LISA CHOW: From Gimlet Media, I’m Lisa Chow and this is StartUp. We ended last week’s episode with Google hiring a bunch of people from the Grand Challenge races and starting its self-driving car project. It was 2009, and Google called this project a moonshot. In other words, it was wildly ambitious, with no immediate plans for profitability. Something that only a huge company with a lot of cash to burn would dare to pursue. And it’s not hard to see why Google was so excited about the prospect of a driverless future. It’s a pretty incredible thing to imagine. Take it from Jim Scheinman, a venture capitalist who could not be more enthusiastic about driverless cars.
JIM SCHEINMAN: I don't even like calling them cars because I think they're going to be something different. They’re going to be autonomous vehicles, they’re going to look different, they’re going to feel different. It’s going to be an amazing experience.
LISA: The kinds of things that Jim imagines can sound a bit crazy to the unconverted.
JIM: We're all going to love pushing a button and having this robot pick us up and we sit back and take a nap before we get to work, or catch up on your Game of Thrones or Netflix shows. Have a drink. You know, whatever you want to do: get get your nails done, get your hair done on the way to the meeting. All that’s going to be possible.
LISA: Jim thinks that with autonomous vehicles, you'll no longer want to own a car, because it will actually be cheaper to get around in a self driving taxi.
JIM: We're going to look back and say, “What were we thinking?” Why would we spend 35-50 thousand dollars on a thing that sits around taking up space for 96% of the time? It’s crazy. And not only that, I only really needed one seat for like 90% of the time I was driving, you know. So if you're going to go on a vacation somewhere and need to be driven and it's three or four in your family, you'll order that kind of vehicle. Most the time, though, you just want to commute, so you just need a one-seater. So the one-seater comes.
LISA: It really does sound like science fiction, what you're describing.
JIM: Yeah. But it's not.
LISA: Jim doesn’t only think driverless cars will change how we get around. He predicts all kinds of ripple effects. In this new world, parking garages go away because these vehicles will be constantly moving from one passenger to the next. But even bigger than that, with autonomous vehicles, you don’t have people drinking and driving or texting and driving. It saves lives. Being able to watch Game of Thrones in your roving one seater, that’s just the small stuff.
LISA: And when do you think this is going to happen?
JIM: I'm definitely much more aggressive than most people. I’d say, in a meaningful way, in certain areas of the United States, within five years.
LISA: Last week, we followed autonomous vehicles driving through the desert in a race called the Grand Challenge. But that race, it really kicked off a much bigger one: the race to actually get a driverless car out in the market and onto the streets. A lot of people are pushing to be the one to make Jim Scheinman's vision of the future a reality. From the companies you’ve heard of—Google, Uber, GM, Tesla, Ford—to smaller upstarts looking for a way to get to the front of the pack. Today on the show, we look at how the spirit of the Grand Challenge is still going on 13 years later. And how the challenge has gotten even grander.
LISA: One of the first people to jump into the driverless space to compete with Google was this guy.
KYLE VOGT: Hi Lisa, I'm Kyle.
LISA: Hi.
LISA: Kyle Vogt built a vehicle for DARPA’s Grand Challenge as a freshman at MIT. And after that, he went on to co-found a company that wasn’t in the driverless space. A company called Twitch that sold to Amazon for $1 billion. By 2013, Kyle was in a pretty good spot. He wanted to start a new company, and he could afford to take a risk. Afford to come up with his own moonshot plan. But he wasn't sure what to do yet. So he gave himself three criteria: his next project had to involve a challenging technical problem, it had to have a positive impact, and it had to be a good business—something that could really grow.
KYLE: I probably made a list of 50 different things, and actually self-driving cars wasn't even on that list the first time. And then it, it really just sort of hit me one day as I was sort of reminiscing with some college friends about the past, that working on the DARPA Grand Challenge was probably the most fun I had had while at MIT. And then I started thinking more about that and then realizing that if you could be successful at building and deploying self-driving cars, it would actually check all those three boxes. Like a 10 out of 10 on the scale for every single box, in terms of technical challenge, impact. And you know being a successful business and having scale. Once that clicked in my head like the very next day I started working on Cruise.
LISA: Cruise, Kyle’s new company. He knew he was up against big competition. Google had been working on self-driving cars for five years at this point, pouring hundreds of millions of dollars into its project. So Kyle figured he’d have to find a totally different approach.
Kyle: I said I'm going to look at this a fresh lens through the way of a startup might approach the problem which is we're gonna bite off the smallest piece of the overall problem that we can solve that really well. And then over time will lead our way into solving the larger problem.
LISA: Kyle decided Cruise would build something that would turn your regular car into a self-driving car, but just on highways. Cruise’s product was a rack that you’d bolt onto your car roof. It would hold cameras, radar, GPS and other sensors, and it would give your car autopilot capability on straight, simple roads. But right as Cruise was getting off the ground, something big happened that changed the way people thought about transportation. Ridesharing was taking off. It was in cities all across the U.S. It was expanding internationally. And its explosive growth was changing the industry that Kyle had just entered. The success of companies like Uber and Lyft showed that eventually people might not need to own their own cars. They could just call one when they needed to go somewhere. Add driverless cars to the equation, and ridesharing looked even more like the future. Because not having a driver means not paying a driver, which could make these services cheaper and available to more people. Investors looked at this…and saw a trillion-dollar market for the taking. And that helped Cruise raise more than $12 million in its first big round of funding. But it wasn't only investors who were seeing huge potential in a driverless future. Uber saw it too. It realized driverless car companies could become its competitors. And so in 2015, Uber jumped into the driverless game. Ex-CEO Travis Kalanick explained the decision at a conference that year.
TRAVIS KALANICK: This technology is coming. And so then, the question for us is does Uber want to be part of the future, or are we going resist the future. Like maybe the taxi industry before us. And really for us, we’re a tech company, and so our choice is to be part of that.
LISA: But not everyone was excited by the vision of a driverless ride-sharing future. It made car companies panic. Because if more people were sharing rides, fewer people would be buying cars. Some car companies decided to face this threat head-on—by joining the race. One of them was General Motors. In 2016, just two-and-a-half years after Kyle started Cruise, GM bought the company for over a billion dollars. Since then, Cruise has grown from 40 employees to over 350. And they’re no longer working on roof racks to make your car autonomous on highways. They’re part of the big race—the moonshot race—to be the first company to develop an autonomous car that can drive through all weather conditions, day and night, anywhere in the world. Here’s Kyle again.
KYLE: Whoever gets there first ,everyone that needs to compete with them is now dead in the water, unless they have a technology to be able to compete with. If one of the car companies comes out with a self driving technology first, the ride sharing companies would be in trouble or vice versa.
LISA: So now the question is: who will be first, and how will they pull it off?
NEWSCAST 1: Riders in Pittsburgh are getting a preview of what it’ll be like to be a passenger in a self-driving car. Uber has become a test program…
NEWSCAST 2: GM is already testing right now in San Francisco, Scottsdale, Arizona...
NEWSCAST 3: In more than a million miles of road tests, Google says only one minor crash has been caused by its self-driving cars.
LISA: A bunch of companies are testing their driverless cars—with safety drivers, just in case something goes wrong. And these cars are out collecting data about how people operate on the streets—data that will help these vehicles navigate the complicated world that we live in, where unpredictable things happen all the time.
KYLE: It's like how do we deal with construction zones while there's a thousand pedestrians and balloons and junk on the street from a parade. And you know mattresses falling off trucks and trees that aren't trimmed properly, all at the same time. Those are the toughest, meatiest problems, and the ones that we spend almost all our time on.
LISA: Cruise, Uber, and Google—they each have advantages that could help them be the first to get a fully autonomous car on the road. Google has the most sophisticated and detailed mapping and navigation system in the world, a huge plus in the self-driving space. Uber has its app, which millions of people have already downloaded to their phones. It also has a huge amount of data about how consumers behave—where they want to go and when they want to get there. And Cruise has GM, a company that knows how to manufacture cars at scale. But what do you do if you don't have these competitive advantages? You’d think that no founder in their right mind would dream of launching a scrappy startup in a space already dominated by giant companies. But that’s where you’d be wrong. Who dares to do it. That’s after the break.
LISA: Welcome back to StartUp. Remember Kevin Peterson from last week’s episode? The brilliant coder who spent six weeks in Nevada teaching Sandstorm how to get around?
KEVIN PETERSON: I really remember being in the desert for the first time, and that was it for me. I really fell in love with autonomous vehicles driving outside.
LISA: That was 13 years ago. Here’s what he’s doing now.
(SOUND OF ROBOT ROLLING DOWN STREET)
LISA: following a robot down a sidewalk in San Francisco. The robot is about the size of a washing machine. It’s white and boxy with four heavy duty wheels. The company that built this robot is called Marble, and Kevin is one of its cofounders. The robots deliver takeout and groceries. Right now Marble is in a pilot phase, operating in just a couple of neighborhoods in the city.
KEVIN: So at any given time it’s tracking everything that’s around it. From telephone poles, to people, to dogs, and it’s constantly making decisions about what the right action is based on what it’s seeing.
LISA: At one point during our trip, we’re crossing the street and a car makes a turn into the crosswalk and stops right in front of us to see if the robot will run into it. The driver leans out the window and asks Marble’s CEO, Matt Delaney, will it move around me?
MATT DELANEY: Eventually, yes. Right now, it’s just hoping, hoping that you’ll go.
LISA: The driver decides not to wait the robot out, says goodbye and keeps moving.
MATT: Cheers.
DRIVER: Later.
LISA: So does that happen a lot?
MATT: No, I think that’s the first time that’s ever happened. A more common thing is a lot people love to see if the robot stops if they jump in front of it, and of course it does, but it’s everybody’s favorite pastime.
LISA: Marble is doing something a bit like what Cruise did with its roof racks. It’s not going for the big goal of building a fully autonomous car. It’s biting off a smaller part of the problem first. Marble's autonomous vehicles, of course, aren't even transporting people. And they move as fast as you and I walk, about three miles per hour. At that speed, a lot of the harder technical problems of self driving cars just go away, because these robots don’t need to see very far in the distance, or react really quickly to things in their path. And a mistake at walking speed probably isn’t going to kill anyone. Kevin says the area where they're learning the most is in engineering the robot’s soft skills.
KEVIN: How does a robot interact with people. How does it deliver to a person. How does it drive down the sidewalk politely, which I think is a really fundamental, interesting question. At some point robots are going to be out there operating in everyday lives across the board. They're going to be in hospitals. They're going to be in your apartment helping you out. They're going to be doing deliveries. They'll be everywhere. And in order to do that we have to be able to interact with everybody in a very general way. So our product is one of the first ones that gets to do that in a broad scale.
LISA: And delivery is a super fast-growing market, since everyone orders stuff online and wants their shoes, diapers, and food now. Kevin believes that by tackling what the industry calls the last mile problem, getting your package to your doorstep at relatively low cost, Marble will learn a lot.
KEVIN: It’s an area where things are about to change. Where the technology is viable, and where the business is really, really interesting.
LISA: Marble’s starting small but its ultimate goal is to automate the transportation of all goods, and that could at some point involve self-driving cars, trucks, planes, drones. Another startup in the space is called May Mobility. Their corner of the market is autonomous shuttle buses. So they are moving people, but they’re doing it in a really limited way. Their business is getting people around places like retirement communities, office parks, and college campuses. Places that are far less complex than busy cities. Last month, May Mobility ran its first pilot. For a week, its autonomous shuttles took employees from an office in downtown Detroit to a nearby parking garage. We sent producer Tyler Scott to check it out for us, and he told us about it.
TYLER SCOTT: It basically looks like a mini version of a minivan. It had three rows of seats in it. In the first row seats, it’s like a typical front row with your driver and passenger seat, except there’s no steering wheel. It’s just kind of like this t-bar thing, it’s more like a bike handle.
(SOUND OF TYLER GETTING IN VEHICLE)
TYLER: You get in and the guy who is in the driver's seat, quote unquote, just presses a button and this electric vehicle takes off pretty quickly and pretty silently.
TYLER (in May Mobility vehicle): The whole top half of this thing is glass, it’s sweet design. And it is cool we’re kind of just rolling through downtown, it’s at night. So everything’s lit up.
LISA: The shuttle drove a one-kilometer loop. It ran only in the evening, from seven to ten, when there isn't a lot of traffic. And it never went over 25 miles per hour. That’s how May Mobility, just like Marble, has found a way to wriggle into the market. Its cofounder and CEO, Ed Olson, hopes they can start selling their service next year.
ED OLSON: The central hypothesis for us is really that there are routes out there that we could do safely with the technology that we developed right now. And of course there are many routes which we cannot yet do safely at high levels of reliability but we'll get there over time. The idea of May is but let's go out and get those, the ones that we can do. And by tackling them, we're going to learn a lot that we think will actually help us develop the technology faster.
LISA: Ed has this analogy that he uses to describe the difference between what his company is doing and what the big driverless car companies like Google and Uber are doing. And bear with me. It’s an analogy about ducks. The big companies are trying to train their ducks to do all kinds of tricks. Pirouettes, waltzes, flying in complex formations. Ed—he’s aiming for something different.
ED: The idea here is to be able to market dumb ducks. Ducks that only know a couple of tricks.
LISA: On the evening that our producer, Tyler, rode in Ed’s dumb duck…
TYLER: It definitely when I was in it like didn’t go much faster than 15 mile an hour. There wasn’t much going on and then all of a sudden we went through this corner and these two kids on skateboards hopped off the curb and skateboarded through the intersection in front of the shuttle.
TYLER (in vehicle): Woah!
TYLER: The van, it might’ve slowed down a little bit, but it didn’t come to a stop.
TYLER (in vehicle): Close call there.
TYLER: A couple inches from like, hitting this kid on the shoulder.
LISA: And there was another problem during the pilot. On the first night, one of the shuttle vans approached a manhole cover with lots of steam coming up from it. The steam was thick enough that it confused the van. Was this a solid object or something the van could drive though? The safety driver ended up taking control of the van so that they didn't hold up traffic. These types of problems, they’re essentially the same problems that every driverless company is puzzling through. And they’re problems that human drivers are pretty good at dealing with. The driverless car industry spends a lot of time talking about how bad people are at driving. But Ed says, it’s actually the opposite.
ED: People drive 100 million miles between fatalities. In other words, you take the total number of miles driven in the United States and divide by the total number of people who are killed, and it's 100 million miles per fatality. That's a lot of miles. So when we look at building autonomous cars that are as good as human drivers they need to be amazingly reliable.
LISA: Safety drivers in autonomous vehicles intervene at wildly different rates, depending on the company and where that company is testing. Documents leaked from Uber earlier this year showed that its safety drivers in Pittsburgh and Arizona were intervening every 50-200 miles to prevent collisions. That doesn’t mean the cars would have killed someone had the drivers not stepped on the brakes or grabbed the wheel. But it does give you some idea of how often humans are having to correct the computer. And it’s a long way from the 100 million miles that Ed points to.
ED: I think a lot of the large automakers are painting very rosy pictures of where the technology is going to be and when it's going to get there. So you hear a lot of automakers talking about 2020, 2021, basically saying we're going have full autonomy before you know it. I think this is a good example of where the marketing is ahead of the engineering.
LISA: How long do you think that road is?
ED: It could be decades. Multiple decades. We need fundamental breakthroughs or eureka moments in sensors, the hardware itself, in the perception algorithms, and in the planning. And so we’re multiple breakthroughs short.
LISA: I asked Kyle, co-founder of Cruise, now owned by GM, when he thought autonomous vehicles would be on the road without safety drivers. He didn’t want to answer the question, But last year, during an interview at a TechCrunch conference, he did give some sense of when he thought things would start to really change on the roads.
DARRELL ETHERINGTON: Future of transportation, what does it look like, let’s say, I don’t know, 25, 30 years from now?
KYLE: 25 to 30 years, I hope at that point we’ve got to the point where we’re starting to pull the human drivers off the road.
DARRELL: Just starting to? It’s gonna take that long?
KYLE: Well, you know, it’s gonna take some time.
LISA: Three years after launching its self driving car project, Google's cofounder, Sergey Brin, held a press conference at the company's headquarters in Mountain View.
SERGEY BRIN: I think the self-driving car can really dramatically improve the quality of life for everyone here in California, in the country, and in the world.
LISA: At the end of his prepared remarks, there was a question from a reporter.
REPORTER DAN SIMON: When do you see the self-driving cars having a real practical application for everyday folks? How many years away?
SERGEY: That’s a great question. I don’t want to overpromise right now. We have some pretty ambitious targets for our team. You can see them stressing just looking at me answer this question. Anyway, you can count on one hand the number of years until ordinary people can experience this.
LISA: That press conference was five years ago. Matt Delaney, from Marble, the robot delivery startup, he says he has a running bet with people in the autonomous vehicle space about the timeline, and he’s noticed that the more thoroughly you understand the challenges, the farther away you believe we are from a driverless future. Building a fully autonomous vehicle is still a moonshot. The technical challenges are still enormous. What’s changed is the marketplace. There are more companies in the race today than five years ago. But the finish line is still far away.
LISA: StartUp is hosted by me, Lisa Chow. This episode was produced by Bruce Wallace, Luke Malone, Simone Polanen, Emanuele Berry, Amy Standen, and Max Gibson. Our senior producer is Molly Messick. We are edited by Annie-Rose Strasser. Mark Phillips wrote and performed our theme song. Build Buildings wrote and performed our special ad music. For full music credits, visit our website. David Herman and Ian Scott mixed the episode. Special thanks to Aarian Marshall, Martin Ford, Matt Johnson Roberson, Mike Isaac, and Mark Harris. To subscribe to StartUp, go to Apple Podcasts, or whichever app you like to use. Or check out the Gimlet Media website: GimletMedia.com. You can follow us on Twitter @podcaststartup. Thanks for listening. We’ll see you next week.
Correction: This story originally identified Kyle Vogt as the founder of Cruise Automation. He is actually the co-founder. The episode and transcript have been updated to reflect this.
We do our best to make sure these transcripts are accurate. If you would like to quote from an episode of StartUp, please check the transcript with the corresponding audio.