a16z Podcast - a16z Podcast: Self-Driving Cars — Where Are We, Really?

Episode Date: January 31, 2018

As cars become more like iPhones and less like just, well, cars — everything changes, from data to mapping to interfaces to security and more. How so? Where are we anyway, given all the hype around when self-driving cars will appear everywhere? And where are new opportunities in the space? This episode of the a16z Podcast, based on a panel discussion from the most recent a16z Summit, features a16z research and deal team head Frank Chen in conversation with various companies doing different things in the autonomous space. Guests include: Taggart Matthiesen, head of product at Lyft, which is developing autonomous car technology; James Wu, CEO and co-founder of DeepMap, which focuses on full-stack HD mapping for autonomy; and Qasar Younis, CEO of Applied Intuition, which provides advance simulation software for autonomy. ––– The views expressed here are those of the individual AH Capital Management, L.L.C. (“a16z”) personnel quoted and are not the views of a16z or its affiliates. Certain information contained in here has been obtained from third-party sources, including from portfolio companies of funds managed by a16z. While taken from sources believed to be reliable, a16z has not independently verified such information and makes no representations about the enduring accuracy of the information or its appropriateness for a given situation. This content is provided for informational purposes only, and should not be relied upon as legal, business, investment, or tax advice. You should consult your own advisers as to those matters. References to any securities or digital assets are for illustrative purposes only, and do not constitute an investment recommendation or offer to provide investment advisory services. Furthermore, this content is not directed at nor intended for use by any investors or prospective investors, and may not under any circumstances be relied upon when making a decision to invest in any fund managed by a16z. (An offering to invest in an a16z fund will be made only by the private placement memorandum, subscription agreement, and other relevant documentation of any such fund and should be read in their entirety.) Any investments or portfolio companies mentioned, referred to, or described are not representative of all investments in vehicles managed by a16z, and there can be no assurance that the investments will be profitable or that other investments made in the future will have similar characteristics or results. A list of investments made by funds managed by Andreessen Horowitz (excluding investments and certain publicly traded cryptocurrencies/ digital assets for which the issuer has not provided permission for a16z to disclose publicly) is available at https://a16z.com/investments/. Charts and graphs provided within are for informational purposes solely and should not be relied upon when making any investment decision. Past performance is not indicative of future results. The content speaks only as of the date indicated. Any projections, estimates, forecasts, targets, prospects, and/or opinions expressed in these materials are subject to change without notice and may differ or be contrary to opinions expressed by others. Please see https://a16z.com/disclosures for additional important information.

The content here is for informational purposes only, should not be taken as legal business, tax, or investment advice, or be used to evaluate any investment or security and is not directed at any investors or potential investors in any A16Z fund. For more details, please see A16Z.com slash disclosures. Hi and welcome to the A16Z podcast. This conversation looks at where we are now in self-driving cars and in the big trend of autonomous and self-driving cars of all levels. From data, to mapping to interfaces to security and more. Moderated by A16Z partner Frank Chen, the conversation, part of our November 2017 summit event, includes, in the order in whose voice you will hear first, Kasser Yunus, CEO of Applied Intuition, James Wu, CEO of DeepMap,

and Taggart Matheson, director of product at Lyft. So on the scale from crazy, 13 years, to myopic, never. When does self-driving start? It's not going to be one of these things where we wake up one day and, like, autonomy is here. You know, every two years, every one year, we're going to just make that incremental move. And I think when the CT6 launches with Super Cruise or the Tesla autopilot, and then maybe 10 years from now, we'll wake up and you have that option to go to the dealership and buy a car that does everything we think autonomy does.

The major OEMs in Europe, U.S., and Japan, other places, if you look there at the timetable, it's about 2020, 2021, they're going to start shipping level four vehicles. I think it's probably happened much sooner than 13 years for everybody to have the capability to use a level four technology. Over the next couple of years, you're going to start to see very small pockets where these vehicles are, in fact, operating at a level four capability. But I think the challenge here is there's a cost aspect. If you look at the sensor suite on these vehicles, it's not economically

feasible. Two, once you get into the world of a large-scale deployment of vehicles, it's not just about the car being able to navigate the environment. It's also just kind of the fleet operations as well, from prepositioning to fueling. It's going to take a number of years before a large majority of these vehicles are autonomous. What you're going to see over the next couple of years is small pockets of innovation, where over time, more and more vehicles are going to be operating in greater situations in terms of level four. And then if you think about sort of the long pole in the tent, do you think it's going to be technology, it won't work, is it economic, you hinted at this, maybe the sensors are too expensive, is it policy, the governments won't allow it, or something else. What will cause it to come more slowly, if anything? Yeah, I keep coming back to the economics and at scale. First of all, no one's at true level five or even level four at this point. And to do that at scale in all different kinds of, of conditions from rain to snow, even in a geographically confined area at a point where

you figured out the economics, I think, is going to be quite a difficult challenge. We'd say infrastructure is a big thing. And I don't mean infrastructure in the traditional sense. We're talking with infrastructure is like software infrastructure and the actual product getting out to the market. Censor suites, it's like a real problem. I think that hasn't been solved yet. And I think I would say, secondly, it's cultural norms. Like, you know, I think The old adage in automotive is like the stop sign came out 10 years after the Model T. And so there's a lot of regulatory hurdles that need to be kind of defined and overcome. For me, I think the major bottleneck at this stage is probably talents in the next few years.

Because suddenly, you need a lot of robotic experts. A lot more computer science, engineers, scientists. And this will be a bottleneck four cards. state. This is the heart of like, you know, making it happen. I think that's what you're really hearing is to make these long-term visions happen, there's just a lot of these individual hurdles which are not trivial. They're non-trivial. So most of the discussion around self-driving cars on the benefit side is around convenience. You don't have to drive yourself in safety, right? We're going to stop killing people on the roads with these dangerous cars. What other ways are our cities

or our lives going to get better aside from convenience and safety that nobody's talking about. about. Crime rate will drop, in my opinion, because basically every car will become a moving sensor, collecting tons of real-time information. It's like a surveillance system, right? And also the housing price is probably going to change because people don't have to like pack together. And also all these parking space will become available space for housing. There's a huge amount of opportunities in software. There's a hundred tier one suppliers right now that make over a billion in revenue a year. I'm from Detroit. I grew up in that ecosystem. I really wonder, like, what's going to happen to those companies? Yep. I think the real bloodbath in this ecosystem

is not going to be at the manufacturers level, because consumers have some affiliation to, like, a Tesla vehicle or like a Mercedes-Benz, et cetera. It happens at the tier ones. But where there's that destruction, there's an amazing opportunity. You have these massive companies which are going to be almost useless in this new economy, and you have this new, huge market. So I think the untapped opportunity is, I think there's still huge, huge potentials for software-specific vehicle kind of applications. And I'll kind of play on top of that. I mean, we see this today with ride sharing. It's getting time back. And so I go down to Palo Alto all the time, and I can't tell you how quickly I want to have an autonomous vehicle drive me. I spend 30 to 40 minutes on, like,

the best day getting down to Palo Alto. But the hard reality is, how can I reimagine that time? I mean, we were talking about Netflix earlier and Stranger Things. The beauty would be Netflix already knows that I'm on season two, episode four. The car should already know that it's, you know, 40, maybe an hour long, and it should just pop in that video for me. And whether it's like something, you know, focused on entertainment or just having that connectivity from a work perspective or connecting with friends, the point here is that you can get that time back. And I think that's going to be pretty magical. I love this idea that sort of personalization, your fleet remembers what you were doing, what, you were using becomes the buying criteria, right? So it's not the door thunk. It's the,

do you remember that I was watching Stranger Things? All right. Good. I have seen now probably half a dozen startups that are working on basically remote driving, whether it's remote driving a car or remote driving a truck on the assumption that the algorithms won't be able to figure out what to do in all cases. So is this going to be a thing? How long will it last? I think it's going to be a thing. If you think about the challenge of self-driving vehicles, it's kind of the unknown, unknown. It's the situation or the scenario that you never tested or was aware of. And so, you know, a car gets into a situation where it's not able to classify or understand the scene around it.

You can imagine a teleoperator quickly going in there and helping, like, to some extent, labeling and letting the car kind of understand the context of the environment to maybe even drawing a line to say, look, vehicle, here's a valid path for navigation. However, I would argue on the teleoperations, you need to be very, very careful here where I don't think they're going to be driving. the car. I would almost argue that it helps the car understand the scenario or the scene and then leave it up to the vehicle to then do that navigation and motion control. The concern I would have is one in terms of latency, so the operator doesn't actually see something that happens in the scene.

And then the other thing just comes, it's yet another vector for potential people taking over the vehicles, security-wise. For sure. To me, I think self-driving is really not try to get rid of the driver. It's actually get people more efficient. Like, instead of driving one car, one person is driving or managing thousands of cars by one person. I think that's self-driving is still a technology, is a tool to enable human being. It's like the DevOps operators in data centers, right? So now, like, one single guy can manage millions of machines. I think teleops is going to be a critical part of that software infrastructure. I think less in the traditional passenger space. I think those edge cases relatively are less than if you're doing like automated

construction equipment. Caterpillar has done things like this in coal mines. We're having a teleoperator kind of always on call managing those fleets, I think becomes really, that's where you're in kind of an unbounded environment. But it's one of these like, I think just chunks that will exist and probably for a long time. Ultimately, I think those companies will become fleet management companies. But yeah, I think it's going to happen without it out. And to add to that, actually, Even we have this kind of operations for planes, for trains, for subway systems, even for shuttles already there. So we were honored to have Governor Doug Ducey from Arizona join us at the conference. And he's on the record saying, look, I want Arizona to be the friendly estate for these types of experiments.

So if you could do one or two things as the governor, what would you do so that we get to this world faster? Well, I mean, I think Arizona, I think Nevada, I think Michigan and Florida, I think there are states who are like very, actively asking that question. And I think they're doing a lot of the things that you would expect. We build advanced simulators for autonomous vehicles. And, you know, we very much engage with regulatory kind of bodies, both the federal and state level, to really answer this long-term question of if the DMV goes away and you have these over-the-air updates to self-driving cars, how do we make sure that we don't have some regression in the software, which ends up creating a scenario which puts the entire industry a step back. That's kind of where the,

I think a real bleeding edge of the conversation is it's not the 10,000 vehicles. It's really the what is it when we have a million vehicles and you have updates and software happening every single day. Once you have thousands of players doing really edge case, you know, essentially autonomous robots, not necessarily only self-driving. That's going to be a big question. The first time we're introducing really intelligent robots into the world. And it's the first time we as a society you have to regulate these robots and in a way that is not constraining innovation

but at the same time not putting the public at harm. So let me push on that a little, which is do you think the government should have a role on over-the-air software updates, right? So today we do lots of over-the-air updates to our iPhones and Android's and to Windows. Government has no role in that. Would you argue that

the government should in the self-driving car world because the stakes are higher? You need to have some basic standards. So whether that's cybersecurity, whether that's, there's multitude of things that we can talk about there. And the professional quote-unquote companies that have brands that have so much riding on it that can't afford to have a real issue, that's not the problem. It's a long deal. It's like fintech regulation. It's not regulating necessarily the Goldman's and Morgan Stanley's the world. It's this individual person who's trying to sell

penny stocks to people who shouldn't be buying that. They're not accredited investors. That's where regulations really come in. It's to protect the uninformed. And so that kind of version of this world is, whether that's construction equipment or that's shuttles or whether that's passenger vehicles, definitely people can die. Maps is a pretty sophisticated product, but we would do, like, updates where we'd have P-0 bugs introduced into, like, the public domain, and we'd immediately fix those. You can't do that in autonomous vehicles. I couldn't agree more on, like, making sure that there are standards.

I think the challenge here is that you don't want a patchwork of different standards. You don't want 51. Correct. Correct. But this isn't something that you should just leave it up to a group of. organizations. I think this is a collaboration between these companies today as well as both at the state and federal level, getting to some level of certification and standards to ensure that these are safe vehicles. I think the government can do two major things to help the self-driving

industry move forward. One direction is to promote cooperation and provide digital infrastructure, either add investment in digital infrastructure building or promote and help companies jumping in to help build the digital infrastructure. By digital infrastructure, I mean like HDMAPs. The other direction is to draw the boundary and limit where we can actually play with self-driving technologies.

For example, we can probably have some certain street paved with a self-driving lane instead of carpool lanes. that's kind of giving people some kind of a boundary where the technology is. I want to turn the line of questioning a little and talk about sort of our personal experiences. So what would you say to the pitchfork people, right? Like I love driving, driving is independence, it's part of myself identity, I identify with my car. What would you say to somebody like that who surfaces those types of objections to this future?

Well, I mean, I'm as much of a car guy as anybody else. I grew up in the Detroit area. I went to GMI for undergrad, worked at General Motors for five years. I love the auto industry. I think mixing those emotional sentiments with what's right for society shouldn't and is not an individual decision. It ultimately ends up being a decision that we all make together by the choices that we make and the way that we allocate capital both as private investors and as individual citizens. Whether you like it or not, it's going to happen. The second thing I would say is the reality is it's not that we're going to take away people's cars. I like to go to the

track. It's a hobby of mine. I don't think that in the duration of my life is going away. the people who are really into cars, I don't think they have anything to worry about. And for everybody else who doesn't want to do a commute, it's a wonderful change. So I don't think it's like this binary, you know, we have to have a revolution to inject this product into the consumers. What's probably going to happen is car ownership is not going to go to zero. What will happen is this really becomes more of a recreational and less of a utility. And so if you want your hot rod or if you want your, you know, muscle car from the 60s, I don't think that anyone's going to stop you from doing that. I think the interesting question becomes over a long enough,

period of time, do states, do cities allow potentially, you know, roadways or arteries for transportation that have fully autonomous to have human drivers in there as well, just from a safety perspective? You can still have your muscle car. In 20 years, can you drive that muscle car in the same streets? I don't know. I think that's an interesting question. Like, you don't see horses going down 101 so often. Just to follow up, but I will point out that for people who love to ride horses, there are still places for you to ride horses today. And also, I think, For self-driving cars, it will heavily rely on data, especially mapping data. So I think for certain space, like rural regions or tough road and people seldom drive,

where you don't have enough data there is probably still need human drivers there for a long time. So this issue of security, and we've been talking around the edges, I'm going to sort of hit head on, which is how do we deal with it? Are we going to have to have open source? so everything is open to inspection. Are we going to need to get government regulations in here, the same sort that would govern over-the-air updates? Like, what's the answer to how do we make sure

that these things can't be hijacked? The biggest correlation here or analogy here would be the problems that we currently have with the web and mobile web. I mean, we have massive data breaches so common that we can't even list them anymore. I think we will have them. I think the key point will be

there will have to be some version of overrides that are still low. local to the car. That's why you won't see this like iRobot type of environment emerging immediately until that infrastructure gets built up. We go through some of these growing pains, stepbacks, and then I think ultimately we'll have that maybe in 20 years, that future that we think of with no steering wheel, no brakes. I don't think anyone here is going to claim that they're going to be able to solve security. I think it's going to be a very big challenge, and anyone looking at this space is absolutely digging into that. And that's one of the reasons why I was talking about

remote operators and how you limit the surface area by which someone can attack the vehicle. The minute you have an open system is the minute someone can get in, but even if you have a closed system, there's the physical area of the car, whether you're going to open up the back seat and maybe there's some access and now you can get into the, you know, the brakes or the steering. It's going to be a problem that we're going to have to solve. And there will be a lot of opportunities in that space for startups. We are focusing on data security. So when you look at the self-driving vehicle, it's actually collecting a lot of data with a super high precision, centrimeter level of precision and also privacy information. So data ownership

and the data security is a major concern. We need to make sure the data should stay, where it should stay, and not get into the hands of people like try to abuse it. All right, my last question. Each one of you is basically working in a company that's trying to bring this vision to life. And so you've made the ultimate bet, right? Your time and energy and talent are headed in this direction. What are you most excited about? Saving lives. I really, 90-some-five percent of vehicle accidents can be avoided.

And so I think this is a huge opportunity for us to change that. I mean, we cannot overstate the amount of people that get, you know, injured and maimed and die every year. That's insane. The number is something like a 747 falling out of the sky every day and everyone dying. That's the equivalent of how many people die. And like, if that happened for three or four days, in any other, like in the air transportation industry,

we would land all planes and say, what the hell is happening that we have 747s falling out of the sky? An automotive, we just kind of like saying, well, you know what, let me just text something to my friend as I get in the car. So I think it's, yeah, that's the biggest benefit by far. Nothing even compares to it. For us, it's saving time.

So basically we want to enable our players, serious customers, serious players in this self-driving space, to get to the state where Waymo or the other, for runners are by providing this HDMAP solution to them. Good. And with that, that's a wrap. Thank you very much for coming up.