June 19, 2019

Life after Driving

I’m working on a three-part series on self-driving automobile technology for Ars Technica. In part one I covered the state of existing self-driving technology and highlighted the dramatic progress that has been made in recent years. In part two, I assume that the remaining technical hurdles can be surmounted and examine what the world might look like when self-driving cars become ubiquitous. The potential benefits are enormous: autonomous vehicles could save thousands of lives, billions of person-hours, and billions of dollars of energy costs.

The article has sparked interesting discussion around the blogosphere. Matt Yglesias has a long-standing interest in urban planning issues, so he did a post about the urban planning implications of self-driving technologies. I argue that by making taxis cheaper, self-driving cars would shift a lot of people from owning cars to renting them. And that, in turn, would dramatically reduce demand for parking lots, which will allow more pleasant, high-density cities. It’s hard to overstate the extent to which the need for parking exacerbates sprawl and congestion problems. Parking lots consume vast amounts of land in suburban areas. This, in turn, means that stuff is farther apart, which forces people to rely even more on their cars to get from place to place.

Matt’s post prompted a number of interesting responses. Ryan Avent chimed in with some thoughts about how self-driving technologies would make urban living more attractive. On the other hand Tom Lee offers a counterpoint: making car travel cheaper and more convenient will, on the margin, cause people to drive (or “ride” anyway) more. This is a good point, and it’s not clear how these factors would balance out. But even if Tom is right, this wouldn’t be an entirely bad thing. Increased mobility is a virtue in its own right.

I think Atrios and Kevin Drum are on less firm ground when they argue that this technology is so far in the future that it’s not worth thinking about. Drum compares self-driving technologies to cold fusion and human-level AI, while Atrios compares them to flying cars and jet packs. I can only assume they didn’t read the first installment of my series, in which I discuss the state of the technology in some detail. The basic technology for self-driving is already here. There are cars in university laboratories that can navigate for hundreds of miles without human supervision, and can interact safely with other cars on urban streets. Of course, there’s still a lot of work to do to enable these vehicles to safely handle the multiplicity of obstacles they would encounter in real urban environments. And after that the technology will need to be made reliable and affordable enough for commercial use. But these problems are nowhere close to the difficulty of human-level AI. Your car doesn’t have to understand why you want to go to the store in order to find a safe path from here to there. If you’re skeptical that this technology can be made to work, I encourage you to read my first article and watch PBS’s excellent documentary on the 2005 DARPA Grand Challenge. There’s a lot of uncertainty about how long until this technology will be mature enough to let loose on our streets, but I think it’s pretty clearly a matter of “when,” not “if.”


  1. These devices will always be deadly in a busy urban environment because so much of the interaction between vehicles relies on eye contact between drivers. They do not need to hit another vehicle to be the cause of the accident and there is no easy way to do the switch to all driverless vehicles overnight. Anyway, I like driving and would not consider paying for a machine to do it for me any more than I would pay a machine to make love to my wife.

  2. Village Idiot says:

    Self-driving cars sound a little like public transportation but with individual pods instead of trains or buses. Cities that matured before the automobile arrived are already much more dense than recently developed sprawlscapes and usually have very good public transportation systems, which seems to already be working as a solution in many places though mostly outside the U.S.. None of my friends living in Europe or New York City own cars and they get around just fine. Conventional public transportation is also always going to be much more efficient overall than individual vehicles for everyone; it’s an economy of scale thing.

    I’d always thought if we’d built robust public transportation infrastructure within our cities and kept cars out altogether, they could be much smaller and more efficient; maybe even approaching sustainability. Cars could be kept in large parking superstructures on the edges of cities; go rent one from the North side parking center if you need to go North, then park it in the South parking center of your destination city and take a train the rest of the way in. For small towns and rural areas, you’d still have your car if you lived there or could rent one if you wanted to get there. This is something we could have done with 1950’s technology. Sigh.

    I’m not sure it’s actual ‘driving’ that people like when they say they like driving, unless they have open expanses of scenic winding roads ahead of them and a sports car. Most of the reality of driving is sitting in traffic or waiting at stoplights (woo-hoo!) so I think it’s more about a (subconscious?) desire to seal the world out of our little private boxes for a while or sustaining the illusion that we are totally in control of our lives than anything else. Driving is an important metaphor after all; I’m driving my point home, he drove the company into the ground, I’m driven to succeed, etc. As far as equating driving with sex and not wanting technology to do either one for us, I’ve seen plenty of marriages get to the point where the husband and/or the wife were ready and willing to buy that special machine Duncan isn’t ready to get quite yet (plenty of people are working hard on that project, too!).

  3. I wouldn’t be so worried about the interaction between podcars and human drivers (eye contact, waving someone forward and so forth), both because the pods can be programmed to drive defensively (don’t think that someone isn’t going to turn left in front of you just because their blinker isn’t on) and because people can be programmed the same way — when you see a podcar coming, you react pretty much the way you would to one of those big “STUDENT DRIVER” signs. Add to that some rules about podcars getting to use HOV and bike lanes, and you’re a fair amount of the way along.

    That said, you still have a serious chicken-and-egg problem with commuting. Even cities with taxis handle this with rail and bus. It’s not clear that eliminating the cost of a driver will make pods really affordable for drivetime, and then, as long as people still have cars they’ll use them for other things. I also wonder how much of the podcar advantage could be gained simply by a much-expanded Zipcar short-term rental system. (Then again, something like zipcars could also serve as a bridge to pods, by weaning people away from personal autos; then it eventually wouldn’t matter whether they drove themselves or were driven.)

  4. I’ve been involved with research for autonomous vehicles for several years, and I think the main obstacle may prove to be non-technical: who will be liable if there is an accident? If the car company could be sued if there were an accident, they probably wouldn’t sell the car in the first place.

  5. I know this is a small part of the overall picture, but it would make a big difference to my wife if she had a self-driving car. My wife has vision impairment and doesn’t drive. It would give her some real independence to be able to get in the car and run errands or just take off for a while.

    I’ll be watching this issue with interest going forward.

  6. osama_been_forgotten says:

    Yes, we can build cars that drive themselves. We could also build a rocket that could take men to the moon and back over 39 years ago. We are not all flying back and forth to the moon on a daily basis.

    The technology for self-driving cars requires a TON of hardware (maybe literally). Sensors, computers, power supplies, cables, software, servos, etc. None of this contributes to the vehicle’s efficiency (electrical power required, cooling systems, as well as sheer weight that has to be moved, and interior volume consumed). And this technology has not been proven in bad-weather situations, either.

    Then, there is the idea of cost. In an industry where a manufacturer will spend tens of thousands of dollars studying whether to use the 80-cent airbag trigger switch, as opposed to the 50-cent airbag trigger switch, the industry that switched to seatbelts and unleaded gas because government had to FORCE them to do it – they’re going to be able to successfully market a car that’s going to have $50,000 worth of hardware added to the sticker-price?

    Will anybody be able to afford to PURCHASE such a car? Let alone pay for maintenance and operation? Give me a break.

    Yes – I am sure that economies of scale, and technological advancement will eventually mitigate these issues somewhat. But again: this is the industry that had to be FORCED by the government, to put a single indicator light on the dashboard to tell you that there was a problem with the engine. (OBD). Look how long it has taken for automotive technologies like disk brakes, fuel injection, etc. to become ubiquitous. Even automatic transmissions are not yet ubiquitous – they’re still an “extra” option on many low-end cars.

    I respectfully submit that anyone who believes that self-driving cars will become ubiquitous in our lifetime, has not really thought this out.

  7. James Hanley says:

    Some interesting comments here. I think the technology is clearly feasible, but it is the cost and personal factors that could delay implementation.

    Osama-been-forgotten makes some errors, though. We don’t fly back and forth to the moon regularly because it’s still highly dangerous, just because there’s no air in space, so an accident has higher probability of causing death, and because the benefits of repeat trips to the moon just aren’t very high. Neither of those apply to self-guiding cars.

    Also, it’s a great myth that the government had to force automakers to put in seatbelts. Automakers invented and began installing seatbelts well before gov’t got into the act, but as with all new technologies they offered them as optional equipment at first. Then the government, in its paternalistic role, decided that we shouldn’t have the choice to have a car without the safety device (and then later decided that we shouldn’t have the choice about whether to use it). You can be damn sure that if seatbelts were highly demanded by consumers at the time that the automakers would have been happy put them in, because whoever didn’t would have lost sales.

    Unleaded gas was an entirely different matter–there was no direct economic benefit for either the manufacturers or the consumers. Car buyers certainly weren’t clamoring for unleaded, in fact they resisted it at first. That was a case of third-party harm–other people’s leaded gas exhaust was harming everyone else.

    Sorry to be a pedant–I just get tired of people who repeat the same old anti-free market bile without knowing what they’re talking about.

    On other matters, In addition to the problem of some people actually enjoying the act of driving, there’s the problem of inaccurate risk perception. People believe they’re safer when they’re the ones driving (studies show a majority of people think they’re above average drivers, a mathematical impossibility of course), and it may be very hard to weed people off that inaccurate mindset. Not impossible, but I would expect it to take more years rather than fewer.

    I think where technology advocates make their errors is not in the technical feasibility of things like this, but in assuming the average citizen has anywhere close to the techie’s “gee, cool, gotta have it” response. That’s not to bash the techies for having that response. It’s just that the majority of the public generally doesn’t, which slows down the adoption of the new technology.

  8. Perhaps even more relevant to this discussion was the 2007 Grand Challenge, which was an “urban” challenge, where the cars had to interact with traffic (light) and obey traffic rules.

  9. in addition to the problem of some people actually enjoying the act of driving, there’s the problem of inaccurate risk perception

    Yes, but isn’t the consensus perception what is going to drive acceptance and commercial success? I think OBF is wrong, mostly because he is measuring machines against some imagined competent human driver that doesn’t really exist. However, if his view is widely held, that’s what will drive acceptance.

    There’s numerous cases where careful measurement says one thing, and we just go on ahead and do something different. A rational, safety-minded driver would never use a cell phone in a moving vehicle, studies show it’s about as bad as being drunk. Nonetheless, people do this all the time. Once, in the spirit of Science, I decided to stop for pedestrians in crosswalks if I possibly could — clearly safe, because everyone drives attentively and with adequate following distance to stop quickly if the car in front of them stops. After being rear-ended twice, I ended that experiment. (And ask yourself, how bad does a driver need to be, to not see a Ford Taurus wagon, in front of them, stopping at a crosswalk, with a pedestrian in it? And I found not one, but TWO such drivers. How sure are we that robots are unable to beat us?)

    One possibility is that robots will use different safety algorithms, and actually obey them. The standard rule, widely ignored, is to maintain a safe following distance. Another rule is to not drive when you cannot see. Another rule is to not drive when you have no traction. People routinely break these rules — they tailgate, drive when sunblinded, or peering out of a little porthole scraped in the snow on their car, or try to drive in conditions so bad that cars are sliding downhill. A robot could simply say “no”. A robot could elect not to drive drunk, and that cuts the accident rate in half. It could obey speed limits, stop for stop signs, stop at stop signs. It could stop for pedestrians in crosswalks. It could stop for stop lights. It could not drive the wrong way down one way streets. It could not cross the double-yellow line. Every day, here in Massachusetts, humans are “unable” to obey these simple rules. I’ll bet we could program a robot to obey all of them.

    Or, how do you define “safety”? It’s not “minimizing expected years of life lost”, because if people were doing that, lots more of them would be riding bicycles (lack of exercise is far more deadly than auto or bicycle accidents — and walking is more dangerous per mile than either). So, a truly smart car, after you programmed in your 1km trip, would tell you “go ride a bike, lard-ass!”

    Speaking as someone who does try to behave rationally, and who does ride his bike at least 50 miles every week, I think robot cars would be great. I am sure that there will be some shape, particular color, or feature that they are good at detecting, and once that is published, I will ensure that my bicycle and all my kids’ bicycles display that thing. It’ll be wonderful. Whenever I ride my bike, I know that the exercise is doing me good, but I also think of those idiots who could not even see a stopped station wagon right in front of them.

  10. Interesting post. I recall reading a science fiction story about this very subject in 1977 or 1978. In this story, the technology was so safe that human driving was outlawed. If I recall the author I’ll follow-up with a post.

    Let’s just say the the technical hurdles are surmountable, (and that seems realistic) here are some scenarios:

    1. The technology is adopted, and it is much more safe than human-controlled cars. Should society then seek to make the much more dangerous human-controlled automobiles pay for the social costs they extract from the rest of society? Perhaps higher taxes on human-controlled cars would be an appropriate policy response?

    2. If your answer to the above question was yes, it would be appropriate for society to make a more dangerous technology pay for its higher costs, exactly why do we have to wait for that particular superior technology to come along before we start taxing dangerous human powered cars? Today we have subways, trains and bicycles, all probably safer than cars. (I am unsure about bicycles, but we have to use the right metric–not just deaths/passenger mile because that doesn’t account for increased healthness of bicycle riders, or the deaths from pollution caused by automobiles, to say nothing of the deaths in wars to get the gas to drive the cars that we don’t need)

    3. I think one item needs to be thought through a little more–what are the ergonomics of such a machine? I believe it would be very disconcerting to be in a car that would automatically drive itself. Do you have any first-person descriptions of what it felt like to drive in one?

    4. My one take on this is that all of the hurtles are so high that you would require a very well-funded research effort AND a regulatory change. In this case, you are in a kind of a chicken/egg dilemma. No one will want to fund the development until they know that the result will be able to be driven, and no law maker will want to change the law until he can see the technology works. How would this technolgy get past this dilemma?

    5. I’d like to note that there is another technology that exists now that, if we were to adopt it on wider scale it would have many of the same benefits, to an even greater extent, as this technology. It would: take less space parking than present day cars, be much better for the environment, encourage much more compact urban development. It is also extremely affordable, nearly everyone in the USA could have one today. It is called a bicycle. So, isn’t there something of the technology for the sake of technology going on in advocacy of this?

    The best solution is sometimes (very) low-tech. I would propose we need to rethink what defines progress, and to do so we just need to dispassionately chose our metrics (criteria) then evaluate solutions. I would propose this technology has been arrived at backwards; it is a solution to a problem that has occurred, in part, because it is such an intellectually and technically fascinating process to create such a machine. But that doesn’t mean that the bicycle isn’t a better solution to the identified problems. My point in this is not to deride those who have created these amazing machines, but to make the point that we need more BALANCE in the development of technolgies.



    • Yes because driving a bicycle for 50 km in 40*C weather or when it’s snowing/raining sounds like so much fun. Never mind the even longer trips, and carrieng all the groceries for a family of 4 or more would be problematic, especially if you have to take toddlers with you to the shop.

      Bicycles are not the solution in the US, from sheer distance, terrain and weather factors. They barly work in Copenhagen, and that is a city that was built with a city wall to defend it in mind and so is quite cramped.

  11. Village Idiot Take Two says:

    Regarding a (very) low-tech solution, consider that not much more than a century ago, our primary transportation operated entirely on domestically-produced biofuels, was already semi-autonomous (if you fell asleep on the way home, you’d wake up at home instead of wrapped around a tree), replicated itself, could traverse very complex terrain relative to our modern passenger vehicles (reducing the need for expansive and expensive highway infrastructure), and made bicycles seem high-tech in comparison.

    While researching unrelated events, I saw ads in the New York Times from around 1908 (if I recall correctly) accusing this historically venerated mode of transportation of being too stinky and how it needed to be fed frequently unlike the new, miraculous, ostensibly not-stinky Horseless Carriage of Tomorrow! They actually claimed cars didn’t stink or need to be fed! Hmm… Horses do have their own unique aroma and need to be fed, but replacing them with cars just meant a different (not to mention very unhealthy) odor following us around, feeding them a toxic and unsustainable ‘food’ instead of grass or grain, and no longer being able to reproduce them at home. Ah, but nowadays we can go get a bag of Funions and a Mountain Dew from a store that’s miles away in just minutes that on a horse would’ve taken over an hour! Well, why is the store is so far away in the first place? The Amish seem to be getting around ok, so why do we assume we need to always be able to move around so fast? What’s the rush, and how has moving faster and faster improved the quality of (some of) our lives? Or has it done the opposite? How can we tell?

    Automotive infrastructure (roads, driveways, parking lots, gas stations, car dealerships, repair shops, tire stores, car washes, the acreage around on-ramps and highway medians, etc.) adds up and promotes sprawl since cars are such large objects to accommodate, sprawl necessitates having a car to traverse the increased distance between everything, everyone having a car then severely congests most of that infrastructure and slows it to an average speed similar to that of horses, and we’re right back where we started except for it costing many times as much and being completely unsustainable and exceedingly unhealthy.

    Whether the cars drive themselves or not seems kind of irrelevant to the central problem of balancing the efficiency of higher population density with the ability to move around with minimal friction due to that density. Eliminating all autos in cities (even autonomous autos) would allow cities to shrink to a human scale and so we’d no longer need cars (or horses) there and the efficiency gains would be vast. Our Funions and Mountain Dew would always be a block or less away, and we wouldn’t need to rush around so much all the time trying to pay for it all (or get our prescription for high blood pressure medication filled).

    I’m also surprised no one has mentioned the obvious privacy concerns. Everywhere we go could be precisely tracked, and our cars could theoretically be programmed to not even go to certain places at all, or they might receive an official request that we answer a few questions and automatically drive us to the police station to resolve some ‘issue’ instead of taking us to our desired destination. Or someone presses a big red button and all our cars simply stop dead in their tracks for whatever reason the one in charge of the button wants, like maybe for a random ID check…

  12. osama_been_forgotten says:

    [quote]Every day, here in Massachusetts, humans are “unable” to obey these simple rules. I’ll bet we could program a robot to obey all of them.[/quote]

    . . . even worse; I figure we could program a robot to daily divert the passenger past a large billboard advertisement for the manufacturer’s parent company. Then a billboard for viagra, singles websites, and credit card company.

    Oh, I guess I was wrong – we will easily be able to figure out a way that the average consumer can afford a self-driving car!

  13. Anonymous says:

    James Hanley wrote: “Studies show a majority of people think they’re above average drivers, a mathematical impossibility of course.”

    Of course not. A majority of the numbers in the collection 1, 10, 10 are above average.

    Enigma Foundry wrote: “I believe it would be very disconcerting to be in a car that would automatically drive itself. Do you have any first-person descriptions of what it felt like to drive in one?”

    Sure — get in the front of a taxi on the other side of the pond and tell the driver where you want to go, then get driven there. 🙂

  14. It also occurred to me, seconds after my rant, that the easy way to deal with the liability issue is simply to be sure that the robot car films and records everything around it, as it surely will. If the robots drive well, their manufacturers won’t lose lawsuits.

    Of course, now we get to wonder what we will do with all of these videos.

    Anyone hoping to save energy and/or cut greenhouse gasses with a bicycle, needs to pay a little bit of attention to their diet. Beef and pork (and lamb) are not energy-efficient at the system level. Pimentel and Pimentel have written on this, there’s more references (and another rant) here. The rule of thumb is 30-50 kcal/mile for a moderate pace; I rode 80+ miles this week, or 2400-4000 calories. You get hungry.

  15. Anonymous says:

    “Of course not. A majority of the numbers in the collection 1, 10, 10 are above average.”

    Consider the distribution 1, 2, 7, 10, 10.

    * The mean is 6; they sum to 30 and there are five of them.
    * The mode is 10; the individual number that pops up the most.
    * The middle of the spread is 5.5 — 1 + 4.5 and 10 – 4.5.
    * The middle number of the distribution, in sorted order, is 7.
    * A majority of the numbers (three of them) are greater than the mean, the middle of the spread, AND the middle number.

    So what’s the “average” here? Is it 5.5, 6, 7, or even 10?

    “Average” can be a tricky beast (though most usually it is treated as synonymous with the mean, so it would in this case be 6).

    Similarly, what is “above average” and what is “below”? A lay person would probably decide that 1 and 2 were below average and the others were “about” average simply because the other numbers are fairly close together. If pressed, they might only claim the tens to be above average.

    When lay people and statistics meet, the results range from amusing to tragic but rarely prove to be especially useful.