Autonomous Vehicles Where We've Been and Where We're Going - Chris Urmson + Mike Volpi

Good afternoon. What everyone's pleasure to be here if any of you have been following the movement around self-driving cars the industry and all the startups. Chris will need no introduction, but I'll give him one. Anyway, the first time I actually met Chris he was on stage and I was in the audience. It was Ted 2013, really and exactly. Yeah. Yeah. You were pretty good Chris gave a talk about self-driving cars which at the time was a pretty abstract concept to say the least

and Chris turns out have been working all ready for six or seven years. I think on self-driving cars. He was famous for the original work done at Carnegie Mellon University at the DARPA challenge where they made the first sword of suburban and urban assault vehicles that drove themselves that was not was If you if I had a photo for you, I could show you one that was flipped upside down. He didn't join Google where he led the self-driving project there. And he was t t o of the program until 2016

and finally after a lot of cajoling from a lot of venture capitalists like me. She decided to launch his own company to do it, which is Aurora. I've had the fortune of being on Chris's board for almost three years now and some of us like to call him to Henry Ford of self-driving cars. She gets embarrassed when I say that but it's okay and so it's real privilege to actually be able to put them on the spot today and ask him some questions. So Chris I'm going to start with this. There have been a lot of promises made about this

technology and when it's actually going to be ready. There are people right now out there who are saying we're giving rides commercial rides to people with it. You have your known for taking a bit more of a conservative position on when the technology is actually going to be ready. So when is this thing actually going to be ready? And when is somebody from the audience here going able to going to be able to flag a uber that has a self-driving system in it and jump in and drive around by Kim Phillips be on stage with

you a lot of fun. So 17 years, which means you're full of sugar can someone smarter and better than that's fun, but I think it's exciting to see if I we've been working at the wrong time. We're now starting to see it on the cusp of being commercially viable. So if you go to Arizona today and you're one of the fortunate few you could get a ride in the car with nobody behind the steering wheel that the team that I helped build a Google. Is it doing that? That's that's

really exciting to see over the next five years. We're going to see hundreds to small thousands of these vehicles out in the world and they'll be parks of fleets in the sleet will be moving people and they'll be delivering goods and that'll be a chance for us to kind of get over that mvp Hertel. I understand you have do we have a technology that works and start to find product Market fit start understand the business model and then get us there to see it happen and then over the second half of the coming decade. We're going to see this technology really start to descale and you

don't have a much broader impacts in people's lives. Now, everybody knows that there's a lot of players in this business. There's Google which is now been renamed way. Uber's doing a program there is a cruise their Zoosk is all these talk a little bit about the philosophy that you took in building Aurora and how you are different than a variety of these other players in the market. So I think the different part comes back to like we know what we're doing. We think that's one of the differentiators import it matters. So we we've been really fortunate to pull

together this amazing group of people who have done this before a different places and so is were building Aurora. We really think about it is version 2.0 of the technology. So I had the privilege of leading the Google team for many years one of my co-founder Sterling. He was at MIT doing this type of research and then went to a couple places went to Tesla and launched autopilot model X for them Drew back now is that Carnegie Mellon professor in machine learning and Robotics? And helped found the Uber atg and help get that program off the ground. So the three of us

can understand what it takes to solve this problem and it's got the Third Kind of scars to prove it as we're building technology. Now, we're thinking about what you have to build to actually make a product not you know, how do I get over the next little hump in the road that we we might discover out there at their testing. So I think that's a big difference here and then we've got, you know, we'd make some really cool progress on Virtual testicles so we can develop much more quickly than we could a decade ago, which is empowering. We've bought this company that makes its really fancy lidar

that unlocks our ability to see further than other people can and not just he further but she better and that we can see not just where stuff is in the world but how fast it's moving towards or away from its instantaneously and we think that really unlock the power of what we're building one aspect. That's somewhat different about Aurora is that you're an independent come And many of your competitors are subsidiaries of big auto companies Cruise as a security of GM or goes a subsidiary afford you have way more that's part of Google but you've chosen a path of staying

independent. Why is that an advantage or maybe it's a disadvantage y-yes oats. We we can build a company from day one as an independent company, right? We didn't build it with smoke and mirrors to flip it we built it because we believe in our mission to deliver the benefits of self-driving technology safely quickly Broadway in the way to do that is to a focus on the thing we can do better than anyone else in the world that is build the driver and to have that kind of focus means we got to do one thing right? We don't want to be part of some tech company that's making money selling

ads or phones or whatever you do. Let let's have a business that's really about the thing we care about and we think that that Focuses is a power and then beyond that that driver that were building we think an impact will broadly I can drive little compact electric cars that can drive minivans they could drive box truck to drive a Class A trucks and so we can build a platform that supports all parts of the transportation ecosystem make him we can partner with our vehicle Partners wasn't making trucks or cars and support them in building their business and and need to get something

from one place to another whether it's a person or a thing. And so whether that's a taxi company or Uber or Lyft we can support them or if it's logistics company or a shipper like a an Amazon or a FedEx or a Google UPS or Walmart and buy being independent. We can actually work across all those companies which we think is really important to have any impact in the world. We want now there's a lot of sort of I call it micro segmentation. Happening in the industry of people saying oh I'm doing self-driving trucks and I'm doing

self-driving Urban delivery vehicles and I'm doing Robo taxis. And so is that a are those valid? Is that a valid segmentation? Is it really that different or is it, you know, you're obviously not taking that approach. So we believe that in so if you can pick up a product really delve into that by itself, then yeah, I can see some advantage to that. We think that the right path is to build a driver because really the hard part of the self-driving problem is not plugging it into the car and then I'll be able to speak Japanese if it's a Toyota and

speaking German if it's a VW, the hard part is understanding the world around the vehicle and figuring out how to move through it. So let's focus on that and then let's let's leverage solving the hard part of the problem by doing a bit of work across these different application. So we think that's the right path. I think is its kind of Painting is and he's not there but one of my pieces is flawed is is Trucking. So a couple of years ago that kind of the wisdom was hey where to go build a self-driving truck company because that way easier

than building, you know something to drive around Tampa Downtown San Francisco. Of course, is that most of the time to write you have to drive from here to La down? I-5 not a whole lot happens. It's pretty easy promise that sometimes it does happen. Something interesting does happen and when it does appear 7,000 lb of load moving to tell me in miles an hour the kinetic energy is profound, which you compare that to two cars bumping each other 25 miles per hour and it in a neighborhood. So it's higher risk. The other thing is most the time nothing happens, and if nothing's

happening, then you're not learning anything and so you're not taken that opportunity to advance the technology. So again, if you take I-5 from San Francisco to LA Right, maybe one thing interesting happens or nothing if we spend time driving down Palo Alto Downtown Palo Alto or Redwood City or San Francisco. Basically once a minute something interesting happen for learning orders of magnitude faster than folks in the trucking space. And so we think as this Aurora driver will will will will learn where it's easy to learn rapidly and safely and then we'll be able to apply that into things

like traffic that makes a lot of sense one of the things that's hard about this business you're in is measuring progress. You know, if your if I'm selling a sass product that has subscription. I can just count the number of subscriptions or the growth rate or whatever but you're working on a pretty fundamental technology which takes a long time to build and so there's a bunch of metrics out there that people site right they talk about numbers of miles driven they talk about this engagement rates from from rack. Drivers and whatnot. Are

those the right ways to measure progress in the industry. How are you doing on this? If you want to comment on that should we be looking at an entirely different way of measuring progress? We don't think those are great numbers to look at I can tell you where they came from. So way back in the day when I was running a Google program. We had a bunch of cars out driving and so we had a bunch of Miles which we said we drove much miles to anybody was excited and and then that kind of became that the de facto way of measuring yet in practice. We think that's a terrible way to measure

because one again take the I-5 example driving on I-5 for the most part is not that hard and so you don't learn much and so you maybe it works really well there but it's not actually going to help solve the broader driving problem. And so at Aurora our strategy is your be super-efficient about with the way we use our testing on road go to the places where we can learn Rapley that means that we're going to end up disengaging more often because it's a more interesting environment and we're learning and then the other is to really use put a bunch of emphasis on Virtual develop into Google

search testing tools. So you have a really interesting number so our vehicles and how can make left turns across traffic. So unprotected left turns is one of these hard things to do because you have to estimate the distance of traffic yet to find gaps in and make the turn well for us. We just got to go down the road and hope and see what happened. We we first had our our human drivers go out and make a bunch of these Maneuvers so we can kind of understand the characteristics of them and then we built a bunch of simulations and different tests that allowed us to understand, you

know, who developed the software again, and in fact, we made something like two and a quarter million left turns in simulation before we ever actually tried it on the real world and so you can imagine We were if it took us two and a half million iterations how long it would take us to trap a car. You don't drive around the block and learn and be by doing in a simulation first. We didn't have to put the public at risk and have to put our team at rest. We could, you know, we can try all kinds of things out. So so we don't think the metrics that are out there really the right one to use

internally. We look at all kinds of things. We do. Do we measure a certain stuff we measure how long does it take us to take software and put it on the car because that does affect the cycle time. We measure, you know, we have this whole sweet simulation tests both for relation of the movement of the car but also stimulation of sensory data, and we asked you know, how does the vehicle form against those we can kind of read about test and then as we're thinking about rolling a feature out have confidence that what kind of what limits we can put on in the real world but

realistically fear an outside Observer cuz a lot of the times that you're talking about there are internal measure my You're not going to necessarily sure they'll be out the rest of the world. So if you're an external Observer to this industry, what should you be paying attention to its extremely difficult right at the other folks face and I have a sense of where they are, but it's hard to measure objectively. I think you look at the quality of the team because that's that's what's the horsepower is going to ultimately a skewed and then you look at you know, the thoughtfulness of what they

put out in the world about how they're solving the problem. I think those are kind of the best way to look at it. You were one of The Originators of the Light Art usage of lidar Technology. I'm going to assume that most of the audience knows what light are as if not, you know, there's always Google for you. There is another very famous person in this industry his name starts with Elon who says that lighters are stupid with eyes supposed to drive with Cam. Yeah,

can you please respond to that response? Thanks Mike. I'd be happy to get right so I disagree respectfully. So so it is very beguiling right people drive around with two eyes. They're obviously not lidar and so there's an existence proof that it can be. I actually latch onto a different quote which was by Alex kaparthi who said lighter is a crutch. It's a way to avoid solving some problems with the camera. I say exactly right like we are not scientists about this right? We're not purest we're actually trying to get a product out into the market to make the road safer to make them

more accessible to make the more efficient and less less costly could get around. So we're going to find all whatever technological advantage we can I think about, you know, we're pretty convinced. Right way to travel is on things with wheels whether their bicycles are cars what turns out I was born with legs really many people here were born in legs and right that's what nature evolve to get us around. So, you know, if we said that we should drive cars because they don't have legs. It would make a whole lot of data. We got 5 minutes and I want to ask you two questions Chris.

The first one is it it's a bigger issue that people often to brush aside when they're talking about Ai and machine learning, but the fact is that the technology or developing will endanger the jobs of people who drive today truck drivers taxi drivers and so forth and that's their livelihood. How do you think about that problem? I think first it's real that we're developing a technology that I deeply believe we'll have a massively positive impact on society.

If you look in the US that 38000 Americans who died last year on our road globally want a quarter million people the status quo is broken. If you look at the accessibility my first six million Americans, they don't have the same privileges of getting around that you are many of you here do that. You can get a vehicle and drive it to giving them access is profoundly important. We can reduce the costs. Right? So traffic actions the us or something like at 800 billion dollar to trillion-dollar drag on the US economy. So there's a massive economic social personal benefit to

self-driving vehicles at the same time. There are people whose livelihood depends on it and And it is a it is a really tough conversation around to the broader good of the many over over the the few and how do we as a society come together to support those people in France at the same time. There's a lot of room for Hope here. So one example where where automation came into industry. I actually increased labor pool is the automated teller. So in the out of the ATM machine was invented the concern was

that suddenly there be this whole class of people bank tellers who would not have jobs anymore. And what happened was because the ATM made it cheaper to open a branch there were more branches and the branches needed more Bank branches that would form or Bank branches to lose more people employed is tellers and the tellers moved from basically the L side of it, right just their job being too effectively take checks in and hand money out to actually creating value for the company's and you have selling additional Prada. What not? So there's now more bank tellers employee today than there

was prior to the ATM be invented in our space by reducing the cost of friction of Transportation by making it safer, you know, they will be inelastic effect will we expand transportation to some degree? And I think that will create all kinds of new opportunities, but we have to find ways to to support these people through that transition the last thing I would end with with maybe two two more points or one is Many of these jobs are pretty pretty tough. Right if you think about truck driving one. There's a 70000 person shortage of truck drivers in the United States.

So it's not a particularly desirable job for most people died in the life expectancy for people who drive trucks is is lower than the average substantially and so we should be kind of asking the question over the long-term. Is this a job we really want people to do or is this a job where we we'd rather find ways to automate it and allowed that part of society to to eat somewhere else. So it's it's a difficult question. It's one that we can solve by ourselves the company we have to engage with the with the broader elements of society and government. That's great last

talked. A lot of people in the audience here are running their own startups you have you have to give a particularly. I would say complex task of running your business because you have a product that is going to take from the Inception of your business. Probably five years before you Ever ship anything that people will use and if there's a large group of people how many people work in Aurora now 444 50th. So you're you're you're managing a large group of people with a product that will take a long time to build and so I want you to talk about your approaches to leadership and culture

when faced with your type of problem of managing this business in a minute where you're going to work together with very low ego. So our problem is not something where we can have kind of a hero hacker in a corner throw together with them overnight and it's only have forty thousand users the next day right as Mike said this is a many year investment. And the problem is Broadway. We we we we are all the different kinds of Engineers, but then you know, we have to work on government relations. Because we have

all those Engineers we have to have a high quality human resource people. So it's really about recognizing that what we're doing is a team sport. It's a one of our core values that we have. No jerks. We have a rule no jerks the company and it matters because we have to be in it for the long haul. We it is a team sport. They have to be able to work together and we've been successful in hiring world-class people across our organization. And so it may be that you're really good at the thing you do and it's different of the person who sat next to you and if you don't

respect them, if your kind of your your ego gets in the way, then they're not going to be able to form at their best and we're going to fail so so keeping the team high-horsepower high quality people, but humble, I think it's been a core part of of how we've been successful so far isn't that having a vision that says, you know, 40,000 people a year die on the roads in the United States in a fix this problem. I know that will be reduced and a half for BLS has got to be motivating to the people to come to work everyday. Absolutely Anna and I think that's the other thing that I think about

when we build a team is you need to have two things with list three things. I think about fruit for employees one is the heat emission they believe in that meaningful in the world II is the they have to have a job that they're excited about great people to work with and the 30s. You have to pay them in a way that's not embarrassing. That's fair because everybody had crappy days until the mission is the thing that gets you through the crappy days, but if everyday is crappy, I it doesn't matter what the mission is. You're going to go find something else to do and then if you know, if you're not

being paid fairly then why you there it's a job after all Chris. We're at a time, but I just want to thank you. It's a it's a true privilege to be able to work with you not just because of the great Mission, but because of the great person that you are and more people should come work for you in case you were with your Can't we are hiring and just thank you for sharing your thoughts with us in good luck on the journey over the next few years. Thanks, Mike.