --> 01:06:45.037 >> Right. >> Um and where where you you can imagine 01:06:45.120 --> 01:06:48.717 as as AI tech evolves here, you you have like this this concept of this video 01:06:48.800 --> 01:06:51.557 game world model, and there's there's like some sort of baseline, let's say, 01:06:51.640 --> 01:06:56.117 data store that represents the real world in somehow, and then and then you 01:06:56.200 --> 01:06:59.157 have a some translation layer that's actually turning that data store into 01:06:59.240 --> 01:07:02.557 something that you can see and and run around in. Like it's It's possible. 01:07:02.640 --> 01:07:05.917 >> the game is a constant It could be the real world, right? As we said, you could 01:07:06.000 --> 01:07:08.877 you could have a complete recreation of the real in the game. Well, this has 01:07:08.960 --> 01:07:11.317 kind of happened with flight simulators, hasn't it? Isn't it? The most recent 01:07:11.400 --> 01:07:15.517 flight simulators are literally is the entire planet rendered accurately, is my 01:07:15.600 --> 01:07:17.037 understanding, at least from the air. Is that right? 01:07:17.120 --> 01:07:20.277 >> Yeah, yeah. And I mean, you really that's where our our bread and butter is 01:07:20.360 --> 01:07:23.517 when we started started the business. We hired so many people out of the 01:07:23.600 --> 01:07:27.157 Microsoft Flight Sim. I'm going to say I'm surprised they didn't you know 01:07:27.240 --> 01:07:29.477 >> But when you fly over New You know, whatever, New York or when you fly over 01:07:29.560 --> 01:07:31.957 Duluth in the flight simulator now, it is the real city, right? 01:07:32.040 --> 01:07:35.277 >> Exactly. But there there's some tricks that they play there. And a lot of that 01:07:35.360 --> 01:07:37.517 is fidelity. You know, the real world, 01:07:37.600 --> 01:07:42.637 the more you zoom in, it stays a certain level of fidelity. And and so they the 01:07:42.720 --> 01:07:46.397 tricks that you play there is you you basically are 01:07:46.480 --> 01:07:48.797 down sampling very very aggressively. 01:07:48.880 --> 01:07:52.997 And then as you get closer, you know, then it becomes more more high fidelity. 01:07:53.080 --> 01:07:54.797 Where the real world isn't like that. 01:07:54.880 --> 01:07:57.477 The if you were to try to rebuild the world 01:07:57.560 --> 01:08:00.837 with this level of fidelity, it would you know, it would take the all the 01:08:00.920 --> 01:08:03.877 energy of the universe, right? It's it's it's it's it's quite complex. And that's 01:08:03.960 --> 01:08:07.677 probably by the way the best argument against us being living in a simulation 01:08:07.760 --> 01:08:12.357 is among the But of course then you would say, well, the simulator we're in 01:08:12.440 --> 01:08:18.120 doesn't follow the laws of physics that were at that point 01:08:19.400 --> 01:08:21.917 Yeah, yeah. Yeah. >> [laughter] >> As far as I know, everything happening 01:08:22.000 --> 01:08:22.837 outside this room doesn't even exist. 01:08:22.920 --> 01:08:26.397 >> Yeah, this is I mean, you know, you know, like Buddhism believes this. We're 01:08:26.480 --> 01:08:27.837 This is a different type of podcast. 01:08:27.920 --> 01:08:30.837 It's like, you know, when you open your eyes, the world is rendered. And then 01:08:30.920 --> 01:08:34.017 you close your eyes, the world That's literally religious. 01:08:34.100 --> 01:08:36.037 >> [laughter] >> I don't see why I don't see why it's 01:08:36.120 --> 01:08:38.607 necessary for it to keep rendering if I'm not there. 01:08:38.690 --> 01:08:40.917 >> [laughter] >> Buddhism from first principles. 01:08:41.000 --> 01:08:43.117 >> Yeah, exactly. That's what you should have That'll get a lot of clicks. That's 01:08:43.200 --> 01:08:47.277 what you should call this. >> [laughter] >> Yeah. Well, just going to be on the 01:08:47.360 --> 01:08:50.757 timeline topic, you know, you gave us timelines on self-driving cars. What 01:08:50.840 --> 01:08:53.757 timelines do you want to give us, if any, on the sort of 01:08:53.840 --> 01:08:56.557 you know, other interesting things that were worth tracking, like perhaps when 01:08:56.640 --> 01:08:59.677 we'll get laundry folded or other you know, things that emerge 01:08:59.760 --> 01:09:02.797 because of humanoids or I think also maybe just just touching a little bit on 01:09:02.880 --> 01:09:05.757 world models, where we see world models going, cuz I think it's it's fundamental 01:09:05.840 --> 01:09:07.517 to all the work we do. >> Yeah, pretty sure. 01:09:07.600 --> 01:09:11.277 >> Yeah, yeah, so so so to ask the the first question 01:09:11.360 --> 01:09:15.237 so laundry folding, it's it's it's not terribly far from being 01:09:15.320 --> 01:09:18.997 solved to to be clear. And there is a lot of interesting research being 01:09:19.080 --> 01:09:25.250 >> And then humanity can rejoice. That's in Proverbs 4:16, I think. [laughter] 01:09:26.480 --> 01:09:30.277 >> Well, here I I I I do think I do think housekeeping is is 01:09:30.360 --> 01:09:32.077 a killer use case for physically I, right? 01:09:32.160 --> 01:09:34.597 >> Peter thinks two that he always talks about in the company. One is 01:09:34.680 --> 01:09:38.317 housekeeping and entertainment. He Peter's long on humanoid entertainment. 01:09:38.400 --> 01:09:40.157 >> What is it like? >> What kind of entertainment? 01:09:40.240 --> 01:09:42.517 >> I 100% [laughter] agree with that. I think entertainment is the robot as 01:09:42.600 --> 01:09:45.317 killer app. I don't think anybody else >> Yeah, these are your admissions. 01:09:45.400 --> 01:09:49.997 >> I mean like >> [laughter] >> So these Midwest white guys are really into this. 01:09:50.080 --> 01:09:52.757 >> I'm just saying, but I think I think >> I just want to know when I get 01:09:52.840 --> 01:09:54.517 Westworld. That's all I want. >> Yes. 01:09:54.600 --> 01:09:57.237 >> No, I actually have an entertainer. I have a little I have a little a tiny 01:09:57.320 --> 01:10:00.077 little Chinese robot dog that's like literally it's just like a little it's 01:10:00.160 --> 01:10:03.157 just a little and it just like runs around and it just like first base a little bit. 01:10:03.240 --> 01:10:06.377 >> you pay to see Cirque du Soleil with robots? Like yes. 01:10:06.460 --> 01:10:11.037 >> [laughter] >> Yes. >> I I want to see I want to see kung fu 01:10:11.120 --> 01:10:14.237 trapeze swinging. >> Spoken by like a compiler's guy. 01:10:14.320 --> 01:10:16.340 >> [laughter] 01:10:17.800 --> 01:10:20.117 >> I know, but but >> I want Westworld. I want Westworld. 01:10:20.200 --> 01:10:22.917 >> I mean the the turn The funny thing is I was I was just saying like well will 01:10:23.000 --> 01:10:26.877 people in like the suburbs of Detroit actually that pass the test? I bet you 01:10:26.960 --> 01:10:30.837 people in Sterling Heights would actually pay to see that. It's actually pro. 01:10:30.920 --> 01:10:33.837 I stand corrected. I stand >> [laughter] >> But but but back on laundry folding for 01:10:33.920 --> 01:10:38.157 a moment it's it's actually not far from being folded if you remove the time 01:10:38.240 --> 01:10:41.317 constraint. And so the the trick that's played in if you look at the latest 01:10:41.400 --> 01:10:45.277 research videos is they'll say like play it at 8x real time or whatever, right? 01:10:45.360 --> 01:10:48.437 And and that's for you to make it watchable. 01:10:48.520 --> 01:10:50.717 So so the question is when can you actually 01:10:50.800 --> 01:10:52.757 achieve human parity of performance? 01:10:52.840 --> 01:10:55.797 That's that's further off. >> When you decouple models from just the 01:10:55.880 --> 01:10:57.437 hardware the hardware can do it now. 01:10:57.520 --> 01:11:00.717 That used to be a constraint. So, the hardware is very fast and accurate now, 01:11:00.800 --> 01:11:03.197 which was actually >> There there's still overheating issues 01:11:03.280 --> 01:11:05.797 that are still being dealt with, but it's it's not it's not terribly far off. 01:11:05.880 --> 01:11:07.717 Like, these are solved. >> I mean, it's far off from like, you 01:11:07.800 --> 01:11:11.317 know, when I was a mechy, that was like fantasy. Like, there's like nothing can 01:11:11.400 --> 01:11:14.197 >> What's the movie that has the most realistic future vision of robots? 01:11:14.280 --> 01:11:16.597 >> Oh, man. Bicentennial Man. >> Is it Okay. 01:11:16.680 --> 01:11:18.917 >> Yeah, he's he's realistic. >> [laughter] 01:11:19.000 --> 01:11:20.157 >> Why that one? I actually haven't seen that 01:11:20.240 --> 01:11:22.997 >> Well, I like I like that scene. I think it's I, Robot when Will Smith jumps in 01:11:23.080 --> 01:11:26.517 the car and uh it's his his you know, whatever his like accomplice is in the 01:11:26.600 --> 01:11:29.477 car and he's like puts the car in manual and she's like, "What are you going to 01:11:29.560 --> 01:11:33.717 drive this thing yourself?" Like, out of like, you know, she's Yeah, she's like, 01:11:33.800 --> 01:11:36.277 "Yeah, are you crazy? What are you going to drive this thing yourself?" Like, 01:11:36.360 --> 01:11:40.357 that's what we're That's the flight of intuition's you know, like goal. 01:11:40.440 --> 01:11:43.637 >> Well, again, by the way, I I haven't seen this movie in a long time, probably 01:11:43.720 --> 01:11:47.037 since it came out. So, my recollection of it is probably a bit incorrect. 01:11:47.120 --> 01:11:48.557 >> Don't worry, the internet will correct you. 01:11:48.640 --> 01:11:52.357 >> [laughter] >> But, I I think I think Bicentennial Man has 01:11:52.440 --> 01:11:57.677 fully self-driving cars. Uh and it also has the housekeeping robot, which which 01:11:57.760 --> 01:12:01.437 is played by Robin Williams. And that's it's sort of like the the friendly guy 01:12:01.520 --> 01:12:05.517 that will the friendly robot that will clean up and also babysit your kids and 01:12:05.600 --> 01:12:09.277 stuff like that. And it seems like it's it's in the not terribly distant future. 01:12:09.360 --> 01:12:13.397 >> I got a different answer. You guys ever see that movie uh Sam Rockwell Moon? 01:12:13.480 --> 01:12:15.517 >> Oh, yeah. >> Yeah. The the setup I don't want to It's 01:12:15.600 --> 01:12:17.477 a great movie. Don't watch the trailer. 01:12:17.560 --> 01:12:21.317 Just to watch the movie. Uh it's the the premise is the tagline 01:12:21.400 --> 01:12:25.557 of the movie is 250,000 mi from home you find who you are. And it's one guy who 01:12:25.640 --> 01:12:29.157 works on an energy harvesting base run by Pliant Intuition run by 01:12:29.240 --> 01:12:34.317 Lunar Technologies. >> [laughter] >> I'm I 01:12:34.400 --> 01:12:36.797 I don't want to be Weyland I don't want to be Weyland-Yutani. I don't want to 01:12:36.880 --> 01:12:41.037 be, you know, that's from from the Alien franchises. And then Tyrell Corporation 01:12:41.120 --> 01:12:44.757 from Blade Runner. No, no. I want to be Lunar Technologies in the Moon 01:12:44.840 --> 01:12:49.557 franchise. Not even franchise. What There's one guy who works on this in the 01:12:49.640 --> 01:12:53.997 base basically runs by itself. And he's just there to kind of mind it when 01:12:54.080 --> 01:12:56.237 things kind of some you know error signal 01:12:56.320 --> 01:12:57.877 >> the the the reason why it's it's I think 01:12:57.960 --> 01:13:02.317 so accurate is because the state of the art for AI systems is like these systems 01:13:02.400 --> 01:13:03.917 they just need the occasional grounding. 01:13:04.000 --> 01:13:06.717 >> Exactly. They'll just go off and do something crazy and then you have to 01:13:06.800 --> 01:13:08.237 say, "No, no, stop doing that." 01:13:08.320 --> 01:13:10.397 >> to like that too. I mean that's that's a coding bot sort of like, right? 01:13:10.480 --> 01:13:14.277 >> And then the reason other reasons I think it's quite accurate that it's 01:13:14.360 --> 01:13:17.357 maybe uncouth now but Kevin Spacey is the 01:13:17.440 --> 01:13:20.477 the AI you know smiley face and he's 01:13:20.560 --> 01:13:25.037 he's just there to kind of placate the human to to assist but to also like he's 01:13:25.120 --> 01:13:28.317 like, "Oh, you're you you seem like you're sad Sam." And like you know like 01:13:28.400 --> 01:13:32.837 that's the the the the but really it's the one running the base. 01:13:32.920 --> 01:13:35.717 And hopefully I mean I shouldn't say we want 01:13:35.800 --> 01:13:38.237 to be lunar technologies because I don't know if they're they're quite a positive 01:13:38.320 --> 01:13:43.157 force in nature in that in that but I think massive energy farm that's 01:13:43.240 --> 01:13:46.277 completely autonomous that's going to be the future. And I and 01:13:46.360 --> 01:13:50.477 I think everyone like everyone reacts to things like that with like fear. And 01:13:50.560 --> 01:13:53.237 it's like guys, that's amazing. That 01:13:53.320 --> 01:13:56.597 means energy costs go way down. Like 01:13:56.680 --> 01:14:00.557 that's an incredible positive thing. I think I think the you know I just did 01:14:00.640 --> 01:14:03.437 this commencement speech at my my undergrad. 01:14:03.520 --> 01:14:07.717 >> Did you get uh destroyed? >> No. You know what? I I >> Unlike Eric Schmidt. 01:14:07.800 --> 01:14:11.077 >> Yeah, yeah, yeah. Yeah, yeah. I Listen, listen. My my 01:14:11.160 --> 01:14:13.477 the situation my my wife started watching and she said, "I feel like 01:14:13.560 --> 01:14:16.397 you're yelling at me. I can't watch [laughter] this." 01:14:16.480 --> 01:14:20.597 So I I I basically I mean I I don't I'm not like I don't I don't I'm not going 01:14:20.680 --> 01:14:23.597 to say which tech leaders who just basically avoided by like punting and 01:14:23.680 --> 01:14:27.837 saying I'm not going to talk about it. I I talk about this stuff. And partly it's 01:14:27.920 --> 01:14:31.517 the General Motors Institute. No one's like these these are I I I don't want to 01:14:31.600 --> 01:14:34.237 I don't I don't want to I don't want to throw judgement on you know the people 01:14:34.320 --> 01:14:38.117 we recruit out of out of MIT and Stanford but I'd say GMI people are a 01:14:38.200 --> 01:14:41.397 little different and they're like you know they're like pragmatic people and 01:14:41.480 --> 01:14:46.277 they under they you know you don't go to a place like GMI if you believe a 01:14:46.360 --> 01:14:50.277 superficial view of what corporations do. Corporations are just people working 01:14:50.360 --> 01:14:52.997 on projects together. And by the way, people working on projects together in 01:14:53.080 --> 01:14:55.997 government, people working on projects together in nonprofits, they all screw 01:14:56.080 --> 01:15:01.037 up. It's And so it's it's too simple to say AI corporations are are terrible. 01:15:01.120 --> 01:15:04.637 It's also you can't say the other side, which is like it'll all be great. So you 01:15:04.720 --> 01:15:07.837 have a role to play. That's basically what you know, what my what what my 01:15:07.920 --> 01:15:10.597 message is. And it's that's that's the 01:15:10.680 --> 01:15:12.997 case. I think if you feel like if if it 01:15:13.080 --> 01:15:18.637 if the the of the the I think the obvious uh you know, uh abundance that 01:15:18.720 --> 01:15:21.037 comes from self-driving trucks self-driving cars and the fact that 01:15:21.120 --> 01:15:23.237 people don't die, which is amazing. 01:15:23.320 --> 01:15:25.477 But then you also get this efficiency of 01:15:25.560 --> 01:15:28.277 cheaper energy etc. If all those things 01:15:28.360 --> 01:15:33.437 don't still satisfy your your fear, you you as a person, it's up to your 01:15:33.520 --> 01:15:37.357 responsibility until really learn about that technology. You you you can't just 01:15:37.440 --> 01:15:39.797 say, "Well, I'm afraid of it and my 01:15:39.880 --> 01:15:44.437 reaction is shut it down." That's not it's simply it's the 01:15:44.520 --> 01:15:46.797 uh and I don't I don't say this just to say that that we're competing with the 01:15:46.880 --> 01:15:49.557 Chinese, but there's a Confucian >> [laughter] 01:15:49.640 --> 01:15:52.397 >> the saying by my Confucius is just no hand can block the sun. 01:15:52.480 --> 01:15:56.357 >> Mhm. >> And [clears throat] the sun is technological progress. And if we as a 01:15:56.440 --> 01:16:00.357 society don't embrace technological progress, we will be left behind. 01:16:00.440 --> 01:16:00.997 >> Somebody else is going to do it. 01:16:01.080 --> 01:16:03.357 >> Somebody else is going to do it. And if it's not the Chinese, who knows? It's 01:16:03.440 --> 01:16:07.437 maybe it's the Uzbeks or the you know, it's another country that is is 01:16:07.520 --> 01:16:10.797 recognizing, "Hey, my citizens are suffering and I'm going to use this 01:16:10.880 --> 01:16:12.757 technology to remove the" It is 01:16:12.840 --> 01:16:17.837 honestly, it's because we're living in such a great society that we can have 01:16:17.920 --> 01:16:21.197 these like >> Yeah. >> I would say stupid conversations. Like 01:16:21.280 --> 01:16:24.557 there still are people who don't can't get food. 01:16:24.640 --> 01:16:26.997 >> Yep. >> And and and and someone will immediately 01:16:27.080 --> 01:16:29.237 equip if they were debating me, they would say, "Well, there's plenty of 01:16:29.320 --> 01:16:32.357 food. It's it's the capitalist system that doesn't" No, no, no. Let's let's be 01:16:32.440 --> 01:16:35.317 very specific. There's plenty of food, but getting that food to those people is 01:16:35.400 --> 01:16:38.477 difficult. >> Yep. >> So that means we should let robots get 01:16:38.560 --> 01:16:40.077 that food to them faster. >> Yep. 01:16:40.160 --> 01:16:41.317 >> That's just that's just how it is. So I 01:16:41.400 --> 01:16:45.077 think I'm I'm And I think like we we we 01:16:45.160 --> 01:16:48.557 as like technologists, I think sometimes we, you know, it's I think an 01:16:48.640 --> 01:16:51.757 inclination just to say, "Leave these people behind." I think you have to 01:16:51.840 --> 01:16:55.077 bring them along. You have to explain it to them. But, we also have to treat 01:16:55.160 --> 01:16:58.717 folks like adults and say, "If you don't get it after I explain it 01:16:58.800 --> 01:17:02.277 a couple times, then you just don't get it." So, there's like a middle ground. 01:17:02.360 --> 01:17:05.677 It's not everyone's an idiot and every or we should just be technology will 01:17:05.760 --> 01:17:08.317 just be perfect perfect perfect. There's a middle ground. Let's have that 01:17:08.400 --> 01:17:10.317 conversation to a point. >> Yeah. 01:17:10.400 --> 01:17:12.917 >> And then we just move forward and make society better. And then the results 01:17:13.000 --> 01:17:16.997 show it. I mean, there's people who still shockingly believe communism is the 01:17:17.080 --> 01:17:21.397 right answer. I mean, I just want to say why And I'm I'm, you know, I I am a 01:17:21.480 --> 01:17:25.157 capitalist. I can't Yeah, I can't I can't not admit that. But, 01:17:25.240 --> 01:17:28.957 there's 70 years of history there. Like, we're not That's not a debate anymore. I 01:17:29.040 --> 01:17:31.517 mean, that that I think it could be a debate if we're in if we're sitting here 01:17:31.600 --> 01:17:34.197 in 1965 and having a debate, you say, "Okay, 01:17:34.280 --> 01:17:37.917 maybe centrally controlled systems work better." There's no debate anymore, 01:17:38.000 --> 01:17:43.037 folks. You know, systems where individuals make decisions on their own interest 01:17:43.120 --> 01:17:46.197 actually work better for society. And so, that doesn't mean everything is 01:17:46.280 --> 01:17:48.317 perfect. And you can't extrapolate. 01:17:48.400 --> 01:17:50.197 That's same thing with, you know, AI. 01:17:50.280 --> 01:17:53.397 Doesn't mean everything's going to be perfect, but net net, it's definitely 01:17:53.480 --> 01:17:56.517 going to be better. And that's roughly what my commencement speech was without 01:17:56.600 --> 01:17:59.557 the booze. >> [laughter] >> They Mark and these guys were booing. 01:17:59.640 --> 01:18:04.197 So, they just cut it out. Eric was booing and throwing stuff. They 01:18:04.280 --> 01:18:07.997 just edited it out. >> You mentioned uh you know, the Japan market earlier. 01:18:08.080 --> 01:18:11.517 When you talk briefly about sort of the global ambitions and how these 01:18:11.600 --> 01:18:13.637 technologies, you know, interplay and and what we're doing here. 01:18:13.720 --> 01:18:15.997 >> So, I think uh America particularly is 01:18:16.080 --> 01:18:20.957 still the most advanced in terms of when you take account of business model. The 01:18:21.040 --> 01:18:24.117 second thing for a company like Applied Intuition, we're an extremely global 01:18:24.200 --> 01:18:27.437 company. We work with everybody uh uh uh uh 01:18:27.520 --> 01:18:31.397 minus we don't have an office in China, but really everyone else on the globe. 01:18:31.480 --> 01:18:36.557 Uh and we're a horizontal company. We're a technology provider. And I think we 01:18:36.640 --> 01:18:39.677 I think more Silicon Valley companies I think can employ a little bit of what we 01:18:39.760 --> 01:18:42.917 do which is work very I would say collaboratively 01:18:43.000 --> 01:18:47.757 with the local economies. As sovereign AI becomes more of a real thing 01:18:47.840 --> 01:18:50.677 we have to uh you know, build businesses that take 01:18:50.760 --> 01:18:53.517 that into account. By the way, we're not the first ones to do this. If you look 01:18:53.600 --> 01:18:56.877 at the history of America, you read the history of Standard Oil, you'll see that 01:18:56.960 --> 01:19:01.117 this is this is that was the history of companies. You'd work internationally. 01:19:01.200 --> 01:19:03.197 Aramco is not a random company, right? 01:19:03.280 --> 01:19:07.797 You you build based on the real geopolitical realities of the time. And 01:19:07.880 --> 01:19:10.757 so I think we you know, we've I think navigated it quite well. I've I've lived 01:19:10.840 --> 01:19:15.077 you know, in Japan, I lived in Germany, I lived in Dubai. So I also being 01:19:15.160 --> 01:19:19.157 Pakistani by birth, I think that's also influenced our company. Peter's only 01:19:19.240 --> 01:19:21.117 lived in Michigan and [laughter] here. 01:19:21.200 --> 01:19:25.157 But he is a German, so 01:19:25.240 --> 01:19:29.837 But so I think innately we're more we think about the globe more. I think when 01:19:29.920 --> 01:19:33.757 I was at both at Google and at YC, I was always surprised at how 01:19:33.840 --> 01:19:36.877 uh kind of almost myopic the companies are. They're always 01:19:36.960 --> 01:19:39.797 looking at the market that's just like within the 30 you know, between San Jose 01:19:39.880 --> 01:19:42.517 and San Francisco. I was like, actually the market is really big. I think 01:19:42.600 --> 01:19:46.437 physical AI, the nature of it being physical, I think we we have to be a 01:19:46.520 --> 01:19:49.757 very international company. And I think we've had a lot of success being a very 01:19:49.840 --> 01:19:51.197 very you know, being international. 01:19:51.280 --> 01:19:54.437 Yeah. >> Cool. All right, that's a good place to wrap. 01:19:54.520 --> 01:19:55.997 >> Okay. >> Peter Gasser, thanks so much for coming 01:19:56.080 --> 01:19:58.157 on the podcast and congrats on big lots of data. 01:19:58.240 --> 01:19:59.557 >> Yeah, thanks for having us. 01:19:59.640 --> 01:20:00.317 >> Awesome. Great to see you guys. 01:20:00.400 --> 01:20:02.880 >> Okay, great.