On with Kara Swisher - Driverless Semi-Trucks Are Here & Coming to a Highway Near You

Episode Date: December 29, 2025

Kara sits down with Chris Urmson, CEO and co-founder of the autonomous trucking company Aurora, and Johnathon Ehsani, a professor of public health at Johns Hopkins University and leading road safety researcher, for a candid look at the future of AI-powered freight transport.  Recorded live at the Hopkins Bloomberg Center, the three discuss the rapid rise of driverless trucking, what it will take to convince a skeptical public that sharing the road with self-driving 18-wheelers will actually make driving safer, the potential for job losses, and how to regulate autonomous vehicles across state lines. It’s a deeply informed look at the promises and the trade-offs of autonomous trucking with two experts. Questions? Comments? Email us at [e-mail] or find us on YouTube, Instagram, TikTok, Threads, and Bluesky @onwithkaraswisher. Learn more about your ad choices. Visit podcastchoices.com/adchoices

I tried to get the car to run him over. Thanks for that, by the way. No problem. And it wouldn't do it, which was a great disappointment for me. Thanks for that also. No problem. Today, I'm talking. Today I'm talking about autonomous trucking with Chris Ermson and Jonathan Asani.

Chris is the co-founder, chairman, and CEO of Aurora, and a true pioneer in the industry. He's one of the first people I started talking to about autonomous vehicles back in the mid to late 2000s. Back then, he was working at Google, where he helped launch and lead Google's self-driving car project, which is now Waymo. He co-founded Aurora in 2017, and last year it began driving commercial freight in Texas. The Aurora driver system has already done over 100,000 solo miles in the real world, and by the end of 2017, Ermson plans to have thousands of driverless trucks hauling freight in the U.S. Jonathan is a professor at Johns Hopkins Blueberg School of Public Health and an internationally recognized road safety and injury prevention researcher. He advises the U.S. Department of Transportation

and has studied the public's perception of autonomous driving technology. I wanted to talk to Chris and Jonathan because I think it's really important to look beyond cars because trucks actually carry most of the food and other stuff we get in our stores, and the innovations in it are critical to our economy. It's also got a lot of big issues, including jobs and safety. It can be scary to think of a semi-truck rumbling beside you on the highway without a human in the driver's seat, but the reality is they're already here and many more are on the way, and they'll actually make our roads safer. Our conversation was taped live at the Johns Hopkins Bloomberg Center in Washington, D.C. as part of the Discovery series on AI, which is jointly presented by Johns Hopkins

University and Fox Media. This is a good one, so stick around. It is all. Chris Ermson and Jonathan Asani, thank you for coming on on, and thanks for joining me here at the Johns Hopkins University Bloomberg Center in Washington, D.C. for this special live conversation. Thank you, audience, for being here too. So, Chris, I've interviewed a lot, but I interviewed it in 2023 the last time. At that point, Aurora Driver hadn't done any solo drives on public roads. Now it's done over 100,000 miles, and you plan to expand exponentially over the next

two years. And Jonathan, you study autonomous vehicles from public health perspective, and you've searched public's acceptance of them. So as driverless trucking accelerates rapidly. Let's talk about the challenges in persuading the public. They'll be beneficial for society. There's, of course, all the negatives replacing drivers, although we don't have enough drivers, as most people don't know. So, Chris, explain what Aurora does so people who don't understand. Yeah, great. Thanks, Kara. Because all the attention goes to Waymos or cars or whatever Tesla's doing. And they're doing pretty awesome things, certainly Waymo. So Aurora, we're on a mission to deliver the benefit

to self-driving technology safely, quickly, and broadly. And today we're focused on trucking. And so We have big Class 8 tractor trailers, semi-trucks, driving on the highway. We use a combination of verifiable AI, interesting hardware, things like LiDAR, radar, and cameras to see the world, fancy computers, and then map data, offline data, to help these things drive safely through the world. As you said, we're the first and only company operating trucks regularly without a person on. We're without a person in the loop. We're paying attention for keeping them safe on the road. And it's kind of off to the races. So we have about a handful of them running today.

By the end of next year, we expect to have hundreds of them running across the Sun Belt, nobody on board serving customers, delivering toilet paper and baked goods and whatever else you put in the back of a truck across the country. And these are on highways, public highways, correct? That's right. So today we're running on I-45 between Dallas and Houston. We also run between Fort Worth and I-L-Pass on I-10 and I-20. In January, we expect to expand that to be out to Phoenix, so at that point is 1,000 miles. And part of the benefit, what we think we can bring with this is, one, a safety improvement.

So today, there's somewhere between 5,000 and 6,000 people killed in accidents with heavy trucks every year. We believe, and we see a fuel economy benefit. So we've done modeling as we expect between 14 and 34 percent proven in fuel economy. And so if you're a trucking company today, that's your third largest expense. So that's a big deal for you as a trucking company. And then for us as a society, that's a really big, meaningful impact for sustainability. And then, of course, because the computer is ever vigilant, because it's looking 360 degrees around it, it doesn't get distracted. It doesn't have an hour's service limitation.

So we can move goods almost twice as far as a person driving a truck can without all of the negatives that come along with that. person having to drive the truck, right? You should be very thankful for anyone who is willing to drive a truck. If you look around this room, there's literally nothing here that didn't move on a truck at some point. But American truck drivers are 10 times as likely to die, which are obviously average American. They have other health challenges. They often have substance abuse challenges. They spend time away from family. It's just a really tough life. And so it's one of the reasons why, you know, we should be grateful that people are willing to do it. But much like, you know, the 20th century wouldn't have occurred without coal miners, it's also a really good

thing. We don't have people mining coal, or the same number of people mining coal as we do today. So that's how, you know, as we think about it. You're looking at it. But they talk about the challenges of the public intersection is you're placing truck drivers, which we don't have enough of, which people don't realize. Yeah. I think that's the thing. As I look to the future here, my expectation is that if you are a truck driver today, you will be able to retire a truck driver if you want to, right? The shortage is that dire. Over the next decade, we expect to need another million drivers in the U.S.

And that was before the aggressive policies of the current administration, which are, you know, eliminating a significant fraction of our driver pool with these non-domiciled CDL drivers. And so if we want the American economy to continue rolling, so to speak, we really need access to drivers. Now, of course, it's a thing, right? That years from now, there will not be as many people driving trucks. Eventually, my expectation is there will be no people driving trucks.

And we have to help manage through that transition. As our technology comes to market, we expect it primarily to be driving in the middle mile, at least initially. This is the between distribution centers. And these are kind of the less desirable of the truck driving jobs. The ones that people want are the ones where they can stay at home every night, where they do more than drive the truck, where they go to the store and they help unload. They are the customer service rep. They're the sales rep. And that's the jobs that I think that what we're doing certainly.

Convincing people that you're not killing off an industry, essentially, or a job. Is that your biggest problem? Not that you're going to be mowed down by an autonomous job. No, I think our biggest problem is our internal execution and just making sure that we continue to execute well. At this point, the technology is working and that we can bring this to market and now we just need to scale it and get it out there. I think it's society's decision as to how we accept this technology or don't. I see my role in that to try to help educate and inform. And then as a company, we try to do our part as well.

So we've worked with Allegheny College in Pennsylvania, Gallatin College in Montana. We've worked with Over the Road Garage in Texas to put in place training programs to help elevate truck drivers into other roles in logistics. So you're doing that part of it. Jonathan, what do you think the public feels about trucks? They're a little better about trucks, correct? Yeah, I mean, trucks are essential to the economy, as Chris mentioned. They're also truck drivers, just to, this is the premises that they're among the safest

drivers on the road. In terms of crashes per mile driven, they're extremely safe. However, when there is a crash, it's catastrophic, often not for the truck driver, but for the person or the object that's struck. Yeah, I've been there. So it's terrifying. Not close to an accident. Yeah.

So I think the public acceptance of trucks is different to autonomous vehicles just because of the physics involved. However, as Chris mentioned, as they kind of grow the platform and demonstrate safety time and time again, then public trust will build and experience hopefully will kind of yield

the expectation that these are safe, and they're part of kind of an integrated transportation system. What do you think the biggest public problem they face in rolling out trucks? Is it the autonomy itself or the job loss, you know, just no one's there, essentially? Yeah, I think it's probably a collection of things, right? In our public policy and kind of public opinion research, we found that trust is an issue. It's the safety, the transparency, all of those become kind of, one of the key pillars. But another one, which I think is often overlooked, is social good.

If the vehicles, if the trucks or autonomous vehicles are being used for social good, we see public opinion double in terms of acceptance of accepting these vehicles on the road. So I think the industry can advance both through forging credibility, through safety, but also through demonstrating that the social impacts are not only going to be minimizing job losses, but also serving society. Right. So through climate or efficiency or they go through the night so they're not as unsafe during the day, that kind of stuff.

Exactly, yeah. Right, because the computer never sleeps, presumably. So, Chris, explain how the technology for people who don't, people are becoming increasingly familiar with autonomous vehicles. I've been in them for a lot. Yeah. I feel very comfortable in them. But most people have not tried them.

A lot in San Francisco, they certainly have gotten. initially hesitant, and we've had our beefs with them and stuff like that. But in general, acceptance is rising. Those are commercial, those are, you know, consumer vehicles, just like a taxi. Talk about the Aurora driver and how it works, so people understand. Yeah. So, and my experience, by the way, has been very much the same as yours, that you can get very skeptical people and you get them in a vehicle for just a few minutes even. And, you know, it quickly goes from, oh, my gosh, I can't believe this, to, that's kind of all it does, to, huh, okay, what's my phone doing? Yeah.

And this happens within about 10 minutes or so. I said, I had someone get in and I said, look, you can text and drink, like, you know, which isn't funny. It's a bad joke on drink driving. But there is a thing underneath there, right, which is our solution to problems like texting and driving, drinking, driving has been prohibition. And the challenge with prohibition is you want to do the thing that you're being told. not to, the benefit with a technology like this is it enables you to do the things that you find more valuable, whether that's the social activity and drinking or whether it's dealing with your text or your email, while also having improved safety on the road.

So explain what the actual technology is. Nobody's in these trucks, right? You had drivers along. So we have a person riding on board as observer, but they're not there to drive. They're not there to keep the thing safe. The way that we're going to be. works is we've got a collection of sensors. So we use laser, radar, camera, and a special laser we've developed in house that can see hundreds of meters down the road. That's looking 360 degrees around the truck, understanding what's happening there. And then many times a second, it's both figuring out what it should do in response to the traffic to get where it wants to go safely.

But then literally thousands of times a second, it's also checking itself, checking the truck to make sure everything is in as safe and operating as it should be. so that if something is broken, it can identify that and come to a stop off the freeway somewhere safely. This is our approach to do, and this is something we call verifiable AI. It's a way of using some of the most modern techniques

that I'm sure you've talked about on the podcast, putting guardrails around them and allowing us to make it interpretable so that we can ensure that the concepts that it's learning are the ones that actually matter driving safely and that it's actually responding to them and that it's constrained in how it can respond because what we can't afford to have

is, you know, what they're gently called confabulations, right? This idea, let's use glue to hold cheese on our pizza, right? Like, that's cute when I read it in a text box and it's disastrous operating at 70 miles and out down the freeway. And so by construction,

we ensure that we can't take some of those actions that are completely, you know, ludicrous. And we have these guardrails. Would it, would it do that? Like, ah, a cyber truck, I just think I shall run into it. No, we do not run into cyber trucks. Gentle tap is fine with me.

And I think this is why you don't work at Aurora. Correct. Yes. But, no, so we do understand vehicles on the road. We understand, you know, people. We understand bicyclists. We understand trucks, Jersey Barrett, like the stuff that matters, we respond to. And by separating the way we understand the world from the way we interact.

with it. Again, we can make sure that it's actually responding to the vehicle or the painted lines and not some random signal that it kind of divined in the background. And that's not like a hypothetical problem. You know, we've done experiments where we're trying to see, okay, how can we reduce the sensor input that's going to this? And eventually we were like, okay, well, let's try it with no sensor input. And it turned out it worked 90% as well, well is when we put the sensor input in. That seems crazy. Of course, you need to see the world to drive safely.

And it turns out that what it had been doing is learning, well, basically, if I look at the last five seconds of driving, if I do that for the next five seconds, it's going to be mostly right most of the time. Right. So there's no such thing as mostly right. That's right. We need to be right. So that would be AI guessing again, right, essentially.

It's, yes, and it's also figuring out what it thought was a smart way to cheat. And we don't want it to cheat. You want it to make sure it's using. Whatever the sensor happens to be at the moment, rather than... Whatever the best signal, best way to interpret the data gets from around it. I think one of the concepts is actually really interesting, what makes these things superhuman is, one, they can look in all directions at the same time. So if I'm driving down the freeway, I don't, you know, many of you have had this experience, right? Your shoulder check, and then he's come around, the vehicle in front of you hit the brakes.

And it's kind of terrifying. The Aurora driver is looking at that vehicle in front of you and the vehicle beside you and the person stood on the sidewalk and the person looking at the corner and is thinking about all of them constantly. And so as a person, we foviate on the thing we think is the biggest risk

and we'll miss other actors. And that's a real opportunity for us as we're going to driver to be safer. But one of the things that you don't want, you wanted to teach other trucks these instances but not rely on that patterns. We certainly want to love. learn from them, and we use that as part of how we train the system, but we also want to

represent things explicitly. So as an example, there's this great study that looked at how often people actually stop at stop signs in America. Never. That's me. You're not wrong. You're not wrong. It's about 11% of the time, right? And so if you just trained the system to look at how people drove, you'd have robot cars driving through stop signs. Yeah. And you know, maybe society would accept that but I don't think they should. And so we want to be explicit about, okay, no,

this is a real constraint. You should stop at a stop sign. Right. Rather than hope it figures it up. Although I have to say, you do notice autonomous vehicles. Like, they used to drive like your grandma, and now they drive like your aunt who's a good driver and a little more aggressive. And I've just noticed they've gotten more aggressive.

And I think finding that right, and I think you see exactly this right. Because then they were going too slow. You're like, really? And that doesn't quite work. Yeah, really. The plastic bag, you're going to stop for the plastic bag. Please don't. And this is the art, right? And this is where some of these modern techniques become really useful, because you can say something like you should stay three seconds behind the vehicle in front of you. It's a good rule of thumb, driving down the freeway. Most drivers don't. But let's say we want to maintain that. And we can do a little better because our reaction time is better than people. But what happens when that

driver cuts right in front of you, right? You don't want to hammer the brakes and instantly create three seconds of separation, but you don't want to wait until next week to have three seconds separation. There's some natural process that feels human-like and is acceptable of creating that gap. Which they're learning as they're doing and through AI. And that's where these kind of modern techniques really come to the fore. Like, we can say at steady state, keep, you know, in our case, you know, something less than three seconds, but roughly three seconds. And in transition, learn how to open that gap. Right, because it matters a great deal more with trucks. So, Jonathan, you're bullish on autonomous vehicles and the ability to

use traffic deaths, speaking of knowing when to stop. Talk about traffic deaths and injuries, because that, to me, is one of the greatest inputs. And what are some of the potential externalities, for example, an increased capacity to haul freight could lead to more emissions, which would lead to negative health, you know, the more we can do it, the more they will, essentially. Thanks. This is a terrific question. You know, motor vehicle crashes are a public health crisis. I mean, we in this country alone last year. Which we've accepted, like gun shootings. Exactly. Exactly. And I think we've accepted it for really, you know, terrible reasons for the cost of doing business. For the cost of getting places, we've accepted last

year, NHTSA of the US DOT's safety agency, estimates 39,345 lives were lost in 2024. That's about 100 lives a day are lost from crashes. The potential for the autonomous vehicles and autonomy in general, is that they can learn from their mistakes, which regular drivers, human drivers, us generally don't. Say, for example, I didn't sleep well last night, or I was taking some medication that meant that I was very, very drowsy, or I was impaired or whatever it may be. Today, we allow anyone to drive, basically, and many people who may not be fit to drive. So I think the role of autonomy in getting to safety is critical. The challenges or the externalities that you point out

is that autonomy will solve some safety issues, but then it will introduce new ones potentially. So right now, when I'm driving a vehicle, there's no privacy or cybersecurity threat. But autonomous vehicles, as a system, are vulnerable. Well, there are some because they can program cars that people are driving in. And there's so much stuff on the cars now.

There is today, right? There is. Yes. And the mix of the vehicle fleet, such now is that that is penetrating the vehicle, and there is kind of cyber and privacy concerns today. But if you look at the traditional fleet, I think it's a mix of kind of old-school vehicles, kind of more instrumented or tech-heavy vehicles and then are fully autonomous vehicles. And so I think as those technologies penetrate the fleet, we introduce a new set of challenges.

they're not insurmountable. We just have to be very clear right about what they are and how to navigate them. And then in terms of once they can use them, they'll use them a lot more, which means emissions will presumably go, unless they're electric, right, or accept that kind of thing. If it's going hand-in-hand, EVs, and although we're seeing a decline in that today, just Ford just took a massive right there. You're raising a terrific question about, you know, the mobility system in general. Like how will it play out in terms of convenience and access, public transit, what's going to happen to buses and transit? trains, investment, and so on, those are questions that we still need to figure out. I think the economic model of the autonomous vehicles themselves is also something that we're only at the very beginning of figuring out.

And so understanding how this is going to scale to operate a society level, it's going to take some time. We'll be back in a minute. Support for this show comes from LinkedIn. You know how important it is to hire the right person for your small business. That's why LinkedIn jobs is stepping things up with their new AI assistant so you can feel confident you're finding top talent that you can't find anywhere else. The best part is those great candidates are already on LinkedIn. LinkedIn says that employees

hired through LinkedIn are 30% more likely to stick around for at least a year compared to those hired through the leading competitor. That's a big deal when every hire counts. Hiring doesn't have to be complicated. And let's be real, when you have a business to run, you don't have the hours to spend on hiring. And you can't just wait around hoping the right person stumbles upon your job. That's why LinkedIn Jobs' AI assistant suggests 25 great fit candidates daily so you can invite them to apply and keep things moving. Hire right the first time. Post your job for free at LinkedIn.com slash Kara, then promote it to use LinkedIn Jobs' new AI assistant, making it easier and faster to find top candidates. That's LinkedIn.com slash Kara to post your job for free, terms and conditions

apply. A lot of human drivers are nervous around semi-trucks for good reason, right? But I'm nervous around every semi-truck in existence, actually. If the car gets hit by an 18-mile highway, passengers don't too well, as you noted. And even though autonomous trucks are much safer than human drivers, there's no perfect system. And often it's a human's fault, typically, right? I know that. I've seen the statistics.

But they have, as we just talked, cyber attacks. So what are the failures that you're looking for now? And maybe explain what conditions are you particularly looking at right now as you move to hundreds of thousands of driving hours? So part of how we've approached this is to build something we call a safety case. Because there's no silver bullet. I can't say do A and you're done. our safety case is really this kind of a holistic view

at both the company and the technology to make sure that the thing that we put on the road is safe so in our case it's got five pillars the first is that it needs to be performance so when it's driving that it behaves the way it's supposed to it responds appropriately

and we put the thing through something like four and a half million tests to have conviction that yep this is going to behave as it should on the road the second is that it has to be fail safe that it's not okay for us to say oh that broke That's why this accident happened.

We need to be able to identify what might break. We do all these kind of systems engineering things like FMEAs and DFMEAs and hazard analyses that allow us to get convinced that, okay, we've thought about the way this could break systematically, and we've used that to design the system so that should that computer fail or this tire blow out or that sensor fall off the truck, that the vehicle is going to be find a way to respond safely to that, not create undefiote. do risk on the roadway. The next is we need to be continuously improving, right? As you point out, we're human and fallible. And so we need to be able to have processes at our company

where we learn from what we see in the field, where we learn from our new experiences, and we feed that back in to refine the requirements, refine the system, and make it incrementally safer over time. We need to be resilient, which means that we need to foresee how people are going to misuse the technology. And importantly, we need to be robust to cyber security attack, or cyber attacks, right? Which you're open. That's one of your biggest challenges. Well, it is, I actually agree with you that today's vehicles have enough, right, have enough footprint that you can do real damage with a cyber attack if you could do it on broad-based way. It's a lot easier to do damage to a vehicle by going up and physically touching it, right?

And so I'm not particularly worried about any individual vehicle because it's just cheaper, easier to go and cut a brake line or stab the tires or do whatever. And so we do have, you know, like a 50-person cybersecurity team. But you're in the south, you're in the southwest, very straight roads, highways, which are standardized around the country, which is good for you. So you have something that's pretty, you know what a highway looks like. Wherever you drive, a highway is a highway. I think that's one part of it. And they're flat and warm, correct? They're not always warm. It's pretty cold in Texas right now. Okay, but it's not heavy snow. It's not heavy snow. But, but But that's not really, so the shape of the road is not really the thing you worry about.

That's like, we solved that in computer science and robotics decades ago. It's really the interaction with other actors in the road. And this is one of the places where I think people have had this misconception about, you know, city driving hard, freeway driving easy. And the truth matter is, you're right, most of the time driving down the freeway because it's kind of straight and it's limited, you know, it's pretty easy. The problem is we don't care about the most of the time. That's not the limiter. If most of the time we could have launched self-driving cars 20 years ago, what we really care about is those events. And what happens is all the crazy stuff you see in the city, you see on the freeway, but you're moving at 70 miles an hour with a hell of a lot more kinetic energy.

And so the outcomes are worse. And so we have to be really careful. We have to push the perception horizon out. You know, we literally see people running across the freeway, five lanes of freeway in Dallas or in Houston, you know, 60 miles. on our traffic person playing Frogger going across the road. We saw, you know, a police chase going the wrong way down the road, right? And so we need to think about those. We need to design the system to be robust to them. Yeah. Even in those situations. People are crazy. Yes, I'm aware. So it is. Appointingly so. It is. So, Jonathan, but people have to decide how

when safe is enough, right? And I'm going to agree with Elon Musk here for a second. But one of the things he said, when one bad thing happens to one of his cars, it gets tons of attention, who's whining again about it. But he's right, like people get in accidents all day long or, you know, there's all this damage by gas guzzling vehicles. But we've sort of gotten used to that. What has to change?

Is it younger people? What are the main factors? Is it just older people have to just die and then, like, younger people are comfortable in them? So I think there's, well, sorry, folks. Hopefully none of those bad things is going to happen. But I think there's, there is a paradox that you've just pointed out that we hold these vehicles to a

highest standard than we do to ourselves. And I think, you know, the truth is that it feels scary to put your trust in something that you don't understand or don't know how to relate to. I think there's also a challenge that we all face, which is something we haven't resolved about the transportation system in general, which is that, you know, we've never answered this question before. Like no one ever sat back and ask how safe is safe enough for any of the vehicles that we have, for any of the systems that we have. So I think autonomous vehicles are forcing us to kind of ask questions that have always been there. These are deep questions. And I think part of this is also bumps up against human

values. I think what we're really feeling the friction on is like, I don't want to be killed by a machine. Well, guess what? You don't really want to be killed by anyone, but today it can happen. And so I think what we need to center is what do we want out of our transmitting? system. We want a safe system for everyone, for all road users. And autonomous vehicles are kind of adding a lot of momentum to that conversation because they're forcing us to kind of come to groups with things that we really haven't resolved for 100 years. Right, for a long time. Besides getting seatbelts and airbags, there hasn't been a lot of that. So let's pivot to jobs, as we said, we talked a little bit, but you basically had trucking jobs aren't good and Aurora will create better jobs, right? But

you haven't been able to persuade the Teamsters Union, which represents truck drivers or the owner-operator Independent Drivers Association. They fought you in court and in state houses and promises of, you know, new jobs. You've got to give them the jobs, right, presumably. So talk about why they're pushing, continue to push back on you. You have to ask them why they continue to push back on us, right? That's their perspective. I'm asking you.

You know, from our perspective, right, and everyone can have their own perspective. It's not a surprise that they're doing this. No, it's not, right? they have an entrenched interest that they're defending, right? Have they been a real detriment to you? Obviously, unions are losing power across the country, but... What I would say is that, you know, we are democracy. We have elected policymakers, and they're forcing a conversation, right?

And I may not agree with their viewpoint. I have a viewpoint, and we're having the conversation. And to date, consistently, the conversation is netted out on the side of progress. right, that they have challenged in many states and pushed for requiring a driver in a driverless truck, and they've lost every time to date, right? And I think that is because there's a compelling argument around road safety. There's a compelling argument around economic growth. There's a compelling argument around sustainability. And importantly, as you kind of alluded to, this is probably the biggest change in transportation infrastructure in a century.

the internal combustion engine, right? That built the 20th century American success, right? That allowed us to have interstate commerce and grew that, that basically built America, ultimately enabled us to win World War II, right? It was kind of the platform that lifted the American economy. This will be the next platform for transportation, and we cannot afford to cede that to our adversaries. And, you know, this is the other winning argument. right? China is pursuing automated vehicle technology aggressively.

Indeed. Although China is always like, China, like whatever they want to make an argument. But there's a reality to it. There's some reality. Yes. In that, you know, I think it was a legitimate concern we would have about having the American Internet run through Chinese hardware. And if we were concerned about the bits, we should be equally concerned about the atoms, right, that we'll move on these trucks. And then, so we've got regulation of the U.S., but this competition is going to play out globally. And I believe... Where you assume you have aspirations of global. We would love to, right? We would ultimately like to be driving equipment around the world.

And I think that having a U.S. underpinned transportation economy globally is much better for the world and better for the U.S. than having a Chinese underpinned economy. There's certainly a lesson in BYD, which is taking over the whole world. Well, we're not. And why? Because we collectively did not get behind electric vehicles in the same way. that the Chinese government did. We'll be back in a minute. So, Jonathan, we have seen this pattern before in transportation, these changes, the automation improves safety, but also eliminated job categories. In aviation, there used to be a pilot, a co-pilot, a commercial, in commercial cockpits.

There used to be a radio operator, a navigator, a flight engineer. But planes are safer than ever, right? for the most part, except in disasters. But from a public held in policy perspective, what happened with that transition? Because it's still a vibrant industry for some of it. It's just not as many people. They don't need

as many people. And someday, you know, they're working on autonomous planes right now, obviously. Yeah, I mean, I think a lot of planes could presumably just kind of fly themselves. You know, you have the pilots there and the co-pilots for almost for, you know, keeping in our own things. As a transportation research, I think

we can definitely look to the past to learn lessons about the future. These transitions are going to take some time. I mean, I think I'm fundamentally optimistic and bullish on autonomy and autonomous vehicles and trucking. But I think, you know, one thing to keep in mind, and perhaps it's not the kind of popular view is that these transitions are going to take time. Right now, the vehicle fleet in this country takes about 30 years to turn over.

And so if every single vehicle sold today was fully autonomous and every truck, it's going to take us really until like, you know, 30 or so years until we see this mixed turnover. So I think we've got time for this transition to take place and to manage these transitions. I think what we need to be clear right about is what is our vision for the system and what's our pathway to get there. And part of our pathway is we don't want big social externalities like loss of jobs. Okay, that's part of our pathway. Let's prioritize that. Let's put that in the center. Right now, it doesn't feel like we've got a clear vision or a coherent pathway to get there. And so I think that's part of the challenge that we're trying to resolve.

Yeah, clear vision and coherent pathway is not really America these days. But the Department of Transportation recently approved or a warning device to replace the reflective triangles truckers use when they stop on the side of the road and took a lawsuit to get there, though. The DOT says it's developing regulations specific to automated trucks right now, but basically there aren't any. what's going to take to get updated federal regulations here? I think getting to the point where there's something to regulate, right?

I've actually been pretty positive about the way the federal government has approached this, that they've put guidelines in place. You know, the idea that there's no regulation is kind of, is not accurate. Well, you're under trucking regulations. Well, we're under trucking regulations and car regulations, and there's always a backstop, right, that both at the state and federal level, if the state authorities or the federal authorities believe you're creating unreasonable risk on the roadway, you can kick those vehicles off the roadway. We've seen that actually in use with Uber back in the day, where they were kicked out of both Arizona and California.

And so there is a stick that's available. But to answer your question directly, I think there is value in federal regulation because this is fundamentally an interstate commerce issue. and that avoiding 50 different regulations would be helpful. Now, we don't need it, that we will adapt and work and, you know, there'll be certain states that are more willing than others. And today, the majority, the vast majority of U.S. states are either open to or enthusiastic about this technology. And when it comes to the warning triangle thing, like, again, it's great that we got this waiver. We would have found a way to drop triangles if that was it.

But it's also common sense, right? The truck drivers walking down the side of the freeway to drop cones is immensely dangerous. And other ways, like other vehicles that are stopped on the freeway, what they do, they turn on blinking lights because that's better than walking down the freeway and putting cones. And, you know, the technology breakthrough we had for this was turn on blinking lights on the truck. Brilliant. So, yeah, that's what we do at Aurora. And that took a lawsuit to you know. Well, and we've got to, we've got a waiver from them.

The lawsuit has moved on. now engaged constructively. So what were you against the, the cone people? Yeah, big cone. Big cone was keeping us down. Yeah. Wow. No.

So there's a rule on the books from like. I even believe you should win on that one. Yeah, on big cone. Yeah, we did too. And ultimately we're kind of continuing through the process. And it, you know, what we see with the current administration is that Secretary Duffy's innovation agenda is pretty well aligned with this. We've seen positive comments from the.

secretary and from the vice president about the importance of automated trucking. We've seen legislation introduced at the federal level by Representative Fong from California, who's a Republican, called America Drives. And so we're seeing some momentum here. And I think that's great. Was that different with Democrats? It depends on which Democrats. Under the Obama administration, that was when I first started talking with Secretary of

D.O.T. about this. Very enthusiastic and positive. under the first Trump administration, enthusiastic and positive. The Biden administration was not focused on this at all. And then with this administration, we're seeing, again, support. Right. So now universities are one of the few institutions that are structurally designed to think beyond political and market considerations. So Jonathan, as an academic researcher, talks to lawmakers, policymakers, industry stakeholders.

Do universities have any role in the web of interest between private industry and government agencies? Definitely. I think I can speak from my own experience with respect to transportation safety in this area is that universities offer kind of a neutral and independent place for data, for ideas, and for kind of arbiters of new technologies and emerging things in the innovation ecosystem. And so I think in that way, they become a kind of a one of the pillars that can advance the innovation in this country and the competition. We see this all the time, actually, and I think one of the, you know, we poke fun at academics for taking forever to write a paper or so. But there's more and more industry leaders and companies coming to universities and wanting to work with us on issues related to safety and autonomy. And the broader AI advancements in public transportation, right? And I'd say beyond that, I think what's most striking about this is that it's often the leaders who are coming to us. It's almost like those who are ready to prove their technologies work and that they're safe

come knocking. And I think in a way it's a test. I think the other role that universities can play in this ecosystem. system is that they can begin to advance the broader societal agenda of the public good that can come of these. So one real opportunity that exists with autonomous vehicles and trucks is that once one catastrophic thing happens, it should never happen again because the entire vehicle fleet should learn from it.

And so the opportunity for something like data pooling for safety critical incidents. Between all the companies. It's not happening right now. Right. But universities could form consortia where these data are shared. There is precedent. It happens in aviation industry already. And everyone can learn from those events.

Right. They're not operating autonomously from a safety point of view. Airlines are. That's certainly a good comparison. Yeah. So I think that in that way, I think universities and research can play a critical role in advancing this and not slowing down the industry. I think that's the other kind of stereotype is that, you know, things are moving quickly.

academics and researchers can move rapidly with them. With them. So, Chris, right now you have no competition, really, that I can see it. But at some point, many other, there's going to be other successful autonomous trucking companies, including from China, as you noted. You'll have to compete on price speed. Pressures will be to cut costs to get from point A to point B faster. Presumably, there at least one company that's willing to cut corners order to gain market share. That's often a way to do it.

Do you, how do you look at, I mean, there's more competition in the, in the autonomous vehicle, the car space than there is in the truck space. Why is that? And you move from the car space to the truck space. We did. We focused, we were intending to operate cars and vehicles, and we decided to focus on trucks because we just think it's a better, more impactful market for us to go after, and one where we think we have some unique advantages. How do we think about the kind of the impact? of pressure for shipping versus safety. I mentioned the company's mission is

deliver the benefits of self-driving technology safely, quickly, and broadly, and that safely is first for a reason. And it's kind of a decision-making paradigm for us, right? Do it safely, move as quickly as you can, think about scale. For us, when we talk to our customers, the trucking companies,

safety is top of mind for them. And so I continue to be transparent, if you want to see something really boring, YouTube.com at Aurora Driver. We live stream the Aurora Driver every day, right? That kind of transparency builds trust. And we think the customers who have an experience with us, work with us, we step the standard for what it means to be safe on the road. It's going to be hard to unseat that.

Unless, like, there's a catastrophic accident, right, with you in the middle of it. And again, part of building trust is also to position yourself for forgiveness, right? That by demonstrating that we're doing these things the right way, that we have taken the steps, that we engage appropriately with the regulators and the policymakers that we're transparent about this, it's how you get through those moments. And I actually think it's warranted, right, that we are doing something new. There should be scrutiny. We should be held to a standard. And what happens, you know, you remember the days, right? the very first days of the first accident we had at Google with self-driving car, national news, right?

And then the second one, slightly less visible national news. At this point, just given the scale of operations, I am sure there is a way more crash every day, right? Not because they're bad, not because there's any kind of problem with the system, just because of scale and people on the road. Well, they had a good one in San Francisco the other day where three of them were facing off. That one made national news. Yeah, that was good, though. But that was novel, right? It was in my neighborhood, yeah.

But what happens is people like, okay, we've scrutinized this, we get it, we understand it, okay. Yeah. Chris and I were joking. That just happened to me the other day with some lady who was texting and driving and wouldn't move her car out of the middle of the road. So it's common for people, but as you said. So last question, we only have just two minutes. Let's fast forward 10 years. I'm driving down a busy stretch of Interstate 95 here on the East Coast.

In December of 2035, I probably shouldn't be driving, so I'll be really old. Be driven by some. Whatever. Whatever. I'm driving, probably. I don't actually like to drive. It's after dark. It's windy and snowing. Driverless freight trucks are all around me. I hate driving around trucks in general, especially on 95. What happens for that to be a totally normal occurrence? What has to happen? Jonathan first and then Chris? The safety first has to be... That I don't feel nervous around trucks when I drive now.

And so the paradigm that Chris articulated, the safety first, is going to build trust and it's going to going to be demonstrating time and again, also the learning that's happening with these vehicles so that they've been in this situation, they've learned, they know how to operate, they know also how to handle the unpredictability of human drivers, because I think part of the challenge here is that the mix of road users is going to create new vulnerabilities. And so I think you may actually end up driving with more confidence in that circumstance than what you would know. Oh, I'd like to get all people off the road.

That's my goal. You know, I wrote a column in the New York Times where I said, owning a car will move as quaint as owning a horse someday, which I didn't think would happen. So you think safety is the critical thing, that people feel safe. What about you, Chris? I think that's it, right? I think as a company, we need to continue to execute.

We're doing a great job of that today, but we need to continue to put safety first. And I think you will be safer on the road. You know, in those situations, I joke that, you know, what a person does is they drive down the road and they can't see anything, so they drive at freeway speeds. It's like faith-based driving. There's nothing there.

And the self-driving truck is going to go, oh, geez, that's a blizzard. I can only see so far, I'm going to slow down, right? And that will enable the roadway to be safer. You'll avoid these 120-car pile-ups that you have in snowstorms today because it just won't happen because these vehicles are controlling their speed appropriate for the conditions in a way that people are too impatient and too self-confident to really do themselves. So when will they be on 95? Certainly before 10 years from now. Before 10 years to now.

So that's the goal, is to get them all over the country. We expect to be driving, like I said, across the Sunbelt next year, and then we'll start to work up through the country. People make a big deal about snow. It's not that big a deal. You know, when you're building an engineering system, you want to kind of find the parts that you can get done and get that out and start getting out and benefit from it. And then you add little bits here and there, and we've done that over the course this year,

from just daytime to day and night to in January, day night and in the rain. Still is just another thing to add at some point. Where's the worst place that you're like, oh, I better not go there with my trucks? I have seen those crazy internet videos of traffic in India. I think we're going to be a little ways off from that. Okay. All right. Well, on that note, thank you so much, Chris Irmson and Jonathan.

Thank you. Thank you. Thanks, Bill. On with Kara Swisher is produced by Christian Castro-Rousel, Michelle Aloy. Megan Bernie and Kaelin Lynch. Nishat Kirwa is Vox Media's executive producer of podcasts. Special thanks to Catherine Barner and Madeline LaPlante Duby. Aaliyah Jackson engineered this episode, and our theme music is by Tracademics.

If you're already following the show, you get a free traffic cone. If not, you'll be stuck watching the Aurora YouTube live stream all day. Go wherever you listen to podcast, search for On with Caroswisher and hit follow. Thanks for listening to On With Caro Swisher from New York Magazine, Box Media Podcast Network and us. We'll be back on Thursday with more. Support for this show comes from Odu. Running a business is hard enough. So why make it harder with a dozen different apps that don't talk to each other? Introducing Odu. It's the only business software you'll ever need. It's an all-in-one fully integrated platform that makes your

work easier. CRM, accounting, inventory, e-commerce. and more. And the best part, O-DU replaces multiple expensive platforms for a fraction of the cost. That's why over thousands of businesses have made the switch. So why not you? Try O-D-Fer free at O-D-O-D-O-O-com. That's O-D-O-O-O-O-O-com.