Pushkin.

In two thousand and three, Chris Armsen was a grad student at Carnegie Mellon University and he was working on a research project in Chile in the Ata Kama Desert.

Beautiful place, very desolate, the closest thing we have to Mars in many ways on Earth, which is why you were there.

We were building a robot to go look for signs of life, and I was developing the navigation system for it, and my PhD advisor came down and said, Hey, the Defense Departments has this competition.

They want to build a robot to drive fifty miles an hour across the desert between Los Angeles Las Vegas.

And I thought that just sounded awesome.

So we bought a Humvy and we cut the top of it, put a big electronics box in a lot of computers.

All of the stuff that's in a self driving car today.

The high performs computing, the lasers, the radars, the cameras, the hd map apps were part of this.

It's kind of the first time when all this stuff came together into a vehicle, and then we developed it for six or nine months, and then we tried to compete with it.

The first darper challenge.

The team I was a technical director for we actually had the highest performing thing there.

It was supposed to go one hundred and twenty six miles across the desert.

It ended up going about seven when it got hung up on a berm and almost literally burst into flames.

And you, I mean, nobody won.

But if anybody won, you won, right, that was more than nobody.

That was farther than Yeah.

The key thing is nobody won.

If you imagine a marathon where you're supposed to go twenty six miles and the best runner went too right, it's yes.

Yes, fair, and then burst into flames and then burst into flames.

Yeah.

I'm Jacob Goldstein, and this is What's your problem?

The show where I talk to people who are trying to make technological progress.

My guest today is Chris Ermson.

He's the co founder and CEO of Aurora, a company that's trying to bring autonomous driving to commercial trucks.

As you just heard, Chris started working on autonomous vehicles more than twenty years ago, and in fact, his career traces the really the entire modern arc of self driving cars, from that first DARPA challenge to a top job at Google's self driving car project, the project that later became Waimo to founding Aurora, the company where he works now.

We talked a lot about the problems Chris is trying to solve in autonomous trucking, and we also talked about the future of autonomous driving more generally.

But before we got to that, we talked about the earlier work he did, work that really helped get the field, the broader field, to where it is today.

So we'll pick up the interview back in two thousand and four at that moment when his team's autonomous car had failed and almost burst into flames.

It was crushing, yeah, right, Like, you know, on the one hand, it was called a Grand challenge, right because you know, a lot of people didn't think it could be solved.

It was impossible.

Because on the one hand, it's kind of like, okay, it was really hard.

On the other hand, you've you know, we're sleeping on the floor in the lab, in the you know, in this garage in the middle of the desert for weeks and months trying to get this thing to go, and then you launch it out there and you have this this epic moment where you're watching this thing that you've built, tear out across the desert and you can just see the top of it, and these helicopters that the military is using to track them as they're out in the desert doing this, and then you hear you know, it's supposed to one hundred and twenty miles, and they're like, oh, it's stuck.

And then you hear like billowing smoke and broken and poor thing comes back looking very sad, and you know, it was crushing.

But at the same time, we did set records.

Right, if you look at the kind of the product of the speed and the distance it went, this was an order of magnitude better than anyone had done before, and so it was a big step.

And I think that was partly what encouraged the Defense Department to say, hey, come back again and try it in a year.

And then you did, and then you did it again and again.

And what year do we go to?

Is it two thousand and seven?

Is that at the next big moment?

Yeah?

Two thousand and five was another one of these in the desert.

Two thousand and seven they did what they called the Urban Challenge.

So two thousand and five we're able to drive it around the desert not hit things too much.

Two thousand and seven, they say, okay, that's great, but now let's stay on the right side of the road.

Let's move in kind of pseudo urban environments and show that that can work.

And that was the last of Grand challenges, because that year we actually accomplished it.

We drove sixty miles around this air base and interacted with traffic.

They hired a bunch of stunt drivers out of Hollywood to come and drive cars around these robots to make traffic for them, and it was a heck of a day.

And then the Defense Department, or at least darpas said, you know, effectively, problem solved.

We've got this close enough.

Now hand it off to industry or whoever to take it from here.

And how did how did your car do?

In two thousand and seven, we won, which was nice.

It was a good change.

So you know, we went and did it catch on fire?

No fire?

This time, it just kind of worked.

We had a little bit of nerves at the very beginning, because you know, our system was pretty reliable, robust, it worked all the time.

We go and bring it out to the start line and we can't get it to boot properly because it can't see this GPS satellites that it needed, and we never had this problem before.

We're like, you know, frantically trying to figure out what's going on.

And then we look over and realized that for race day they brought a JumboTron and they've set that up right where the robots are at the starting place.

And it turns out they don't really check these mobile jumbotrons for like how much interference they're generating, I guess.

So we had this panic moment we said, you know, can you turn off the jumbo tron?

And sure enough they turn off that and suddenly everything works again.

So you know, it was it wasn't without its challenges, so so but it.

Works right, So it works, and it works in a like quasi real environment, a pseudo real environment.

So I mean when you at that moment, when you were thinking about the future, the future of self driving cars, how did it look to you?

What did you think?

Yeah, I think at that point was starting to believe there was a thing that could happen here, right, and probably you know, certainly naively thought oh yeah, we've got this right.

This maybe everybody kind of thought that, right, Like.

Yeah, I think, And that's been part of what's made the journey really fun for me.

It's just there's just so many things to learn, so many new problems to overcome along the way of taking something from a it worked on this day in a race, to hey, we can go do something transformational that will save lives and improve the economy and you know, do all these other great things.

It seems like that the distance between it worked in a race and we can do something transformational has been longer, certainly than a lot of people thought.

Then a lot of like smart, well informed people.

Thought, yeah, me among them, right.

And you know, it's one of these things where I think anyone who sets out and does something interest and exciting, often I think if they understood how hard it was going to be before they started, may not have actually taken the first step.

And so I feel a little lucky to have not understood that, because it's caused me to kind of stick with it and really enjoy the journey along the way.

I mean, entrepreneurs basically always say that thing that you just said, but in the case of autonomous driving, I really believe it.

Right, It seems like extra true.

Like two thousand and seven was basically like the year the iPhone came out, right, like, which is to say, in technology long time ago, not to mention that there was like no computer vision as we know it today, right, which has become important, I mean so much.

No, it is.

You know, the waves of technological evolution that we've seen since then are profound, and we've been able to kind of surf those as we've been building it.

And it's been Yeah, it is, it's been a long time.

So then you go to Google, right, you go to Google a couple couple of years after that two thousand and nine.

So you're at Google from twenty nine to twenty sixteen, Right, what do you and the team figure out?

And then yeah, when do you leave?

So we get there, and we originally want to show that we could go from what we've shown at the Darker Challenge to something actually worked on public roads, and so we had this challenge of driving one hundred thousand miles in public roads and miles of really interesting, challenging roads in the Bay Area, things like driving down Highway wand from San Francisco to La and across all the Bay bridges and things like that, with the idea that if we could prove we could do these things, then yeah, this technology probably has legs.

And it was kind of a dread pirate Roberts setup right in that, you know, we were told that you've got you know, two years to do this, and you know then we'll probably fire you the day after, right, and you kind of this was kind of with us for the time.

And so we worked with urgency, and it turns out we did in eighteen months, and through that we had some really interesting technological evolution.

You know, we started to see some of the you know, early legs of deep learning started to show up in that point.

During that time, so this is like two thousand and ten, twenty eleven, yeah.

Exactly those times, yep.

Yeah.

And then but I think almost more importantly really started we really started to understand what this would mean to people.

Right that we you know, as somebody who had worked on this because the technology was cool and initially and enjoyed the challenge of it, and seeing the benefit to the military as we thought about you know, as we started to get into the space and started to talk to people in the public sector about the number of lives lost on America's roads, you know, forty plus thousand people every year.

The challenge that people, whether because they have physical limitation to prevent them from driving or because they've aged out no longer feel comfortable driving, the implications that has for the quality of life, and how we can make that better for folks, you know, the accessibility that we could bring to transportation that's kind of lost with the public trans system we have in the US.

All of this was like, oh wow, there is not just it's not just a cool, interesting technology.

It's not that I get to work with good people.

It's not just that you know, this could be valuable.

It's that, like, this is socially really important to do as well.

And then, you know, the course of the program, we got to the point where we launched for the first time a vehicle operating on public roads with nobody in it, well, actually with somebody at it, with Steve Mahon, who is a blind gentleman who'd worked with us on some of the early kind of concepts and understanding of this, and he took a road to Austin, Texas by himself and seeing the emotional impact that had on him of you know, once again having the freedom to get from me to be without having asked somebody else.

It was just like, wow, this this actually matters.

So that was very cool.

So so you mentioned the machine learning piece starting to come in.

It's I think I was looking back at I think it's twenty twelve that alex net comes along, right, the first thing.

I think that's right, kind of big computer vision breakthrough.

And so it is interesting.

I mean when I I think as a sort of lay person about autonomous cars, now, I think of computer vision as like central like, how could you even do it without that right, without computers being able to quote unquote understand what they see.

But this is actually emerging just as you're building this car, right, this system, So how like tell me about that?

Tell me sort of building in parallel with this technological enabling wave.

Yeah, there was a bunch of these things.

So Moore's law was continued in a march along and the amount of computation we had in the vehicles improving the cell phone was, believe it or not, was a big boon to us because it drove the performance of cameras.

So, oh, interesting thing about like all those selfies everybody just wanting to make better cheaper cameras and the processing for images, yeah.

One hundred, but even just the physical imager itself, right, went from you know, a one megapixel three twenty two forty camera, which is what if you bought a top the line you know, Kodak digital camera back in the day, was now my iPhone's got forty eight megapixel, right, and so, and the quality and color performance and everything that's improved dramatically.

And then light ar is another thing that has really evolved over the course of the last twenty years, going from things where we could get a single point along the line to now full three D fields of view and allows us to combine the camera data with the light our data with the computation.

And then the final big thing that's moved forward is in the sensing side is radar and the automotive radar.

The quality that you can get out of that, the price point that comes in at now, you know, when you combine that set of advances, there's orders of magnitude improvement in both the quality of data we can consume and the amount of data that we can perform computation over that really is unlocked our ability to understand the world around these vehicles.

So you leave in twenty sixteen, like, are people getting rides in wherever?

It was?

Chandler, Arizona yet at that time?

Like, is that is it out in the world yet when you leave?

Okay, No, we've been doing demos and we were doing development.

I think we had vehicles in Chandler at that time, and we're starting to operate, but they were still with a person behind the steering wheel, and we were still we were building towards what ultimately became the initial channel launch.

Why'd you leave?

It was time?

You know, I had been there.

I really valued and left the experience, but at some point I kind of lost confidence that we were going to make the decisions that we needed to make to be successful.

And as a person leading the organization that you know, Google had been incredibly generous to me, give me this incredible opportunity, and if I just didn't believe it, it wasn't my place to lead the team.

And and so you know, I'm a big believer that in business, you you know, kind of you have three options.

You if you see something you don't like, you try and fix it.

If you can't fix it, you get in line.

And if you can't get in line, you get out of the way, and you know, I did what I could to try and you know, kind of move things away that I thought they needed to go, and ultimately said, you know, this company's been incredibly galous to me.

Let me get out of the way and let it gets on its own path.

When you say you lost confidence that they were going to make the decisions they needed to make, I mean, obviously we should evaluate decisions based on the information available at the time, but it is the case that you know, you can get in Awai Mo in San Francisco in one minute and have a beautiful ride right now, Like, were you wrong?

I apparently yes, right, Like you know, like you think, there's no way to judge it other than they they've ultimately been successful.

You know, maybe would have taken a different, shorter path, but you know, I think I think it's it's yeah.

I don't know if you're a parent, but one of the things you hope as a parent is that you you you embody your kids with things and then they go off from the world and succeed.

And the same is true when you lead teams or build something, is that you know you do your part and you hope it outlives you and that you know, outperforms what you might have hoped for it.

And I think it's been awesome to see way most succeed.

So you started in twenty seventeen, right, uh, And as I understand it, you didn't start focused on trucks, Like, tell me about what you're thinking when you start the company?

Yeah, how you get to trucks?

So you know, I didn't leave Google to start the company.

I left Goog because it was time, and I then spent some time trying to figure out what to do and kind of came to the conclusion that, you know, it was worth taking another shot at this and trying to build something.

And so found two great co founders in Sterling and True, and we wanted to build a driver and we wanted to apply that driver and passenger cars and trucks and wherever it could go.

And we thought at the time, we didn't see how it's going to be possible to do trucking because based on the experience I had and others had, we looked at how far you had to see down the road to do that safely, and we came to the conclusion you just couldn't do that with the technology that exists all the time, and you have.

To see farther just because it takes longer for a big truck to stop.

This is physics.

There's just way more kinetic energy.

Right, They're heavier, and they're moving at seventy miles an hour instead of in a city you're moving at fifteen miles an hour.

So the combination of the speed and the way it means that it just takes longer to stop farther distance down the road.

And then you add to that, it's much easier drive around a light vehicle than a big truck, so you don't need a special license.

They're much smaller, you don't need special thing maintenance programs and whatnot.

And so we're like, okay, let's focus on light vehicles.

We see there's a real opportunity there.

If we can crack the can you see far enough problem, then trucking to be a great application to go to.

Yeah.

Yeah uh.

And so part of what I spent my time doing during the first couple of years of the company was looking, you know, basically turning over every rock we could with with Bart, one of our early team members, to find a technology that would allow us to see for far enough that we could actually do trucking.

And ultimately we found that in this little company in Montana called Blackmore that had built this really special kind of laser range finder that because of the way it did measurements and what's called frequency modulated continuous wave it it could see basically twice as far as the rest of the light our stuff could see.

Uh, And so we're like, huh, we now we we have some really interesting driving capability.

We've now got this pseudomagical light our sensor that can see far enough trucking for like a really great application to go take on.

Let's talk about a few of the things you had to figure out.

So you decide you're going to do trucking.

Yeah, and presumably by trucking, Like I mean, I know where you are now, Like, so presumably by trucking.

You don't mean like the bread truck that goes to the little market by my house.

You mean the great, big truck that goes on the freeway.

Yeah, that's where we thought there was the biggest opportunity was working what we call class eight tractor trailers, big you know, semi trucks.

You know, they drive a lot of miles.

We don't have enough people who want to do that job, and yet it's absolutely essential to American and worldwide way of life, and there's a real need.

Right as we've dug into it, we realized that there's five thousand people killed every year in heavy truck accidents, five hundred thousand people injured.

And then we can move goods more efficiently, which you know is important, particularly in the you know, this age of e commerce where people expect things next day or you know, be able to move those goods efficiently the customers is really valuable.

Okay, this April, if I've got my dates right, you actually have a truck drive for real stuff on the freeway without someone behind the wheel, right.

That's exactly right.

It was awesome, right.

We've as a company been building towards that for almost eight and a half years, eight and fourty years, and you know, we take doing these things safely, very seriously, and really a lot of the last couple of years has been making sure we could have confidence that when we let the thing go, that it was actually going to be safe.

And then we got to a point where we're like, yeah, this is safe.

We can go send this out in the daylight and dry weather and have it drive back and forth between Dallas and Houston.

And I had the privilege of right along in the back seat.

And it was awesome right in that it worked.

Incredibly boring, right, It turns out it's like it's.

Wanted to be so boring.

You wanted to be very boring.

But but it just kind of worked.

And I went down and back that day in the truck and were.

You hauling stuff at that point?

Was it was it?

We were?

We were all the histories.

Yeah, oh pastries, Okay, pastries both ways, that would be a bit.

There's some pastries.

It just that they want to know.

I think there is a pastry imbalance.

I think we for the first trip, I think we may have actually been empty.

But on the way back, I'm pretty sure from for whatever reasons, the pastries diffused from Houston to Dallas.

So then what happens in May and may you put a person back behind the wheel, right?

Tell me about that.

Yeah.

We so as a company, we've really focused on do what we do best, and what we do best is building a safe, capable driving system.

The oral driver and we worked with other companies like Peterbilt at Pacar and Volvo Trucks, and our partners at peter Bolt called up and said, hey, we'd like you to have an observer in the seat because there's some prototype parts in our truck and it matters to us that you put an observer in.

And so we said, okay, uh, you know, we had a conversation.

We respect them, and we ultimately put the observer in the seat, and we have a person sat on board who is sat in the driver's seat, but they're an observer and they're really just there because our partner asked them to put them there.

The VIHA presumably they are a license to drive an eighteen year It's not just some n week watching them.

What's going on.

That's the truck driver.

You know.

We we we don't want to put anyone in place where you know, the high control has to debate whether you know, it's just like, okay, let's just put a truck driver there.

They sit there.

If you go to YouTube dot com slash or row driver, you can see a live video between eight and five o'clock every day of our trucks drive down the road and you can see that you know, this person sat there and sometimes they're twiddling their sums, sometimes they're eating Fredo's, you know, just observing.

I mean, at least from a narrative standpoint, it's a bummer to have to put the person back behind the wheel.

Surely, absolutely, from a narrative standpoint, it's a bummer, right, And it creates a nuance that is, you know, certainly in today's.

You don't want a person buy the wheel.

It's the whole point.

It's the whole point.

It's you don't want a person buy the wheel, right.

Totally agree.

But in terms of the development and delivery of what we're building, it just doesn't matter, right, it's a complete nothing.

Well, I mean that's some margin.

It matters at some point, it matters, right.

The whole premise is you can do it without a person there, and so like at some point, like relatively soon, I would imagine you need to do it without a person there.

That's right, and uh, look forward to doing that soon.

We'll be back in just a minute.

Yeah, So where are you now, Like, give me where we're basically to the present now after our twenty year journey, Like what's like, are you are people paying you to haul stuff.

Yeah, absolutely, So to like today is an incredibly exciting moment.

So we have a technology that can drive a truck.

It has the skills necessary to drive on the freeway and drive to sites off the freeway in these industrial park areas, and it does that safely.

And then internally we've got the capability say, okay, what are the new features we need to make this more useful for our customers.

And so we launched in April.

About three months later we were able to show, you know, be able to launch the next version that now doesn't just operate day in the daytime, but operates at night as well.

And of the course the rest of this year, it's really about taking and building the you know, in enhancing the skills that it has to add the few new skills it needs to be able to drive between Fort Worth and Phoenix and El Paso.

And so at that point, just.

To be clear, the route you have now is Houston Dallas, Is that right?

That's right?

I'm sorry, We drive tween thousand Houston and by the end of the year we expect to be driving between Fort Worth and Phoenix.

And what's exciting about and fort Worth and El Paso and El Paso and Phoenix.

And what's exciting about that is fort worth of Phoenix is a thousand miles and a person is not able to legally drive that in a day, And so we'll be able to start doing things that are superhuman in more and more dimensions.

Presumably the value preposition is really long drives, right, Like it makes absolutely no sense for whatever one hundred miles or something.

And the longer it gets, the more the economics makes sense.

Is that right?

I mean, that's think it makes sense across all different kinds of scales smiles.

But I think we're the biggest impact we'll ultimately is on these long call trips.

And like, I get the freeway part in Broadway, How does the not on the freeway part work?

How constrained are you as to where you can go when you're not on the freeway.

Yeah, So the capability we're building is to be able to drive everywhere.

Sure, but like now, so today we drive from a place that's off the freeway onto the freeway, and then when we get in at the Dallas end of the route, and then at the Houston end, we drive about five miles from where we get off the freeway to our terminal side, and along the way we drive past customer locations.

So we're making the choice today to go to our terminals, but it's relatively straightforward to stop and go to a different terminal instead or a different See.

Like, how hard is that off the freeway part, the sort of marginal you know, mile when not on the freeway?

Not, right, would be the short answer.

Right, it's you know, we've put a lot of for for a lot of time, the primary focus what we're building was the freeway part, because, as you pointed out, that's where the most kinetic energy is.

That's where you know, that's where the bad things could happen.

Yeah.

I mean it's also simpler in certain ways, right, Like it seems like turning a big truck around or just getting a truck around.

Like I live in New York City and I see eighteen wheelers in New York City, and I don't know, maybe that's maybe that's not that hard to engineer for autonomy.

You tell me it seems hard.

It looks hard.

Yeah, I think New York City is hard, right, And I think there'll be a time where we get to that where I'm really focused and we're really focused on in the near term is how do we get to the place where most freight moves right, And a lot of that will go to these industrial parks in regions of a city.

And the reason why it's hard to drive a truck in New York City right, and it's slow.

And so if you're trying to build a distribution center for these Class eight tractors and trailers, you put them in places where it's convenient because you're trying to run a business, not build a technology showcase.

So what can you not do now ya?

So today we can't drive in the rain.

Well, in fact, that's not actually true.

So today we drive in the rain really well, but we haven't validated that fully for driverless operations.

And this is really an important idea, right, is if you came down to Texas and got a ride in one of our trucks and it was raining out, it would work, and you would go, huh, why if you guys not launched this, and we would say, yeah, it basically works.

But that's not good enough.

It's not good enough that it feels like it works.

What we want to have done is.

All basically it makes me nervous with a ten roil truck.

Exactly and it should right, And so we put a lot of effort into making sure that it doesn't just seem like it works, but we have conviction that it will work right.

And you know, as an example, two of the last things that we we got checked off when we launch, where it's one really high speed motorbikes running red lights in the city.

So we had done a bunch of testing, I think up to let me say, I'm gonna say some numbers here.

These are probably not quite right, but I'll use some numbers, you know, at ninety miles an hour, because we figure, you know, ninety miles an hour in a surface road, nobody.

Need the motorcycles going ninety.

Motorcycles going ninety miles an hour and they're going to run a red light in front of us, and we want the truck to not hit that motorbike.

And then we're on the freeway and we see a motorbike doing one hundred and fifty miles an hour, and we're like, okay, that is let's go make sure that that works right, because you know, at some point we just want to have comps.

And so we went through and we found out that when motorbike's going that extremely fast.

First, we found that our simulation wasn't quite up to snuff to deal with that, so we had to go fix that.

And then we found that our perception system wasn't able to quite track it as well as we'd want, so we had some things to fix there, and so ultimately we got to the point where if you're a motorbike, please don't go test this run in a red light at one hundred and fifty miles an hour.

We're going to see you and assuming you know, we can actually see you, because there's not a building in the way, we're going to stop and you know, not have an event with you.

So is the rain analog of that, like more rain than I can imagine and flooding and it's dark or something.

I mean, yeah, maybe frozen frozen also.

And so this is this is the answers.

We we have some good ideas of what these challenges are, and so we're in the process of putting place the last of the tests that we need to say yeah, across the things that might be a problem, we actually handle those in a way that's safe.

Once you got rain, are you good?

I mean, I guess you have the whole southern US without snow and ice?

Right, Well, think it not freezes in Texas?

Right?

Does it freeze in?

Does we get we get snow in Texas?

And but the good news is, like the reason why a lot of freight moves wrong the sun belt is because trucks don't like driving in snow.

Right, It's just, you know, much like people in general don't like driving in snow.

But there is snow.

But if you've ever been in Texas when there's a snowstorm, it basically shuts down.

Yeah, so you don't you don't have to solve snow.

You don't have to solve snow for now.

For now.

But we will solve snow, right because all of the stuff that works for rain mostly works for snow.

To me, all of these problems we can stipulate will be solved by somebody, right, Yeah, plausibly by you, if not you, by somebody else.

But like I feel like there's some sense of urgency.

Like I'll say so interestingly, just purely coincidentally, I talked to uh Worris softman?

Is it softman?

Yeah?

Have his name is Med and so he said, and we were not talking about you.

This was not a dig on you.

It was more talking about why he's doing what he's doing.

He said, I don't think Open Road Autonomy is a startup game.

Uh.

And he basically you probably know he thinks, I mean, he basically thinks it's too expensive.

Yeah.

In short, too expensive and too hard in that way for a company that is capital constrained, for a company that you know doesn't have a monopoly on search.

Let's see.

And so I do feel like in talking to you like I, you plausibly may solve them.

But there must be some clock for you, right, You must have to get to a point where you're making money relatively soon.

Is that right?

Yeah, Well we've been at the game for a while.

You know, we went into an eyes open right and looked at the you know, having spent the time with Waymo, having the history I have, I was like, this is not one hundred million dollar problem or a ten million dollar problem.

This is a multi billion dollar problem.

And so as we've built Aurora, it's been Okay, we need to set ourselves up with the staff, we need to set ourselves up with the capital partners, we need to set ourselves up with the technology partners that allow us to take that run.

And I think we've been very careful about how we've done that, and we've put ourselves in a position to succeed.

And I agree with with Boris this is you know, I hear people throwing out, oh, we're gonna, you know, spend one hundred million dollars to do this.

I think, you know, you haven't even met the ante, right, And so for us, we look at say, driving in the rain, we expect to be solving that by the end of this year, right, And at that point, as you point out, we've basically unlocked the Southern freight belt.

The amount of truck travel that's there is gigantic, and so you know, from now where we have a small number of trucks in the road, by the end of next year, end of twenty six, we expect to have a few hundred trucks on the road, and then the year after that we expect to have one thousand plus trucks on the road, and then we'll built from there.

And so it's you know, the conviction we're starting to build internally, like yeah, this is just just turn over the kind of the way we want it, and we expect it is happening now and so it's not a hypothetical.

It some day we'll solve it.

It's like, no, we're you know, four or five months kind of time frame.

So, well, three four months timeframe.

What's the business model?

Yeah, so our business model is to work with the ecosystem that's there today.

So today, if you're a truck and company, you buy a truck and then you pay somebody to drive that truck for you.

Uh, and so that's what our company's going to do.

You'll if you're a company and wants some Automas truck, you're gonna go to pac r and peter Bilt, or you're going to go in Evolve and say I want to buy a truck, and they buy the truck stuff and then it'll roll off the line from our partner and it'll have the row driver installed on it and you'll pay a subscription to Aurora to drive that truck for you.

I see.

Okay, So so you guys get paid by the by the mile or by the by the hall or something like that.

Yeah, that's exactly right.

You know, it's at the risk of at a service as a services you know, it's driver as a service.

That's the way we were thinking about the business.

And the trucking company is buying the truck, So that's good for you.

You don't have to When you say you'll have a thousand trucks out there, you don't mean you're going to own a thousand trucks.

You mean you'll have essentially a thousand autonomous drivers in trucks that other people own.

That's right, And this is because that's what our customers want right that there Also it.

Seems easier as a business, and I still not have to get into the trucking business.

I agreed again back to the philosophy of we think we're really good at building the driver.

And I look at our partners, whether it's Werner or Hirshbock or Federal Express, they know what they're doing.

Why don't we help them make their business a bit better and work together?

Can we talk about autonomy a little bit more broadly, just since you're at it so long and I'm curious what you think, So, like, what do you think are the important constraints on autonomous vehicles right now?

Sort of broadly technologically, you know, in terms of policy.

I think it's primarily technological right now.

Okay, right, And I've said that.

You know, people for for the better part of a decade have been you know, when when it's convenient hiding behind you know, regulation or policy.

And you know, the US policy of regulatory environment is permissive.

It's one of the things that's allowed innovation to flourish in this country, certainly in the automotive space.

So what are the what are the important technological constraints right now?

Like, I mean, obviously WEIMO is thriving where it's living, but like, also there are no autonomous cars where I live and people aren't buying them, and so like what's the what's what are the important bottlenecks right now?

Uh?

Execution madwidth?

Right?

Like, So so at least where I sit at a roar today, I don't out and see, Oh my gosh, I don't know how we're going to solve that.

It's hey, we have this many people, we have this much leadership band with we just need to go execute right.

Like thinking industry wide and not to be like petty or something, but like why aren't autonomous cars everywhere?

Right?

Yeah?

Because it's really hard, right, And I think the right metaphor for this is commercial aviation, like the physics are really in grasp and like you can both things together.

And yet you know, we basically have two commercial aviation manufacturers globally, right, we have Boeing and Airbus.

Because it's hard to build big jet aircraft.

It's hard to build them, it's hard to have conviction they're going to be safe, it's hard to produce them right.

And when I look at automated driving, I think it's in a very similar space in that a lot of things have to work well.

You have to have the discipline to be able to work across all of them, make sure they all fit together, and then you have to have the processes.

And process in Silicon Valley is often kind of a four letter word, like you know, you've got to move past and break things, but when you're building something safety critical, you actually have to get that stuff right, and the art is getting it right, we'll also be able to do it quickly.

So so when you think about the spread of autonomy from here, again not just within your company, but sort of more broadly, like how do you think it'll play out?

I think it's going to usher incredible new age all right, that we should be able to dramatically reduce traffic fatalities.

The fact that you know, it had been a decade without a fatality on civil aviation, and yet we had forty thousand people killed on the road every year in the US.

Like, we can bring a technology to bear to help solve this, right, and I think that's profound.

I think we can have a huge sustainability impact all while actually kind of power up the US economy.

And so for us, you know, we've got this mission right now that's focused on trucking, but what we're really building is this capability to release safety critical software and systems and this ability to understand the world.

And so you combine those two things, the places we can go with that are profound, right, whether it is you know, last mile delivery or ride hailing applications, or mining or farming or aviation.

Like there's it's just going to be a lot of fun over the next decade.

So yeah, I'm psyched for it.

Next decade sounds fast given the pace so far.

I was surprised when you said next decade at the end there.

Yeah, I think that like anything, there's a point where you have built a foundational base and then suddenly you're like, oh, now I have this power.

Now I have the superpower to go do these things.

It's basically the package of hardware plus software at a reductive level, and you can sort of put it on other things and it'll more or less work.

More importantly the process and the data and again does not sound sexy.

But the data sounds sexy, Like the more you talk to machine learning people, the more sexy data sounds.

Right.

Yeah, And you know, and we joke internally, you know, if you if we have three kind of big artifacts, being our software, our data, and our process, if you held my gun in my head and said delete one of them, I'd say delete the software, right, Please do not delete our software.

Yeah, that's really interesting.

Yeah, we'll be back in a minute with the life around.

We're gonna finish with the lightning round.

I appreciate your time, and we were almost done.

Cool.

Thanks for the conversation.

So right that your father was a prison warden he was, in fact, Yeah, did you learn any management and were parenting tips from him?

Lock them up?

Uh?

No, absolutely not.

You know.

One of the things that that stuck with me is, you know, it's it's a difficult job being a prison warden because you have the staff and then you have the inmates and right, and and they're often at odds.

And one of the things he said was, you know, make sure you treat everyone with respect.

Right that he would, you know, these are people who had made a mistake in life, you know, and you treat them with respect.

Uh.

And you know, I think that's really powerful and important.

What's one thing you learned working at Google?

I think big, right, I think I think that was, you know, Larry and Sergey never lacked for vision, and I think you can often find yourself constrained by kind of your own kind of sense of what's a limit, and they just thinking like, no, what, why can't you do that?

Why can't you do that?

Yeah?

I remember, just as a consumer, just as a as an ordinary person in the world, when they started what became Google street View, I was like, yeah, surely you can't take a picture of every bit of every road in America, and actually you can and they did.

Yeah.

Right.

Like I think that that sense of like you don't need to be intimidated by this, like you need to think rationally, you need to figure it out.

But you know, I think it's a power one of my professors at Carnegie Mellon said, dream like an amateur, execute like professional, right, And I think that's just a really kind of profound sentiment.

What's one thing your time at Google taught you not to do?

Oh?

What it teged me not to do?

Uh?

Maybe kind of adjacent that.

Like, so when I went to Google, I thought I was a pretty good programmer, you know, I was, Yeah, Carnie Melon, I'm just good at what I did, and I'm not a bad programmer.

But like, there are truly exceptional people out in the world, right, And so I think, you know, maybe don't overestimate yourself and kind of again kind of think, kind of think big, don't underestimate what's out there.

Chris Sumson is the co founder and CEO of Aurora.

Please email us at problem at pushkin dot fm.

We are always looking for new guests for the show.

Today's show was produced by Trinamnino and Gabriel Hunter Chang.

It was edited by Alexander Garretson and engineered by Sarah Burgher.

I'm Jacob Goldstein and we'll be back next week with another episode of What's Your Problem.