The Operating System for Self Driving Cars (and Tanks, and Trucks...) With Qasar Younis and Peter Ludwig of Applied Intuition

When will fully autonomous vehicles see widespread adoption? According to Applied Intuition, that future is closer than you may think. Applied Intuition’s CEO, Qasar Younis, and CTO, Peter Ludwig, talk with Elad Gil about how now is the best time to both work on self-driving vehicle technology and monetize it. Qasar and Peter discuss the advantages of developing their own OS in-house for their autonomous applications, self-driving technology’s potential to drive re-shoring of vehicle manufacturing to the United States, and how best to gauge the bar for safety in autonomous systems. Plus, they explore how self-driving technology may reshape the designs of not only vehicles, but cities themselves. Sign up for new podcasts every week. Email feedback to [email protected] Follow us on Twitter: @NoPriorsPod | @Saranormous | @EladGil | @qasar | @AppliedInt Chapters: 00:00 Qasar Younis and Peter Ludwig Introduction 01:28 A Primer on Applied Intuition 11:08 Applied Intuition’s Customers 12:04 Impact of Chinese Vehicles Manufacturers 15:44 EV Policies in the European Market 20:49 Can Robotics and Automation Re-Shore Vehicle Manufacturing? 21:53 Training Models for Autonomous Vehicles 26:41 Gauging the Bar for Autonomous Vehicles Safety 32:03 Timeline for Large-Scale Autonomous Vehicle Adoption 36:28 Rethinking Urban Design for Autonomous Vehicles 38:47 How Applied Intuition Uses AI for Tooling and OS 42:09 Designing for User Experience 43:31 Applied Intuition’s Hiring Strategy 45:01 Conclusion

Speaker names marked ? are inferred from speaking time and the show’s known hosts — they are not voice-verified.

• 0:05Speaker 00

  Castor Peter, thank you so much for joining me today on New Priors.

• 0:08Speaker 01

  Thanks for having us. I came as casual as I could for the San Francisco.

• 0:12Speaker 00

  I was worried about you. You're wearing something nice. That's kind of, that's concerned. Is everything okay?

• 0:18Speaker 01

  Yeah, it is. It is. Two things. One is, you know, I wanted to fit in with the San Francisco vibe, you know, Sunnyvale. I usually dress more like Peter, to be honest. I think that's my outfit.

• 0:29Speaker 00

  Because you're just.

• 0:30Speaker 01

  Secondly, you know, I got this Carhartt, you know, thing, and I don't know if you guys know, Carhartt is suddenly cool. I got the memo from the Cool Club newsletter. It's a Detroit brand, if anybody doesn't know. We're representing Detroit and Silicon Valley, so.

• 0:45Speaker 00

  Yeah, I just thought, I thought.

• 0:47Speaker 01

  You're not impressed. It's okay.

• 0:48Speaker 00

  I'm extremely impressed. I thought it was one of those things where, I think you guys just raised at a $15 billion. I thought it was more like, kind of, you're done. You've just checked out now.

• 0:57Speaker 01

  No, no, no, no. So last night, we were having dinner with, with one of the top three global OEM CTOs. And he says, you know, I can't reveal who it is only because it's active, active at negotiations on a deal. And he says, well, you know, congratulations on the fundraiser. How do you feel? And I said, well, you know, honestly, I feel a little nervous. You know, we have, we have, you know, big goals ahead of us. He said, don't be a coward.

• 1:21Speaker 01

  Attack. And I was like, okay, I will send that message to my team. Don't be cowards. Attack.

• 1:28Speaker 00

  I mean, I guess in general, when I think about you also, I've known you for over a decade.

• 1:32Speaker 01

  Yeah.

• 1:33Speaker 00

  I think I've led two of your rounds, and I feel like you're the most successful, most quiet company in AI. You've, you're now at over 1,000 people. You're in the hundreds of millions of revenue. You've been profitable the whole time. So you haven't spent a dime, I think, of any of the money that you've ever raised, which is pretty insane from a capital efficiency perspective. Well, we're trying to change that quiet part, by the way. Yeah, but you've been super quiet.

• 1:54Speaker 01

  We came up to San Francisco and wore a jean jacket.

• 1:57Speaker 00

  Yeah, welcome.

• 1:57Speaker 01

  See, mom?

• 1:58Speaker 00

  Yeah, exactly. Could you tell people a little bit about more generally what you do? I know you have three lines of business around engineering, tooling, autonomy. And then sort of in car related product. Could you kind of break down the origins of the company? How you got started, what you focus on and just kind of give a primer? Cause I think again, you've, you've accomplished an enormous amount. There's still a lot ahead of you, but my gosh, you've done.

• 2:17Speaker 01

  Yeah. I mean, by the way, this is a directly for founders more than anything else is that there's a huge value in especially when the company is young, not to be constantly out there. I'm sure there's downsides as well, like people don't know you and it's harder to recruit or whatever. That's never really been a huge issue for us, but the advantages you get is you can operate. You know, the moment you say, we do X, there's an expectation you do X, even if X ends up being wrong. And so I think that like, it's like, you know, keep your identity small kind of view. And so, yeah, the company, Applied Intuition is a $15 billion, we just raised $15 billion profitable AI company. And with that, what we really do is we're, we're in, we build vehicle intelligence.

• 2:57Speaker 01

  So it's a broad category of how do you take this, you know, all the positive things about AI that we're seeing in, you know, in L L Ms and in chat, and you take them into the real world. So how do you put intelligence into cars and trucks and tanks and fighter jets? And, and as roughly as those three business lines. So we originally started with engineering tools in order to build and deploy that type of, you know, intelligence into vehicles and test it and validate it. Because unlike your laptop, these are safety critical systems.

• 3:29Speaker 00

  This was initially stuff that was built, I think for the self-driving world, right? You were doing simulation environments.

• 3:33Speaker 01

  Exactly. And then we expanded to all software in the vehicle. That's broadly what, what the company does in terms of what's unique about the company. I think is we've always thought pretty deeply about building products that are going to be used quickly. If you're in the AI universe, especially in eight, nine years ago, in the autonomy universe, a lot of research happening and research can be, it's super exciting and actually quite, what's, what's the word, like losing huge like metric tons of money without actually, you know, you can convince yourself you're doing some really, really impressive things. How do you describe, how do you describe the company?

• 4:14Speaker 00

  Just underlining engineering tools, our vehicle operating system, and then autonomy and applications. And so we started in engineering tools, but you sort of reach a point where in order to make your engineering tools better, you actually need to be developing applications yourselves because that sort of informs you and if we're really getting to the next generation technologies. And then it comes to, okay, well, I want to run these applications on vehicles and well, we need a great operating system and we think that we could do something better than anything else in the world. And so that's sort of how we ended up with those three business areas.

• 4:44Speaker 01

  I mean, there is a company that executed this strategy is Microsoft.

• 4:47Speaker 00

  Yeah.

• 4:47Speaker 01

  Seventy-five to eighty-two was doing tools that people don't remember. That was Microsoft's region. Then they went to operating systems and they went to applications like, you know, Office and Windows and stuff like, Office and Word and stuff like that. We're doing the same, except the hardware is not PC manufacturers, it's cars and trucks. And I think what's, you know, one of the old Peter Thiel things like, what do you believe to be true that other people don't believe true? When we were starting the company, if you remember, there were a lot of startups in 2016, 2017 doing self-driving. And our view was there's no path to autonomy. There's no path to like this next generation because without the manufacturers in the loop. Today, we're almost like a Tesla minus the hardware. We do all the stuff that a Tesla has or some other companies, but we put it, we didn't partner with manufacturers to bring that technology to them.

• 5:31Speaker 00

  Yeah, it's really interesting because if you look at the self-driving wave, to your point, there's like two dozen different companies or three dozen, you know, tons of companies. And then if you look at the ones that are, have arguably been most successful in the US, it's two incumbents, right? It's Waymo, which is a subsidiary of Google, and then Tesla. And then you guys are sort of providing that same sort of stack more generally for any automotive provider to sort of adopt and use. And I think it's not only the self-driving or the autonomy side that you folks really focus on, but I think the OS is really powerful. You sell, to your point, defense, construction, but a lot of your business is the giant automotive companies around the world. Could you talk a little about what you're providing through that OS and why it's beneficial and what does it actually do?

• 6:10Speaker 02

  When we refer to the vehicle operating system, we're talking about the full software stack that runs on these embedded systems that run on these vehicles. So at the lowest level, right, we even do bootloaders. Because if you want to do very reliable updates to embedded systems, you have to control the bootloader. And then you go above that and we can talk about the actual, the technical gory details of the operating system itself. So think about like a true operating system. But on top of that, you have to have the middleware, which is responsible for some abstraction, but also some really important safety critical data transport aspects of it. On top of that, you actually have the applications that are running and also many, many layers to those, of course, but they end up really controlling the vehicle, but then also displaying information. And then information could be displayed to somebody who's in the vehicle, or it could

• 6:52Speaker 02

  actually be displayed to someone who's outside of a vehicle, let's say in the mining example.

• 6:56Speaker 01

  What, I mean, Peter worked on Android at Google. That's where we met. We worked together at Google. What is the big lessons from Android? Especially on hardware diversity, right? That's kind of the, I think, the big, big innovation of their.

• 7:10Speaker 02

  Android, it's an incredible story from where it started to where it is today. I think it's true that Android is the number one OS in the world. Like it runs on more devices than, than anything else by, by a healthy margin. The big thing that Android figured out was just how to run applications uniformly on a huge variety of hardware. And how to do that in a way that the user experience is actually consistent. And there are a lot of lessons, both at the technical and non-technical level, on how to actually enforce that. So there's this thing called the compatibility test suite, which is super important. And it, it's this enormous set of tests and test infrastructure that allow these hardware makers.

• 7:50Speaker 01

  I think it's like a north of.

• 7:51Speaker 02

  It's like millions, I've heard, yeah. It wasn't even as big when we were working on it back then. But what, what do you get? You get this OS that can be used by billions of people on like many thousands of different types of devices. And so certainly in our work, we've taken inspiration from some of those techniques and practices to make sure that our, our technology is applicable to such a wide variety of vehicle types and chipsets and all of, all of the details there.

• 8:14Speaker 00

  It seems like one key insight there as well is that if you look at the supply chain for automotive. There's lots of different manufacturers for a lot of different modules. There's sort of embedded intelligence, release embedded systems on each one of these devices. And it's really hard for traditional car OEM to actually make use of some of the capabilities of these things because they don't really have a good either API or interface to interact with them. And I think you all have kind of built out.

• 8:36Speaker 01

  Exactly.

• 8:36Speaker 00

  That full layer to hook into all these different.

• 8:38Speaker 01

  Yeah, and I would adjust that the, you know, that's not really embedded intelligence. A lot of the stuff that's the software that's on these.

• 8:45Speaker 00

  It's embedded, not intelligence.

• 8:46Speaker 01

  Yeah. A lot of just I O, you know, it's, it's the seat warmer turns on and off. And you can actually pull all of that into a, into a central, you know, CPU. And that gives you a lot of efficiency, not only in just the ability that now all the signals are centrally processed and you can do more interesting things like Tesla does. But it's also just cheaper. You're, you're taking all these like redundant systems, which are all kind of, you know, poorly managed and built with different software stacks within each of the subcomponents. And, and you're removing wiring harnesses, thousands of dollars of physical hardware can literally be removed to get more functionality. So.

• 9:25Speaker 00

  Yeah. So you're basically streamline lining the guts of a car and you're replacing hardware components with software.

• 9:30Speaker 01

  Yeah.

• 9:30Speaker 00

  And then removing a lot of the.

• 9:32Speaker 02

  And I would say that people who grew up, let's say purely in the Silicon Valley mindset or purely in the mindset of I can go to the store and buy this computer and I can, I can write software and this computer runs whatever operating system and it generally works. You're so abstracted away from some of the very complicated details of how that hardware works. It's overly simple. When you actually get into safety critical systems, there's so much complexity in how reliable this hardware needs to be, how long it needs to last when it's actually deployed in a field. And then just the cost constraints that you have, like fundamentally, in cost matters a lot, especially in embedded systems. And you have all these constraints and we still want to do really advanced things. And I think something that we've done really well is figuring out how to do some really

• 10:17Speaker 02

  advanced things, but actually in a cost effective way.

• 10:20Speaker 01

  Sure. In the kind of, you know, current zeitgeist of Silicon Valley where you have humanoids emerging, a lot of these questions are not answered on that side. Now it's, it's better in the sense of. A lot of these companies are kind of verticalized, you know, they're doing the hardware and the software and the system is way, it's more complex and more simple than a car. But in trucks business, car business, tanks, jets, you're talking about hundreds of companies, sometimes thousands of companies, working on an individual product. And they all come with very different views.

• 10:49Speaker 00

  And nobody's really stitching all that together, except you.

• 10:51Speaker 01

  Yeah, and we don't, since we're not, we don't have a wiring harness business, or we don't have a, you know, we don't make chips, we can, or we're not a cloud provider, we can really come to the manufacturer and say, hey, you actually need a new way of, you know, operating this vehicle.

• 11:07Speaker 00

  So that's one area that you have, are you able to announce any customers that are working with you on this?

• 11:12Speaker 01

  Yeah, I mean publicly, we have a bunch of customers, publicly, our, the hero customer that we'd come out with was Porsche. So I think everybody knows, they're generally considered to be the most competent OEM on the planet in terms of, you look at a Ferrari, a Ferrari will make more dollars per vehicle because it's truly a luxury good, but Porsches are the most profitable cars on the planet. They, it's thirty to forty thousand dollars in profit per vehicle. So they found that sweet spot of high volume and, you know, ability to charge a lot because of the brand. And I think, you know, a company like that is getting pressure from a Tesla that says, you know, people are looking at those, even though they're very different products, people are looking at those. And so I think we're helping them offer, you know, offer a consumer experience which

• 11:57Speaker 01

  is at par, if not better.

• 11:58Speaker 02

  And I would say it's a pressure from Tesla and also the upstart Chinese companies. There's a lot of interesting stuff happening in China.

• 12:04Speaker 00

  How do you think about that? So I think one of the big shifts that's happened globally is this rise of Chinese manufacturers like BYD, Xiaomi, actually launched a car, I think, within five years, which is pretty amazing, because I think they were like smartphones and other sorts of hardware, but they never really did anything in automotive. The claim is that these systems are actually pretty good on the self-driving side and in a variety of other ways. What do you view as the global impact of these Chinese car manufacturers rising up?

• 12:30Speaker 01

  Lots of nuance here. So, number one, they are good. So, the car business is extremely international. And what I'm going to say next applies also to truck construction and mining. But in the 80s and 90s, the big boogeyman was Japan.

• 12:47Speaker 00

  Yeah.

• 12:47Speaker 01

  And if you grew up in Michigan and Detroit like us, that's all it was. Japan's coming, they're buying Rockefeller Center, and we're all be speaking Japanese soon. Obviously, it didn't happen.

• 12:57Speaker 00

  Yeah, yeah.

• 12:58Speaker 01

  Then it was the Koreans. So, 2010 was Hyundai's coming and all the conglomerates there.

• 13:03Speaker 00

  Kia and etc.

• 13:04Speaker 01

  Yeah, exactly. Genesis, etc. That obviously didn't happen. Right now, the newest version of that is China. There will be another one, but there'll be Vietnam or India or something will come after China. So, what the upstart wants to do is they look at the industry, they don't have any legacy platforms, and so they can enter the business with a blank slate. And you get a lot of advantage of that. You don't have all of these, you know, you have hundreds of millions of vehicles out that you're servicing and maintaining, and brands that already have some legacy to them. And that's allowed them to, with this EV shift, introduce, they being the Chinese Communist Party and broadly the Chinese ecosystem, to introduce lots and lots of brands at lots of different price points that all have pretty

• 13:45Speaker 01

  impressive products. Though what it's not, so the autonomy stuff, you know, we go to China regularly, and we test drive these vehicles, super impressive, and better than Tesla, to be like very, very clear.

• 13:56Speaker 00

  On autonomy or other features?

• 13:57Speaker 01

  On autonomy and other features.

• 13:58Speaker 00

  Yeah.

• 13:59Speaker 01

  All around, all around, like, super impressive. I think if you look back at Elon, some of his statements about, he's seen the same thing over the years. And then you can just honestly look at the sales of Teslas in China, it's not that impressive. It's because if you go there and you do the comparison, the local stuff is really good. Now, the stuff that's not talked about often is there is subsidies that are happening. The Chinese consider this to be a national asset. They look at, you know, the car business as a, as we look at defense.

• 14:26Speaker 00

  Yeah.

• 14:26Speaker 01

  And they're willing to subsidize it, because fundamentally, the car business, in some ways, and in these manufacturers, are there like jobs programs.

• 14:34Speaker 00

  Yeah.

• 14:34Speaker 01

  So, you have hundreds of thousands of, you know, people who work in these industries, and then service these industries, and they create entire economies around cities. We think about like a, you know, one manufacturing plant, let's say it's a billion, $5 billion investments, the large OEMs have Dozens, sometimes hundreds of manufacturing plants. And these are, when you talk about global industry, you don't get any bigger than in the industrials and then automotive. And so China's like, okay, if we're going to be a superpower, we need to have a real industry. I think you fast forward five to 10 years, you're going to see a heavy consolidation. You're seeing the early version of that we saw, you know, already, which is the like the Huahues of the world are starting to become this like new generation supplier. I mean, if we could be a company,

• 15:17Speaker 00

  Yeah.

• 15:17Speaker 01

  we would be Huawei. Like it's a super impressive company. Not on the, everyone thinks about Huawei on the mobile phone side. We're talking about their automotive business. They provide everything and they provide a platform, a reference vehicle. And allows the OEM then to really focus on what they're good at, which is manufacturing, marketing, branding, distribution

• 15:35Speaker 00

  Yeah.

• 15:35Speaker 01

  and consumer experiences. And so it's, that's a really, it is a really interesting ecosystem to keep an eye on. It's the most dynamic ecosystem on the planet.

• 15:43Speaker 00

  How do you think that impacts sort of the economies of some of these countries? So if I look at the US, I think we're very lucky. I have a Tesla, I think it's a fantastic car. And I think it's almost like a local champion in terms of EV and autonomy.

• 15:53Speaker 01

  It's an American car company, by the way. People forget that. People only think the American car companies are in Detroit. Tesla is an American car company.

• 15:59Speaker 00

  Tesla is an American, is an amazing American car company. In Europe, it feels like they're much more threatened by the Chinese OEMs in part because they're letting them enter the markets, right? In the US, there's heavier tariffs around it or, you know, other means to sort of prevent access in Europe. It feels like certain markets are pretty wide open and you see BYD and others gaining share really aggressively. Is that a, how do you think about that from a policy perspective in Europe? And is that, you know, not going to really hurt the economies there?

• 16:24Speaker 02

  I think there's

• 16:25Speaker 01

  A resident European.

• 16:28Speaker 02

  I mean, there's always a question of right, trade and trade deficit. And so, I mean, European companies as well as American companies, they benefited a lot from the Chinese market, right? General Motors, Volkswagen, just as an example. So they've, over the years, they've made a lot of profit in China. And so just from, let's say a fairness perspective, like you can see it just by trade, right? It makes sense that there's, there's some balance that that can be achieved there. I think in the long term though, absolutely these, these questions always arise. And you hit a point where the volumes become high enough that countries will demand that you manufacture there. And then once you start manufacturing within the country, oftentimes those cost filters actually go away. Like whether, no matter what the brand is, if it's a local brand or an international brand, if it's manufactured in the same place, you usually end up with a product that's going

• 17:13Speaker 02

  to be of similar cost.

• 17:14Speaker 01

  Yeah, when you talk about like, if you look at, so if you open your car, you can look at the content of the car in your door frame. It says, this is, where is it made and how much of it's made where? This is not a new thing. In Michigan and Detroit, this is a 50-year-old, 70-year-old debate. And because their real implications are if you just buy all the subcomponents from foreign countries and then you assemble them in Detroit, that doesn't mean it's made in Detroit. My slightly more caustic view or aggressive view on Europe is Europeans are a little bit of asleep at the wheel. In the, in the sense of, I think if I could inject something into the brains of the, of the leadership of whether it's the EU commissioners or, or the, the industrial families and leads

• 17:57Speaker 01

  is, it's like what I was told, you gotta fight.

• 18:00Speaker 02

  Yeah.

• 18:01Speaker 01

  And there's almost this like, oh, it's gonna be, you know, the Chinese are so cheap. The reality is, it's not. There, there's a finite amount of dollars it requires to make something. And when these factories are so automated, that labor arbitrage, which historically was the reason why China was really, you know, cheaper, goes away. So a fully automated factory in Romania versus, you know, China, they're not as different as you think. I, you know, I fall into the category of you for, at least for America, being an American citizen is, we can't just be a consuming state. We have to build. Because with that building, you also provide jobs and expertise.

• 18:37Speaker 02

  Yeah.

• 18:38Speaker 01

  And there are countless case studies of American companies that offshore are manufacturing. This is, you hear the Detroit annoyance here, right? Offshore are, are manufacturing to other countries, including Mexico. Then the knowledge of building things is there. And then if you're Chinese, you're Mexican, you're Vietnamese, you're Indian, whatever you are, then you're like, also I just need a little thin layer and this vacuum company is now.

• 19:02Speaker 02

  Yeah.

• 19:03Speaker 01

  I get all the profits. I don't need the 25 employees sitting in San Diego marketing this thing. Growing up in Michigan, I remember this very, because I was entering the workforce at the time in the late 90s. And the view was at the point was like, oh, this is globalization and it's okay. And it's like, we have to have a bit more strategy because. The reality is, everybody on the planet has some strategy.

• 19:25Speaker 00

  Yeah.

• 19:26Speaker 01

  Thailand has some strategy. Brazil has some strategy.

• 19:28Speaker 00

  The US basically made a conscious choice to allow its industrial base to leave, even though it was phrased in a different lens, but that was a conscious choice.

• 19:36Speaker 01

  It's our naive view that the system will just take care of itself. It doesn't, because everyone doesn't play by the same rules globally. And I think in industries like us, if you're an AI company in San Francisco, and you supply developers that are mostly based, that doesn't matter. For us, we're a truly global company, so we think about a lot of these things all the time. The advantage that we have, which is extremely significant, is we have the best technical talent still. You hear about DeepSeek, you hear about these other things, they're absolutely real. It's not, again, you can't discount it. And by the way, it's not that for America to win, China has to lose.

• 20:13Speaker 00

  Yeah.

• 20:14Speaker 01

  The UK won, Germany won, Japan won, and the US still won. We can live in a world where everybody is winning. We don't have to have this conflict, but we do have to agree on the terms of engagement. And I think that's where my view always is, is like, if we're on the same playing field, in that context, Silicon Valley has some huge advantages. The best of the world come here. So it's like, just as much as we complain about China, Mexico, Germany, we want to attract that talent and to live here and to build here. And we're still, I think, the best in the world.

• 20:49Speaker 00

  So if we think ahead and we say, okay, we went through a period where parts of our manufacturing and industrial base were effectively exported to other countries under labor arbitrage effectively, or cheaper labor. A lot of know-how went out of the country and kind of stayed there. And in some cases, it almost feels like we've lost some of that know-how over time. Does robotics and autonomy and the automation of factories allow a moment in time where we can bring that back? A factory arbitrarily anywhere. Is cost competitive based on automation versus labor? Is this a moment in time? And how should people act on that from a policy perspective? Or how should, how should we be thinking about that more broadly from the perspective of starting companies or innovation?

• 21:22Speaker 01

  We don't know policy. We're not, we didn't go to, you know, Kennedy School, went to engineering schools. So, but I think broadly speaking, I think it is a huge opportunity. And you see companies like Rebuild and Androl are taking advantage of this reality. So I think if I'm a founder, it's an extremely inspiring time because everybody is seeing this opportunity and willing to fund it. You know, these companies like Rebuild Androl are funded by classic venture capitalists. They're not funded by some PE shop that's doing a roll up in New York.

• 21:51Speaker 00

  Yeah, makes sense. And then I guess the other piece of it is you all spend, I mean, you're a very profitable company, which has always been very impressive, given how much you've scaled the team and scaled your efforts, but also you spend a lot on models and on the development of different AI-based tooling. Could you talk a little bit more about what sorts of models you've been training? How you think about the world there, where that's heading?

• 22:10Speaker 02

  Yeah, so we do a lot of work in autonomy, right? So we talk a lot about our work in L4 trucking and we've some really interesting things there right now, largely in Japan. That itself is a huge, a huge part of this and there's an awful lot of data and model training that goes into that. We also do interesting work in autonomy in aerial and maritime as well. And so fighter jets and drones and boats and, and all of these things, there's a huge data problem. Data is not nearly as easy to collect in some of those domains as it is, let's say, on the on-road domain. So we've had to do a lot of work over the years on how do we actually collect this data and make this useful and, and, and put it into formats and such that you can actually get good performance out of the autonomy models that you train.

• 22:53Speaker 00

  Is there anything you can share in terms of some of the approaches you've taken there?

• 22:55Speaker 02

  Yes, there's a few things I think that we, that we've done that have been super advantageous over the years. So we've been investing in synthetic data for quite a while. And the details of how that works has evolved and actually gone through several generations in our own tech stack, right? Traditionally, you had a very computer graphics heavy approach. Now you have approaches that use things like Gaussian splash and diffusion models to do really interesting things with, with synthetic data. And you can extend synthetic data into A bunch of other domains, including into like classified domains for for defense, where you can actually get some some really interesting advantages. And then, I'd say broadly in autonomy, you always have this thing where if you have interesting data, as the machine learning techniques change and as let's say the the new research papers come out,

• 23:40Speaker 02

  as long as you have that corpus of data, you can do really interesting things. And we we've seen our own technology evolve in that direction as well. Where we were actually able to use data that we collected even years ago to do things and get levels of performance that would have been previously impossible, just using newer techniques applied to the data that we have.

• 23:58Speaker 01

  I think one technical strategy that's been quite advantageous for us is, I think we always wanted to wait in the way, I mean explicitly, we talked about this, wait in the wings until like the autonomy ecosystem converged on a handful of techniques and this kind of post transformer boom. And then just seeing like you can ride in V13, the Tesla FSD product, right? Or you can go to China and ride in these and you're like, this is it.

• 24:22Speaker 02

  Yeah.

• 24:22Speaker 01

  This like can't be, there was a, I mean, you're not talking about three years ago. There's a huge debate on if, you know, end to end, in the many ways, and then this marketed. It's a marketing, you know, phrase as much as a a technical phrase. End to end camera heavy systems where they are, they're gonna, are they gonna be able to perform? I think the big, big thing, so now we're on the same page that autonomy is definitely gonna happen.

• 24:48Speaker 02

  Yeah.

• 24:49Speaker 01

  And when you're, you know, playing in the ecosystem, you also know kind of which way it's gonna happen.

• 24:54Speaker 02

  Yeah.

• 24:54Speaker 01

  And so now the question is, who's gonna monetize this? And who's gonna take advantage of it? And, you know, the Waymo stuff is super impressive, but it is worth putting an asterisk, the business model hasn't been figured out yet. The Tesla business model has been figured out. And so people kind of interchange these and this and they often talk poorly about the Tesla system by saying, oh, it's dangerous and it's not as good as a Waymo. Yeah, but the like much more likely to continue to exist. And the big story of the many dozens of autonomy companies actually isn't, I mean, they know as much as us and we know as much as them and simply because we recruit many dozens of people from, you know, I think we're probably Waymo's biggest employer outside of Waymo and Cruise and Tesla, all of these these organizations.

• 25:39Speaker 01

  So it's not like there's this false. View that there's some secret in how to build self-driving tech. Actually, within the business, everybody knows how to build it. And it's just like, you know, in these podcasts where you're always going to stay fairly high level. You're not going to get into that, you know, into that depth. But generally, the ecosystem's conversion on what those techniques will be. That was not the case just four years ago. A lot of debate. And then as we saw, okay, there's a convergence. This is when you really, you know.

• 26:07Speaker 00

  Yeah, I think so. You folks basically waited for that moment in time where the techniques in the purchase crystallized. You're like, okay, now's our moment in time to really enter the market.

• 26:13Speaker 01

  And stay alive until then.

• 26:14Speaker 00

  Cause we don't have to do the research or we do different types of research, but we can kind of wait for state of the art and just jump straight.

• 26:20Speaker 01

  I mean, we're really, you know, in automotive terms, there's research, advanced engineering and production. We've always kind of hovered around advanced engineering where it's like, you don't want to just try things and have a group of 50 to 80 researchers. We are publishing papers, but at the same time, you can't just be like some system integrator.

• 26:40Speaker 00

  One of the things that I find really interesting about the adoption of generative AI is the bar is often higher for these AI systems than it is for people. In terms of how good the output has to be. And it's kind of striking, right?

• 26:54Speaker 01

  If you ask somebody like, how should I do a series A pitch?

• 26:56Speaker 00

  Yeah.

• 26:57Speaker 01

  You'll take whatever they say, because they say it's a confidence. And then like, you know, you do it in chat, you'd be like, no, actually this isn't right.

• 27:03Speaker 00

  You look at self-driving and you look at the safety record. The claim for many of these companies is that you decrease death and injury dramatically by just adopting self-driving. And yet from a regulatory perspective, there's all sorts of hurdles to adoption. What is the right bar for safety for autonomous systems for, you know, the like, should it be at human level? Should it be dramatically better? Like how much better?

• 27:26Speaker 01

  Yeah, there's a colloquial and then there's like a regulatory answer. And the colloquial answer, my view is in that many people in the industry is most of the systems are pretty good right now. Like they're already certainly way more. If you just like, you use a proxy metrics, like meantime to disengagement on a way more, that's tens of thousands of miles.

• 27:47Speaker 00

  Yeah. It's in 30,000 miles. Also like, I guess people measure the number of accidents, injuries, fatalities per mile, right?

• 27:53Speaker 01

  Yeah, exactly. So like all of those. The Waymo is way better, not even kind of close. And frankly speaking, a lot of these L2 plus systems are a lot better. We're in a capitalist environment and there's liability involved. And so the question really fundamentally emerges is who's liable for what? And so in the historic debate of roughly what you have is like, am I the driver who's responsible? Or is the vehicle responsible? And that can be actually boiled down to, is there somebody in the driver's seat?

• 28:21Speaker 00

  Yeah.

• 28:22Speaker 01

  When you're, when, you know, the Waymo has taken that view and a Waymo engineer would always say, well, FSD is nothing like us. You can't have a Tesla with nobody in the driver's seat. Elon is trying to, trying to change that.

• 28:32Speaker 00

  But there's a liability piece of it, but separate from that. I mean, the media headline is always Waymo causes an accident versus Waymo saved N lives.

• 28:39Speaker 01

  Yeah, exactly.

• 28:40Speaker 00

  Or Tesla caused an accident versus Tesla. And so.

• 28:42Speaker 01

  I think it's human nature as well. We can't be afraid of our own shadow. Progress requires some of that risk. And I think part of this having.

• 28:51Speaker 00

  The key thing is the risk is lower. In other words.

• 28:53Speaker 01

  Way lower.

• 28:54Speaker 00

  It seems like a lot of the media headlines are actually the opposite, where there's a purposeful amplification of danger, even if the danger is decreased, right? And we've seen that a number of other instances too. So it does, it does feel like a purposeful approach by the people writing the news.

• 29:07Speaker 02

  I think part of the issue is that sometimes the mistakes that the autonomy systems make, they're just not the same mistakes that a human would make. And so then when it makes the, the news, a human will see that and be like, well, that's, that's completely crazy. Like, how could that happen? But it's, it's more of a result of the technology itself.

• 29:25Speaker 01

  Yeah, but I, I think, you know, just going back to the journalist kind of media view, we are what our incentives are, right? And the incentives there, a journalist writing nine people died of pedestrian fatalities last night in America. Well, that happens every single night.

• 29:38Speaker 00

  Sure.

• 29:39Speaker 01

  It's not interesting. And so it's just, just trying to get clicks and the clicks are, you know, the Uber ATG car in Arizona killed somebody.

• 29:46Speaker 00

  Yeah.

• 29:47Speaker 01

  Is just, people talk about that all the time, happened years ago. And it's like, you know how many hundreds of people have died in America from human driven vehicles? But that's on the, let's say on the cultural colloquial side. On the regulatory side, the issue that you, we have in broadly speaking, mature, developed economies is the bureaucracy is, the bureaucracy's goal is to, you know, create more bureaucracy.

• 30:09Speaker 00

  Yeah.

• 30:09Speaker 01

  And so a place like the national, you know, a transportation authority is, is going to, Department of Transportation is going to. Say, hey, for airbags, we did this.

• 30:20Speaker 02

  Yeah.

• 30:20Speaker 01

  So this new technology comes and we definitely need to regulate it in this way. Take a more extreme example. Take a country like Spain or Italy, where the government is really, you know, very bureaucratic.

• 30:30Speaker 02

  Yeah.

• 30:31Speaker 01

  Imagine you take Italy and just remove the government. I'm not saying, you know, a revolution, but remove the government and say, we're going to build these government institutions for the things they need today. Like that defragging and that the garbage and maintenance, that garbage collection never happens. And these systems, so it's like, a lot of it is just illogical.

• 30:48Speaker 02

  That makes sense. And you're talking about the programming term garbage collection.

• 30:51Speaker 01

  Yeah, exactly.

• 30:52Speaker 02

  And yeah, I would just add that.

• 30:54Speaker 01

  No garbage collection.

• 30:58Speaker 02

  These autonomous systems, they, they will be significantly safer than human drivers and provably so. And actually, a lot of the technology that we develop as a company contributes to this across the industry. And we do a lot of work in verification validation, which is around this like proofability aspect of that something is safe. But it's never going to be perfect. Like you can fundamentally, you can easily design traffic scenarios where the autonomous vehicle gets into an accident by no fault of its own. And so you can't have that bar, right? It's, it has to be.

• 31:26Speaker 01

  Imagine this alien shows up to the planet and it's like, there's two ways of transporting people. One is this, this one that the computer does and it virtually is perfect. And the other one is this like half asleep 18 year old teenager.

• 31:42Speaker 02

  Yeah.

• 31:42Speaker 01

  Going to Starbucks shift after partying all night.

• 31:46Speaker 02

  Yeah.

• 31:46Speaker 01

  It like no, no objective, you know, would, would, would pick the, the teenager. The same thing, I think, even you look at like AI in the, in the, like, why is chat GPT so loved? Search exists, there's all this legacy looks like, it's just a better product.

• 32:02Speaker 02

  What do you think is a timeline to large scale adoption of autonomy? So obviously Waymo is growing really quickly right now. Tesla is continuing to push forward on different autonomous systems. Like, how many years away do you think we are from?

• 32:13Speaker 01

  General availability?

• 32:14Speaker 02

  Yeah, I don't know. And frameworks changing from a regulatory or other perspective. So it's just kind of very, very nuanced question because it, it, it's a different and different country. So let's, let's kind of bound it to America for now, and then we can talk, talk with us to the world.

• 32:28Speaker 01

  Every vehicle manufacturer in America that ships a real amount of cars is building some version of a Tesla competitor. So I think general availability on a FSD-like system over the next five years is going to be common, just like navigation systems. Now, how good those systems are and what the price points are, those will all define the breadth of those features. Waymo will continue to expand cities. I think we're seeing Waymo in 2025, what was promised in like 2018, 2019, which is we're going to do 40-city rollouts and that's going to be the next, you know, two or three years. The example really to think about this is, let's say we're sitting in 2007, the iPhone just launched, and you asked me, "How fast are we going to see the iPhone in lots of people's hands?" It's like, slow, slow, slow, and then suddenly everybody has it.

• 33:12Speaker 01

  I think there's going to be some version of that. One, you know, again, anecdotal example is, I go to LA and, you know, it's not Silicon Valley. Everybody knows about self-driving. That was not the case like 12 months ago. And why is that? Because the Waymos are everywhere. And so it's like, for the first time, when I explain applied intuition, people are like, "Oh, I get it." Before, they were just like, you know, all these like, almost like Luddite kind of responses. And it's every person who sits in a Waymo, they just, they convert over to, okay, self-driving is a thing. The natural next step is, okay, why doesn't my car have some version of this in my passenger car? So I think like, the whole, the whole window of expectations for self-driving was like 20, let's say, 15 to 2020, and that's really like 2025 to 2030.

• 33:56Speaker 01

  But I think general availability is coming, coming quickly.

• 33:59Speaker 02

  I just add, like, as an engineer, I think the next five years are probably the most exciting period imaginable, because whether it's automotive, defense, so think planes, ground vehicles, boats, all, general robotics, like all of these things, right now, we don't see a, a real rate limit to how fast we can advance technology. Like, we're just making progress every single day on all of these things, and we don't see a point where that progress is going to slow.

• 34:28Speaker 01

  Yeah, it's definitely like, it's funny from like, an engineering perspective. It's, it's, it's sad and somehow that like, all the people got excited like, when it wasn't. Like, this is the best time to work on self-driving. Self-driving, like it's ready, it's ready to go mainstream. And frankly speaking, to monetize all the billions spent on Cruise and Argo and those companies, all of that didn't go for nothing. That's in people's heads and then those people continue to work on the technology. So I think it's a super, super.

• 34:55Speaker 00

  It's kind of a magic moment in time in the industry. And that that's true of autonomous systems. And obviously it's true with all the sort of language models and chat GPTs and all.

• 35:04Speaker 01

  Like maybe, maybe extrapolating from that question is like, when will this stuff be completely commoditized? Where it's like your expectation for self-driving is zero dollars.

• 35:14Speaker 00

  Yeah.

• 35:14Speaker 01

  I expect it for free. Like right now, when you get in a car, you're not like CarPlay, I need to pay like three dollars a month. No, it's my phone is just being projected in the heads up display as expected to free. I think at 2030, 2035, you're going to have deep downward pricing pressure. Like to where it's like, the expectation is going to be close to free. And I think the, not to pick on Waymo a lot, but it's just the easy one. Cause we're in San Francisco, see all these cars around and it's, it's, it's continued to, to still exist. That's a big question, right? You spend 12, 15 billion dollars on developing this technology and it commoditizes before you monetize it. That's a super, super scary situation.

• 35:55Speaker 01

  And I think Waymo's public offering and Waymo's business model are going to be big, big questions. If they pull it off, it's phenomenal.

• 36:05Speaker 00

  It feels very transformative in all sorts of ways in terms of, you know, the magical experience of getting in the car and, you know, everything else. I think they've done a really good job and.

• 36:13Speaker 01

  I mean, I would go on a limit, say like the self-driving revolution will have as much of an impact as the LLM revolution is happening. Like we, because not everyone grew up in Detroit, you don't appreciate these little things. The entire, all the way where everything's designed, where hospitals are, how pregnant, it's all based on cars. Cars are this, like it's this invisible thing that it's like electricity. It's invisible things around you impacting every single thing. Can you imagine going for a month and not interacting with a car?

• 36:42Speaker 00

  Yeah.

• 36:43Speaker 01

  Or a week?

• 36:45Speaker 00

  Yeah, no, it's interesting because I remember when all the first wave of self-driving stuff was happening. There'd be these dinners and conversations around self-driving adoption. At the time, people were really worried about truck drivers being displaced. And I remember meeting with different congresspeople to talk about that specific issue that they're worried about. And to your point, I think people have kind of underthought the degree to which urban design is, in the modern world, designed around cars. It's where you put the parking lot.

• 37:09Speaker 01

  Everything. Grocery stores, how far they are from neighborhoods, all of it. How big can they be?

• 37:14Speaker 00

  Yeah.

• 37:15Speaker 01

  How small should they be? If you look at a New York and why New York is the way New York is, it's because actually the car is not the main thing of your life. It's actually the subway and walking. And it's, it's, there's no other place like New York.

• 37:27Speaker 00

  Yeah. And eventually what you should end up with is lots somewhere outside of the city with a bunch of cars. And then when you need a car, you push a button and it comes in and it grabs you and it takes you wherever you want. Or, and you probably need fewer vehicles.

• 37:36Speaker 01

  Yeah, I think, I think, well, that's where there's a debate. I think you see this with mobile phones is as calling became free. Remember, calling used to be minutes and texts used to cost 10 cents a pop.

• 37:45Speaker 00

  Yeah.

• 37:46Speaker 01

  Texts and calling haven't gone down. They've gone way through the roof. Orders of magnitude higher than we ever communicated in 20 years ago. Only 20 years ago, not a long time ago. So there's a version that in 20 years from now, we have an order of magnitude higher. Imagine if everyone had a private jet.

• 38:01Speaker 00

  Mm hmm.

• 38:02Speaker 01

  Do you think people are flying less around the world? More, they're flying a lot more. So I think there's a version that you have a lot more actual miles driven.

• 38:10Speaker 00

  Yeah. And the inside cabin of the car can change dramatically, right? You can work out in the back if you want to, if you're in the right.

• 38:15Speaker 01

  But that doesn't necessarily mean there's more traffic. It doesn't necessarily mean you need more parking lots. You could actually, I'm, you know, I'm in the optimistic view of the world. And this is the, again, my view, the issue with, with Europe and sometimes is, instead of the knee jerk reaction being no, the knee jerk reaction should be yes. This will usher more productivity. It brings up, you know, we shouldn't have huge percentage points of our cities just empty parking lots. You make those, make those parks is really nice. And I think self-driving can, can help it beyond the just, it's literally less humans die.

• 38:46Speaker 00

  So you've kind of laid out your vision in terms of economy and sort of the reworking of cities and some really large scale things that are very exciting that are coming and how you guys are going to help power that for the industry. How do you think about using AI in other parts of your business?

• 38:57Speaker 01

  Yeah, and specifically on tooling and, and, and in the vehicle OS that, or let's say in cabin experience.

• 39:05Speaker 01

  Number one, the way software is being developed, we see this through all the, you know, the coding assisting tool, assistant tools, all of that needs to happen in automotive and these other industries. So we're at the heart of that. That's why we're changing our name to Windsurf Automotive.

• 39:22Speaker 02

  On that, there is a lot of very interesting constraints to the problem when you talk about building software for vehicles. And so I think we, as a company, we're in a really interesting spot where we both have the AI technology and we also have the expertise and all of these really interesting constraints on the software that's actually being developed. So I think you'll see some interesting stuff from us.

• 39:44Speaker 01

  Yeah. And then on the, on the vehicle OS side or in cabin side, you know, the vision there is a pretty straightforward, you know, which we're working on now, which is you take a, you know, a mining operator and he walks up to his Komatsu dirt mover, this giant machine that, you know, moves dirt basically 24/7. That machine should know who that person is. And as they enter, that person can have that, you know, conversation with the machine. The machine sees it, it sees the world around it, it's multimodal and it can sense it. What really happens on a mine in terms of safety is a little bit of intelligence goes a long way. And so what you're starting to see is emerges this teaming. And I think like that

• 40:22Speaker 01

  type of AI experience is, I think is more fundamentally important than even just getting, you know, Waymo going or just getting out to, because that's the hidden part of the economy. Whether it's writing software for the engineering tools that write software for safety critical systems, or the stuff that happens beyond your eyes. And that's the stuff I think we're, we're super excited about without even going into defense. Defense is its own universe. The short version of what, you know, what we're seeing in, in defense is, has been talked about by others, but we're certainly experiencing it as defense is moving from a one person, one machine to one person to many machines.

• 41:00Speaker 02

  Uh huh.

• 41:00Speaker 01

  The way to think about intelligence within these other domains and dominions isn't just, well, how can you use an LLM?

• 41:07Speaker 02

  Sure.

• 41:08Speaker 01

  You know, there. It really is like, well, what does intelligence look like for a warfighter in the field who has a couple hundred Drones.

• 41:15Speaker 00

  Yeah.

• 41:15Speaker 01

  And how does he say, okay, I need this there and I need it fast now to get information from other, you know, other parts of the force and other forces and other countries to actually make that decision very, very quickly.

• 41:26Speaker 00

  Yeah, then the problem space is all about autonomy for the individual system. It's about collaborative autonomy for the swarm. And then it's about the the comms and the RF between all of these systems and and how the data is actually moved between them. And so really, really interesting problems. And yeah, we've got some really interesting.

• 41:40Speaker 01

  So like, you know, we think vehicle intelligence trademark is is like this. Yeah, I, you know, it's funny, you know, as founder, for the founders who are listening, it's super rare to actually work on stuff that you, you know, you really love. And it's no disrespect to people who work on like the dentist CRM.

• 41:56Speaker 00

  Sure.

• 41:57Speaker 01

  But it's like a lot more fun than that to CRM. Because it's like taking AI in a way which is not just the chat box.

• 42:05Speaker 00

  Sure.

• 42:05Speaker 01

  Which is great for all the reasons it is, but it's it's like beyond that.

• 42:08Speaker 00

  So, you know, one of the things that I think is unique about a pod intuition is you also have a very large and sort of deep bench on the design side.

• 42:14Speaker 01

  Yeah.

• 42:14Speaker 00

  And you think deeply about these sort of forward looking experiences, both in cabin and more broadly. Could you talk a little bit more about both that team, but also that future that you're imagining that's coming?

• 42:24Speaker 01

  Yeah, I think if you, you know, in the in the chat GPT world, there's some design there, obviously, and it's significant. But once you're truly like multimodal in the in cabin consumer passenger vehicle, there's lots of screens and just the way you interact, it's not just a button and a voice. And so it's still some, some of that work is under wraps, but, um, you know, we, we're doing real work there. If people are particularly interested in that, especially from a designer's perspective, and you're tired of doing like login screens for Google like this. This is like, or like, you know, spending six years on just changing the shade of the sign in. So we're, we're really working on fundamental experiences, like HMI in the way that I think a lot of people at CMU when they're doing.

• 43:08Speaker 00

  So very deep experiences where you walk into the car, it somehow recognizes you, the seat moves into place for you versus somebody else who drives it sometimes, like everything auto adjusts.

• 43:17Speaker 01

  Yeah.

• 43:18Speaker 00

  Your playlist comes on, whatever it is, it kind of logs you in.

• 43:21Speaker 01

  Yeah, and just like how it interacts with you. Design isn't only. Pixels, it's an experience, right? Um, yeah. So I, I think design is, design is a huge area for us as well.

• 43:30Speaker 02

  And then I guess from a, a team or hiring perspective, to your point on great things to work on, are there specific types of profiles you're looking for, or just hiring across the board? Or how are you thinking about that right now?

• 43:39Speaker 01

  I'm fortunate to be hiring across the board. The, the, the, uh, obviously I think, I think we've, we hire from all of, let's say, the, the spectrum of like pure researchers all the way to folks who are working on the other end of just like infrastructure implementation. I think we have like over a hundred roles available.

• 43:54Speaker 02

  Yeah. I mean, I, of course, everyone, ourselves included, is, is always looking for great AI talent. So that, that goes without saying, but, um, we also have, I would say, an extremely deep appreciation for just really strong software engineers that are also just deep in the, the operating systems and, and let's say, systems engineering realm. Um, we, we work on stuff that's very important and very, very technical. And so, yeah, people who like hard problems like to work at a project.

• 44:17Speaker 01

  Yes, our, our company is, um, it's something we're proud of, is like 82% software engineering. This is, is extremely concentrated. It's a, it's a very, very technical company. Um, and we build products, you know, we're not doing services and things like that. So that, that allows, you know, if you like that kind of stuff, you like building products, um, you know, we're, I think, I think we're an interesting place. We, we might not be that cool though.

• 44:41Speaker 02

  You got the jean jacket.

• 44:42Speaker 01

  I got, I got the, I got the car jacket. That's about it. Last time you'll see this. It's like a costume.

• 44:51Speaker 01

  The best evidence it's a costume is I'm uncomfortable. It's like I'm wearing like a clown outfit.

• 44:57Speaker 02

  Great. Well, thank you so much for joining me today.

• 44:58Speaker 01

  Yeah, thanks for having us.

• 44:59Speaker 02

  Thanks for having us.

• 45:00Speaker 01

  Awesome.

• 45:02Sarah Guo?

  Find us on Twitter at NoPriorsPod. Subscribe to our YouTube channel if you want to see our faces. Follow the show on Apple Podcasts, Spotify, or wherever you listen. That way you get a new episode every week. And sign up for emails or find transcripts for every episode at no-priors.com.