Russ Roberts

Benedict Evans on the Future of Cars

EconTalk Episode with Benedict Evans
Hosted by Russ Roberts
paper TOWELS or PAPER towels... Where has all the Parking Gone...

driverless%20cars.jpg Benedict Evans of Andreessen Horowitz talks with EconTalk host Russ Roberts about two important trends for the future of personal travel--the increasing number of electric cars and a world of autonomous vehicles. Evans talks about how these two trends are likely to continue and the implications for the economy, urban design, and how we live.

Size:30.9 MB
Right-click or Option-click, and select "Save Link/Target As MP3.

Readings and Links related to this podcast episode

Related Readings
This week's guest: This week's focus: Additional ideas and people mentioned in this podcast episode: A few more readings and background resources: A few more EconTalk podcast episodes:


Podcast Episode Highlights

Intro. [Recording date: July 27, 2017.]

Russ Roberts: Today we're going to be talking about the future of the car, based on a very provocative and lengthy blog post that you wrote on the rise of two things that appear to be transformative for that industry--which are the electric car and the driverless car. And what I loved about the post--it was a beautiful example of one extremely important aspect of what I call the economic way of thinking, and that I associate with George Stigler and Thomas Sowell. And that is: And then what? That is: Something gets put in motion. Something happens. Something changes. And a lot of people think, 'Well, that's the end of that.' And, what a good economist does, and what you did in this blog post, is start thinking about, 'What are going to be the implications for a much wider range of stuff?' In particular about the consequences of more electric cars or more driverless cars--what you call second- and third-order effects. So, I want to get started with electric cars. How might they change things?

Benedict Evans: Well, I think there's sort of, there's two sets of things to think about here. The first is that the electric car doesn't so much as get rid of the gas tank as kind of rip out the spine[?] of the car. So, it's not that you get rid of the gas tank and replace it with batteries. You get rid of the internal combustion engine and all of that's[?] associated systems. And you get rid of the transmission system and the gear box. Or most of the transmission system. So, you probably have between 5 and 10 times fewer moving parts. And, that obviously has an awful lot of consequences inside the car industry, which are kind of the first-order effects. It has fairly obvious effects on kind of the supply chain; and also on things like companies making machine tools--which is a big part of German industry. But then the [?]--

Russ Roberts: Companies--I'm sorry. Companies making what?

Benedict Evans: Machine tools.

Russ Roberts: Oh, machine tools. Sure, the work on the cars. Yeah.

Benedict Evans: Yeah, like the people who make old stuff--the machine tools that make all those moving parts inside the gear box have a problem. But then you sort of start [?] thinking, 'Well, what about things like gas stations?' So, there's 150,000-odd gas stations in the USA. Gas is sold at almost no margin. They make their money from everything else. [?] they would base it, you mean [?] salt, sugar, and nicotine, in kind of shiny plastic packaging. And some portion of that is an impulse purchase. And, if you are never going to a gas station again--basically you'll only go there if you want the salt, sugar, or nicotine--you won't [?] go to get gas any more. So what happens to sales of those? Something over half of sales tobacco in the United States, say, are sold in gas stations. Some portion of that is an impulse purchase--as [?] sort of suggests studies of who and what pricing changes and what availability in packaging changes due to tobacco consumption. So, um, I thought that was kind of an interesting consequence. There's another, um, perhaps [?] more directly related to cars, around repair. So, as far as I can make out, something around half of repair maintenance expenditure in the USA on the stuff[?] that's already related to the internal combustion engine--like the oil change and the transmission and everything else--the rest is like, you need tires or body work or the age-fact[?] rates, or something, so there's other stuff that won't be in sync[?]. But, again, you go--you have many fewer moving parts; you will have many fewer failures. You won't need an oil change because there's no oil. The radiator fan belt won't fail because there's no radiator. So, you get a radical simplification in the mechanics of the car; and that's what a lot the maintenance expenditure you go through [?]. And of course that is actually the economic support for a lot of the dealer network as well. Um, that's where they make their money. So, you've got these sort of rippling-out effects around the stuff that's sort of the support infrastructure around the gasoline car. Which will go away. You know, the adoption of electric cars is really a question of when rather than if. It's a function of battery pricing. And, battery pricing is kind of function of scale. So there's a circularity there, or virtuous circle. We are now at the point that we have expensive, un-economic electric cars. We will get to the point in the next 5 or 10 years that electric cars become cost-competitive with gasoline. And then it's[?] just the question of time, [?] or basically cycles out.

Russ Roberts: How confident are you that it's a 5-10 year process?

Benedict Evans: Well, so there's two processes here. So, there's: How long does it take to get to the point that, um, an ordinary, boring car is cheaper to buy as an electric car--it's cheaper to buy an ordinary, boring electric car than to buy an ordinary, boring gasoline car. So, that's how long--and that's a question of battery pricing, really: How long does it--and scale? Then: how long would it take before all new cars on the market are electric? How long does it take before all new cars on the market are electric? How long does it take before all the old cars cycle out of the system? And that kind of depends on public policy, because it depends on what kind of incentives you put into government [?] to do that. But, that feels like, you know, a 20-, 30-, 40- year process, depending on how aggressive you are, while going, you know, from the $50,000 electric car to the $10,000 or the $20,000 electric car; and how what you think the lifespan of existing vehicles is, and so on. So, it's not something--it's not likely when it will be done in 5-10 years. It's more likely it will get started in 5-10 years.


Russ Roberts: I guess--is there any issue--there are some things that are simpler about an electric car. Which means that there won't be the repairs that are necessary for an internal combustion engine. Are there some things that are more complicated in an electric car?

Benedict Evans: Well, there's more software. So--and then--and then by extension, there's more computing. But, that's kind of solid state. It's not moving parts. So, you know, there's a different cost structure to a--to a [?] car. So, if you look at--an UBS[?] today, a tear-down of a Chevy Volt, it's a kind of a normal gasoline car--and, you know, the propulsion part is a lot cheaper. The electric motor and the battery cost less than having an internal combustion, a complete internal combustion engine plus the cost of gasoline. But then you've got the--the electronics in there add a significant amount of costs. But that's kind of a transitional issue. You know, that will shrink down over time.


Russ Roberts: Your point about the margins of dealerships: I have heard that before, that they don't make much money on the cars. And they make increasingly less as information becomes more widely available and people shop more wisely. The Internet has, I think, really brought those margins down dramatically. And, so, the claim is, then, that they make their money on the repairs. Of course, if the repairs go away, it's possible that the price of cars will be higher: that those margins will have to adjust to make that worthwhile--if there are still places that people show up to buy cars at.

Benedict Evans: Yeah. I mean, it depends what the purchasing model looks like, doesn't it? I mean, I[?] will see Tesla is trying to go direct[?]. It's not clear that everyone will be able to go direct.

Russ Roberts: Right.

Benedict Evans: Now, of course, there's now there's another effect, which [?] what comes on to autonomy--which is if there are fewer accidents, there will be much less repair work, as well.

Russ Roberts: For sure.

Benedict Evans: So, and that doesn't need full autonomy. We can maybe go on to talk about, you know, different kinds of autonomy. But, you know, the holy grail is the car that doesn't have a steering wheel, or the truck that doesn't have a human cabin. But, you don't need to get that far before you start reducing that accident rate really significantly; and also, you don't have to have very many vehicles that never crash into anything before the accident rates are going down.

Russ Roberts: Yeah. We will get to that. I want to stick, for the moment, to electric cars. What do you see--and you speculate a little bit in the piece on this--what do you see as the model for how people are going to be charging the cars? Obviously, one of challenges, at least right now, is the charging phase that place over time. So, you are not going to show up at a gas station with an impulse purchase opportunity to gas up, to charge up your car. You are going to have to plug it in somewhere for a relatively longer period of time. Is that going to change, do you think? And, how might people charge their cars in a world where electric cars are more common?

Benedict Evans: There's no question there's a problem here, or a challenge, depending on whether, how you want to talk about it. And yes, you don't charge quickly. And we're not on a--there's not a kind of an imminent timeline to be able to kind of fully charge your vehicle in 5 minutes the way you do with, you know, with gasoline. So, there's a bunch of options. One is that you charge at work: You have charging stations in car parks; you have charging stations in supermarkets. How far you scale that is a challenge. This is actually, I think, this [?] becomes an urban density question, because if you live in--if you have somewhere where you have your own driveway or you have your own garage, then there's just--well, you've just got to pay somebody a couple of hundred dollars to put a charging point in your home. Where I used to live in London, I didn't own a car. If I had a car, I don't think I would ever have parked it in front of my front door. It would have been somewhere within 5 minutes' walk if I could find the space. And even if it was parked in front of my front door, it would be parked on the street, across the sidewalk. You know, 500 yards from a pub. Where, in my view, people come walking past. And I kind of, might kind of might collect drunks on the sidewalk, if I had a power cable strung out of my front door into my car. So, you know, yes: Real, practical question around what charging looks like. Related, frankly, to the real, practical question, the random walk[?], fueling looked like 100 years ago.


Russ Roberts: Yeah. I guess--I guess what we want is the same thing I have on a trip, when I use--I worry about my iPhone not lasting an entire day. I carry a small battery that I can recharge it with in the middle of the day if I need to. It's hard to imagine that will happen for a car.

Benedict Evans: Well, we do actually have that in a traditional[?transitional;?foundational?] sense. A moment. So, you can have cause, have a gasoline range extender. So, like the BMW I3--it does, as it might be, 100 miles; but then you can get an option to have a gasoline pack that will give it another 75 or something--I forget what the number is. So, there are all sorts of kind of transitional solutions to this. But, I mean--I just, I have in my mind--the kind of image is of people kind of doing kind of long-distance endurance rallies in automobiles before the First World War. And the photographs of these cars, they've all got bundles of like 3 dozen tires strapped to the side of, each side of the vehicle. Because, tires were not very good then.

Russ Roberts: Right.

Benedict Evans: So, these things get solved. That's the thing to solve. But it will get solved.

Russ Roberts: So, it could be, like, the Kindle. Instead of going for 300 miles, it will go for 3000 . And I'll only need to charge it once every 3 or 4 months.

Benedict Evans: Well, I think that would be--I don't think we are any place in battery technology that would do that unless you are thinking about [?] nuclear. But, I mean, there is also a psychological issue here. So, you also have to ask, 'Well, yes, your car could drive maybe 200-300 miles between being fueled.' But, how often do you actually need to drive 300 miles? You know: What is your actual average--even your average weekly consumption? And, you know, back to my example: If I did, most of living in Kendishtown[?] with a car, I would never drive--I might drive 300 miles once a year.

Russ Roberts: Right.

Benedict Evans: So, what happened with phones is that you went to having to charge once a week to having to charge once a day. And as long as you can charge overnight, you're okay. If it's more than overnight, then that becomes a problem.

Russ Roberts: But it's also your point about where that charging station would be. If it was once a month, you wouldn't mind driving across town to a charging station. And it would charge it in 10 minutes.

Benedict Evans: Yeah. I don't think we have anything in sight that would get us charging once a month. I'm more kind of thinking at the other end--that we [?] went from charging once week and to charging once a day, that actually that's turned out to be okay most of the time. And the same thing for cars. If it can't go for 300 miles, it can only go for 100 miles without being recharged, so actually it would be refueled, well, maybe that's actually okay for most people. You know, your gasoline car can go for 3 or 400 miles on a full tank; but actually don't do that, or very rarely do that. So, maybe you'd be okay with a vehicle that actually only does 100 miles on a charge. Because, actually, you'd never go more than a hundred miles.

Russ Roberts: Yeah--I'm thinking, though, about your urban density problem, now. If I don't have a driveway, the once-a-day problem might be a barrier.

Benedict Evans: It is. You kind of backed my point. So, imagine a--[?] once a day. If you have a collar[car?] that can only go a hundred miles on a charge, but actually you only drive 50 miles a week, then charging it once a week is kind of okay.

Russ Roberts: It's true. Good point.

Benedict Evans: And particularly, that's going to correlate with people in a high-density area. Where, if you live out in the boonies, yes, you might drive 50 or 100 miles a day; but you are going to have a driveway.

Russ Roberts: Yep. I like the point.

Benedict Evans: There's a bunch of variables.

Russ Roberts: Yeah.


Russ Roberts: What's the Elon Musk vision of the battery in my house that you allude to in that piece? What do you think he has in mind?

Benedict Evans: So, well, so there's two answers to that. One is, Elon Musk wants to do anything he can to get battery volumes up, in order to get battery prices down. And so, anything, any way he can think of to get people to buy more batteries, is good. And, that applies both to going from a low volume car to a high volume car, to talking about trucks potentially; but also trying to sell you another battery. And frankly, the solar[?] thing applies to that as well: Why buy a battery? Well, if you want to buy--having solar panels gives you an incentive to buy a battery in the same way. So, that's a large part of the story of your solar company. Um, so that's just kind of a scaling point. I think there's kind of a slightly more interesting conversation longer term when you think about what charging infrastructure looks like. So, you know, we talked about charging these cars. You have to think about, 'Well, what does that do to, like, power-generation requirements[?]?' and also the local power distribution grid?

Russ Roberts: Yeah. That's my favorite example. Because, I like to point out, Prius is really a coal-powered car--in some dimension, in many parts of the United States. Except for the fact that if you charge it at night, you can use excess capacity and you are not really increasing the amount of coal powered electricity generation. But, as more and more people adopt cars that are electric, that won't work.

Benedict Evans: Yeah. Well, it depends where you are. So, if you are in France, it's a nuclear-powered car.

Russ Roberts: ah-ah[?]

Benedict Evans: They tell you there's a question around excess--what the loading of the power-generation system looks like and when you are charging it--so, if you are charging it overnight, for example. That also is kind of comes to the point of the battery, because the battery can draw the excess power and then give it to your car when you need it. And indeed the battery can also store solar if you've got solar on your roof. So, the battery is kind of an intermediary buffer for that. Um, I mean, there are also kind of [?] power company questions around, 'Well, what does it actually mean if park the car in our neighborhood, if suddenly half the houses in this neighborhood are suddenly charging an electric pickup every night? What does that mean to the [?] substations we have?' And, 'What does the actual total load look like in that street all of a sudden?' So, that's a kind of transitional point. There's a bunch of analysis that's been done on, you know, what loading would look like, when people would using it, what the impact would be. And it kind of depends when and where you think people charge, and what time. Of course, that's kind of the crucial point. But also, what is the power infrastructure even look like in that country; and that is that point about France. Well, it depends. And that's kind of the renewable thing as well. If you've got a whole bunch of [?] in your power system, well that's not necessarily much good if people are charging overnight unless people have batteries.


Russ Roberts: Let's go back to this question of relentlessness of this process and the likelihood that we are going to move toward this world away from the internal combustion engine. Right now, electric cars are subsidized by the government. I'm against that, but that's neither here nor there.

Benedict Evans: Well, so are gasoline cars.

Russ Roberts: Yeah, well they are also--well, in which dimension?

Benedict Evans: And cars?

Russ Roberts: In which dimension are gasoline cars subsidized?

Benedict Evans: Well, it depends where you are. But, I don't want to dig into that. I think that's kind of a short term question.

Russ Roberts: Well, I mean--there are certain things--I always think it is funny when people say we should have a carbon tax. It might be a good idea. But we have a carbon tax. You could argue it's too low. We tax gasoline quite a bit. And you point out--I thought, also, cleverly in your piece that one of the impacts of this is going to reduce revenue for highway-building and other sorts of things that people use with gasoline taxes. But, the point is, is that right now there's a pretty heavy subsidy to the purchase of an electric car. If that went away, do you think that the technology improvements in battery pricing that we might see going forward are going to be sufficient to make an electric car with an internal combustion car in the next 5-10 years?

Benedict Evans: No question. You can argue about the timing but the cost track of lithium batteries is pretty clear. Quite when it reaches quite what price in order to be cost competitive with gasoline is still kind of a matter of opinion; there's a spread of estimates. But clearly having a subsidy on top of the purchase price of a car helps that. But, the underlying trend is the same.

Russ Roberts: So, that's going to be very interesting. If that really happens, the geopolitical impact is going to be quite dramatic. Large parts of the world--we've been mostly focused on, I think most of us in this conversation, you and me, the two of us--I've been thinking about the United States and the drivers and gas stations in the United States. But, of course, the nations that supply the oil to the world would have a very different future in a world where the internal combustion engine was not relevant. And one would expect, then, to play economist--one would expect that their behavior would change dramatically if they thought this was imminent. So, maybe they would--it doesn't seem to be changing. The price of gasoline is lower right now than it's been in a while. But, you'd expect it to be maybe lower still, because, just to take the most obvious example, Saudi Arabia is sitting on enormous reserves of crude oil, still. And, if it turns out that the internal combustion engine is not going to be a viable product in the marketplace, the value of that is going to be lower.

Benedict Evans: Yeah. So, there's a bunch of things to think about here. The portion of global oil production that goes towards is something like 40%, I believe, from memory. Now, removing that demand over a period of anything between 20 and 40 years, maybe longer, and of course removing it where--maybe a different rate in different countries--so there's a relatively--it's not going to happen overnight. And there are other uses for that production as well. So, it's not like there's going to be a guillotine. Also, to really play economist, of course if you were to look at Venezuela you could argue, 'Well, what good is oil today?'

Russ Roberts: Yeah, you can mess it up. You can have something valuable and not use it wisely.

Benedict Evans: Or, well, I mean there's a whole host of these problems--the curse of having assets, of having natural resources. So, you can certainly make and argument that [?] maybe it might do more harm to people selling luxury apartments in New York City than it might do to many of the people in those countries themselves.

Russ Roberts: That's a great point.

Benedict Evans: That's an economic argument--the development argument, is not the[?] only argument. I think there's another--on that point [?] it's worth looking at the other side. You'd think they are not so much Saudi[?] but if you look for example at Nigeria, the relevant petrochemical for a lot of people in Nigeria isn't gasoline: it's kerosene. For lighting. And, so, there, there's a solar[?] story. And so, the growth of solar and emerging markets as cheap, healthy energy--because you are sitting in a hearth[?] burning animal dung or burning kerosene--and that's how your kids are doing their homework--well, now you can have a solar panel. With an early day light[?delight?]. And so that's a way in which chemicals affect people's lives, and again to be changed by in fact some of the same technologies. There's a lot to do with cars.

Russ Roberts: Yeah.


Russ Roberts: Let's move on to driverless cars. Which, I have to confess I'm somewhat endlessly fascinated by the possibility of it and the implications. One of the things I really liked about your article is that, I think most people have misunderstood the full scope of the implications of driverless cars. They get the idea that there are going to be fewer accidents; they get the idea that there might be fewer people working as taxi drivers, Uber drivers, or truck drivers. And then they have some idea about, maybe that people won't be buying as many cars. Which, I think is much more complicated. But you spin out quite a few more interesting implications than those simple ones. And some of those implications reminded me of some ideas that I think people misunderstand--which we'll get to--not just don't notice but I think get it wrong. But, I want to start, in talking about driverless cars, with the different levels of autonomy that you alluded to earlier and that you mention in the piece.

Benedict Evans: So, that's a good way to frame this. So, the industry currently talks about 5 levels of autonomy. So, Level 1 to Level 5, or L-1 to L-5. Level 1 is the mechanical cruise control that came in in the late 1950s and early 1960s: so you put a switch on a stick, and the car will go at 69 miles an hour right into the truck in front of you. So, it's just a purely mechanical thing. Level 2 is that you add a little bit of radar to that, and maybe, or a [?] sensor to that. And so you will slow down if the vehicle in front slows down, and you will maybe get a warning if you start straying out of your lane. Purely based on looking at [?] what drives on the road. This is also based on the mechanistic as being done in software. But there's no intelligence to it. This is what you get if you buy a high-end German car or if you buy a Tesla. And, of course, Tesla, until very recently was buying exactly the same technology off the shelf, and an Israeli company called Mobileye that BMW (Bayerische Motoren Werke) and Mercedes were buying. And a Tesla which is calling[?] autopilot, which is the same thing. Um, and yeah--that will drive straight into the back of the truck in front of you slightly less often, but it's slightly less intelligent. That's Level 2. Level 3, it actually has some sense of its surroundings. It can look around itself. It has some 360 awareness although of a basic kind. You can give it a direction and it will take you there. But, you need to be sitting in the driver's seat with your hands about an inch away from the steering wheel at all times, because at any point it might get something wrong, or it might just stop and said, 'What to do now?' That will fall--you can read a book. It might stop and don't know what to do. But it almost certainly won't. Level 5 is just a question of Level 4 but with more lines of liability. So, Level 5 is the point where you are confident that you can take the steering wheel out of the vehicle, and that you can potentially drive--for example, you can design a commercial vehicle without a human cabinet at all. Now, I think the interesting thing, as you move along that progression, is that a Level 2, Level 3, is basically a safer car but it's still a car. Level 4, Level 5, you use the term 'self-driving cars.' I prefer the term 'autonomous.' And the reason I don't like the term 'self-driving car' is that's very like saying 'horseless carriage.' That, you remove the horse from the carriage--and if you look at early automobiles, early vehicles from the early 20th century, they've taken the horse off, but it's still a carriage.

Russ Roberts: Yup.

Benedict Evans: And in the same way, you remove the person, but it's still a car. Well, that's not how it works. And, you know, removing the gas tank just as electric isn't about removing the gas tank, autonomy isn't about actually about the car driving itself. It's about the you getting rid of the person. And it's about changing everything else about that vehicle. And everything about the city around it. In much the same way that removing the horse wasn't just about removing the--it changed everything else about vehicles and everything else about it[?]


Russ Roberts: So, let's start with the question of--you want to speculate, just for fun, how far away Level 4 and 5 might be?

Benedict Evans: Um, so, opinion varies on this. The most optimistic people will say we'll have sort-of Level4/Level5 in 5 years. The consensus probably edges more towards 10 years. The very ability within--and that's for the first vehicle. Then, just as for electric, then you have a whole transition question of: Well, how long does it take for the first one to all vehicles being sold, new vehicles? And then, how long does it take till all vehicles in an area, or in a city? And, do you have, like, segregated lanes, or park-and-ride? And what incentives do you have? And so on. But, 5-10 years for the first one. The very ability in that is really about, like, the last couple of percent of difficulty. So, if you had a city where there, if you have a like a, if you have a place where there are no pedestrians and no human drivers, an automatic car is really easy. The hard part is accounting for what other people are going to do, accounting for pedestrians, accounting for the child that might run out from behind the parked car.

Russ Roberts: Yeah.

Benedict Evans: And so, like, we may well have like Level 4 on highways like soon. Having Level 5 in Naples [Italy] might be a bit more difficult.

Russ Roberts: Right.

Benedict Evans: Like, what exactly is that hand-gesture telling you to go do? So, there may be that, you know, there are situations and use-cases where it takes a lot longer for this really to work properly.


Russ Roberts: Is it--do you think it's feasible? And one of the glorious prospects for this is, um, zero accidents. Obviously, most, many accidents are caused by human failure right now. There's going to be--

Benedict Evans: It's over 90%.

Russ Roberts: And, of course, an autonomous car is going to make a mistake now and then. It already has. And tragically it's killed some people.

Benedict Evans: Well, that was an autonomous car.

Russ Roberts: Oh, that was a--

Benedict Evans: That was a Tesla.

Russ Roberts: That was a 2.

Benedict Evans: Yes. That was a Level 2.

Russ Roberts: So, when we go to 4 or 5, is the only accident going to be the deer darting in the road that you mention in the piece? Or, will there be the potential for--in other words, here's the thing. When autonomous cars first got proposed--and I see them driving around my neighborhood: I'm in Palo Alto for the summer, so I literally see them in front of my house, cruising around--you know, it's kind of amazing. It's like a dog on his hind legs. Amazing--it's not at all. Is it really feasible that it will be accident-free?

Benedict Evans: Well, it depends on what you mean by 'accident-free.' Um, yes, there will be the deer that runs out, and there will be the tree that falls down. There will be the acts of God, so to speak, where it is not an error. Will there be errors? Realistically, perhaps. Will they kill people or will they be something else, where the car just kind of sucks? You know, the error I can expect to see more of is like 2 cars at a junction and each of them keeps waiting for the other one to go, and neither of them go and they stay there all day.

Russ Roberts: Yeah, it's a Buridan's Ass problem, for you philosophers out there. But, go ahead.

Benedict Evans: Yeah. So, I think that feels more probable to me than the vehicle that just kills, that's just killing people. Although you can't really talk in absolutes here. On the other hand, 35,000 Americans were killed in road accidents last year; and we just kind of shrug that off as part of life. So, there's also a kind of degree of psychological acceptance and understanding I think. And the number of people--you know, how many people are killed by tobacco in the United States every year? And yet we don't ban tobacco. Car accidents killed 35,000 people in the USA last year. Tobacco killed half a million. Tobacco is legal. So, there's a real question of psychology and consumer consciousness and consumer--how people think about this stuff, [?] that. But, yeah: I don't think--there's no, like--let's put it another way. There's certainly no theoretical barrier to getting to Level 4. Which is the accident-free place. The questions around the Level 5--which is the--okay, now you can have a vehicle that doesn't have a human driver. And that's a part that all the economics change, and the cities change.


Russ Roberts: Yeah. And it's more fun to talk about. So, we're going to talk about Level 5. But before we do, one more question: What's the hacking risk here? If we move to a Level 5--I'll make the segue here. If we move to the Level 5, where the cars are coordinated in a way--so we could think about cars going 100, 120 miles an hour or faster, in a chain, 2 feet apart because there's no risk of coordination among them that cause traffic or death. What's going to be the potential--and I don't know how smart the grid is going to have to be, versus the car itself. Because, there's certainly going to be some gains from going from it--cars being smart to streets being smart. And if we go to that level, what's the hacking potential?

Benedict Evans: Well, so, first of all, I don't think there is a consensus that you would have a smart street, so to speak, as opposed to small cars. There are kind of two levels to this. So, one level is, in an on-demand world, you would absolutely have coordination of placement of vehicles around cities: so as to optimize the efficiency of traffic. So, there was always a vehicle, there was just the right number of vehicles to pick everybody up, and no more in any given area. And, you know, optimized routing of those vehicles. And you may well have optimized routing, optimization of the routes the vehicles take around the city. So, it's to kind of make sure you don't have one city where every vehicle decides--where every vehicle decides to go down the same road. So, at the moment, you drive around using vehicle maps. You can clearly see--you know, this 5-element[?]--you are going down this kind of random suburban side-street and you can see the [?sci-fi?] level cars who are following the same Google routes. So, you expect to see coordination of that. You expect to see coordination of cars within the city.

Russ Roberts: Yep.

Benedict Evans: Whether you would have coordination of vehicles on a one-to-one basis talking to each other to make sure they don't hit each other--I think that's a whole lot of question. I think a lot of people would say, actually, 'No, you won't.' Because, you know--they may--it would be much more like the relation of you driving yourself versus Google Map. So, the car might be being told what votes go down. But it may be entirely autonomous as to how it goes down those roads and how it stops and accelerates. Opinions vary on that. So, some people think you would have everything kind of managed, done, centrally. But I think most people think it would be a car itself that would be making that decision. Now, clearly there's a threat of the car itself, because itself being hacked. You know, they are [?] network-connected computers designed by human beings. And, you know, every time you make something idiot-proof, God creates a better idiot. So, if you've got one, you can't guarantee that. But, at the same sense, I don't think one can just say we won't do anything, just in case something goes wrong.

Russ Roberts: Yup. I agree. I just wonder--you know, I think the--there is a temptation, I think among technology people, to just sort of wave their hands and say, 'Well, this problem will get solved.' Because it's always been solved; and we're just smart and we'll think about it, and we'll work it out and we'll fix it. And then there's the similar, anti-technology people who say, 'Well, then, a bunch of evil people are going to mastermind all the course crashing into each other at once.' And, neither of those is quite capturing what actually goes on in human innovation. I'm just curious about how worrisome you think that latter problem is. I know it's not the movie version, where an evil genius whispers the wrong word, and all the curs[?] hear it through their censors; and it ruins their computer systems, and they all drive off cliffs. But is there some in-between scenario that is a little bit frightening to you?

Benedict Evans: Um, so, there's a whole lot of conversation around how, like how the sort of computer security is threatened by [?]--that it's no longer, you know, a teenager walks into your building with a [?] on a floppy disk, or, you know, somebody hacks in from outside. It's now 300 people in a building in China who know what school your kid goes to and send you or your executive assistant an email that looks like he comes from the head of--it looks like an invoice from the school or something. So, there's a--you know, that's a kind of other question. So, yes, it's a concern. It's kind of like, it's a problem to solve once we've worked out if we can ever make this work at all, though, I think.

Russ Roberts: Could you say that again?

Benedict Evans: Well, it's a problem that we'll have to solve once we've made a car that can drive itself. If you can't make a car that can drive itself, anyway--

Russ Roberts: it's moot--

Benedict Evans: we'll put the [?] on afterwards.


Russ Roberts: So, one of the things that people forget about, is your point about the horseless carriage. There's going to be a very different design of the physical car. It's not just going to not going to have a front seat. So, talk about some of the changes. Or everything will be the front seat. But there will be no place for a "driver." It might look like, I don't know, a circular lounge, or a table with 10 seats around it. Who knows what it will look like? It'll look more like a teacup at Disneyworld floating around. What are some of the changes? You point out an interesting and not-so-obvious fact that cars are heavy--for safety reasons. So, what else might change?

Benedict Evans: So, there's a bunch of kind of basic assumptions that change. The first one is that the vehicle is going to be in collisions. And, obviously, again, there's a transitional period where you still have human-driven cars around; and then there's a period where everything is totally automatic and even more stuff changes. But, in a fully automatic world there are no collisions; therefore there are no safety cages, no crumple zones, no airbags. Or maybe much reduced; or you design those with completely different thinking in mind. And so that obviously changes the weight, and the cost. It can also change the physical design that the vehicle might look like--I mean, the design that we have now where to kind of have a sloped front and a slope to the back and trunk and so on--the durables[?] can be in different places. You know, the windows can be in different places. Also, you don't have a loads of[?] mechanical lump[?] of an internal combustion engine there, which you can get rid of. That's part of the point the point of looking at a Tesla is you wouldn't actually know if it didn't have a gasoline engine inside it if you didn't know in advance. You can design, and with both electric and a 0-accident world, you can design vehicles that look quite differently. Secondly, think about, if you think about on-demand as well, is that you could design vehicles that would only go at 20 or 30 miles per hour. So, today, vehicle you design has to be able to go on the freeway because you might need to go on the freeway. But, in an on-demand world, the system would know where you were going. And, if you weren't going on a freeway, it wouldn't necessarily have to send a vehicle that was going to do that. So, you really could have, you know, "pods" that aren't particularly streamlined because they are not going to get over 25 miles an hour.

Russ Roberts: So, you are talking now--when you say, 'on demand,' you mean, like, an Uber-type service without a driver.

Benedict Evans: Yes.

Russ Roberts: So, you are thinking about--let's think about what this would be. So, I'm going to go to the grocery. The grocery is about 3 miles away, so I don't want to walk.

Benedict Evans: Presuming that you'll ever go to the grocery yourself again, of course.

Russ Roberts: Well, that's true. Of course not.

Benedict Evans: That's a separate question.

Russ Roberts: Good point. I forgot about that. I'm just in a wistful--

Benedict Evans: [?] but that makes another point: which is, how does this change thinking about delivery? Particularly, again, if you are in the suburbs, you could very easily imagine a vehicle that kind of came and dropped off a canister at your front door, when you were there, that didn't even necessarily have a human being in the vehicle.

Russ Roberts: Correct. Yeah. It would pull up and it would be like an Amazon Locker: I just go in--

Benedict Evans: Yeah, exactly.

Russ Roberts: and take it out. Or--I want to be there, though. That's a huge problem. It needs to eject it onto my front lawn, or it would probably need to stick its arm into a box that I had on my front lawn.

Benedict Evans: Yeah. There's a bunch of other stuff that would need to change before you could go to, like, before you could do all of that, those kind of things. But you know, test the assumption that you are going to take yourself to the grocery store, because of course that, the driving is a cost of delivery in the other direction.

Russ Roberts: So, let's take--

Benedict Evans: I mean, I was going to say, the embedded point within this, of course, because it isn't clear, is that if you remove the human driver from a vehicle, you take out at least 3/4 of the cost.

Russ Roberts: Yeah. That's huge.

Benedict Evans: So, an on-demand ride that costs $10 bucks today would cost $2 or $3 bucks. And if you remove the insurance--if you are in a fully autonomous world with no accidents--then the insurance goes down as well. So it goes from $10 to $2. And so your calculation about whether you own a vehicle, or whether you own one vehicle or two, is going to change a great deal. And your calculation about where you might come in--going to the grocery store-- [?] to me that's like [?] to me that's like a less-interesting example. The example I really like is: You are going to go out, you are thinking about going out to dinner, in central, in Manhattan or central London, on a night in November. And it's cold and dark and raining. And you live in the suburbs. So, you can walk 10 minutes to the train station, and then get a train for 20 minutes, and get the subway, and you'll arrive at this restaurant an hour later. You could call for an existing car, like a taxi--and that will cost you $20 each way. You could drive yourself; and then you are going to spend 25 minutes looking for parking and pay for parking; and of course, one of you can't drink because you've got to drive back. And then you've got to walk back to where you parked the car. Well, now, an on-demand ride will get you there and back for $3 each way. Or, $4 each way. So, the whole way that you think about a city changes.


Russ Roberts: We got started on this because you were talking, making an observation, which was very interesting, that's not in your piece, about the fact that a car's design would be different if it never had to go above 30 miles an hour. And in that case, though, I'd probably want to go a little faster.

Benedict Evans: Yeah. In this case, you might be going fast. Yes. Or, the car might have a bar in it. Of course, there will be no such thing as drunk driving.

Russ Roberts: Right. No, it would have a bar, or a vending machine, or a--who knows what--a TV.

Benedict Evans: It will have a mini-bar, and that will be the expensive part of your trip.

Russ Roberts: Yeah. You know, I'm thinking more of a surround screen of, say, the whole front half of the car, which would wrap around you so that you could be immersed in YouTube for the 20-minute ride rather than talking to anybody or thinking independent thought. It's a little bit of a frightening world. But, I think it's coming. Right now, we drive, and our mind wanders. Which is an amazing thing. You can drive pretty well when your mind's not looking at the street, really. Your brain is doing about a thousand other things. Now you are going to be able to decide what your brain does in those times. You are not going to be just on your cellphone. Because, I assume driverless, autonomous cars--without drivers, autonomous, on-demand cars--are going to have entertainment options. I assume there will be some competition; and some that will have a person.

Benedict Evans: So, I actually disagree with that.

Russ Roberts: Okay. Fire away.

Benedict Evans: So, certainly, I've seen, like, people in the phone[?] industry sort of presuming that they are going to be able to sell Netflix subscriptions. My presumption is that it's your phone. And maybe the screen in the card connects to your phone. And so, your phone sends video to the car[?]. So, it's AirPlay. It's AirPlay or Chrome, or music. So the car will have a screen and speakers, yes. But I don't think that that will be stand-alone. I think that will be part of your broader account. Like, maybe you'll log in to the phone. Maybe it will be this Apple Car--you know, you'll log in to it, and it will have the same stuff that you have a new iPhone, or maybe it's just a dumb luck. You know, what's going to happen with TV, is TV is going to be a dumb screen; and that speakers will be on speakers and they'll just accept stuff on the phone. I think that's probably more the model.


Russ Roberts: So, in this world--let me try a different version. I know you've written some interesting things about virtual reality and augmented reality. So, suppose, instead of a car, I'm on, like, a Segway, maybe? Some kind of--little, tiny--instead of being like a limo, I'm in like a platform, just a rolling platform that goes at various speeds. And, I'm using my augmented virtual reality phone/headset/embeddedship to entertain myself. The car itself isn't going to have much of anything.

Benedict Evans: Yeah. Absolutely. And of course that would also be the case if it were a shared vehicle. And so this is, again, this is another tangent, another interesting case to spiral off in. Right now, you've got a kind of binary distinction. Set trains aside as a separate conversation. But you have a bus, and then you have a car, which is highly your own; or you have a taxi. And, as opposed to a bus--which is this big, heavy thing that carries 50 people. And, um, a 4-seat vehicle that is completely automatically controlled--so, [?] back--so, I'm living, I'm sort of speaking to you from Silicon Valley, you are in [?] Palo Alto as well. There's whole parts of the world where by density of people it means there are a lot of cars around, but the density also means that you can't support a bus network that runs on a fixed route and a fixed schedule, stopping at fixed points with a vehicle that's a certain size and that has a human driver. Now, one of the things that autonomy does is it sort of breaks down that binary distinction. So, if I step out on the street and I press the Palo Alto Public Transit App button on my phone and say, 'I need a ride,' and a pod that's two blocks away and has two people in it already, and I get in, and I take a third seat; and then it goes off and drops someone else [?] and then drops me where I want to go and picks up somebody else on the way: Well, is that a bus or a cab? I don't think that's a meaningful question. It's something else. But is it public transit? Is it a cab? Well, it's changing what that conversation[?] looks like. You are kind of unbundling the paths[?], but you are also in a sense re-aggregating taxis. And so you are changing how you might think about how you might move around a neighborhood like that.

Russ Roberts: Yeah--

Benedict Evans: Because you can--you can [?] reach it dynamically. Because you don't have the cost of a driver.

Russ Roberts: Do you think there will be drivers? Not drivers, but co-passengers for socializing? Do you think someone will want to ride with someone who is not driving, just wants to chat 'em up?

Benedict Evans: Well, it's just [?] the liftline[?] is the new tender. So, yeah, maybe. Things like, you, your selections might be a bit limited. A little bit too random.

Russ Roberts: Yeah, that's true.

Benedict Evans: Depends, you know, depends what part of the country you are in, whether that's going to work or not, I think


Russ Roberts: So, you mentioned that there's not going to be any drunk driving. There might be a bar in the car. You certainly don't have to have a designated driver. Have we seen, right now, in places where Uber is fairly friendly and doing well: Do we notice that there is more drinking going on? I feel like there is. But there should be more drinking. Is Uber big enough--and Lyft--ubiquitous enough, that we can observe fewer accidents and more drinking?

Benedict Evans: So, short answer, Yes. I don't have it immediately to mind, but there's definitely been studies showing declines in drink-driving. I think--and actually, I think the reverse in Austin when they shot down Uber and Lyft.

Russ Roberts: Well, I know there's going to be more drunk--there's going to be less drunk driving; I wonder if there's going to be more drinking overall? In other words, you and I are going to go out, and one of us is going to have to not drink--so maybe not have so much fun, maybe. Is it now the case that, since we can both drink, that we go more often?

Benedict Evans: Um, I think that's one of those like [?] consequences it's really hard to predict. I mean, certainly something that I experienced--you know, I was at university just before mobile phones happened. And they kind of happened as I was in my late, as I was in my kind of, as I was in my early 20s in London. And there was a fundamental change--suddenly people started making [?], because before mobile phones if you were going to meet you had to agree at like lunchtime where you were going to meet. And they were all going to meet here at this time. And if you didn't get there, then that was it: You didn't see your friends that night. And then mobile phones happened; and suddenly people don't make plans; and stuff just kind of spontaneously organizes, as you drift through, you know, 10 blocks of the city. So, I don't think anyone was sitting in 1995 and saying, 'Gee, mobile phones will mean that people will go to different, will go to 5 bars instead of 1 bar,' which is kind of what happened. I don't think you can predict those kinds of changes.

Russ Roberts: Yeah, it's fascinating, really. I didn't think about that planning thing. Although I still think--it's interesting how social norms have to evolve at the same time. We live in a world where everybody has their phone with them now. And most of us don't make phone calls. It's a remarkable thing how little you use your phone to call somebody. You use it to text; you use it to email; you use it to whatever else--whatever social media you use to interact with people, especially younger people. But I find it fascinating that this device we call a 'phone' is so rarely used for calling.

Benedict Evans: Well, how often do you dial it?

Russ Roberts: I don't think I do it more than--

Benedict Evans: You haven't actually used a telephone dial in probably 20 years.

Russ Roberts: Oh, for sure, but I was thinking you--

Benedict Evans: And you haven't hung up a phone in probably 40 years.

Russ Roberts: For sure--

Benedict Evans: These terms linger on long after we've forgotten what they actually mean.

Russ Roberts: But the only phone calls that I make on my cellphone--90% of them or maybe higher are people who live out of town: my parents, my siblings, my friends back in Washington, D.C. when I'm away for the summer, I might call them. But, for the people I'm living around, I don't call them very often. Mainly text, and say, 'See you soon,' or 'Can you meet me here?'

Benedict Evans: I wrote another blog post looking at sort of the ways that we make, like, the wrong predictions about the future. And one of the things that I put in this was a report from a firm called [?], which is a telecom consulting firm in 1990. And I'll just read you the opening paragraph. This is 1990, so we are just starting to get, like ISDN [Integrated Services Digital Network], like digital connections. Very few people have a PC [personal computer].

The telephone is going through a metamorphosis, that black Bakelite crystal is becoming an electronic butterfly, but what kind of butterfly? Could it be that the Fax has given us a glimpse of what lies ahead? [?] the world there may be a world where millions of people [?] workstation displays and electronic documents through cross-border telecom circuits.
Circuits. And so this then-report then goes on to talk about how the government treaties that regulate how phone companies pay [?] for cross-border traffic will have to be reworked, because will be making many, many, many more international circuits through phone calls to exchange information with each other. Maybe color faxes; or faxes that don't print out--that just have screens. But you'll still be like[?] actually dialing an actual telephone number to somebody in Japan to see what's in their computer. And so, you can--this is all kind of the problem about predictions about the future: you make linear predictions. So, you look at the thing that's going to happen now and you extrapolate it into the future without realizing that the character of it is going to change into something else.


Russ Roberts: So, let's talk about a few more examples of that, in the driver/autonomous car world, because they are some of the more exciting ones. One of the things --there are two things that I think about. And one of them I didn't think about correctly; and your piece reminded me to get it right. Which is, if you don't buy a car, as a consumer, and you rely on on-demand autonomous cars, then you don't need a garage. Which is good, because in Palo Alto nobody puts a car in a garage. They use it as storage. But, you don't need a garage for storing your car. You don't need a driveway. Roads don't have to be as wide, because you don't have to worry as much about human error--going outside the lane. And so, cities could be very different. And in particular, one thing you focused on is parking. Parking becomes a very different thing. But one thing you didn't mention much--I mean, you alluded to it earlier in our conversation--is where the driverless cars, the autonomous cars, are going to be hanging out while I'm eating dinner, wandering to go see a friend or whatever it is. And I think in most people's minds, they are just wandering around, like Uber drivers do now. But, they probably--

Benedict Evans: They'll be plotting.

Russ Roberts: Yeah. They wouldn't be wandering around. That would be not the right thing that's probably going to happen. The other thing that I just want to mention, because I think I've gotten this wrong before--and this is the one thing I've thought about I think that's right now--is, a lot of people I think misunderstand thing that, right now, your car is sitting in your driveway doing nothing. And with autonomous, on-demand cars, it's going to be in use all the time. But that just means it's going to wear out sooner. It's really your driveway that's wasted, not the car when it's sitting there. When the car is sitting there, it's in the elements. It can get rained on. It can rust a little bit, and other things. It's not always good for it to not be used for a while. But the real effectiveness of this issue is going to be on the roads and the driveways, not so much the cars. They are going to need more maintenance. Even if they are electric, they are going to need more--their tires are going to wear out sooner if they are used 24/7 or close to 24/7. And, where are they going to hang out? That's the other thing I'm thinking about.

Benedict Evans: So, I think there's--well, there's--so, there's a bunch of things to think about here. One of them is, like, certainly, even if it's your own vehicle that isn't on-demand, it doesn't need to wait[?] distance. So, particularly in city centers, you can re-think most obviously on-street parking. And so if you look at like any New York, or any kind of European city, the center of the city, you've got cars parked down most sides of the road, and that like halves the width of the road. So, you don't need on-street parking. Crime[?prime?] and sort of building parking space in new buildings, in high-value real-estate areas--again, you don't need to give up 6 floors of Park Avenue for parking. Something like 25% of the surface area of LA County [Los Angeles County] is parking. So, there's a real estate question here, first of all: that the parking can move. There is a issue[?] that a real [?] your car isn't parked but it goes away and then it comes back, so that's more traffic. But, on the other hand, you don't spend 25 minutes driving around and around looking for a parking space, so you have less traffic. And the roads are twice as wide because you don't have people parked on both sides of the road. So, you know, you get pros and cons here.

Russ Roberts: Yeah. Sidewalks are going to be a lot more interesting, because they are going to be a lot wider, potentially.

Benedict Evans: Yeah, they could be. Yes. Or there could actually be an extra lane, depending on where you are. There's a utilization question is kind of interesting, which is, like, it feels like, to me it feels a little bit crude to say, like, your car is used 93% of the time, because, you know, 8 hours of that is in the middle of the night when no one else is going to want the car, either, then.

Russ Roberts: For sure.

Benedict Evans: And, there's a whole bunch of other--like, how many people actually want to be driving around at 3 o'clock in the afternoon on the Thursday? So, I think it's a bit problematic to look at the total because, actually most cars are used--yes, they are in use 7% of the day, but it's all the same 7%. And actually unless people stop working or work completely different times, that's not going to change.

Russ Roberts: Yup.

Benedict Evans: So, I think it's a little bit tricky to talk about utilization rates overall. Then, of course, you see this now--Uber and Lyft, they actually have to get more drivers, because actually there are actually lots of unused cars driving around that are available. I think the--there's a kind of--the broader kind of parking question around, like--okay, you see your house has a parking space and a garage. Okay: you won't fill the house with a garage. In fact, no houses in the United Kingdom have been built with garages, because of course the weather doesn't require it. There is a period when you actually need to park your car indoors in the British weather; and now you don't. American weather, you still need to put your car indoors to get [?]. Um, but, where they go, it depends. The point is they don't have to go within walking distance of where you are now. So, you could optimize it in ways that you can't optimize it now. The answer might be, might just stay in place, sometimes. You know, Walmart, presuming there is still a Walmart, or Walmart out in the suburbs, if people are still going there themselves, well the car is just going to stop in the parking lot. That's probably the best place for it. But in central Manhattan, or, you know, the expensive part of LA or something, they might have a very different calculation. So, you know, it's complicated.


Russ Roberts: I guess the interesting question is: How many do you need? And I think you need quite a bit fewer--except for the fact that delivery and the fact that the price of an on-demand ride might be very low, as you suggested earlier. You might see people going out a lot more, because it's going to be so cheap.

Benedict Evans: You might. Again, you might go out a lot more, on demand, you might have one car instead of two. You might have a car for the weekends but on a weekday you might previously have driven to work, but now you go to work on a bus, but because a bus takes half as long because there's no congestion. Or, maybe, you previously, or maybe you are in, like I mentioned earlier, you are in a 4-person bus. Like, if you live in Palo Alto, previously you had to drive to work, because there was no other way to get to work. Well, now you might summon on-demand ride that might be carrying three other people. Maybe that's 4 cars that have become one car. So, there will be an awful lot of different variables in quite what this does to traffic, quite what it does to vehicle utilization, quite what it does to real estate.

Russ Roberts: I just want to say, we are recording this in 2017; and it of course is immortal. In 2027 and 2037 and 2047, people will of course be able to listen to old EconTalk episodes, maybe just by saying, 'Okay, Google: Benedict Evans, EconTalk 2017,' and then they are going to look back on this and say, 'What a genius he was!' Or, are they going to say, 'How clueless they were back in 2017! They didn't have any idea.'

Benedict Evans: It won't even be that, because it will be, 'Find me a podcast with somebody with a British accent in 2017 talking about autonomous cars and getting it wrong,' because by 2040, you'll be able to do that.

Russ Roberts: That's true.

Benedict Evans: Maybe not to that extent. But you'll certainly be able to say that, you know, this is a whole lot of conversation around machine learning. But, audio and video will be indexed in the way that text is indexed now, as a result of machine learning. And it's machine learning that is also enabling autonomous cars. So, you will be able to say, 'Show me a cool car chase on YouTube,' and it won't be because somebody typed in 'This is a cool car chase.' It will be because YouTube has indexed the video and it knows it's a car chase.

Russ Roberts: As a podcaster, I'm longing for that. It's, right now, one of the few frustrations about podcasting is the fact that they are hard to search. We have a semi-transcript--we call it the Highlights--that allows people to go back and find old points and old episodes. But I can't tell you how many times listeners have said to me, 'I can't find your interview with so-and-so.' And I always say, 'Well, that's because I've never interviewed him.'

Benedict Evans: To the point--just expanding a little bit on what I was just saying: of course, there won't be any car chases[?].

Russ Roberts: Oh. True.

Benedict Evans: Think about how many [?] from the past don't work now because of mobile phones. Well, there won't be any car chases. So, you couldn't make Heat again, for example.

Russ Roberts: Or The French Connection. The best--

Benedict Evans: The French Connection. All sorts of, you know--the world changes. There was a time when a movie plot point was a car breakdown. That--cars don't break down. There was a time when the movie plot was you couldn't reach somebody. Well, you can always reach people. So, movies will change.

Russ Roberts: For sure. The question is, there are people going to enjoy, just like we like looking at period dramas about, set around the Middle Ages, or British royalty, Henry the VIII--is it going to be a nostalgic thing to watch a movie with a car chase? Probably. I would think so.

Benedict Evans: Well, maybe it will be, well, yeah, there's a joke that every British actor needs to know how to ride a horse, and every American actor needs to look like he can shoot. So, maybe there will, in the future, be an actual need to look like, to pretend to look like know how to drive a car. No one will actually need to know, but there will be all these actors who will kind of be waving the steering wheel as though they know what they are doing.

Russ Roberts: Yeah. It's like smoking in a movie, right?

Benedict Evans: Exactly.

Russ Roberts: So often, a character takes out a cigarette. And I'm thinking, 'Why do that put that? Oh, it's supposed to show that he or she is cool, or has this savoir faire about them.' But, yeah, it will be the same thing. It used to be--of course, it used to be you drive a stick shift. And that would show that you were a person of the world. Now, maybe it will just be that you know how to start the car.

Benedict Evans: Well, yes. It's similar to that scene in Star Trek, maybe, where the engineer is shown a Mac and he talks to it; and they say, 'No, you have to use a mouse.' And he picks up a mouse and speaks into the mouse. So, it will be the same thing: People get into a car and stare at the steering wheel and just say, 'Hey, car, take me to work.' And the car doesn't do anything.

Russ Roberts: Yeah.


Russ Roberts: Well, we're almost out of time. I want to close with an observation that you close your piece with, which is another thing I hadn't thought about. And, we'll of course put a link up to the piece. It's really a thoughtful exploration of these issues. Which is the fact that, these driverless cars or autonomous cars are driving around with cameras. Twenty-four/seven. And so we're going to have a video footage of the world, the urban world for sure, available to law enforcement, to the government, to the NSA [National Security Agency], to--you name it. And we think about--I remember the tragedy of the Boston Marathon bombing--how much footage was available for people to paw through and pour over, that was publicly available. And a lot of really clever speculation was done. And it turned out to be totally wrong. I thought the wisdom of crowds was going to solve the problem because people had stitched together so many Flickr photographs and so many public-available ones. But the law-enforcement people had, of course, some access to other stuff that the public didn't. And it was very powerful; and I'm glad they caught the people; and they appear to have caught the right people. But, there's going to be some serious privacy issues with driverless cars because they are running video all the time.

Benedict Evans: Yeah. So, every autonomous car is capturing high-definition, 360-degrees, 3-D video all the time. Now, whether it keeps that, and where it puts it is a slightly different question. It's actually quite problematic to store all of that data from every car--like, there's actually a question of where you'd physically put it: Do we actually have enough storage to store all of it, from every car, would be, might be possibly [?] for a while. But, yes: You know, they, somebody's been killed and the cops say--they don't just pull the footage of all the cars in the neighborhood. They say, 'Did any car see anything strange?'

Russ Roberts: Yeah. 'Round up the usual suspects' will be 'Round up the cars,' first. To try--

Benedict Evans: Well, yeah, but it's not just 'Get me the raw video and I will watch it.' It's, 'Did any car see anything unusual?'

Russ Roberts: Right.

Benedict Evans: So, um, I mean, I sort of--you know, [?]. But, there was a murder in Britain sort of 15 years ago, where the police got the CCTV[?Closed-circuit television?] from the buses that were driving down a road nearby. And the bar[?] camera is just on the inside of the bus. But you just see out of one window, like, first bus, second bus, third bus--the second bus is a white van by the side of the road. And in the first and third bus, it's not there. So, they, you know you are looking for a white van. And that was kind of the break in the case. Well, you know, those buses will have 360-degree video, now. It won't be the white van. It will be the license plate and the model and the guy standing next to it.

Russ Roberts: Who will be identified immediately by--

Benedict Evans: Yeah. And: Were there any sex offenders in the area? Because you'll have that; you'll have all of that, all of the images of all the faces.

Russ Roberts: And I just want to, I just want to emphasize: the older you are, the weirder and creepier and alien a lot of this seems, of a world of surveillance, a world of cameras, a world of autonomous cars, a world where your does blah-blah-blah. Younger people are just going to find this normal. And, it will be the way that they are used to things. And culture, I assume, will evolve to accept most of this. You know, just the thrill of going shopping--we talked earlier: I happen to like going to the grocery store. I don't so often, but when I do go, I enjoy it. And, that fun's going to go away, I suspect, if delivery becomes cheap enough. And all these things are going to cause culture changes that we can't imagine.

Benedict Evans: Well, the changes that we can't imagine--the funny thing about this stuff is that we actually can't imagine the changes that have happened, because we don't have the perspective on what it was before.

Russ Roberts: Yeah.

Benedict Evans: So, to give an example: Like if you were a wanted man in a provincial town in Europe in, like 1890, the police--all the police had to do was put, like, one guy on each of the three roads[?routes?] out of town, and that was it. And they'd put a guy on the train station. And that was it. Like, you couldn't get on at the train station, so what are you going to do? You are stuck in the town. You can't go anywhere. And then cars come along and suddenly people can escape; and people can--you know, that degree of control just disappeared completely. There's a thing from the early 20th century called the Bonnot Gang in France, who were a gang of anarchist terrorists, who stole cars and stole rifles and went around killing people and shooting policemen. Police had no vehicles. Police had nothing--couldn't--struggled to work out what to do about this because they were stealing cars and could travel over 50 miles an hour. But the time that the police could catch someone by just checking the railway station is just kind of unimaginable now. But that was the way the world was.

Russ Roberts: And now they can check everything almost. And they will be able to--

Benedict Evans: Exactly. And again, we'll get other kinds of freedom and other kinds of restriction on freedom.

COMMENTS (23 to date)
Jim writes:

Another perhaps unexpected upside: decreased interaction with law enforcement, since, presumably, vehicle autonomy = no speeding or other moving violations. Lately, who wouldn't want to both allow LEOs to focus on serious crime and avoid unnecessary contact?

Martin Brock writes:

I doubt that level 5 vehicles will replace personal vehicles any time soon, if ever, because usage of a car is more varied than Evans imagines. To get from point A to point B, I may prefer a fully autonomous car 90% of the time, even 99%, but people will want a personal vehicle with a manual override for use cases too unique to be automated. I have used my car to pull a bush in my front yard out of the ground. Will autonomous vehicles have a "follow that car without being noticed" mode? How about "drive slowly around the neighborhood and stop when you see a cat"?

Earl Rodd writes:

In this very interesting discussion, there was one frightening line from Mr. Evans: When asked about "hackers" and more generally about security, he answered to the effect that "first you have to build it, then figure out the security." I'm confident Mr. Evans knows more about cars than computer security, but sadly, far too many devices and applications are developed by people who have this attitude - one which has lead to numerous security disasters. The only proper way to implement a device which must live in the dangerous world of computer connectivity is to build in security as an integral part of the design from the very start! Significant flaws have been found in automotive computer systems, thankfully none that have lead to widespread safety problems yet.

Max Zombek writes:

As a car nerd and a frequent listener to your podcast this was a very cool episode. But I am not sure if Mr. Evans takes the spontaneous nature of change fully into account: It is sword that cut into both sides. Of course it is possible that battery technology will be able to provide far more distance in the near future (unfortunately he not talking about the limited nature of resources like lithium which is probably needed), but he oversees imho the possible advantages of the internal combustion engine. Mazda for example is developing an engine currently, that combines the fuel efficiency of a Diesel with the characteristics of a petrol engine.

Despite all the euphoria for a new one, we should not forget that the dynamics of change are not calculable and this knowledge should be brought more strongly into politics.

[links and html reformatted because they did not work. Please make sure your links work by Previewing your future comments.--Econlib Ed.]

Gary Goubeau writes:

This episode seems to beg for fact checking.

Maybe the guest is fundamentally correct about the coming of the electric, but several of his statements were inconsistent with my admittedly limited understanding and amateur knowledge about cars, ICE and electric.

My skepticism was aroused when the guest said electric cars require more computing power than ICE cars. Maybe, but I doubt it.

Every event of the ICE is now controlled by an on-board computer - every burn, every transmission shift, the braking, and even the operation of suspension components. With electrics, it's mostly battery management.

He is certainly correct in saying the electric is mechanically simpler, and this simplicity requires less electronic control for operation.

Electrics have lots of appeal for the greeenie and geek, and even for the driving enthusiast; they do not for the accountant. They are not now cost competitive, even with subsidy from both the manufacturer and the government, and many are not convinced the gap is closing.

Batteries are getting cheaper, but the ICE technology is also improving. And the world is awash in oil.

This assumption that the cost of the electric will soon drop below the cost of the ICE vehicle needs a more rigorous examination. (And there's that range anxiety thing.)

The guest also said a 220 home charging outlet could be installed for a few hundred dollars. Would be more like a few thousand. A comparatively minor issue, but it makes this listener wonder about everything being said.

I suspect there's more to question, but I haven't finished listening.

Ben writes:

I loved the deep dive into the automated world of self-driving/navigating vehicles.

I would like a follow-up episode to go deeper into the economic implications that will inevitably follow:

* What impacts will this have on the travel industry? Hotels, interstate restaurants, and even air travel? When individuals can sleep in a vehicle overnight and wake up at a destination, the ripple effect this sends through the industry is incalculably disruptive.

* This was touched upon, but I would like to know how deep this forthcoming revolution disrupts the job market? Professional truck drivers, delivery drivers, taxis, insurance agents, car salespersons, etc.

Thank you!

J Scheppers writes:

Interesting episode:

As a skeptic of 'cars get huge subsidies', Russ superbly and politely defend his point regarding subsidies.

The future of transportation has so much speculation, and the inevitableness of electric cars seems the most questionable. Batteries are still by far less efficient at storing energy compared to fossil fuels. This could change but tech progress with fuels could easily be equal to that of batteries. The environmental cost with either option are not fixed nor known.

I was quite impressed with the theory of minus the driver and minus the weight then on demand rides cost could be reduced 75% from today's prices. There is quite a bit of conjecture there, but if even 1/3 true this could be a substantial innovation. This one statement made the whole episode valuable.

I like the mutual agreement that almost all that was said would be wrong, but I thank both Dr. Roberts and Mr. Evans for prognosticating.

My theory is that Tyler Cowen's analysis of chess should be applied to car autonomy. It is the mixture of the machine and human that produce the best result. Level 3 is better than Level 5.

David McGrogan writes:

There is a certain utopian strain of thinking in the discourse around driverless cars that I find a bit concerning. Yes, it's true that 35,000 people are killed on the road in the US every year by cars today. It could be the case that with driverless cars the number would be 0. But let's wait and see what happens when there are 300 million of them on the road.

Kevin writes:

One thing not well addressed was that while batteries are getting cheaper, their energy density is not improving very rapidly and has not for a long time. Batteries do not advance incrementally they tend to advance in big jumps. We may be due for the next big jump which makes all of this possible, but lithium is not up to the competition with oil. I am hopefully this might improve, but until it does electricity faces the problem of the huge energy density of its competition - gasoline. Gasoline engines don't fall to 80% range after 2-3 years, batteries being regularly drained do. Replacing the batteries is more expensive than replacing the internal combustion engines in most cars. They are serious costs.

How much weight in batteries do I need in an SUV to jump in and pull the boat to the lake 350 miles away for a day of fun? How many batteries does a Ford350 need to pull lumbar and a trailer full of tools all day without time for hours of refueling. The modelX experience pulling a trailer suggests we are not close to this being a good tradeoff for the majority of people outside of cities.

Electric cars have a fantastic use - commuting to jobs that are in dense population centers. So far they totally fall apart in other situations. All these problems might be solved, but its not clear they are about to be solved soon.

I think robot cars are the super future. I cannot wait to get my lost time from commuting.

Bill Brown writes:

I wish you would have pressed him on the oil subsidies. Everyone just says it as if it's common knowledge but rarely can elaborate when asked for specifics. Kudos, at least, for asking.

Much of driverless car promotion features the sort of hand waving that Mr. Evans displayed. It's far more complicated than they let on and is akin to AI in that regard.

One thing that I haven't heard discussed at all is modding. Lots of people mod their cars nowadays and there's no reason to think that they wouldn't do it to their autonomous vehicles—the problem with hackers isn't always and only external.

In fact there's a lot of incentive to mod your car to go 120 mph and be "inconsiderate" knowing that the other drivers' cars will react and route around your aberrant behavior. Two of the problems with human drivers is that they take this sort of thing personally and that they can't react to bad driving fast enough, so most people drive carefully.

This is exciting. I hope I live long enough to see what evolves.

I'm skeptical about the electric cars, though. I can think of too many scenarios where the battery limits are a big problem, especially on that trip to Disney.

Of course, I'm sure that there are still many things that horses can do that automobiles cannot.

Matěj Cepl writes:

It seems to me that however this interview was awesome (and it was) it is still in the danger of being the same kind as the prediction about colourful faxes.

I think we are standing on the edge of the end of the Industrial (or Modern) age and something new is awaiting us. One particular aspect which relates to the transportation and cars is in my opinion that employees are moving back from their offices. With the advancement of Internet and other higher forms of communication it makes less and less sense to commute an hour or so to some other building and work there, when everything required for work can be at home. Why to build specialized expensive buildings just for working?

The idea of “leaving home to work” was completely strange to people before year approx. 1800 and I am not sure whether we can be estranged to it again.

Of course, there is zillion problems with it, but it seems to me that this could be as big revolution as was the Industrial revolution. And yes, everything will have to change: our homes (too small to spend most of our times there), our cities (too many now useless office buildings), our way of life (socializing won't be done primarily in the context of work). Perhaps we may see again rise of deurbanization and people moving out of the large metropolis to smaller communities. Who knows?

But certainly, such process would completely change our expectations put on cars and transportation.

More on this on my blog

Jim Wildman writes:

Very interesting episode.

I laughed out loud at the naivety of "we'll make it secure later". No, we won't. We may be able to make it more secure than it originally was, but those will be bandaids. If the systems are not designed to be secure from day one (and they are not), they never will be secure. You can't go back into systems with millions of lines of code and make them fundamentally secure after the fact. So that thought train is fundamentally flawed.

Mr Evans is also very behind the times as to the prevalence of Level 2-3 features. I travel extensively and rent cars (mostly GM, Nissan and Chrysler models). I just picked up a Nissan Rogue (entry level small SUV) and it has full Level 2 features as do many Chevy Impalas and Nissan models. (ie, lane change, blind spot and collision avoidance/braking systems). I've driven fully decked out Chevy Tahoes/GMC Yukons (luxury SUV's) which have steering override features. If you attempt to change lanes w/o signalling (or drift toward the line) the vehicle steers you back into the lane. It feels very similar to hitting a ridge of snow and is pretty unnerving at first. One of them also had a vibrator under the drivers seat which goosed you when you started to drift, as well as audible and visual warnings. It would very aggressively brake (dropping out of cruise control) if you were closing too rapidly. In order to run into something, you would have to ignore multiple tactile, visual and audio feedback stimuli. So no need to buy "a high end German vehicle" to get those features.

I loved the discussion of 2nd, 3rd and 4th degree consequences. I've opined that autonomous/self driving cars will

  • Kill sales based on performance. No more sports cars because there will be standards set for minimum and maximum acceleration, etc. If the goal is to reduce accidents, all cars perform the same
  • Introduce a whole new level of regulation. Cars and their software will have to be certified (licensed if you will)
  • It follows that there will be no more self repair, tinkering (though as noted, a lot of that goes away with the lower maintenance requirements).
  • The camera issue is scary
  • A side effect of no car escaping other legal issues. "I'm sorry Mr Roberts, I'm unable to take you to the grocery store because you have unpaid taxes. Please swipe your credit card to resume your trip, or I can deliver you to the nearest police station at 1040 Nofreedom Lane, Bigbrother, CA" Ditto for any other crime. And THAT system will be absolutely abused. Patriot Act on steroids.
  • The rural areas are going to become even more of a separate world than the cities. I have lived most of my life in Ohio, Texas and Florida, in mostly rural areas. I and my wife have averaged (over the 35+ years of marriage) over 40,000 miles a year, typically spread across 2 vehicles. Just raising 5 kids and doing our stuff. I don't think there is any electric car on the drawing board with the range/recharge time that can handle that. And I have definitely used my family vehicle to pull bushes out, tow other vehicles, RV's, moving trailers, etc, etc.

Love your show in general and recommend it to everyone that I talk to who is a podcast listener.

tt31 writes:

Something missing from this discussion is that it seems to have focused on the implications for middle and upper income groups, but when talking about wholesale reshaping of cities, suburbs, and rural areas, all groups should be taken into account. For example, do we think there will be a single equilibrium of on-demand vehicles that apply to different income ranges? If not, what would the different equilibria look like and how would they affect each other? I appreciated the episode’s wide-ranging speculation but this seemed to me a notable blind spot. It’s worth noting that the internet and cell phones have impacted all segments of society, but the way different segments have experienced those changes have differed. And while those impacts are interrelated, the interrelations are largely below the surface. But in how vehicles behave around each other, or where people move, the interactions may be more at the surface.

Jeremy writes:

Great talk! Below are a few comments

What about the people who enjoy driving? A massive amount of money is spent on high-performance automobiles, motorcycles, etc every year. The thrill and satisfaction of 'control at speed' has value to many consumers. Also, assuming individuals still view material objects as extensions of themselves, you are going to have humans driving side-by-side with driverless cars which (among other things) increases the computational workload and potential for random events. Are the algorithms going to be able to react efficiently and effectively given all the changing variables? Netflix movie prediction > Family safety.

What about switching of roads to full bike and pedestrian walkways? What are the implications for housing and community policy when multiple neighborhoods move towards road removal / greenway conversions at the expense of a minority of people still dependent on roadway access?

When someone in my office starts talking about driverless cars, I always bring up the fact that poor data quality and variation in cell coverage still needs to resolved. On multiple occasions, I've observed map software (vendor omitted) with incorrect information which has negatively impacted driver choices and in some cases extended trip duration (ex roads marked closed which are actually open). Are you going to put that next vacation or the lives of your family in the hands of an algorithm with bad information speeding down the highway at 70mpg? If the reply is "we just need the right data" please forward a copy of

As others have mentioned, the discussion of autonomous powered surveillance seemed like a horrible honey pot which will attract the government under the umbrella of public safety AND criminals looking to profile individual movements with the intent of hacking the growing number of location based authentication methods.

At some point, we will have electric at 'range parity' with gasoline powered automobiles (say 800-1K miles per charge or the maximum amount a single person could drive in 24 hours w/ load in all weather). Assuming the cost is the same, you could imagine many people taking the electric option. My guess is that will occur before of any mass adoption of driverless cars...

bogwood writes:

Enjoyed the episode and the discussion. I worry about diesel and debt.

Gasoline cars are the cheesecake, diesel engines are the main course. Add diesel's role in nuclear and hydro, diesel is the only source of energy surplus and thus wealth. Diesel subsidizes everything including government. We are in a unique historical period in which agriculture is an energy sink. Without cars the gasoline might have been flared off as a waste product. Without gas engines the cost of diesel would double. There could be a debate with Alice Friedemann("When the Trucks Stop Running").

Someone else mentioned earlier years. Looking back a vehicle was self-amortizing, a farm wagon was also a ride to church, a chariot captured some prize. Gasoline cars are unproductive. The physics of carrying around 4000 pounds of metal is bad enough, the economics of borrowing 40,000 dollars for a depreciating asset is worse. The full cycle cost of autonomous cars may be equally expensive, considering networks, write-offs and infrastructure. At some point a trip has to have a productive element, otherwise walk or bike.

A lot depends on the future of diesel supply, I lean toward the under.

Mark Sundstrom writes:

It's clear to me that electric cars (coal-powered cars as I refer to them here in Ohio) can be useful for short- to medium-length trips in urban and suburban environments. I don't see how they are going to be efficient for my application. I tow a small travel trailer on a few trips per year and often will drive 200-400 miles per day, sometimes further, and sometimes in hilly conditions. I would like to hear about electric vehicles that can efficiently handle that task. I am not holding my breath waiting for them.

Niels Bauer writes:

Though I am skeptical that electric vehicles will completely replace ICE vehicles, they will continue to increase as a percentage of vehicles sold and have an inevitable impact on the following businesses:

  • Oil change shops along with makers of oil filters
  • Muffler repair shops along with the muffler manufacturing and supply chain
  • Transmission repair shops along with the transmission manufacturing and supply chain
  • Gas stations
  • Gasoline distributors
  • Drivers of gasoline trucks
  • Manufacturers of gasoline trucks
  • Gasoline pump manufacturers
  • Auditors from Weights and Measures
  • Auto tool manufacturers
  • Auto parts manufacturers
  • Auto parts stores
  • Auto mechanics
  • Smog testing stations

Next, with the transition to fully autonomous vehicles many common features will no longer be needed:

  • Review view mirror

  • Side mirrors

  • Brake lights

  • Turn indicators

  • Horn

  • Speedometer

  • Perhaps even headlights

Also as mentioned, there will be no more robbery getaways, no police chases, no movie chase scenes, no speeding tickets or fines, no traffic courts, no traffic school, no driving tests, no drivers education, no street racing, etc.

JRo writes:

I enjoy thinking about this topic and it's especially interesting listening to imaginative experts talk about it. But then I get on the airport bus in 'high tech' Seattle and am reminded that it still needs a driver. More depressingly, the on-board ticket machine can't even make change.

Later, looking out the window of my aircraft, I watch the ground support equipment in action. Every piece with an engine in it has a well-paid driver. "Take this luggage 2000 yards on the highly controlled and regulated flight line to the forward starboard door of that 747 parked at Gate 25." That seems so simple compared to "Pick up a drunk at Bar A somewhere in Manhattan and deliver him safely to his apartment somewhere in Brooklyn." When I finally see an autonomous catering vehicle delivering its precious cargo of inflight meals to the flight line, I'll start seriously looking forward to autonomous taxi/buses driving human passengers where ever they want to go.

Jakob writes:

the whole "car on call" model seems to be a very vulnerable system that makes me rather uneasy. Even if each car is not vulnerable to cyberattack, the system will depend on working computer networks and communications. How will that work after a major natural disaster, war, terror attack, etc? Hard to see people just jump in and get the cars out of there...

For example: For a while, I used an app on my phone to buy tickets for my commute. Works very well... until it does not. Like a server or network outage, at which point you have buy tickets at 2x the price at the station rather than the cheaper bulk on the phone.

One other potential joker in the transportation game are the electrically-assisted bikes. Electrobikes. They seem to be quickly outselling cars in sheer numbers, and offers a very clean and compelling solution for a 10km commute. They can replace a lot of use cases for short-haul car rides.

CowboyProf writes:

Interesting conversation, but something bothered me in light of the recent spate of hurricanes. The guest noted that most people don't "need" a car that will drive 200 or 300 miles, so the charging issue is that problematic.

Perhaps true in ordinary circumstances, but allow me to posit a Nassim Taleb "black swan" question:

If we were in an age of fully electric cars that needed to be recharged after 200-300 miles, how would an evacuation ahead of Hurricane Irma in Miami work?

Graham writes:

The discussion of 'range parity' assumes that motorists have a universal preference for electric. I would argue that actually many motorists would pay a premium for a gasoline/diesel vehicle with the benefits of a 500+ mile range and a 5 minute recharge anywhere at any time.

Simon H writes:

An interesting episode as usual and I wish there had been more discussion of some of the regulatory implications of a zero accident world. I assumed that that world would require zero human drivers, which would then seem to require the government outlawing the use of human piloted vehicles, at least on the roads that carry Self-driving traffic. Would human driven cars then exist in the way that off-road only vehicles do now?I wonder which politician will put forward Legislation banning human powered vehicles from most roads. I don't want to seem paranoid but this transition seems like kind of a big deal from a regulatory and civil liberties perspective.


Return to top