Autonomous Vehicles Special: Andrew Miller on Self-Driving Futures

Beatrice: I'm very excited to be here today with Andrew Miller. Andrew, I'm going to let you introduce yourself in a second, but I wanted to say that today's episode is a bit special; it's a themed episode, more so than we usually do. We're going to focus on autonomous vehicles because you just wrote a book about self-driving cars called The End of Driving.

I did want to preface this episode with a personal experience I had going into a Waymo in San Francisco. I think if you've ever been to San Francisco, you've seen that there are actually self-driving cars there now, and you can take them kind of like you can take an Uber. When I went into my first Waymo a few years ago, it was a really just profound and cool experience. And surprisingly, I think it was the experience where I most felt like, "Oh, the future is here." It was also really interesting just feeling like, "Oh, this makes sense." Why are we all driving around in cars and stuck in traffic when there's technically a much better solution?

So today, we're going to talk about this and check in on the current state of autonomous vehicles. Are we about to get them anytime soon? And what are the biggest thresholds? Technologically, they actually exist, but there are obviously a lot of other things that it takes to integrate this into our society. But Andrew, why don't you just start by introducing yourself a little bit and telling us about this book that you wrote?

Andrew: Thank you, Beatrice. It's a pleasure to be here with you. Thank you for the invitation. So yes, my name is Andrew Miller, and I am a writer and thinker about innovative mobility. I cut my teeth working for Alphabet's subsidiary, Sidewalk Labs, the late, lamented smart city company that was going to build a city of the future in Toronto. I was the mobility lead in Toronto, thinking about conventional transport—what you would want in a city, whether that's transit or bikes and so forth—but also thinking about driverless vehicles and what implications they would have for what a city should look like.

And so I got very interested in the topic. I've been writing about it and thinking about it ever since the pandemic. Some of that writing is in my weekly newsletter—I do a Substack called Changing Lanes—but some of it is in a book that I've been co-writing that is coming out in August of 2025 called The End of Driving. The thesis of the book, which we'll get into I'm sure today, is that driving automation could create several different kinds of futures. The one we get is the one that we act to get. But right now, if we do nothing, the default path is to a bad outcome where this technology makes life worse, which we don't want. We should act to make it better, and the book explains how that might work in practice.

Beatrice: It's really exciting. You do a backcasting from these scenarios, and it's something that we've done with the Existential Hope program quite a bit with world-building, so it's really nice to actually see that very much in action. I think to start with, it would be great if you could give us a snapshot of where we are with this today. I mentioned Waymo in San Francisco, but are we already seeing any other real-world deployments of driverless cars? One thing I'm curious about is trucks; it seems like it would make a huge difference if we could also have driverless trucks delivering goods. So where are we today with driverless vehicles?

Andrew: To quote William Gibson, "The future is here, it's just not evenly distributed." If you live in San Francisco or LA or Austin, then robo-taxis are just a part of your life, as you've described taking a ride in one. There are several US cities, all in the Sunbelt, where you can hire a robo-taxi today. If you live in China, it's also a reality in parts of Beijing and Shanghai. There's a gigantic robo-taxi deployment there that we don't talk so much about in North America, but it is quite real.

And then in the middle of the Sunbelt, in Texas, there is a lot of automated trucking, as you described. There's a whole "AV triangle" there between Houston and Dallas that is poised for what they call "24/7 lanes" later this year. They hope a 24/7 lane would be one where it's not a pilot; you can have driverless trucks in that lane anytime without any sort of permission or restriction. But if you don't live in the US Sunbelt or China, then you've probably never seen self-driving, and you don't have access to it. So it's a question of how fast it'll take to scale up from the 5,000 robo-taxis worldwide to the rest of the developed world.

Beatrice: That's really interesting. I had no idea that it was already happening in China, for example, and the trucks in Texas. It's on track, but unevenly distributed, like you say. So, based on this, do you have any take on the realistic timelines for more widespread adoption? And also, how would it be different across the US versus Europe, Asia, or Africa?

Andrew: Well, if I was going to be glib, I would say "2035" and lean back. But no, I don't want to be glib because it's a question that my co-authors and I wrestled with when we were writing The End of Driving. We decided to make it very clear that we weren't going to give a date. And the reason for that is because there's a hidden premise baked into the question, "When is it going to be available everywhere?" Well, that implies that the adoption of this technology is inevitable. It's not. It certainly is not. It'll get here faster or slower, or not at all, depending on what choices your country, your state, or your city chooses to make.

If we want it, we have to create the conditions necessary for it to exist. If we want it in a particular form, like trucking or robo-taxis, we have to set up the conditions that will allow that to happen. We can see there are some moves towards fresh regulatory frameworks. The United States just in April released a new regulatory regime for automated vehicles nationwide. The EU did something similar earlier this year. Canada is not doing anything at the national level, although the province of Ontario, where I live, is doing some pilots. It all depends on what policy choices we make. That'll tell you how fast it comes, or if at all.

I suppose I could say more about that. The first, most basic policy question is, "Are such vehicles permitted on the roads?" I'll use the Canadian example. In Ontario, you can, although you have to apply for a permit and meet various conditions. In the province of British Columbia, it is utterly forbidden. You may not have a driverless vehicle on the roads anywhere in the province. There are no exceptions, no path forward. It is flatly prohibited. And then of course, in San Francisco, you can operate robo-taxi services and make money doing it, provided you've got the right permits. And in Texas, with the trucks, when they get to 24/7, as long as you can prove safety, you'll be able to use those lanes. So, every jurisdiction is making different choices, and the choices they make dictate the results they get.

Beatrice: That's a very good point on creating the future. So, maybe it would be worth taking a step back because you and I seem to be on board that this is an exciting and desirable thing. But it would be interesting to dive into the actual upside. Could you give a summary of the main benefits we could gain from having autonomous vehicles? And also, I'd be really curious to hear if there's any data on what it would look like having driverless cars in traffic versus humans, for example, in the reduction of deaths and traffic accidents.

Andrew: Well, Beatrice, you've put your finger on it right there. Principally, overwhelmingly, the chief benefit is the safety that we would get as a result of automated driving at scale. Just earlier this summer in May, Waymo published a study. Now, it was Waymo, but it was peer-reviewed, so there's reason to believe it's reliable. They found that over 56.7 million driverless Waymo miles, compared to humans, there was an 85% drop in serious injury road incidents and a 92% drop in pedestrian incidents. That's compared to a human driving baseline. If you extrapolate that globally, you're talking tens of thousands of human lives saved every year. As far as I'm concerned, that's enough of a benefit right there—just the amount of human devastation and harm that would be saved if we took our hands off the wheel and let the computers drive for us.

But there are many other things you can say. They're more speculative. With trucking, there has been and will continue to be a shortage of truck drivers pretty much everywhere: Canada, the United States, Australia, Western Europe. If we could automate that sector, how much more robust would our goods movement sector be? How much cheaper would deliveries be?

There are others that are more speculative. If we replaced most private cars in cities and everyone just took robo-taxis, there would be efficiency gains. Downstream of that, there are all sorts of things we could get. We could design cities around ride-hail rather than around the private car, which would have lots of benefits too. Those are certainly valuable, but they're more speculative. There's no question that safety is the undeniable benefit that we would gain.

Beatrice: And maybe on the contrary, it would also be interesting to consider if there are any downsides. I think for a lot of people, what comes to mind is job loss. Many people provide for themselves by driving taxis or trucks. What do you think are the best arguments against driverless vehicles?

Andrew: There are good arguments against, it's true. I think we can deal with the job loss one quickly because I don't think it's the strongest. No doubt, some jobs will be lost. In long-haul trucking, as I said, there's a shortage of truckers there, but the more this technology scales up, the less demand there will be for humans to do that. As robo-taxis scale up, there will be less demand for ride-hail drivers.

But of course, jobs will also be created. If we really turbocharge long-haul trucking with automated driving, short-haul trucking, in the near to medium term, is going to be something that's much harder to automate. It's easy to automate a truck rolling down a highway; it's much harder to have it rolling through an urban core to a mobility hub. So there'll be more work for short-haul truckers near depots and within cities. As we get rid of ride-hail drivers, there'll be more jobs for teleoperation—people who sit behind a computer with a video-game-style steering wheel and can log into a robo-taxi that's run into an issue it can't handle to guide it out. There are going to be people who need to clean and maintain these vehicles.

So, as is always the case, jobs will be lost, but other jobs will be created. The key thing is that those jobs won't be immediately transferable. We will need to do some work to make it easier for the people who used to be driving for Uber to become teleoperators. But this has been true of every technological change. There's been friction, and it's the role of government and industry to find a way to ease those frictions.

But it's not just about jobs. There are other reasons to be wary, especially if this technology rolls out in a poorly conceived way. The most obvious way is that if you make anything simpler and cheaper—which driving automation will do—it'll make taking a road trip simpler and taking a ride-hail cheaper. When you make things simpler and cheaper, people consume more of it. So we've got the driverless cars, but the roads aren't changing, not in the near term. And that means more congestion. A lot more congestion. Every road trip is going to take longer because there'll be more vehicles on the road. So commute times get longer, and traffic jams get worse during peak periods, along with all the ways that congestion makes life less pleasant for everyone, like reduced livability in cities and more tire particulate matter in the air. Congestion is bad, and driving automation is poised to give us much more of it.

Beatrice: That's really interesting and not at all anything I had considered. But this is actually interesting in relation to your book because you argue that we should aim for this "robo-taxi future" as opposed to a future where everyone owns their own self-driving car. Could you maybe expand a bit on why you think that's a better model for society?

Andrew: This is an insight that flows out of the last one. If everyone is being chauffeured in self-driving vehicles but the roads are not getting any bigger, the way that you can retain efficiency is by having fewer single-occupant vehicles—or actually, fewer zero-occupancy vehicles. Right now, the one thing you can say is that the average number of people in a car is higher than one because there has to be one person operating it. Once that's not true anymore, the average occupancy of a vehicle might be something like 0.7 as you've got driverless cars with no one in them moving from point to point. That won't work, especially in cities where road space is particularly scarce.

So what we need is a world where there's more than one person, more than two people, more than three people in a vehicle. The more people we can get in a vehicle on a road, the better. In other words, we need shared fleets. How do we do that in an urban setting? That is shared rides in robo-taxis. If you can get multiple people in a vehicle, you unlock scale economies and you put less pressure on curb space. But it's hard. The thing to emphasize is that we could satisfy all of our mobility needs with fewer vehicles if we were able to share them. People don't want to do that, so we need to create the policy conditions that will make shared mobility attractive and accessible in a way that, absent some sort of intervention, it won't be.

Beatrice: Yeah. I think another interesting point that you made in the book, that I really hadn't considered before, is that one of the biggest problems is curb space and having space for pickups and drop-offs. Could you maybe expand a bit on why that's a really critical challenge to solve?

Andrew: The interface between the road and the city is the curb. That's where trucks pause to have their drivers get off to deliver packages. It's where taxis pause to let people get in and get out. It's where private owners will just park their cars. Up to now, we really have not managed the curb very well. We had parking meters, and sometimes we have officers ticketing people, but we generally don't monetize that space or police it very well, and consequently, it gets abused.

That is a real problem now, and it's going to be intolerable in a robo-taxi world. If we have a world where people are taking many more robo-taxis instead of driving their own cars, they won't be going into garages to park. Everything will be pick-up and drop-off. As night follows day, if you want that world, you have to think very hard about curb management and seeing curb space as valuable real estate that you have to control. You have to use price signals to make sure that it is being used for the best possible purposes.

There are firms that are aware of this, firms that are monitoring what city curb space is available and trying to provide services for monitoring, monetization, and marketing. Those are going to have to go very big. We're already seeing in San Francisco the stress that robo-taxis, as they scale up, pose on urban systems. San Francisco really is a laboratory for the rest of the world, I think. Atlanta, Waymo's coming to New York now—all these other cities should look closely at that and try and get ahead of it, because otherwise, it's just going to be a world of pain for everybody.

Beatrice: Yeah, that's really interesting. The thing that I've noticed in San Francisco is that the Waymos are really careful—they're obviously programmed to be—and so sometimes they'll just get stuck. I guess this is where you could actually have humans come in and help sort out the situation. Do you know if that's how they solve it now?

Andrew: Yeah. Both Waymo and Tesla—although they don't really like to talk about it much—both firms have what we call teleoperators. It is still the case that there's no automated driving system that can handle everything. They get confused when they encounter situations that are outside of the training set, or situations that a human would intuitively understand. Like when you and I are driving, we understand that it's okay to cross over the yellow line into the oncoming traffic lane if there's no oncoming traffic and there's an obstruction. We don't even think about it; we're careful, but we do it. It's much harder for machines to do that. It can lead to situations where they just don't know what to do and they pause.

How do you handle that? Well, if you've got a safety driver in the car, they can just take over. But often, Waymos don't have those. The way that you get around it is that someone else in a building across town logs into their computer, sees what the machine sees through its sensors, and then gives it a trail of digital breadcrumbs to follow. They guide it out of its jam and into a situation where it can pilot itself again. I suppose in the very distant future, when the technology has really elaborated, we'll perhaps need one teleoperator for every hundred or one for every thousand robo-taxis. But in the near to medium term, the ratio of cars to teleoperators is going to be low. So, teleoperation of vehicles is a growth industry. It's a job that kids in school today might be doing in 10 years.

Beatrice: So, it would be really interesting to hear your best-case scenario for a future with driverless cars. And also, do you think that the best case is 100% driverless cars, or will there still be some cases of humans driving?

Andrew: The best-case scenario, I would say, is a world where long-haul trucking is automated. Trucking is a really hard job, and that's why there are trucking shortages. If long-haul trucking is automated, goods movement is cheap, which means wider prosperity and wider access to consumer goods. It's a world where robo-taxis in urban and suburban environments are ubiquitous. A realistic vision would be that everyone today who owns two cars owns one and uses a robo-taxi for all those other trips. Everyone today who owns one car owns zero and uses a robo-taxi for everything because a robo-taxi is always available, cheap, reliable, and safer.

In a world like that, everyone gets to take more trips. I'm an old-school liberal at heart. I don't want to tell people what the good life is, but no matter what your vision of the good life is, it's easier to achieve if you can travel more. Whether you want to travel to a sporting event, to work, or to visit your sister for dinner, cheaper, more reliable, safer travel makes it easier for you to achieve that.

And there are more speculative gains. Particularly in cities, if more and more people use pickup and drop-off space, we need fewer parking spaces. That's space we can reclaim. More speculatively, because they drive better, they don't need really wide lanes, so we could narrow lanes and take back space for other things, which in urban environments is highly valuable. Now, infrastructure is hard, and that'll take decades, but it's something that would be on the table in an all-automated world. And above all, there'd be more of us. Fewer people would be hurt or killed, so fewer people would have to mourn the death of their loved ones in road incidents, which is something that these days we just take for granted, and we don't have to.

I don't think you'll ever see a world where there are no human drivers. There will always be enthusiasts who enjoy driving. There'll always be people in the countryside. Self-driving is going to come to rural areas last because there's less data in the training set and more edge cases. Ironically, you'll see driverless tractors on farms faster than you will see driverless trucks going from the farms into town. But I don't see a world where humans are no longer behind the wheel. Then again, it's still the case that people ride horses. Some people do it, we've all heard of it, but almost no one knows someone who uses a horse as a mode of transport as opposed to recreation. I think that's the world we're ultimately going to get. I don't think I'll live to see that world, but I welcome its coming. I think it's better for all of us when it arrives.

Beatrice: Just out of curiosity, you said that people who own two cars might just own one. Will that car also be driverless?

Andrew: So, there are two ways we are getting to fully driverless systems. One is the Waymo or, more precisely, the Zoox way. Zoox is the other big robo-taxi company, an arm of Amazon. Zoox's robo-taxis have no steering wheel, no human interface to drive the vehicle at all. That's one way to go. Another way is what we see Tesla has been doing. Tesla started with its Autopilot which, despite the name, is a driver-assist system. It increasingly takes the mundane tasks and just delegates the trickier ones to the human. But every year, the set of things it is incapable of handling shrinks a little bit. So you gradually move to a world where there's less and less for a human to do.

Ultimately, private cars will also have... they'll have a steering wheel, but more and more, there'll be a metaphorical button on the dashboard that you can press and the car will drive itself. That is actually the competition to a "robo-taxis everywhere" world. That's a world where it's exactly like it is now: everyone owns as many cars as they do now, but that car can drive itself much or all of the time. In that world, we still get increased congestion because the marginal cost in attention of driving is lower, but everything else stays the same. There are no changes in land use, no changes in household costs. Everyone just drives more and we have more congestion. That is the world that my co-authors and I deplore. We want to try and get us off of that track. That's why we wrote the book.

So how do we get there? Well, the book has a lot to say about that, but very briefly, we need better regulation now. We need to be moving forward quickly with the idea that this is a technology that could save lives. There should be a presumption of urgency. Of course, we don't want to sacrifice safety, but the bar to prove safety ahead of time should be lower than proving it after the fact. Use the insurance subrogation process; penalize companies that have unsafe products to the full extent that you can, but if they think their product is safe, make it easy for them to deploy.

Secondly, cities in particular should be doing everything they can to get curb management going. It would be good in the near term just because curbs are abused today, but the earlier we get that started, the easier it will be for robo-taxis to fit into that framework rather than trying to impose it retroactively.

And then finally, everything to do with safety should not be considered commercially sensitive. Right now, there are in some places barriers to companies sharing information about safe operation because it could be regarded as weakening their intellectual property. We need to make sure that there are no institutional barriers to that, that all safety best practices get shared as widely as possible as soon as possible. There are other things, and the book goes into detail for all of them, but for headline goals, it is better regulation faster and sooner, not just of vehicles, but of curbs.

Beatrice: Yeah, I think that about summarizes it. Thank you so much. I really recommend people, if they want to explore this more, to check out your book, The End of Driving. It was a really interesting read, and it's so obvious that we have the technology, but there's so much more to think about before we have it in society. The driverless car is definitely something that we should keep our eye on and treat with urgency, because regardless, it's going to take a lot of time. Do you have anything else that you want to add?

Andrew: Well, I would say that if you are interested in these issues, but the prospect of diving into a full book is a bit daunting, I write a twice-weekly newsletter on Substack called Changing Lanes. I delve into all of the ideas in the book in smaller, essay-length chunks there. So if you want to know more before committing to the book, the newsletter is a great entry point for that, as well as other things to do with innovative mobility. What does public transit look like in a driving automation world? What does land use look like in the suburbs? I get into all those things there, and I'd be delighted if your listeners wanted to look at the work that I've published at Changing Lanes.

Beatrice: Sounds perfect. I'll make sure to also link that, as well as the book, when we publish the episode so people can find it easily.

Andrew: That's very kind of you.

Beatrice: But yeah, Andrew, thank you. Thank you so much for taking the time.

Andrew: My pleasure.

‍