Inside the AI Garage Powering the Real World

0:01Qasar Younis:Applied intuition. Applied intuition. What does applied intuition do? We have defense. We have auto. We have trucking. We have mining. We have agriculture. We're on the sea. It's like, what the hell?

0:32Kasser and Peter, we are here at the Physical AI Day.

0:37Qasar Younis:Yes. And you're going to show us around, I guess you call this a garage? Yeah, this is actually, it looks really nice today, but it's actually a working garage. We have a couple of large garages that we do all of our work in, working with both physical vehicles and obviously testing and all of that stuff. But yeah, this is the garage. It's the only place where we wear our shoes. I saw that. There's a station over there. everybody's shoes are in cubby yeah yeah yeah that's from the early days in the company but we've been doing it for a long time now also to give credit where credit's due i think gusto also did this in parallel but i'm not sure they still do it so uh but yeah we we were in a house when we started the company and you know we literally had people including peter living in the house uh and uh and then probably a year is when we actually got our first office and we're like let's keep the cleanliness going and also i lived in japan and in the office in japan you You take off like your clothes and you have like a uniform, a work uniform.

1:33Qasar Younis:And then you have your, you know, work. What are we going to get? Some applied intuition work wear. We have work wear already. But we don't force everybody to wear it all the time. But if you go to Japan and you like go to like Honda, every employee is wearing like the same uniform. Engineers. And then they're wearing a flat 80s hat that says Honda on it. It's fantastic. I feel like if we do that and we start doing like calisthenics in the morning, it's gonna be a little too off-brand for Silicon Valley. You'll only fully appreciate that have you been to Japan. Yeah, it's all true. It's all true.

2:06Qasar Younis:Yeah, and it's fun. Yeah. Okay, so we're here in the auto center. Oh, yeah. Big sign. Let's talk about this. So today we had our first ever physical AI day. Historically, we've never invited people in to our garages and things like that. And so when we thought about, okay, how do we show what we're doing? We have all these different sections. And this just highlights a little bit of what we do in automotive. All of our major product lines are also available in automotive. So that means the, let's say, the developer tooling you would use to write software and deploy it onto vehicles. That's one of our products.

2:40Qasar Younis:Vehicle operating system that abstracts hardware and software away from each other. Kind of like you'd have on a laptop or on a phone. That type of operating system is also something we do in automotive. And then, of course, self-driving. We do self-driving in automotive. And we do it from L2++ all the way to full autonomy. But yeah, this is the auto section. And if you look behind you, you'll see kind of the guts of a car. Peter, you want to talk about this a little bit? Yeah, so here we're just showing this is a production vehicle that's disassembled and just has all of its electronics laid out.

3:12Qasar Younis:Hence it says before. Yeah, before. A traditional vehicle, exactly. And you can see just the mass of complexity and wires. This is really what a car is in the guts of the car. It looks like a mess. It is a mess. Yeah, yeah. And I think for the astute viewer, you'll see all of these cables and you're like, why do you have so much wiring harness when typically it can be much more simple? That's a real legacy of the existing supply chain. If you're a company that sells, let's say, a backup camera, you're not only going to sell the backup camera, you're going to sell all the wiring that goes to some compute.

3:49Qasar Younis:And then you have to do that for your windows that go up and down, the sunroof, and that's why all this is complex. But in the after, what we do is we actually take all of that compute that's disparate around the vehicle into essentially a contained one box. It's not precisely a one box, but for all intents and purposes it is. and then you can run very few cables to that one box. And that box has higher performance compute in it and has all the signals from the vehicle. So then you can do really interesting things and including just abstracting the hardware and software away from each other, which allows then developers to develop apps for the vehicle in a way that you would like for a laptop or for a phone.

4:30Qasar Younis:And so making a vehicle with an old architecture fully autonomous is a lot harder than doing that on a modern architecture. And so when we're working with our customers, one of the goals is, well, you want to modernize and simplify the architecture, and then you can add autonomy and all of the AI capabilities on top of it. And so who are some of your customers you're working with that are now modernizing their vehicles? Yes. Earlier today in our presentation, we talked about Solanus as one of those customers, and there's a bunch of others globally as well. There's different amounts of things that we're allowed to say about them because...

5:01Qasar Younis:Probably what we could say is... Yeah, we're in the background a lot. And the companies obviously compete against each other. And this is like a part of their business as well. Of the top 20 global automakers, 18 are our customers. So vast majority of the industry is using some components here to modernize their actual products. From just tooling to the full operating system to intelligent cabin to autonomy. And so what's the difference if something goes wrong in an old car versus something goes wrong in one of these? So in an old car, right, you might have, let's say, an airbag sensor or something.

5:43Qasar Younis:I used to work in airbags. It was my first job as an engineer. And maybe you get like a little icon on the dashboard that you don't really know what it means. And maybe you're looking at the manual like, okay, what's the problem here? And on a modern car, you can have a much more sophisticated diagnostic system, a much more sophisticated way of servicing the information to the user. And then also with AI, we can also then surface that to the manufacturer and they can understand across their entire fleet, what are the sort of the aggregation of the problems that they're having and root cause those things.

6:13Qasar Younis:And I think we won't necessarily go into the deep details, but these are some of our R &D and showcase vehicles. And the middle vehicle here is a showcase vehicle for what's called our vehicle operating system. And then this is a self-driving vehicle. This can drive on urban environments and highways. It can park itself. And it's a demonstration. Maybe we talk a little bit about what's our approach. Yeah. So we started in 2017, initially building simulators, which were used for developing self-driving technology. And we sold the simulators to the whole industry. And then a few years later, we really got deep into building the technology ourselves.

6:49Qasar Younis:And what's unique now is we're able to use what's called the end-to-end technology. So we can take raw sensor data in and actually put out vehicle control signals. And then we have a large-scale data collection across a bunch of industries. So not only automotive, but also trucking and mining, construction. And we can combine all of this data together to create these self-driving models. Yeah, I think the thing that's maybe less obvious is that actually what this vehicle does and uh you know what we see in like that boat over there uh and the cars that you saw and the trucks that we have and they're actually fundamentally the same thing because they're machines that interact with the real world and so there's you know physics is consistent and so we've built a platform that can run across all of these all of these different industries and bring that same level of like for the lack of a better word teslafication to a machine like this um which is self-driving, intelligent, easy to work with.

7:54Qasar Younis:Yeah. And so the same operating system and platform really runs across all of these vehicle types. So like literally on this tractor, and then over here, if you take a look, this is actually we're showing in real time, we have actual construction and mining equipment doing autonomous operations at our test site. Yeah, this is being streamed in. This is real. This is real. These are real time video feeds. Yeah. Let's see what he picks up. Yeah. Yeah. So this is what we're doing, what's called a load haul dump cycle, which is basically the movement of Earth. And this is all happening autonomously.

8:27Qasar Younis:And these operators can control a whole fleet of vehicles remotely here. So what does this mean for the future of industrial work? How many people will be actually on site? So the future of mining and eventually construction is going to be these fully autonomous operations, where labor shortage is a real problem in the industry. And so this allows a very small number of people to control a huge number of vehicles. Yeah, I think, you know, the AI kind of problem and consternation and heartache and kind of hand-wringing that's existing right now with what happens with white-collar workers, with these super intelligent agents that are running and what's going to happen to jobs.

9:12Qasar Younis:it's quite difficult in the physical AI community because farmers are aging. People don't want to work in mines in faraway places. These are dangerous jobs. And so AI really in these industries is being kind of pulled out of our hands because they're real problems that they're dealing with. And technology can help solve those problems, whether it's labor shortages or safety or whatever it might be. A lot of people talk about the negatives for AI, but this is obviously a positive. And because the reality is there's not a new group of people who really want to become farmers. Yeah. It's just not there.

9:57There's only so much propaganda we can put on X for that.

10:00Qasar Younis:Yeah, exactly. Exactly. All right. It's a little bit darker in here. Yes, it is. Where are we now? What is this, Peter? Yeah, so this is what's called Applied Edge. This is our mobile command center. And so all of the development tools and infrastructure that we are typically able to use to build interesting defense technologies in an office environment, all of that is packaged up and can run completely offline, off the grid. In a shipping container. Exactly. And so this can be transported on a C-130. It can be transported on a ship. It's a standard container, which is easy to move. And it's fully self-sufficient.

10:45Qasar Younis:So it has its own power, has its own cooling, and it has all of the software and server technology all built right in for operations. And here we're actually showing a real operation. We have actual boats out in the ocean, which we are remotely controlling and doing patrol operation. Does it come with the people in it?

11:05Qasar Younis:They're laughing, but they don't want to look back.

11:10Qasar Younis:She's been controlling this boat for probably a couple of hours, just going in circles. But yeah, it's a real boat being really controlled. Sorcery is brought to you by Brex, the financial stack trusted by more than 30 ,000 companies, including one in three venture-backed startups in the U.S. Nearly 40 % of startups bail because they run out of cash. Rex is literally built to help founders avoid that. Unlike traditional banks that let your money sit idle, chipping away at it with fees, Rex is designed to help you spend smarter and move faster. Their all-in-one solution combines checking, treasury, and FDIC protection into one powerful account.

11:49You can send and receive money globally at lightning speeds, get 20 times the standard FDIC coverage through their partner banks, and even high yield from day one. with same day and even same hour liquidity. Access your funds anytime. Companies like Scale AI, DoorDash, Service Titan, HIMSS, Anthropic, Flexport, Robinhood, and Plaid trust and use Brex. Start today at brex.com slash sorcery. That's B-R-E-X dot com slash sorcery. Turing is training the next generation of AI with tasks that require real expertise and real world judgment. That's why companies like NVIDIA Anthropic, Salesforce, and Gemini partner with Turing.

12:33Turing builds realistic reinforcement learning environments and data systems based on real operational traces, the kind of infrastructure frontier labs need to train superintelligence. Visit turing.com slash S-O-U-R-C-E-R-Y.

12:48Qasar Younis:So this is a kind of our defense area. Yeah. When did you guys get into defense? We originally got into defense pretty early, actually, like 2019, somewhere right there. So pretty early in the company's history. Our work in defense accelerated when we acquired a company called Episci. This was about now two years ago. Yeah, something like that. But yeah, we work with all the major branches of the military. We also work with primes. There's projects we sub on. There's other projects they sub with us on. But we're really everywhere in the defense ecosystem in the U.S. This is a Ford Raptor, an F-150 Raptor.

13:25Qasar Younis:So it's a it's a rugged off-road vehicle. We use this actually as our development vehicle for off-road autonomy stack. And we also do some really interesting production applications of this. So recently we showed a collaboration with Sierra Nevada Corporation where we actually had like an anti-drone missile launcher on the back of one of these trucks. Yeah, what's roughly happened in defense is if you're using commercial technologies like a Ford F-150, those are also very dangerous because they're not actually made for combat. And so then putting humans in there becomes not really an option. Is that a normal thing?

13:59That's normal?

13:59Qasar Younis:No, it's something that's new within the administration, which is let's use lower cost products, but then you don't put people in them. And they can run autonomously out to the battlefield. And that's actually just in terms of affordable mass, that is the right strategy. That's a correct strategy. And I think that age of really expensive monolithic systems, I think, is gone. I think our dual use strategy, and it's truly a dual use strategy in the sense of we are taking commercially available products, which we sell in all these other verticals, and we're also selling them in defense and we're optimizing for the defense use cases.

14:38Qasar Younis:But we can frankly deploy more robust systems because they're already in field and in production for a fraction of the cost that actually get the same mission accomplished. I think one way to think about this in defense is imagine if the Department of War said, let's build a chat app. How long and how many dollars would it take to compete with WhatsApp or Signal or something like that? This is extremely difficult. So there are real advantages in the commercial ecosystem. We get disciplined by the commercial side of the business because we have to compete against everybody. We have to compete against the low-cost vendors all the way to the high-cost vendors in the West and the East, everywhere.

15:20Qasar Younis:And because of that, like our technology has to be the best. It can't be just the best that's selling to the DOW. It has to be the best in the world. And then also the best selling to the DOW. Yeah. And this is an autonomous boat. It's actually the same kind of boat that we actually have in the ocean right now. And we're showing in that command and control center. It's not this one with like some AI or something. That's a real boat. We have a fleet of these. But obviously this is unarmed in this setting, but there's a lot you can do with this kind of vessel. Yeah. Should we keep moving to the other areas?

15:54Qasar Younis:Yeah.

15:58Qasar Younis:Yeah, let's take a quick peek outside and then head back in. Should we go to the big mining truck? All right, how are we going to get there? Are we going through? We'll go through this way, yeah. Okay, yeah. Before it gets dark. Yeah, before it gets dark.

16:19Qasar Younis:What's most surprising to you? For this? Yeah, as you walk around. Honestly, I just think it's kind of crazy. You guys have been under the radar for so long and then you open up the garage doors and it's like we have defense, we have urban, we have trucking, we have mining, we have agriculture, we're on the sea. It's like what the? How did that? When? Okay. It's a real company. Look at these guys getting their, uh, he's actually our first He's our first employee and he's back seeing the... I think he's really looking at this like, should I really bank on this stock for the rest of my life? I think he's going to.

16:53Qasar Younis:So they're doing self-driving car rides so you can get driven around town. Oh, fun. All right, look at this beast. Big. Yeah. And so this is a Komatsu truck. So funny enough, you might think this looks like a big truck. This is actually a small truck. Yeah, this is the smallest truck we could get in this parking lot. Yeah. So this is actually technically not even a mining truck. This is a construction truck. So we do work also in mining trucks, which are much, much larger than this. But yes, the same operating system and autonomy technology that we do in automotive and other fields, we also put onto this kind of vehicle.

17:31Qasar Younis:And yes, it's obviously super valuable in the industrial domain. Yeah. What's going on over here? Yes, this is an autonomous skid steer, which is an incredibly useful vehicle. So if you're not in construction, you might not be as familiar, but skid steers, there's like 100 different tools that can be placed on the front of these. So they can do almost anything. And a typical worksite will have a whole bunch of these things. And in this example, we're autonomously moving around pallets, but just what you can do with this technology is almost unlimited. And once it's autonomous, you can do anything with it.

18:06Yeah, yeah.

18:07Qasar Younis:And I think the interoperability of our platform is really important because we can go from a mine, then, you know, to self-driving trucks, which often they go to a port. And we think automating all of that is really where there's going to be a huge unlock in productivity. I think automating some of these individual machines is positive, but unlocking all of them and then them working together is really powerful. When do you think you're going to complete an end-to-end autonomous mission? Yeah, you know, we're on that path right now. I think we are working in queries and mines around the world.

18:46Qasar Younis:We have driverless trucks running in Japan right now. We obviously have driverless cars as well. I think in the near future, in the near future, yeah, it's only a matter of time I think that we have like a mine of road and a port and all connects and we have you know we have automation in all three of those. Just recreating the industrial fabric of the world. Yeah absolutely absolutely. So let's walk around this way and we'll go into the trucking section. Yeah so we have a public partnership with Isuzu which is one of the largest commercial manufacturing manufacturer of vehicles in the world. They make trucks among many things and we've been automating these.

19:33Qasar Younis:We run driverless trucks in Japan right now. Peter, do you want to talk about our Gen 2 vehicle? Yeah, so Gen 2, I mean you can look, it's it's really nice. It's refined. This is basically one step before this is fully productionizable. And so they will produce this. This is being commercialized in the next year, which is super exciting and uh and japan's an awesome market for self-driving trucks uh because a very aging population yeah and there's a there's a big need for for this so it's it's sort of the and there's also support from the government so it's phenomenal yeah exactly yeah what countries are you not working in you guys have extreme global reach we're not working in china um china is the only country that i think we probably won't be in maybe ever but certainly anytime soon Sometimes people say that's because we do defense work.

20:24Qasar Younis:That's part of it. Sometimes they say, well, you know, IP restrictions. That's part of it. But yeah, all of that combined equals we don't have, you know, we don't have a presence in China. But everywhere else we do. And most of the areas of the world, we already have some presence. Our most recent kind of expansion has been Australia. We opened also in India and UK offices last year. about 18 sites globally. I think as we kind of look at the future, you know, Australia could be one of those, you know, countries where we're doing driverless trucks and we're doing a bunch of different, you know, ports and suddenly you have that full connection happening where that efficiency is unlocked in a real way.

21:07Qasar Younis:But I think what I want to really do, if we can do it, it's a little bit of our goal, is once we get into automating, you know, parts of ports, then we could go to a LATAM country and we don't have anything in Latin America right now so maybe a Brazil or enter that region through mining and so you can see we're kind of hopscotching like a board game throughout throughout the planet I think probably the furthest out is Africa we don't have any business in Africa right now but I think one day we'll get there and then probably the closest to the Middle East we're expanding into the Middle East basically in real time which has been the fastest to adopt and expand?