Self-Driving's Breakout Moment — With Kyle Vogt

0:00The CEO of Cruise comes on to talk about its rapidly expanding self-driving car efforts and how quickly autonomous driving can take hold. That and more coming up right after this. The truth is AI security is identity security. An AI agent isn't just a piece of code. It's a first-class citizen in your digital ecosystem, and it needs to be treated like one. That's why Okta is taking the lead to secure these AI agents. The key to unlocking this new layer of protection? An identity security fabric. Organizations need a unified, comprehensive approach that protects every identity, human or machine, with consistent policies and oversight.

0:37Don't wait for a security incident to realize your AI agents are a massive blind spot. Learn how Okta's identity security fabric can help you secure the next generation of identities, including your AI agents. Visit Okta.com. That's O-K-T-A dot com. Capital One's tech team isn't just talking about multi-agentic AI. They already deployed one. It's called Chat Concierge, and it's simplifying car shopping. Using self-reflection and layered reasoning with live API checks, it doesn't just help buyers find a car they love. It helps schedule a test drive, get pre-approved for financing, and estimate trade in value.

1:17Advanced, intuitive, and deployed. That's how they stack. That's technology at Capital One. LinkedIn presents.

1:29Welcome to Big Technology Podcast, a show for cool-headed, nuanced conversation of the tech world and beyond. I've been tooling around San Francisco in self-driving cars for the past couple weeks, and I can't wait to talk to you about the experience and bringing on someone who can tell us just how quickly this stuff is going to make its way into cities across the U.S. and the world. We're joined today by Kyle Vogt. He is the CEO of Cruise. Kyle, so great to meet you. So great to have you here. Thanks for coming on the show. Honored to be here. Thanks, Alex. So let me just start by saying my observation after trying this stuff out is that it really feels like we're at an inflection point with self-driving.

2:07I was a tech reporter in San Francisco 2015, 2016, when everyone said this stuff is right around the corner and it's just taken a little bit longer than expected, but it does feel like this is a moment where this is starting to really take hold and go mainstream in a way that I think is really underappreciated. I think you had a stat on LinkedIn that you're increasing your eyes like 49 % every month. So just tell me a little bit about whether I'm right about that observation and if so, why? You are. I mean, you see it in all the announcements we put out and in the growing size of the fleet and the number of people who have ridden in AVs.

2:46But taking a step back, the reason for that, and for context, I've been working on this at Cruise 10 years now, so a full decade. Why are we just now seeing this? The answer is this isn't a product you can put out if it's like halfway there or partially there. There's no like launching early kind of thing. We set the bar, you know, the safety bar really high inside of Cruise. And we wanted to be convinced that by deploying this service, we're going to have a positive impact on day one in terms of road safety and benefits to society. And so, you know, it took the very first prototype self-driving cars, they maybe go around a couple blocks, or if you've tried driver assistance systems, they feel like they work.

3:25But that's the difference between, you know, working 99 % of the time or 99.999, however many nines you think is appropriate. And that, you know, those extra nines of reliability are what take the majority of the work. That's polishing the technology, making sure the safety performance is good, making sure the rides are comfortable. And then also making sure we have systems in place to handle with the reality of operating vehicles on public roads. So that's interacting with first responders, detecting and pulling over for emergency vehicles, handling pop-up construction zones. There's a really long list of stuff you have to do to even reach the baseline performance of a human driver.

4:02But once you do, the floodgates are open. Once you've reached that point at which, you know, you outperform humans in aggregate on a safety basis, at least with regards to collisions, as we've measured, and you feel like you're able to provide a service, then the constraints become very different. It's no longer this science and technology problem. It's how do we scale? How do we build more infrastructure? How do we manufacture more cars? And we finally, after 10 years, reach that point. So would you say the floodgates are open now? They are at this point. Well, I I shouldn't say completely open.

4:35They're open in ways where we can launch in some cities of the US in some areas. So we put limits on our technology today, things like top speed, the weather that it will operate in. And those are tightly validated and controlled, the ability to manage our fleet within those constraints. But that actually opens up a lot of cities in the US where we can operate a reasonably sized fleet. And as new technologies and improvements come along, like the ability to operate at higher speeds or in tougher weather, that will just further open up the areas in which you'll see these vehicles operate. So can you describe a little bit like the magnitude of what we're seeing?

5:10I mean, I quoted one of the stats that you had shared, 49 % increase in rides month over month. But like, can you give our listeners just a sense as to like, because like, it's one thing to say, like the magnitude is increasing, but like, I'd love to hear like any context you can share around that. So, you know, we have number of rides, but like number of cars, whatever stats or a number of cities, like talk a little bit about like what this inflection point looks like on your end. Well, it is growing exponentially. Like, you know, from an engineering standpoint, I like to think in terms of orders of magnitude.

5:45So each time you 10x the scale, you know, you've got some new technologies you need to build or new systems to handle that increase in load and volume. So for context, last year, we're operating tens of AVs. We're currently operating hundreds, you know, almost 400 concurrently at peak. Next year, it'll be thousands. And then it'll kind of continue at least 10x growth every year for their foreseeable future. So that's a big increase, you know, if you're just to plot that curve. But it's still relatively modest compared to the number of, you know, total vehicles in rideshare or the total vehicles, you know, operated on the road in the U.S.

6:20today. And our goal, look, if you can put a self-driving mile out there that's safer, it's cleaner because it's powered by electricity instead of combustion engine and probably more fun and more convenient for people, we should be doing that for as many miles as we can. So it'll start with kind of this robo-taxi market and then expand eventually into personal vehicle ownership. Okay, I definitely want to get into that. And also I see 10 ,000 driverless rides per week. So talk a little bit about the scale. like what was that at the beginning of the year? I don't know off the top of my head, like on a per weekly basis, maybe, maybe hundreds or maybe a thousand or 2000.

7:00Wow. So it's already 10 X just from the beginning of this year. Yeah, it has. And that's driven by a number of factors. Obviously we started by operating our AVs, you know, late at night and overnight and early in the morning. And there just aren't that many people that want to take rides at that time, but that is the least risky environment to start in. And so we wanted to gain confidence and make sure everything operated as we expected it to before expanding into the daytime hours. But you take the same set of AVs, if they're only available in the middle of the night, and then you make them available during the daytime, just that same number of AVs is going to see 10 times as many rides.

7:34Right. And I think we should talk a little bit about like what the experience is like, because there's going to be some listeners who've actually been in these self-driving cars, but it's actually very tough to grasp unless you've actually ridden it before. And I think it's worth like reflecting a little bit about what this is actually like. One of my first observations here is just the progression is so interesting where you first get in, you're a little bit nervous. You're a little bit shocked that this thing has happened. Then the second ride that you take, it feels a little bit routine. Then the third ride you take, it's almost as if you're not paying attention to the fact like it becomes normal so fast by the third ride.

8:13And I think by the fourth ride for me, it's just like, I don't ever want to drive in any other way. The ride feels safe. It's private. You know, you know exactly where the car is. You, you know, you can't get into a fight with the robot. Not like I often fight with the drivers, but like, it's just, there's something about it that it's just, it really catches on and seems like it makes sense so quickly. So I might, do you think I'm capturing it? Right. I mean, I think I'm saying a lot of nice things about like what this experience is like, but it's for a reason. Like it's, it feels safe and smooth and easy.

8:46And I don't know, it's like, and like the future. Yeah. Well, look, I'll start by being humble and saying, you know, we still have some percentage of rides where the AV will do something awkward, like pause a little longer than it should in front of a construction zone or whatever it may be. But, you know, in addition to each individual's personal journey, where you're going from skeptic to, you know, okay, this is interesting to how could I ever live without this? That is a very common journey that all our users go on. But the other thing to keep in mind is that the rate at which these systems are improving is extraordinary.

9:21You know, we push out software updates to the fleet every two to four weeks. And each one of those has substantial improvements across, you know, various areas of the product, whether it's the pickup and drop-offs or the routing or the smoothness of the driving, those are getting continuously better. And so your experience now versus six months ago would have been very different. And like now, you know, several months in the future, it's going to feel like a, you know, a much more confident experience driver, even than, than what, um, you've seen today. So, you know, we're excited about that. I think it's, you know, that there, there will be public perception will always lag the true state of this technology because it's moving so fast by the time, you know, know, people develop a consensus opinion, it's already out of date.

10:02And I think that's, I think that's kind of fun because people will constantly be surprised by the difference in the actual rides versus, you know, their perception based on, on what they see or hear from others. Yeah. And I mean, I do, I use rideshare and also drive myself. And I think it's interesting because on the rideshare front, like if you drive, I've been using these cars in San Francisco. and if you drive in an Uber or Lyft in San Francisco, these are tough roads. You go up and down and there's curves. You tend to get nauseous, I think, by the time you finish your ride. There's a lot of stop signs.

10:37It's just tough roads to drive on, but the roads with an autonomous vehicle feel much smoother. So you talk about these cars understand how to drive more smoothly. Then there's when you drive on your own, there's this constant... Anyone who tells you they're not distracted while driving right now is, you know, either it's some sort of saint or lying because there's always this stuff pinging at you. And to be able to drive without, you know, without any of those forces impeding on you is pretty wild. I think it makes, I mean, it's common sense that, you know, a machine is going to do a better job driving than a human when you would eliminate all the distraction, you know, assuming it's a decent driver otherwise.

11:17But let's talk about, you know, right now is kind of like a flawed social experiment, you know, gone wrong with with human driven rideshare. You know, and there's so many reasons for that. One is, you know, we hear from women who ride and cruise all the time, especially at night, that is so refreshing to get into a car and not have to worry about getting into the car with a complete stranger late at night, you know, and be worried about what their intentions are. And that's not all rideshare drivers, but it's certainly some. And that peace of mind, that's something that we should have never introduced in the first place.

11:54This horrible anxiety of stepping into a car. And you mentioned the smoothness of the ride. So we did some research on this and the cruise AVs are tuned very carefully to minimize motion sickness. And how you do that is, you know, there's acceleration of the vehicle. The derivative of acceleration is called jerk. and like there is a very narrow window of jerk, I guess, settings that you could think of where you minimize the forward and backward motion of the human body sitting in a vehicle. And we tune for that. And so you can, like if you're in a rideshare driver or even driving your own car, a passenger in a human driven car, and you're staring at your phone or on your laptop in that car, you're almost certainly gonna get motion sickness.

12:37But I found, you know, I ride in Coop's cars all the time. I can stare at my phone and do work emails and actually be productive in a car without having to worry about getting sick. And that's not an accident. That is by design. Yeah. And the cruise makes you, I mean, it's so smooth. You don't even feel like buckling your seatbelt and cruise won't start until you buckle. But it is, that's just for people listening. And that's sort of how these rides operate. And, you know, I know I sound like I'm gushing here, but really that it's just like how the technology operates. So I feel like, you know, this show, we'll talk about things when they're bad and we'll talk about things when they're good.

13:09And we're not like bound to like the skeptical or the booster angle. And this is certainly like the truth of the matter. You mentioned, so we're talking about ride sharing. First of all, these, you know, I've used the cruise app. I've used Waymo app. You call the car, it comes to you, it drops you off. It doesn't seem like Uber or Lyft really have a future at all if we move to this future. Like, you know, initially the way that Lyft talked to me about this a few years ago, and we just had David Risher on the show, the CEO of Lyft. So, you know, folks can, I guess this is like ride hail and self-driving car month on the show.

13:47But, you know, the way that they used to say it, they would be the networks on top of the technology. And you would effectively use those ride hailing apps, which have mapped out the cities and routes so well to be able to get your cars and get to where you need to go. But right now with the cruise app or the Waymo app, it's becoming very clear very quickly that that stuff is unnecessary. So what's your reaction to that? Is this something that replaces those apps? I mean, at this point, it's hard to say. There's a lot of nuance and complexity in existing ride-hailing apps and services. I think the consensus view on how the industry would evolve over the last few years has evolved.

14:28Originally, at one point, there were like 60 companies in the state of California testing AVs. And so the view was that maybe half of those or a third of those will start shipping robo-taxis. And you'll have 30 different robo taxis, just like there's 30 different, you know, brands of cars, you know, often sold in the US. And then something like, you know, a rideshare app would be this nice way to have one interface to all of those. What has actually happened is this problem has turned out to be way more complicated to get to do right, way more expensive, way more time consuming. And the field has thinned down to, you know, one or two or maybe three players.

15:06And at this point, I think, you know, that assumption that you need an app to sort of like fan out to all the different types of robotaxis is less of a guarantee. And so I think it remains to be seen, you know, whether AV companies and rideshare companies end up partnering or mixing and matching together to form alliances and whatnot. But I certainly think there are more options now. And, you know, Cruise obviously has a standalone app today because we think there are benefits to iterating really quickly on that app experience simultaneously within car experience. So we can give the customer the best ride possible and not have to sort of translate that experience through a third party rideshare app.

15:44But we'll see. We'll see how that evolves in the future. So could you envision a future where you would partner with a Lyft or an Uber and just make these cars available through those apps? Well, I would say nothing is off the table, but it would have to make sense for both parties and ultimately for customers. Speaking of Lyft and Uber, so I guess this is going to be the last of the ride sharing questions for you. But I've been emailing with the ride share guy, Harry Campbell, and he had a question that he wanted to ask about this. He says, explain the economics of how a cruise vehicle with technology, R &D, etc.

16:16will ever be cheaper than a human driver who gets paid$20 an hour to use their own car. Like, what is the break even point? Because I'm not seeing it. I mean, I think that's a good question. What's your perspective? Uh, well, I mean, I would flip it around. I would say, why wouldn't a vehicle where basically the dominant cost is just the fuel or electricity? Um, you know, once you're amortized, the technical costs, like, you know, versus paying an entire human livable wage, which is going to go up over time. Um, I think as, as society demands, you know, um, you know, higher minimum wage and other things.

16:51And so that's going in the wrong direction. like it's going to increase over time i think the the human wage whereas the cost of the technology is dropping dramatically um both on a upfront cost as in the cost to manufacture the vehicle with all the technology but also on the operating cost per mile so the overhead we pay to monitor these vehicles to develop software for them all of that um basically gets way cheaper on a per vehicle basis as you add more vehicles right so the unit economics aren't great at hundreds of vehicles, they start to get pretty interesting at thousands. And then, you know, above that, the advantage versus a human rideshare driver is pretty dramatic, actually.

17:30And that's both, again, due to decrease in operating costs, but also we have two generations of newer AV sensors and compute technology in development that each one of those has a huge drop in the cost of, you know, the sensors and compute to the point where it doesn't really add that much to the cost of the vehicle a few years out. Right now, your cars are full self-driving. There's another company that talks about full self-driving, and that's Tesla. You mentioned there's one, two, three companies. I mean, two I can think of right off the bat is you and Google with Waymo. But Tesla is an interesting case because, okay, their full self-driving isn't exactly full self-driving, but they have more cars driving more miles autonomously than I think anybody.

18:16So how do you view their effort? Well, I think that's a, that's a common misconception. And I think one that has been played up quite a bit to, to sell more cars, frankly, and that, you know, I understand why that is, but there's a huge distinction between a driverless car, as in you can sit in the back seat or it can go pick you up without anyone inside or pick up your kids from school or whatever it is. And one where you have to sit behind the wheel and you're constantly monitoring and ready to take over at a moment's notice in this like high alert mode. And so I describe it as, look, either the car works for you, that's a robo-taxi, that's driverless car, or you work for the car.

18:53That's what you're doing when you're sitting behind the wheel of... Yeah, but you don't think that training, that's helping to train the Tesla to the point where they could actually be fully autonomous? Well, look, if data was the only constraint to making a driverless car, I think there would be a lot more driverless cars out there than there are. It's a lot more than data. Like Cruise is not data starved, even with, you know, hundreds of vehicles on the road. We've got we're drowning in data. It's really the ability to turn that data in insights and drive improvements to the product. And, you know, a strategy we took early on is we start with well equipped vehicles.

19:26They actually have all the redundancy and telecommunications, the three different modalities of sensors, all those things so that we could solve the technical problems first, which is the hardest thing to do. And we've done that at least in, you know, in our early days. And then the easier part is in driving down the cost. Once you have a working functioning system, because you can go in and optimize all the pieces that turned out to be, um, overbuilt versus what, you know, you now need. Um, but again, there's no, there's no parallel. They're not the same thing. Like a system that's driver assistance where you have to take control versus one where you have cars driving around without a driver.

20:03and that really doesn't come to light until you actually try to do it and realize you need all these degraded states. Like if you have a malfunction right now, if a camera gets covered up on a car with driver assistance systems, it just turns off. If you have a robo taxi driving down the street and a sensor gets blocked, it has to detect that, use its other sensors, pull over to the side of the road safely. And then you may have situations where there's remote assistance needed to figure out what to do with that car. You don't just want to leave it blocking the street. There's so many nuances to going from driver assist to driverless that it's quite silly to me to even put those in the same sentence.

20:41Okay. Well, glad that we could clear that up. Speaking of companies, I mean, look, we got to talk about it, right? So there's one more company that we have to talk about in terms of ones that could end up in that hierarchy, which is Apple. No one seems to talk about their self-driving efforts. I'm sure you're more cued into what they're doing than most. Do you view them as a legitimate competitor in this space? Well, they're pretty secretive. I actually don't have any, you know, inside intel on that. But what I would say is that, you know, Apple has a history of entering industries late, but with a really compelling and oftentimes better product.

21:14And so I wouldn't put that past them. But I will say that unlike a consumer device, which you can test really well behind closed doors, you know, building driverless AVs is very interactive. You have all these feedback mechanisms between, you know, how AVs react with other users on the road, training first responders to interact with vehicles, working with local governments. It's very hard to just show up on day one and drop driverless AVs everywhere. I think it does take a measured and deliberate approach to scaling. That's very collaborative, working with regulators and local, state and federal government.

21:47And so I think there's a long path ahead for anyone else wanting to enter the market, you know, beyond Cruise and maybe a couple others that are already doing that. And my understanding is that they've, I mean, they've been working on this project for a long time. It has been, had a lot of fun names like Project Titan, but inside the company, they've, they've culturally struggled to build it. Their iPhone strategy of, you know, figure out what the hardware looks like and then build the software inside. It doesn't really work for self-driving cars. Like you can't bury the sensor. Like, you know, the cruise, the cruise vehicles, they've got the sensors and the cameras on top.

22:20And if Apple wants to make it look sleek and, you know, potentially risk some of the data collection because of the look, then that's what's that's what's held them back in the past from the reporting that I've done. Well, you know, it takes it takes about four years to, you know, go through the automotive development cycle. And so what you see on the road with cruise robotaxis was four years ago technology. The stuff we're developing now, you know, is a lot more sleek and integrated for future vehicles. But, you know, I guess with a company like Apple, the fear is that cruise gets disrupted. Like there's an iPhone moment when they drop something new.

22:50So what we tried to do is disrupt ourselves. And we have our own iPhone moment coming with the cruise origin, which is our six seater vehicle with seats that face each other. There's no driver's seat. There's no steering wheel. It feels like, you know, your own little like party box on wheels, lighting and screens and sound systems and all that. It's, it's pretty cool. And I think that will be the moment that people realize that a driverless car is not just a car without the driver. it's actually a totally new mode of transportation. And the origin is out in Austin at this point. Like what's the expansion on public roads currently in Austin and actually also in San Francisco.

23:28And then it's in San Francisco now, not every day, but we are doing testing on public roads in San Francisco. Yes. And you'll see that, that ramp up quite a bit later this year. I can't announce too much right now, but we're putting, I would, I think it's fair to say we're putting the finishing touches on that vehicle and it's going to be pretty awesome. Now, the origin. So the origin basically looks like this pod really right out of science fiction movies, two benches facing each other. Is that something that you're going to have one person ride in or is your vision for it to be rideshare? Well, so talk about another social experiment going wrong.