Three counterintuitive scenarios for driverless vehicles

The standard story is that traffic deaths will dwindle, cities will spread out magnificently, and you’ll all be reading MR on your morning commute rather than fighting the traffic.  Maybe so, but what other options are at least worth considering, if only out of contrarian orneriness?:

1. Driverless cars are not actually much better than the really good German streetcar systems.  Those come closer to door-to-door service than many people realize, and of course they have lower energy and congestion costs.

2. The need for exact mapping of streets will restrict driverless vehicles to well-known, well-trodden paths, much like bus lines.  There is nothing wrong with that, but ultimately it won’t do more than save the cost of the bus driver.  Or worse yet — some automobile lanes may be turned over to municipal driverless vehicles in a way which makes traffic problems worse.  It will end up as a way to push cars out of the picture, without building up the broader mass transit network very much.

3. Driverless cars will give governments a chance to “redo” the whole driving side of American life.  Is this so great?  (Imagine if we had to write a new Constitution today.)  Just think, with driverless cars and laissez-faire there will be so many car trips, a city might collapse under the weight of its own congestion.  So a quantity-rationed system will be introduced, and ultimately all of driving will end up more controlled and more regulated, based on licenses in fact and no I don’t mean drivers’ licenses.

Most generally, your predictions for driverless cars should depend heavily upon: a) will there be rational congestion pricing?, and b) how rapidly will cities rezone to take advantage of the new opportunities?  I am not sure we should be especially optimistic about either a) or b).

Or put it this way: the absence of congestion pricing in most major urban centers means we are already bad at running roads, for whatever public choice reasons.  So maybe we’ll get a bad version of driverless cars too.

For a conversation related to this post I am indebted to Alex Tabarrok and also Joe Bous.


I was about to comment that the tech has already surpassed #2, and then I noticed that your paragraphs are numbered 1, 2, 2. Then I just wanted to comment before Thomas gets here and ruins everything. Will it work? I shall click and find out.

By 2, "exact mapping?" It seems from casual reading that driverless cars are getting better at purely visual solutions (driving on an express lane, getting into your garage), but I see that more as chipping away than ready for a random, rainy, mountain road.

Still progress is surprisingly fast. It could be that we need to worry about policy earlier than thought. But not quite yet.

Upon further reading I probably jumped the gun.

Tangential: The best thing about driverless cars, in my mind, is taking the 3000-lb sledgehammer out of the hands of the worst 20% of drivers. Until the tech is standard or mandated, the benefit will largely be novelty and consumption value. In the meantime, I think there's something to be said for tightening license eligibility and letting the drunk/elderly/hypoglycemic ride a scooter or take the bus.

Driverless cars are an old technology, as you point out. Subways, trains, buses, etc.

The driver car is a problem only because the generally lone passenger must drive the car and deal a hundred issues unrelated to getting from point A to B:
does the car have fuel
how do I get where I'm going
where to I put the car at my destination
is the car maintained
where do I go to get it maintained
is all the legal work taken care of
am I insured
why can't I do something useful
oops I need to deal some issue but how
I'm bored
I'm tired and sleepy, or drunk

Yes, but subways, trains & buses only work so well in Germany *because* of the compact city structure.

The streetcar in Tyler's #1 is a mass-transport. Cars, even if driverless are not. Driverless cars in a non-compact city are better than cars for energy & congestion but yet no where near how good streetcars in a dense city are.

'Yes, but subways, trains & buses only work so well in Germany *because* of the compact city structure. '

Not precisely - you need to check into Karlsruhe/KVV's dual system streetcars - they run both in the city and over the normal train rails (the 'dual' comes from the fact that the power supply is completely different between city overhead power and the rail system's). That is, these are trains that do a fairly good job of covering 20 or 30 towns (with populations between a couple and ten thousand people) between two cities. And really, the streetcar/train station at Baden-Baden is extremely inconvenient in terms of door to door use - luckily, Baden-Baden is well run enough to have a nice bus system to compensate for the fact that the train station is poorly located.

The issue with KVV at the moment is that those streetcars are so slow, since they stop literally everywhere. You're often much faster on your Ebike or in a car, especially since KVV reduces inner city congestion. It should all go a bit faster once the tunnel is up, ofc.

Berlin, compact?

Driverless cars don't address those. Nobody has proposed a self-fueling car, or a self-insuring one. Maintenance would need to be done manually.

But it is a different picture when your car can drive itself to the mechanic while you work. And for self fueling, see Teslas robotic recharger arm. Self fueling is simple compared to driving.

Gated neighbourhoods and offices will expand - although car journey times will decrease, corporate campuses will consume some of the benefit by becoming larger; your car will drop you off at the edge and let you walk to the center. Partially this avoids GPS-guided congestion through minor roads, but it's also the case that minor roads in general should shrink, and arterials should become larger, to accommodate the larger numbers of cars that can go home by themselves. This is self-reinforcing - if your home/office block is separated from the next block by a eight lane highway of Googlecars, it'll seek to develop its own services and infrastructure.

In legacy neighbourhoods, this will be achieved by altering minor roads into pedestrian-only streets, and widening arterials by streamlining traffic: pavement, parking lots, turning lanes, etc. all can go. This, too, is self-reinforcing: as traffic is streamlined, speeds increase, necessitating further barriers between traffic and pedestrians.

We already have individuals or developers pay to have their properties hooked up to municipal water and electric grids. Why would it be any different to have people pay to have their cul du sac or driveway or whatever surveyed to sufficient level of detail to have it accessible to driverless cars?

I'm curious to see how this comment section shakes out.

The Constitution is kind of like the Bible. People still go around saying they live by what the Bible says, but you can usually count on them to jump through whatever mental hoops are necessary to rationalize away the parts that are inconsistent with modern liberal society. Selling ones children into slavery comes to mind as an example. Similarly, there is a whole legal apparatus required for a developed country to function properly, and the Constitution is really just not equipped to handle it. There's just no getting around the fact that it was written for an 18th-century society of farmers and craftsmen and such in a context of mostly self-governing states. So by practical necessity we've gotten around this by aggressively interpreting clauses such as the Interstate Commerce Clause. A law needs to be a pretty egregious violation of an explicit statement in the Constitution to get shot down.

While I don't trust our current crop of politicians to write a new one, I for one wouldn't mind a constitution that was more explicit about what was allowed and what wasn't. Instead the process is more like, "All right guys, every other country has laws about X. Let's find a way to get this done somehow without setting off any obvious constitutional tripwires." While I often agree that X needs to get done, a lot of those X's are unconstitutional without some serious legal massaging, if I'm being honest about it.

I feel like this is going to be misinterpreted as a libertarian rant about big government, or a leftist call for a totalitarian dictatorship. It's neither, so if that's your takeaway, ur doing it wrong.

P.S. I mean, not only do we have disagreements about the Constitution, we still have meta-disagreements! We literally argue over the right way to deal with ambiguous words. Do we use an 18th-century dictionary or do we try to infer the Founders' intent? It's a mess, and the Constitution doesn't even specify how to resolve these ambiguities.

Another interpretation would be that it was a rhetorical question whose obvious answer is "no," and that the post is not in fact about the Constitution.

Did Nathan W. pick a new handle?

Joint venture between Jan and Nathan I think. Marrying obtuseness with length.

Jan is guilty mostly of putting a baffle in the echo chamber. Somebody has to call bullshit every now and then.

"A law needs to be a pretty egregious violation of an explicit statement in the Constitution to get shot down."

Well, that's assuming the law is in accordance with the Supreme Court's ideological leanings. If it's not, they'll find a way to claim that a 1868 amendment totally bans states from maintaining the old definition of marriage and somehow it took until 2015 for anyone to notice.

2. This assumes that street markings will be unchanged.

Is the current default level of mapping in say Google Maps not sufficient for driver-less cars? Just curious.

No - 'Google often leaves the impression that, as a Google executive once wrote, the cars can “drive anywhere a car can legally drive.” However, that’s true only if intricate preparations have been made beforehand, with the car’s exact route, including driveways, extensively mapped. Data from multiple passes by a special sensor vehicle must later be pored over, meter by meter, by both computers and humans. It’s vastly more effort than what’s needed for Google Maps.'

"It’s vastly more effort than what’s needed for Google Maps" -- the implication is that if they can pull this off, that effort will be a huge barrier for entry of competitors (or produce lots of licensing revenue).

I hope they can pull this off before I get to the stage where I shouldn't drive. Just to hedge my bets, our new dwelling is on a bus route and .9 miles from a train station.

I wonder how specialized these data needs to be. If you were to, say, take the GPS data from every trip by every car equipped with On Star over the course of 5 years or so, and throw enough computing power at that, what percentage or roads would you be able to map out well enough for driverless cars?

I can see the marketing campaigns now: "Google, with more Sleep-While-You-Drive(tm) coverage than Nissan and Volvo combined!"

I would volunteer several hours of my time to help Google map my commute if in the end I could get a commute I can sleep through.

#2 assume the google model of massive, laser guided millimeter mapping of a route.

Everybody else is just using autonomous software and assuming driverless cars are highway only.

We might get to the point that city to city transport is better left to cars again, as opposed to planes.

Since you have 2 #2's, I'll assume you are really tired.. read this post and thought, "what are you even talking about?"

re: 'not much better than a bus, minus the cost of the driver'.. what? If by 'not much better' you mean getting your own individually sized bus, whenever you want it, picking you up exactly where you are, minus of course the cost of a driver, then yes. I think uber w/o the driver is a much better comparison. There is the issue of requiring mapping, but most ppl live within walking distance of bus routes, and within the route/map, cars would have none of the other downsides of a bus.

re: congestion, why would things get more congested? In the SQ pretty much everyone owns a car anyway, and they aren't very good at driving them efficiently, or at sharing rides w/ neighbors. I find it unlikely that more cars will be put on the road at any one time. I find it very likely that there will be substantially fewer however- w/ an automated system controlling the car, coordinating carpools will be much easier and more reliable.

And by coordinating carpools, I mean in real time, and you will have to pay extra to not have someone else picked up by your car along the way.. because.. you won't need to own your own car, if one can pick you up wherever/whenever. It'll be a subscription instead. (oh the savings!$!$!)

and re: regulation, I'll give you that one. Lots of regulatory schemes I'm worried about... but laissez faire isn't one of them. I welcome deregulation of the driverless car market. All available evidence says it's much safer than the wackos that currently cruise our streets (even with the wackos still out there)

If you've ever lived in a city centre where you never have to wait more than a few minutes for the next ride, and never have to walk more than a few blocks to find such a bus or streetcar, then this shouldn't be too difficult to digest.

#1. Maybe Munich is backward or something, but I don't think trams are as important to people as bikes and cars. Certainly I live near a tram line, but I use the U-Bahn much more often. And my bike more often still. Cars are more common than bikes. Driverless cars won't be different, unless #2 is true.

#2. The need for exact mapping is not an eternal technological hurdle, if it even exists now. I can navigate my way around a city with only the vague memories of a map and some dead reconning. A car computer can do better.

#3. I think this is correct. Driving will become far more regulated. Whether you think this is good or bad probably depends in an obvious way on your existing poiltics. For my part, I agree we already bad at managing roads, and will continue to about as bad.

Never thought of Munich as a typical German city. Too wealthy, too yuppie.

'The need for exact mapping is not an eternal technological hurdle'

But it is an eternal challenge to keep up to date and accurate. As pithily illustrated by '"the map is not the territory"' or its predecessor "the map is not the thing mapped" -–territory_relation

"A car computer can do better."

If computers could do better than humans, don't you think Google would save a ;little money and skip the millimeter mapping? I will grant that Google, a Big Data company, is attacking driverless cars as a Big Data problem. However, the current crop of driverless car apps are woefully behind humans in the ability to handle unexpected situations in a way that would make them much better than a tram on its pre-laid track. I have yet to see an approach that doesn't look like a dead-end inspired by someone's graduate dissertation - kinda like XML :)

Because they are scared of lawsuits.

If we were rational, then as soon as self-driving cars got better than the worst quartile of drivers, we would let them on the road, accepting the fact that they will kill a thousand people a year. That would still save a few thousand lives a year, but that won't be seen in the courtroom.

It's a pretty graphic case of seen-versus-unseen.

It is perfectly legal to have a self-driving car now as long as you accept liability. I don't think it would make a difference in court - either you are liable or not, and the damages would be compensatory in nature.

In most traffic accidents, the driver is clearly at fault, and is of limited pockets, even with insurance.

When a Google car gets into a crash, we all know that the lawyer for the deceased's families is going to argue that Google is playing God, that they are putting their dangerous technology out there, that they are playing with the lives of our children, that we have to send a message. Even if the Google car had stopped 1000 accidents that would have otherwise occurred: those won't be in the courtroom. In the seen-versus-unseen, all that will be seen is that the Google car killing little Billy because of their arrogance.

(Even here on MR we had a commentor be openly proud of being on the wrong side of the seen-versus-unseen debate over in one of the FDA threads. Do you think the lawyer would actually be so noble to not bring this up?)

@P_A, @ScottF. So I guess there is a difference in assumptions. Yes I beleive that cars will be able to sense the road well enough that they can go around the normal buisness of locomtion much like humans and other animals do -- and if they are not doing that now, then that is only temporary.

If you don't need mapping for that part of the job, then I don't think it is such a big deal in itself, at least if your only goal is to drive about as well as humans, because all you need is a rough map, and knowledge of your approximate coordinates relative to the goal, and you will be able to choose a direction, even if it is not an optimal route.

But satnav companies are already doing better than that, by investing in laboriously updated maps.

Wow a lot of kooky theories both in the comments and in the original post.

To address Tyler's issues:

1. Driverless cars are better in that: they actually go door to do, driverless cars go between almost all two points, they are not over crowded, and the passenger always gets a seat with a good measure of elbow room. These are all valueable. We'll see how valueable in the next 5 to 10 years. I would wager that the market will show, resoundingly, how valueable.

2. The cost of the bus driver is a major cost in busses and is dominant in cabs/limos. It is very important. Further, as noted above buses are not directly comparable. Further, maps will be expanded, quickly, as more revenue goes through the system.

2. There will be more car trips but the occupants will, probably, be under less pressure to arive at their destination in a timely fashion. Further, driverless cars may well in time be less susceptible to the behaviorus that cause congestion than humans,

More interesting, IMO, since driverless cars will be engineered for safety will human drivers continually game them causing such unfairness that they do not actually make forward progress in congestion? Will we create legislation so that people that game driverless cars are ticketed, since driverless cars will have full telemetry of the event? Related: how will all cars having a tremendous number of sensors be leveraged by law enforcement? Will there be a backlash to driverless cars akin to google glass? Of course there is a tremendous of software engineering left to be done in the area, so a lot of this is, somewhat, idle speculation.

" the occupants will, probably, be under less pressure to arive at their destination in a timely fashion"" - care to explain your logic?

Many people hate driving in traffic and want the experience to end as quickly as possible.

I can think of several family vacations that would replace "drive there on the first day" with "drive there on the first night while we nap in the car."

I think the idea that driverless cars will ease congestion may be false. We currently pack more cars onto rush hour roads than is optimally safe. While some stupid-driver induced congestion will be avoided with driverless cars, speeds will decrease and spacing will increase. This will put a very hard limit on the carrying capacity of the roadways. Geography is a bitch! I suppose that is why Tyler is introducing congestion pricing (yet again!)

I've got no beef with congestion pricing, it's elegant and would be made much easier by having all cars plugged into the internet of things. I'm still optimistic that congestion will decrease though-- think about how much better it would be to take the bus, if busses came every minute, instead of every 10 or 15- and if there were three times as many routes available.

I think that's the future w/ driverless cars. No figuring out bus routes, no waiting 10 minutes, just plug your location into the app and driverless vans will come and pick you up, make a set number of stops to pick up others (more $ from you means fewer other riders)- and drop you at your destination, or close to it. Most of the congestion we see currently is from hundreds of single car drivers crowding the road. Making half those car multi-passenger would solve congestion.

Can I rent a driverless car right now anywhere in the world?

Yes - and just about any Legoland is likely to fully satisfy your need to be surrounded by the sort of people you approve of, at least when measuring those visitors by appearance.

It all depends on how much boomers are motivated to pay for "liberty". Sooner or later seniors won't be able to drive anymore. During that transition we'll know what seniors prefer: driverless cars or senior transportation services.

New York and London both have very good public transport but they still have plenty of people driving plus many taxis and now Uber. So I think driverless cars will still be very popular even there. Of course this is even more true out in the country where there might be one bus a day if you are lucky.

I actually did take some buses when I was in London recently for the first time for 30 odd years. Both the ability to quickly find a route on google maps and contactless card technologies have substantially improved the performance for occasional users like myself. So maybe public transport will be like sailing where substantial improvements in technology occurred when the steamship appeared.

'Driverless cars will give governments a chance to “redo” the whole driving side of American life. Is this so great?'

As compared to how those same governments essentially created the whole driving side of American life?

Indeed, one little-appreciated element of US history is that Manifest Destiny was in the main an effort to spread the population across distances that would require the as-yet uninvented automobile. Another is that American politicians and civil servants frequently have careers spanning centuries, thanks to funding and secret medicines from Koch/Mercatus. Wake up sheeple.

Thank you, historian. Those who forget the past can't remember the future.

Even the best public transport systems still require one to wait for a bus or train to arrive and wait at certain transfer points. There is no comparison to having your own vehicle which will take you point-to-point with no waiting time. This is also extremely valuable when it is raining heavily outside or when it is extremely hot or extremely cold and there is no climate-controlled, sheltered waiting spot. On the other hand, I think the technology will need to be refined a bit more before cars can also drive around the block looking for a spot and parking themselves. Or failing that, entering a parking garage on their own. So public transport will still be valuable in central New York, London or San Francisco where parking is scarce and self-driving cars will be more valuable in sprawling cities and suburbs.

'There is no comparison to having your own vehicle which will take you point-to-point with no waiting time.'

Obviously, you do not commute in the DC metro region, in terms of that 'no waiting time.'

How dares Ricardo to write about his personal experience and not yours? Simply outrageous.

Of course, I meant waiting for your conveyance of choice to arrive. Trains running on dedicated tracks will always be able to beat traffic but streetcars or buses that run in ordinary lanes of traffic contend with the same traffic you deal with in your car plus the time spent dropping off and picking up passengers.

The city I am most familiar with is San Francisco and I can say having a car if you are traveling between two points outside the city center is very helpful. There are a lot of trips that simply become drawn-out affairs with waiting time, transfers, and a hike or two up some big hill to get to your destination if you use the (by American standards) excellent public transit system there. Of course, if you are going to Oakland during evening rush hour (or D.C. to Arlington, etc.), you'd be crazy to drive.

Even if all of the above are true, we can run certain interstates at 150 MPH+ speed limits. For the US it will be far easier to segment existing road structure as super-fast automated car only rather than build out high speed rail. You should see the emergence of super-metros where several mid-size metros at intermediate distance become a single economic unit. E.g. Tampa+Orlando, Austin+San Antonio, Pittsburgh+Columbus, LA+San Diego, Chicago+Milwaukee, etc.

Since economic productivity scales logarithmically with metro size, this should create huge economic gains. Particularly in the medium density Midwest and Southeast where you have a number of potential super-metros that currently are just not quite close enough.

Fuel efficiency gets a lot lower at higher speeds. But I guess your time is worth a lot, so maybe it'd be worth it.

@Nathan W, drafting could be made safe by making vehicle communicate, and this would help greatly to reduce fuel consumption at higher speeds.

Going by what I know from auto racing, drafting won't help much at all. It might significantly increase MPG at 150mph, but MPG would still be terribly low. Even if you increase efficiency 30%, increasing MPG from 5 to 6.5 isn't going to do much.

some of us are, as we speak, cheerfully reading MR on our morning commute (though perhaps there's little else cheerful about riding the MBTA). it doesn't take magical tech to get drivers off the road, just meaningful investment in existing infrastructure. as to a guess on the coming regulatory structure, my bet is 'intense' and 'tilted towards corporate interests'. I wouldn't be surprised if the licenses were restricted to commercial entities, and the days of private vehicle ownership put behind us.

Right. It's not as if driverless cars will not require investment in infrastructure, whether it be embedded transmitters in road surfaces, communication channels between cars or meticulous mapping of shrubs and mailboxes.

The benefits of driverless cars in terms of safety and capacity (speed regulation, platooning etc) can only be fully realised through government mandate - i.e. banning cars with human drivers in the same area. As this isn't going to happen on a widespread scale the driverless car environment will likely be limited to city centres which are a) easier to control and b) support sufficient levels of demand. This will give rise to three outcomes, all of which can already be seen in embryo:

a) Driverless cars will have a pro-wealth bias, which is why actual and envisaged versions look a lot like upscale golf-buggies. In fact, these are essentially futuristic sedan chairs.

b) We will reintroduce the city gates of old, so that we can segregate the driverless from the drivered (that may be a neologism). Many cities already have the infrastructure for this in the form of cameras and markings. e.g. London's congestion charge zone.

c) Pedestrianisation will give way to mixed-use in which non-threatening vehicles can use the same road-space as walkers. Given that urban vehicle speeds will be no greater than 9mph (see Smeed's Law), we may see the hard distinction between road and sidewalk/pavement gradually disappear in cities, becoming the characteristic of the "outside", much as the sidewalk/pavement was once the defining feature of the city centre.

a) personally owned driverless cars will have a wealth bias, to be sure. But think about all the expenses of human driven cars-- the car, the insurance, the risk of an accident, maintenance, parking and night storage.. uber without the cost of the driver would be extremely cheap, much more so I think than owning your own car. Especially if you tolerated the car picking up 1 other person on your way to work.

I suspect we'd see passenger-less cars driving around big cities as people try to avoid expensive downtown parking.

Think about all the extra downtown real estate that would be freed up by not needing parking garages. Big bonus. And no reason for those cars to not be out picking up other ppl.. daytime cabs would be super cheap while the cars wait for the 9-5 crowd to get off

Even with no congestion pricing, you aren't going to want your car to drive in circles around town. You are going to tell it to drive 5 miles out of town and park there.

Now, on pickup there could be an issue. Someone gets out of their office at 5:25pm, but tells their car to arrive and circle the building at 5:15pm "just in case." Add this up, and toss in having everyone else say "well I need to arrive 20 minutes early because of all the other cars circling the building," and soon you get some real colossal collective action blunders.

Of course, driverless cars (and spaceships to Mars) are a distraction from the real work of the whiz kids in Silicon Valley, to fool us into believing that they are doing something worthwhile rather than real work of building the foundation for a giant security state. Cowen can prattle on about driverless cars (and spaceships to Mars), and readers actually take it seriously. He must be amused by readers' comments. Driverless cars (and spaceships to Mars), how ridiculous! Who knew that the Silicon Valley whiz kids could distract people so easily. The next thing they will be promoting is a web site where people voluntarily reveal the most intimate details of their lives. Oh, wait!

The only thing that would be worse than Facebook is writing an autobiography

3. I think you overestimate the government's power to change people's behavior. I'm guessing one argument is "The government built the Interstate Highway System and that changed people's behavior." But really that just gave people what they already wanted.

Had they instead built a network of high speed rail lines, would you argue that they had simply given people what they wanted, if it was being well used by people to get across the US? You can't argue they wanted an interstate, unless there is a viable alternative to using it that they have elected not to use.

Of course government can change people's behavior. Their preferences, on the other hand, not so much.

I wasn't referring to getting across country. I was referring to moving to the suburbs.

Then why mention the interstate?

People started moving to the suburbs before mass car ownership, once the railway lines were in place. That doesn't tell us that they wanted to drive to and from the suburbs. That they do so now, when it's the only way to actually get to and from the suburbs, tells us nothing about their preferences. If the government had instead favored railway lines, would you argue that people prefer trains to cars?

Do you not live in the U.S.? The interstate system created giant highways that ran through cities and quickened the development of the suburbs.

No, I do not live in the United States. I have visited. You appear to lack the concept of town planning.

But you haven't responded to my point. There is nothing to indicate that people *wanted* the interstate system. Also, what effects do zoning laws have on the development of suburbs? I find it difficult to believe people have the desire to spend 1-2 hours a day driving.

Driverless cars strike me as a patch for bad design. Much like how WhatsApp exploits a flaw in carriers data plans, or Uber relies upon a regulatory loophole.

"I find it difficult to believe people have the desire to spend 1-2 hours a day driving."
No, we don't don't. But we are willing to do so in order to live in a place where we can afford a house where all our children don't have to share a bedroom, a yard where are children can play, and where they can attend school with children from socio-economic backgrounds similar to their own.

"Uber relies upon a regulatory loophole."
Huh? How does that make it a patch for a bad design?

I think these predictions sound more medium-term. When the technology is still new and relatively expensive (hence why it is most economical on small busses and vans, or the cost is spread over several carpooling passengers).

Along with corporate campuses and tightly controlled areas, driverless shuttles are probably one of the first areas this technology will hit as it will fill the gap between public transportation and normal cars.

I think Tyler is being too pessimistic for the longer term. All it takes is some successful pilot projects or cities that implement this well to show the potential benefits.

Yes, the US in general is not great at land usage issues and properly valuing urban areas, but I don't see driverless technology as a partisan issue or likely to become a partisan issue.

Also, urban centers waste A LOT of space on parking. That gives them plenty of wiggle room to experiment, not get thing right, and still improve transportation on the whole.

1. Self-driving vehicles will first make their impact not on congested city streets but driving but on uncongested expressways. An expressway is a much simpler, predictable, controlled environment. Initially, self-driving vehicles won't be driverless -- the driver will be required to take over at any time if conditions become too difficult for the autonomous system (bad weather/visibility, construction zones, unclear lane markings). This phase of self-driving vehicles is sort-of already here (a few luxury cars can take the wheel for brief periods) and it may last indefinitely. A system that can handle much of the driving during road trips would be a popular feature and would sell very well, but would not be revolutionary.

2. The mapping problem isn't that maps as we commonly think of them are so hard to build and maintain, but that the 'maps' that self-driving vehicles require must include a mind-boggling level of detail in order to make up for machine vision weaknesses. Traffic lights must be mapped or the cars can't recognize them. Stationary objects are identified by being permanent map features rather than relying on error-prone real-time visual recognition of objects. But restricting them to only certain city streets would mean that they could provide door-to-door service, which would severely limit their appeal.

3. This point is way too city-centric for the U.S. Most Americans live in suburban/ex-urban/small-city environments. The car trips they take (even for commuting) are neither particularly lengthy nor unpleasant and there is generally ample parking available (mostly free of charge) when they get to their destinations. Self-driving cars would make little difference in the number or nature of trips they take. The exceptions would be children, seniors and others who cannot drive themselves. And that presumes self-driving cars getting to the point where the driver is never expected to take over, and there I remain skeptical. But this is the direction that Google has gone with their slow, friendly little buggies that seem more like autonomous enclosed golf-carts than cars.

"The maps have mind-boggling level of detail" doesn't mean "they are super-expensive to create."

Some of us in the planning world are thinking about this already - keying land use to both driverless and transportation network companies like Uber.

If and when driverless cars arrive, I think they'll displace private ownership, via Uber. When any idle car can do quick service, many more will be offered. All those cars that sit 8 hours in a workplace parking lot will be freed.

From an urban standpoint, this means a little more traffic and much less long term parking.

I'm a bit skeptical that a driverless car can work in the near future in a highly congested place where it has to share the road/compete with really aggressive human drivers, or antisocial types who are going to carjack it for the tires. Might work fairly soon in the suburbs, but not so sure about Manhattan or Newark.

The German streetcar systems are excellent, but let's eliminate the idea that we can do the same right now.

Germany is the size of California but has more than twice the population density of that state. If one looks at the former West Germany, which is really where public transportation is at its best, that country was the size of Oregon with more than TEN TIMES its population density.

Germany's population density is seven times greater than the US. Granted, much of the US is vacant land but you can see where I'm going. The success of a public transportation system, including its efficiency and quality, depend crucially on ridership, and the US will NEVER be as efficient as Germany because our population centers are less compact.

Any discussion of public transportation should also recognize the massive costs of exercising eminent domain to build them. The price government forces land owners to accept is seldom close to what that land would actually fetch on the market, particularly when one considers gamesmanship late sellers.

All that said, the benefits of driverless cars in the US would likely be far greater than in Germany precisely because of our less dense population centers and longer commutes.

As shown on Mythbusters, congestion alone can reduce commute time immensely, ironically because of SAFE operation. If only everyone could coordinate to step on the gas at the same time, congestion could be alleviated. Driverless cars will improve congestion travel because of more accurate data and faster response time than humans. But unless driverless cars are controlled by a central authority, congestion will reduce average speed because each vehicle is optimizing its own safety cushion with only a few other vehicles in mind. An optimal system would control all cars, putting aside security, computer malfunction, unexpected events, and computer self-awareness.


The optimal transport solution is tightly linked to population density.

I didn't grok your last point though: Why would driver-less cars increase congestion beyond what is status quo? Agreed they may not be as globally optimal as a centrally controlled system but still the seem better than human driven cars?

I didn't communicate it well.

Driverless cars have the potential to alleviate congestion relative to human operated cars because the computer and sensors will make more timely and better decisions about heading, bearing, speed, distance, etc. The driverless car would have a limitless attention span and no fatigue. Humans have the advantage of being able to recognize tail lights, see through the windows of cars, recognize sounds, and anticipation from pure experience. So driverless cars will take a while to surpass human capabilities, but other technologies demonstrate this superiority is inevitable. One example is the computer control of a fly-by-wire stealth aircraft. Human inputs would not be able to keep the plane controlled because its stealth shape creates un-airworthy flight.

However, the gains from driverless cars cannot be maximized without central control. For example, one driverless car will have sensors giving it telemetry data on all the cars in a 360 degree circle around it, but not much further beyond that. To maximize the benefit of driverless cars, we would have to have centralized knowledge of every car on the road. For example, if the car in front of you is slowing down, your driverless car would slow down. But if the car in front of the car in front of you slows down, there will be a reaction lag with the car in front of you and then you. Ideally, your car would know that the second car in front of you is slowing, and all cars behind it would slow at an equal rate. The same logic applies to speeding up. Without coordination, you get what's called a 'slinky effect' of traffic compressing and extending. This is not optimal. Optimality requires all cars to speed up and slow down together.

Driverless cars could reduce following distance to nearly the minimum safe distance based on road conditions, which also alleviates congestion.

Re: However, the gains from driverless cars cannot be maximized without central control.

But everyone is going their own way, and people do want, and sometimes need, to make unscheduled stops. No, control will remain localized, though information about road conditions can come from a central source.

"An optimal system would control all cars, putting aside security, computer malfunction, unexpected events, and computer self-awareness."

Yes, lets put aside all the problems. Trivia really in the New Driving Order.

My statement was a classic 'ceteris paribus' argument.

If you don't know what that means, either learn it or get the Hell off an economics blog.

I was not making any value judgment on the merits or demerits of centralized control, but by listing some potential demerits, I was demonstrating my awareness of them. Idiot.

I love your last sentence. It is optimal if all problems are merely wished away!

Classic central planning fantasy.

"congestion will reduce average speed because each vehicle is optimizing its own safety cushion "


'and the US will NEVER be as efficient as Germany because our population centers are less compact'

Well, except in the real world, that is.

Using .4 to convert between sq km and sq miles, the following American cities all have higher population densities than the most dense German city - NYC, SF, Boston, Chicago, Philadelphia, and Miami. (Some definitions are different, making this a less than perfect comparison admittedly.)

Does that detract from my general point that the success of a public transportation network, especially with fixed infrastructure, depends on population density? Don't NYC, SF, and Boston have well developed public transportation networks? Are they "successful" disregarding the inefficient pricing?

Do you see prospects for a successful commuter rail between SF and Los Angeles? Does the failure of Amtrak provide any evidence?

I'm not even discussing the problems with rent seeking by labor unions and riders that put most American transit systems in deficit.

Counter intuitive prediction: People actually like driving their cars more than a small number of intellectuals living in a couple of cities think.

"People actually like driving their cars more than a small number of intellectuals living in a couple of cities think."

Good lord yes. This fantasy of mass driver-less cars is appealing to childless people who live in about 3-4 cities.

'This fantasy of mass driver-less cars is appealing to childless people who live in about 3-4 cities.'

Yep, those people commuting two hours a day in the DC metro area love sitting in traffic, especially those drivers with children.

I think the elderly and people who do have kids and, as a result, live way out in the suburbs and exurbs with long commutes wouldn't mind it. A lot of those childless people who live in 3-4 cities already have public transportation and Uber and, in any case, have more flexibility to live near their workplaces.

Why do you think its only the childless who are pushing for driverless cars? Me and my wife want to be free of driving our kids places. And if we could skip them ever needing to drive themselves (teenagers are especially dangerous behind the wheel, because of maturity moreso than maturity) that would be an even bigger relief.

Are you talking teenagers, or younger kids. I don't see people in our risk adverse culture sticking five year olds in cars and letting them go with no adult accompaniment. Look at how criticized (or worse) people are when they let younger kids ride buses and subways, or even just so play out of line of sight.

I'm thinking the activities that 10- to -16-year-olds get involved in.

Although now that you've put the idea in my head, we *could* put our youngest in the self-driving car to go to school. Because even right now, with a human driver, there is an adult right there to help him out. The car would be internet-enabled so as parents we could monitor exactly what happens.

Older kids, sure, We even let 16 year olds drive after all. Bu ti can definitely foresee laws that will specific a minimum age to travel without adult supervision.

You're missing a crucial detail--(some) people like driving their cars when there are no other people around. When there's traffic, which there is for most trips, driving is miserable.

Well, technically I missed all of the details , on purpose, because I dint need them. For whatever set of reasons people will, on average, or at whatever percentile is necessary for The Dream to come to pass, prefer driverless cars, or they wont. I merely stated it would be counter intuitive, as all disagreement seems to support, if people actually didnt prefer them.

With respect to traffic - Traffic sucks, and no one likes it. It is simply assumed, however, that people will somehow like it more if they have one less thing to do. I am not sure they will.

BTW, I am not stating my preference. I have never been in a driverless car, so I cant possibly know.

It isn't just driving that's miserable in congestion-- it's the trip, period. It sucks as a passenger too, one reason buses aren't popular. There's no way to sugar coat that.

I suspect that one of the things that will come along with driverless cars is automated, precise enforcement of traffic laws. Driving will probably be a lot less fun once you get a speeding ticket every time you drive 1 mph faster in a 25 mph zone. (and many people might end up eventually losing their licenses anyway if every traffic law infraction is caught).

This would indeed make driving impossible. I have know idea what the likelihood is.

We could enforce more laws now, for example on toll roads and chose not too. I would think most traffic laws would be repealed (which is the direction things are going anyway) and people would mess with the automated systems as needed (by assigning 10MM tickets to each rep who voted for the enforcement). Easy hack.

"People actually like driving their cars more than a small number of intellectuals living in a couple of cities think."

Do the elderly and disabled fit into your worldview at all?

I am not expressing a world view. I am quite literally making a counter intuitive prediction. Of all the possible impediments to this Awesome New World everyone is so sure will come to pass - technology just isnt there, political will, regulatory failure, economics, crap I cant even think of, etc - wouldnt it be funny if the ultimate downfall is due to the one thing no really Smart Person would have ever guessed? I do.

If driverless cars work as well as the European methods of public transportation, I will be quite happy with them.

My argument is much simpler: I hate driving. So driverless cars will be extremely beneficial according to my values.

This is not Europe. People do not use public transportation in most places and what there is horrible.

2 point
1. I am right now (figuratively) trying to pry the car keys from my 89 year old mother, and boy would a self driving car help!
2. For my self, I do not mind driving to work but would love to have a car that can self drive on the highway so I could take long trips at night while sleeping.

Self driving city vehicles could be tiny. I like he BMW c1 200.
If city roads get 2 congested why couldn't the gas/millage tax be raised and multi-story roads built?

Here is another very small vehicle:

Why does it have to be self-driving to be tiny?

"Why does it have to be self-driving to be tiny?"

It doesn't have to be self-driving to be tiny, but self-driving will greatly help cars to be much smaller. For example, if you buy an electric car, you might buy a car with 100+ mile range, even if 90+% of your daily driving was less than 100 miles. But if you're ordering a self-driven car to travel say, 10 miles, the car that came to pick one up might just have a 15 miles left in the batteries, because the car would know you're only going 10 miles.

Perhaps even more importantly, the car that you got might only have a single seat, with little or no trunk, if the car knew you were traveling alone and had no luggage.

Also, once computer-driven cars are alone on the road, all the extra safety things like bumpers, crumple zones, and air bags can be eliminated, and the weight can be dramatically reduced.

Finally, in a situation like this one (I think in NYC), you can see the insanity of all those large cabs carrying single passengers. Imagine cars that are only single-seat wide, and less than 6 feet long. I'd bet in that situation the road capacity, in terms of *passengers* per hour crossing a particular line across the road, could easily be 5-10 times as high.

Too dangerous otherwise

Alternate take on autonomous cars:

While this is often predicted as a panacea for suburban commuters, I predict that self-driving cars won't turn out like this. I see the technology being deployed first in long haul trucking. From there it first comes to car commuters as an expensive option to be used for expressway driving. This is more like using autopilot for the boring parts rather than something that will take you door to door. Like a really fancy cruise control. Not exactly a game changer.

Because self-driving will come gradually, the effects will be gradual. Perhaps in 20-30 years the technology will be A) fully implemented for door to door self-driving and B) in enough passenger vehicles that this reality will come to pass.

In that case I think it will be a real boon to cities (and their immediate suburbs). It will mean driverless taxis. Poor Uber drivers will lose work to these cheaper (and safer) driverless cars. It will make living in cities more attractive. It will make taking a bus something that is not needed. So cities will cut back on bus service and bus congestion. It also means taking trains will be more attractive as you can just take a commuter self driving taxi to the station. People will still take trains because they are cheaper than taxis and interurban trains will still be faster than driving, which will get slower still because of increased congestion. It will also mean less need for parking so you will see future development becoming denser, as parking requirements is a large driver of sprawl. Expect a lot of urban activists to lament the end of buses because it will mean less transit options for poor people. But conversely it will save cities a lot of money. Some cities will propose driverless buses. Unions prevent this from happening, citing safety concerns.

Forward thinking cities will put beacons on all street signs that communicate to driverless cars information like construction zone, detour, one way, etc. Backwards cities will not, and this will be something people talk about when evaluating cities.

So there is my alternative "hot take" on driverless cars: Good for cities, and it will make suburbs denser.

All this assumes, of course, that all the thorny issues with AI are worked out. I don't believe all the talk about Moore's Law will bringing AI. Frankly, I think those comments are ignorant. It is not a hardware issue but rather a software problem. Unfortunately until machines become intelligent we are stuck with human software engineers who are quite limited in their ability relative to the challenge. So there is a Catch-22 that may or may not ever be resolved.

"Unfortunately until machines become intelligent we are stuck with human software engineers who are quite limited in their ability relative to the challenge"

How are software engineers with there "limited" ability going to write the software to make machines intelligent? It's Catch-22 squared!

I think it will take decades for self driving cars to be come prevalent and there will multiple different approaches. Any predictions made now are certainly wrong.

These are reasonable predictions.

In the descriptions I've seen of driverless systems, there are networks of location and control nodes - essentially road lines, signs, and signals.

Mythbusters demonstrated the superiority of a roundabout to the four-way stop with human drivers. Assuming their conclusion was correct, I wonder if having an intersection without stops would be even faster. By that, I mean the driverless cars in all four directions never stop, but rather they adjust spacing and speed to weave through each other like threads in a fabric.

This is probably true but it will take a long time before driverless cars become enough of a minority to make these sorts of changes.

Forward thinking cities will put beacons on all street signs that communicate to driverless cars information like construction zone, detour, one way, etc.

You don't need to put this information on beacons. You just have your city planner mail the construction information to Google and Apple and Tesla, who broadcast it out to their cars automatically.

City-placed beacons will still be very useful for positioning within cities.

Beacons are better because emails require a large degree of bureaucratic cooperation. Beacons can also be used for impromptu things like a brief lane closure or rerouting traffic for 2 hours because of a concert or accident.

"I see the technology being deployed first in long haul trucking."

I think you're right about that being a potential first entry point. Computer-driven trucks can go 24-7, as compared to whatever legal limits there are on human drivers (to avoid excessive fatigue). And the capital cost of a tractor-trailer is pretty high anyway, so the autonomous driving adds less cost. Finally, highways are easier to drive.

There's no reason why a driverless car need be a single-user vehicle; doesn't Uber already offer shared rides? Expanding shared-ride vehicle size to passenger-van or minibus size might provide a more flexible and lower cost approach to public transit than the present massive, bureaucratic, union-ridden systems used today.

Presumably some regulation would be required for public safety (although regulated vehicles might co-exist with unregulated gypsy jitneys), but if Uber could crack the ancient cab cartel in city after city then perhaps entrepreneurs will find a way to challenge long-ossified public transit authorities?

"There’s no reason why a driverless car need be a single-user vehicle; doesn’t Uber already offer shared rides? Expanding shared-ride vehicle size to passenger-van or minibus size might provide a more flexible and lower cost approach to public transit than the present massive, bureaucratic, union-ridden systems used today."

Yes, let's look at alternatives when going from Raleigh, NC to Charlotte, NC at present, and 30 years into the future. The distance is approximately 170 miles.At present we have (approximate mid-point pricing taken from the website)

Train = 3hrs 30 min, $44, once per day
Bus = 5hrs 10 min, $38
Car = 2 hrs 47 min, $22 (Editor's note: This seems ridiculously inexpensive...the cost should be approximately 50 cents per mile, or $80. The $22 is probably just gasoline cost.)
Air = 3 hrs 3 min, $200

Obviously, the car looks best from the standpoint of time and cost, using the ridiculously low value of $22. But let's say it was a more realistic $80 (about 50 cents per mile). Then the choice would be less clear. At present though, there is the significant problem of getting to and from the airport/train station/bus station, which the car avoids. (But driving the car also allows no free time.)

Now, 30 years into the future, we might have:

Train = 3hrs 30 min, $44, once per day (but no problem getting to and from the train station).
Bus = 1hr 45 min, $15 -->Yes, under 10 cents per mile might be realistic, and an average speed of 100 mph would be realistic.
Minivan = 1hr 40 min, $35-->Faster than a bus, because no stops, and shorter passenger loading/unloading times. At 20 cents per mile, cost is between a bus and a car.
Car = 1hr 30 min, $50-->Probably the fastest, at 120 mph average speed. But more like 30 cents per mile, because going this fast really hurts energy efficiency.
Air = 3 hrs 3 min, $200-->Conventional airline.
Air = 45 min, $140-->A VTOL aircraft not using either RDU or CLT airports, but instead taking off from the parking lot of a deserted shopping mall.

When looking 30 years into the future, a bus, minivan, and car are all close choices, and the train and non-VTOL airplane are left in the dust.

You have unrealistic expectations of the speed of the bus, and assume that the train remains the same as today - which is woefully slow for an intercity train as it is. If it was being run by Virgin, using Pendolino's, and assuming the track is maintained and there is signaling, the trip time would be under 1.5 hours. On the flip side, you'd probably be paying about $60, but that would still make it by far the best option.

Now, redo your model, allowing for a modern train service and without assuming speeds appropriate for rail will be used on the highway. It's quite clear that trains and VTOL aircraft leave everything else in the dust. Amazing what a modern rail service is capable of.

"You have unrealistic expectations of the speed of the bus."

Why do you think an computer-driven bus can't average 100 mph between Raleigh and Charlotte? It's reasonably flat land, so there aren't mountains and curves as there are in the Appalachians. Pretty much *everybody* is going to be going 100+ mph on interstates that are truly between cities on relatively flat land.

"...and assume that the train remains the same as today – which is woefully slow for an intercity train as it is."

What is going to be revolutionary in U.S. train travel in the coming, say, 10-30 years?

By 10 years (2025), I expect at least a few mid-priced Level 4 (completely autonomous) cars in the U.S. And computer-driven intercity buses may make even *more* sense than cars, because the buses could drive almost exclusively on highways. And in 30 years, I envision virtually every vehicle on the roads being completely autonomous.

I do see that there are ongoing attempts to restore double-tracking between Raleigh and Charlotte. From wonderful Wikipedia:

"NCDOT has worked with NS, CSX, and the NCRR to restore the double-tracking and make other incremental upgrades, a process that reduced the travel time between Raleigh and Charlotte by 35 minutes from 2001 to 2010. Since 2010, work has been proceeding under the Piedmont Improvement Project, funded by a $520 million grant under the 2009 ARRA stimulus.[6] The PIP projects include restoring complete double-track from Greensboro to Charlotte, straightening several curves, closing crosses, and building bridges to separate train and highway movements, and are all scheduled to be completed by 2017."

Other current parts of the Internet say the time from Raleigh to Charlotte is about 3hr 30min, so apparently the time in 2001 was more like 4 hours.

"If it was being run by Virgin, using Pendolino’s, and assuming the track is maintained and there is signaling, the trip time would be under 1.5 hours."

This seems much more speculative than my expectation of a computer-driven bus (within a nearly entirely autonomous vehicle fleet) taking 1.75 hours, unless you know something about Virgin's plans. But I do expect computer-driven cars might help out intercity trains, by eliminating the hassle of taking a car to the originating train station, and having to find a car at the destination station.

I'm not suggesting Virgin are planning to move into the American rail market; to my knowledge, they wouldn't be able to if they want to, because you don't have a railway franchise system. I'm merely pointing out that trains would be competitive if the US had a railway system on par with other developed countries.

On a related note, how would things stand if rail was subsidised to the same extent as roads; that is, if the government(s) spent the same amount on maintaining and expanding the rail system as they do on the road system?

At the current state of affairs, what would be revolutionary in US rail would be the US catching up to the rest of the developed world. Other than that, automation (which has already been developed, it just needs rolling out) that allow day to day operation to be done by a few people in an office watching screens and ready to respond if there's a problem, and general improvements in the efficiency of the carriages themselves. Maybe a program to have small city runabouts for rent at each station as well, which may or may not be self driving.

"I’m merely pointing out that trains would be competitive if the US had a railway system on par with other developed countries."

The question is, "What other developed countries are comparable to the U.S.?" The U.S. is a lot less densely populated than most other major developed countries. It's been 40 years since I lived in Germany and Spain when I was a kid. But I definitely remember cities being much more compact. There weren't any real suburbs, or "bedroom cities" (small cities that are mostly residential, aka, Cary NC, Farmington Hills MI, etc.)Plus, even going out to the country level, there are more people per unit area. Even with the former East Germany, the combined Germany is 230 people/sq km, versus the U.S. at 30 people/sq km.

"On a related note, how would things stand if rail was subsidised to the same extent as roads; that is, if the government(s) spent the same amount on maintaining and expanding the rail system as they do on the road system?"

It seems to me the better comparison would be, "How would things stand if railroads were subsidized to the same extent per passenger mile and freight ton-mile as the road system?" I don't have any idea what the answer would be to that question, but I think it's a better question. (Although there's probably been a significant historical bias in favor of the road system.)

"Maybe a program to have small city runabouts for rent at each station as well, which may or may not be self driving."

Self-driving runabouts eliminate the hassles of insurance (bad drivers wrecking the car) and eliminate the need to return the car to the station. I really do think self-driving cars will help trains, but self-driving to greatly improve cars and buses, too, so all forms of transportation are going to need to "up their game." And I see passenger computer-piloted VTOL as being very dangerous to conventional planes and trains in the 200-600 mile range.

So I could see how trains might never get a firm foothold in the U.S. (That's path dependence for you.) Except for freight, with which they're already #2 to trucks.

Automatic parking first. Then freeway driving.

Eventually people get so used to it, we finish the rest off.

Nobody is worried about planes on autopilot anymore, but I bet they were when it first started.

Big empty sky. Big difference.

Related: Should A Self-Driving Car Kill Its Passengers In A “Greater Good” Scenario?

Hah, the Trolley Problem!

Good one.

Hi Tyler,

Hi Tyler,

Lots of ridiculous claims in this post. Needless to say, I beg to differ. ;-)

"Driverless cars are not actually much better than the really good German streetcar systems."

Is this comparing driverless cars as they are now, or a driverless vehicle *system* as it will be in about 30 years? A driverless vehicle *system* will absolutely clobber a really good German streetcar system. Let's say 70,000 folks want to get to the Allianz Arena for an FC Bayern Munich Fussball (aka, soccer) game one fine Saturday?

Alianz Arena

You really think that the *streetcar* system is going to get them all there in the best manner? Fahgeddaboutit! The streetcar system isn't even available on all the Bundesautobahn 9 and 99 that are right near the stadium. Instead, picture autonomous double-decker buses coming from Ulm, Augsburg...even Austria, to deal with the crowd (because on that particular day, there are available buses in Ulm, Augsburg, etc.). And these buses travel easily 100+ mph on major roads. What streetcar travels even at 50 mph? Plus, the buses deliver and pick up from ***right at the entrance to the arena***. In 30 years, 10s of millions of Germans are going to over 80 years old, and getting and off a streetcar and getting to and from the Arena to the streetcar can be a huge deal for a person who has trouble walking and standing.

It's not even close. Next objection.

"2. The need for exact mapping of streets will restrict driverless vehicles to well-known, well-trodden paths, much like bus lines."

Again, are we talking about today's driverless vehicle, or what they will be in only 10+ years? In 10+ years, the cost of a *terabyte* flash drive will be under $50. (Probably well under, but I'm too lazy to check for an exact number.) And cars with *terabytes* of memory will be able to transfer that information between each other in less than a minute. Further, since people won't own cars, the fleet managers can make sure that all vehicles only drive on the roads that they "know." And by "know," I don't mean like a human "knows" a road. I mean details down to inches.

"Or worse yet — some automobile lanes may be turned over to municipal driverless vehicles in a way which makes traffic problems worse."

No, kicking all human drivers off roads can't come soon enough. Imagine cars passing by each other at 90 degree angles at intersections, because there are **no stoplights**. There's no need. Every car knows exactly where every other car is within 500 yards, down to the foot. Imagine cars traveling at 50+ mph on streets, separated by less than a car length, in complete safety (because all the cars know *exactly* what the other cars will do). And I'm talking about street that now have speed limits of 35 mph or less.

"3. Driverless cars will give governments a chance to “redo” the whole driving side of American life. Is this so great?"

Well, do you like deaths going from 33,000 per year down to less than 3,000 (and probably less than 300)? How about a reduction in the average cost per mile from roughly 60 cents to under 15 cents? Congestion cut by 90+%? Everyone having the option to work or play in their vehicles? (Not "their" vehicles...because virtually no one will own a vehicle...people will just tell their cell phones where they want to go and at what time, and the vehicle will arrive at their door.)

"Most generally, your predictions for driverless cars should depend heavily upon: a) will there be rational congestion pricing?, and b) how rapidly will cities rezone to take advantage of the new opportunities? I am not sure we should be especially optimistic about either a) or b)."

Governments, schmotherments. Technology will rule the future, not governments. (Trust me...I'm an engineer. :-))

P.S. If you want to talk about the "bad" part of driverless vehicles, it will be the absolute obliteration of brick-and-mortar stores. See that nearby Walmart, Target, Kroger, Home Depot, Lowes, Walgreens, etc.? They'll all be gone in 30 years. And as Bruce Springsteen so poignantly sang, "Foreman said, 'These jobs are goin' boys, and they ain't coming back...'"

Why would driverless cars eliminate brick and mortar stores? These cars could eliminate or reduce inefficient parking lots, reducing fixed costs. They would also reduce the burden of driving through traffic to shopping centers.

I can think of several reasons off the top of my head why driverless cars save B&M, and one where they are worse off: reduced delivery costs for non-B&M stores. But couldn't B&M restructure for in-house shopping and home deliveries?

"I can think of several reasons off the top of my head why driverless cars save B&M, and one where they are worse off: reduced delivery costs for non-B&M stores. But couldn’t B&M restructure for in-house shopping and home deliveries?"

This is an extremely important question. It's important not just nationally, but globally. Globally, literally tens of millions of jobs and probably trillions of dollars will ride on the answer. This is why I think online shopping with delivery by computer-driven vehicles will eliminate brick-and-mortar stores of all types:

1) The time savings for customers. Essentially every week, I spend about an hour (total time, including driving to and from the store, walking through the store, and waiting in the checkout line) doing grocery shopping. I could literally do my grocery shopping in a few minutes with online grocery shopping. I could even have a button with, “This is what I think you want this week, based on your purchases of the last weeks” and hit that button in seconds.

2) The reduced capital and operating costs of stores.Look at the well-lit and glass-doored frozen foods aisles of grocery stores. Look at the store ceilings that extend tens of feet above the products. Feel the air conditioning and see the lighting. All that could be totally with a warehouse that is serviced by robots. No wide aisles, no lights, no air conditioning.

3) The energy efficiency, and transportation efficiency.It makes much more sense for goods to go to a spartan warehouse (in which only robots can really work, because it’s hot or cold and dark. Then the goods are delivered from that central location to the neighborhood, rather than individual people coming to the store.

4) Personalized choices.You might think a big store with a large inventory is the way to go for personalized choices. But how many times have you shopped for clothes, and the store had only the wrong sizes? With warehouses, they can even deliver multiple sizes that they think are close, you try them on in your home, then return the ones you don’t want. Plus, they always have something very close to your size, even if it has to be shipped from another warehouse hundreds of miles away. You wouldn’t go hundreds of miles just to shop for clothes or shoes.

Those are some reasons I think driverless cars will completely eliminate brick-and-mortar stores (and all the jobs that go to building and running them).

You're neither a teenager nor a woman, right? For those people, grocery shopping may be a hassle. But they frequently go to the mall purely for its entertainment value, like we men love to repair our vehicles for fun.

No, but I'm 'N Sync with all that stuff. I know all the latest bands: Old Kids on the Block, Lady GooGoo, Taylor Perry-Swift.

Seriously, I go to the food court at the local mall sometimes, and might pick up something (very occasionally) while doing that. The question is, what is the total amount of shopping done at malls, and how much of even that shopping might instead be done online?

My guess is that the total amount of shopping at malls, versus other stores that aren't "meeting places" (for instance, Kroger/Walmart/Walgreens/Home Depot) is pretty low. So I'm not saying every single brick and mortar store will be gone, but I could easily see 80-90%+ being gone in 30 years.

The things white people will dream up to avoid public transportation ...

The things white people will dream up to avoid public transportation.
It's not the public transportation we are avoiding but you knew that.

I lol'd. But to be fair, I don't think most of us are avoiding the minorities on the public transportation, most of us are avoiding being associated w/ those minorities. Subtle difference.. if we could ride the bus and be sure that our own status wouldn't suffer, being around the other bus riders would be no issue.

I know some white enviro types that take the bus.. but they make sure to broadcast WHY they do (not because I'm poor!) very loudly.. in some cases I think it even captures extra social status for them...

The Google Bus's biggest PR problem is that it doesn't pick up poor people.

Since no one will address the elephant in the room we have to work around it.

I, like many, get carsick if the roads are twisty. But, like the majority of those who get carsick, only if I am not driving. If I am driving, no problem. Where I live (Santa Cruz)most of the roads are twisty even in town.

The fundamental point that governments do not manage transport well hints that it will likely screw up the autopilot era, too. The good news is that it will be harder to do because of the inherent autonomy involved. Uber is breaking government control already.

More specifically, some 1/3 of urban driving involves parking. That can go away. Convoying and other close packing can reduce highway congestion. We only need a little luck.

Instead of 'Busses without the drivers' I'm not sure why we shouldn't expect 'Uber without the drivers'. All it takes is either a) one big player with a large fleet (aka Hertz) or a combo of lots of little players with some market maker such as uber

Even if all we get is a reduction in traffic deaths and a big increase in congestion, it could still be a substantial utility win. I do not see a strong argument against the proposition that traffic deaths will decline precipitously.

I am consistently puzzled by the concept that 'Uber reduces energy demand', the advent of Uber is almost certainly increasing demand for fossil fuels (at least first order, we could argue about saved energy costs second order if Uber reduces the requisite aggregate vehicle fleet or encourages sub-urban to urban migration). Uber reduces a lot of things, it reduces the number of vehicles required in an economy, it reduces the amount of time I spend trying to hail a cab.

But, at a very micro-level, Uber is increasing vehicle miles traveled in urban setttings for developed economies; it is no accident that U.S. gasoline demand has seen is strongest YoY growth in a decade at the same time Uber is reaching a critical mass.

I am the marginal consumer of gasoline, I live in a city and have a utility curve determined by cost / time of travel / convenience between three modes of transport, a bike, a bus and a cab/uber (in order of fuel consumption), By reducing the price to me and increasing the convenience of the most fuel intensive transport option I am now choosing to uber/cab far more frequently than I would have when economic moats limited the cab supply 5 years ago.

1. There's more to public transit than door-to-door service or close to it. How good are we at stamping out harassment -- by creeps, beggars, freelance preachers, etc. -- on public transit?

Comments for this post are closed