In praise of driverless cars, don’t regulate them into oblivion

My column is here, one excerpt is this:

The benefits of driverless cars are potentially significant. The typical American spends an average of roughly 100 hours a year in traffic; imagine using that time in better ways — by working or just having fun. The irksome burden of commuting might be lessened considerably. Furthermore, computer-driven cars could allow for tighter packing of vehicles on the road, which would speed traffic times and allow a given road or city to handle more cars. Trips to transport goods might dispense with drivers altogether, and rental cars could routinely pick up customers…

The point is not that such cars could be on the road in large numbers tomorrow, but that we ought to give the cars — and other potential innovations — a fair shot so that a prototype can become a commercial product someday. Michael Mandel, an economist with the Progressive Policy Institute, compares government regulation of innovation to the accumulation of pebbles in a stream. At some point too many pebbles block off the water flow, yet no single pebble is to blame for the slowdown. Right now the pebbles are limiting investment in future innovation.

A few points:

1. I couldn’t fit it in the column, but it is an interesting question why there is no popular movement to encourage driverless cars.  Commuting costs are very high and borne by many people.  (Here is Annie Lowery on just how bad commutes can be.)  You can get people to hate plastic bags, or worry about a birth certificate, but they won’t send a “pro-driverless car” postcard to their representatives.  The political movement has many potential beneficiaries but few natural constituencies.  (Why?  Does it fail to connect to an us vs. them struggle?)  This is an underrated source of bias in political outcomes.

2. In the longer run a lot of driverless cars would be very small.  Imagine your little mini-car zipping out and bringing you back some Sichuan braised fish, piping hot.

3. If a traffic situation gets really hairy, the driverless car can be programmed to pull over and stop.  Oddly I think that perfecting the GPS system might be a trickier problem than making them safer than driver-run cars.  Computers don’t drink, but they will drive around the same block forever and ever if they don’t understand the construction situation.  Even the best chess-playing computers don’t very well “understand” blockaded positions and perpetual check.

4. This isn’t a column about driverless cars at all.  It’s about our ambivalent attitudes toward major innovations.  It’s also about how the true costs of regulation are often hidden.  A lot of potentially good innovations never even reach our eyes and ears as concepts, much less realities.  They don’t have tags comparable to that of the driverless car.

5. Via Michelle Dawson, here is a list of driverless trains.  Here are links on robot-guided surgery.


I can see the benefits, but driverless cars might also greatly increase sprawl in major urban areas as well as car and gasoline usage. For example, any daily commute over 10 miles into DC can be pretty hellish but would be less so if you didn't have to be actively engaged the entire time. I would hate to see urban revitalization as well as the accompanying environmental and other benefits stalled by something as simple as easier commutes from the suburbs.

Yes - driverless cars will enable radical exurbanism, but isn't sprawl what consumers have been demanding for 60 years? Also, driverless cars are more likely to benefit urban consumers first, with rapid proliferation of cheap, driverless taxis -- won' that make living in the cities easier and more attractive? And on gasoline usage, driverless cars are a great candidate for electric power: they can go charge themselves somewhere after dropping you off.

Yes – driverless cars will enable radical exurbanism, but isn’t sprawl what consumers have been demanding for 60 years?

No, it's what zoning boards have been demanding for 60 years. Lots of consumers would love to live in dense areas near good transit, but it's illegal to build them. The current status of housing is neither free, nor free-market.

Other major drivers of sprawl include tax avoidance, and an attempt to re (class) segregate schools by segregating housing (this is part of why suburban zoning codes require such low density -- it makes for more expensive housing that keeps the wrong people and their children out of the school district).

Oh, and I left out parking, which driverless cars need much less of (or can have it put anywhere in a reasonable distance, rather than right next to the home/office/other destination).

"Lots of consumers would love to live in dense areas near good transit, but it’s illegal to build them. "

-- you don't really have any basis for this Richard Florida-esque assertion. Demand for "Americana" -- big house on a big lot in the burbs with safe neighborhoods, good schools, "ice cream socials", and all that is very high.

Sure you do. Look at property values in walkable urban areas.

Exactly. High demand is generally demonstrated by high prices... basic economics.

Sprawlsvilles are cheap. Low demand.

"Sprawl" is just a term of prejudice against people being able to live comfortably. We're entitled to all the sprawl that the market will provide. The need for more and better homes and more enjoyable living hasn't ended just because you've got yours.

Willing to abolish zoning laws so we can have all the urban highrises the market will provide?

Didn't think so, hypocrite.

Just some speculation on my part:

A great benefit of the self-driving car, is that it should bring an end to the 'self-owned car' - which is a horribly inefficient way for society to distribute expensive machinery.

Already, the internet has allowed companies such as i-go or zipcar, to be profitable in urban areas - despite the numerous inefficiencies of the system. Currently you have to know you are using a car hours or days in advance - then you have to somehow get yourself to the car.

Now imagine a city of 30k, and add 8000 self-driving cars under the zipcar model. You decide to take a trip - so you pick up your iphone and hit the "request car button". The phone gps knows where your car is located, and by the time you've put on your shoes, a car has driven around the corner and is sitting in your driveway. Put in your destination on the car gps, and you're on your way.

The efficiency gains will be spectacular. Almost all American families own at least one van/truck/SUV - despite the fact that 98% of the time, the extra space is entirely unnecessary. Under the shared car model - some number (perhaps 80%) of the vehicles would fit one passenger only. You can easily request much larger vehicles, if your needs demand it.

These developments could also reshape cities. Parking centers will barely be necessary in downtown areas (electronic cars can position themselves into extremely dense clusters - since we don't care about what order the cars come out in).

Yes, there could be huge efficiencies from an automated "Zipcar" model. Instead of jack-of-all-trades car, you could call for precisely the car you needed for a particular task. This will be a boon to electric cars that recharge on their own and "specialize" in the vast majority of trips that are less than 50 miles.

I agree with Tyler that the biggest barrier to adoption is regulatory (and cultural: see

Adoption will be much faster if proponents can identify an incremental adoption path rather than all-or-nothing. For instance, auto-parking, very intelligent cruise control for highways, driverless lanes on frequent commuting routes, etc.

Regulation is the biggest challenge, but what about the regulatory challenge from cab driver unions?

I'm in favor of driverless cars, but I'm also in the 15% of drivers who know themselves to be worse-than-average. So I don't worry about the computer getting into an accident I wouldn't have gotten into myself.

I don't consider myself worse than average. (My insurance company agrees with me, so I'm comfortable with my opinion.)
However I don't consider myself better than a computer. So I'm fine with the idea too.

I'm not sure how I normally rank, but I know that when I am sick and/or tired I am below average.

I would have no problems trusting my life to a computer-driven car.

1. How 'bout ending aging? It's killed more people than government. (see #4)

"It is an interesting question why there is no popular movement to encourage driverless cars."

Perhaps there just aren't that many people who are intellectually insane. It is, after all, perfectly possible to be very, very smart and yet love a few very stupid ideas.

I think Americans do not realize how costly car accidents are to society. Between injuries/death, loss of work, medical bills, rehab, car damage, and encouraging lawsuit culture it adds up to quite a bit. It would be significant savings if we could have driverless cars, or better yet, start working on improving transit options.

Could it be because we're suspicious of driverless cars (cf. illusion of control)?

Quite possibly, yes.

Guaranteed. Thanks to the fact that people underrate risks where they're "in control" and overrate risks where they're not, and *even more* than that, people overrate risks where *no humans* are in control and underrate risks where *any humans* are in control, driverless cars are not going to be generally accepted for 50 years or more.

Look how long it took people to accept driverless *trains*. You still can't build them unless they're fully grade-separated.

There's no chance driverless cars will be accepted until they're 10 times better than very good drivers, because of these psychological behaviors.

There are so many things that are being regulated into oblivion. This is a nice, visible example of something that is just around the corner, but as you say, there are lots of other, unseen things, too.


our ambivalent attitudes toward major innovations

Most things hyped as major innovations are bunks. Given human nature and the way our economy works, there is no way around it. So it makes all the sense in the world to be skeptical of major innovations.

Long time reader, first time poster.

Strongly agree with Tim, so much that I wrote my senator and congressman about this this morning.

So no one has yet pointed out the simplest and most obvious solution to long commutes - having people live close to where they work. Concepts like Traditional Neighborhood Design, New Urbanism and walkable neighborhoods would elimnate the need for long commutes and improve the myriad problems associated with commuting detailed in the Slate article - obesity, lonliness, anxiety, depression, divorce, etc. Commuting wastes vast amounts of time and, more importantly, energy, on usless activities. How could any economist interested in efficiency not see this as a disaster? We have the most wasteful arrangement of space in the world. Numerous studies by urban planners over the years have shown the superiority of New Urbanism and commentators like James Howard Kunstler have tirelessly argued the merits of such concepts in print.

Driverless cars are a ridculous solution to our true problem - they will continue to waste our time and keep us dependent upon dwindling oil supplies. I think it would be better for an economist to take a good hard look at all the ways in which we subsidize suburban sprawl and make it easy for people to live vast distances from work and shopping -issues such as how we fund our schools (local property taxes so that 'good' schools are in the periphore), to subsidies for road builders making it easy to live far away. Running utilites farther and farther out year after year is incredibly wasteful. It is well-known that suburbanites do not pay the true costs of their way of life - it is borne by all of us. This, I think, is the real issue worth examining from an economic standpoint.

Driverless cars probably are invevitable if the technology is there. In theory, this is probaly a good thing - less human error and such - but as usual, economists have no answer for all the workers displaced by this innovation, just as they ignore all automation, insisting that some sort of 'innovation' will come alone and provide jobs for displaced workers. How did economic 'science' become simply looking at what happened in the past and assuming without evidence that it will always happen again? Why aren't economists dealing with this, and if they do not, what good are they? In economics, everything is interrelated; it would be nice if economists started looking at the bigger picture.

1) If the driver need not focus on the drive she is free not to waste her time. She can catch up on calls, do email etc.
2) Work is becoming more temporary and short-term and so living close to one's workplace is getting harder. In principle you are right but clean energy plus self-driving cars is a more realistic solution.
3) I don't think we should create jobs that are unnecessary just because society is structured so inefficiently that those people can't be retrained. Freeing people up to do more useful/interesting work is a good thing. The fact we think sometimes it isn't is a fault with the way we organise our societies.

Your first paragraph confuses the concept of shifting a curve and simply moving to a different point on an existing curve. Already people make fully free decisions on where to live based on what they should: relative prices. If you want to live close to the city, you can certainly do so - at a cost (both monetary in the sense of higher housing prices and non-monetary among many dimensions relating to quality of life). Each consumer makes this decision, and those living in the suburbs have implicitly chosen that the additional cost of commuting. The point is that introduction of driverless cars completely *shifts the entire curve* - the trade-off still exists, but the cost of commuting has gone down significantly. You're saying that this is what might cause even more urban sprawl, which may be true - but this is somewhat circular, since if the costs are the real problem with urban sprawl, and the costs are going down, then the whole negative concept around "urban sprawl" has to be re-evaluated.

I think it rather cavalier of you - or anyone - to simply make proclamations about what people should and should not do just because you don't like seeing urban sprawl. I'm all for eliminating subsidies (and also substantially increasing carbon/gas taxes), but that's for correcting for externalities, not my imposition of lifestyle choices on my fellow citizens.

I'm not even going to attempt to address your concerns about displaced workers. This is the same kind of reasoning (or lack thereof) which supports protectionism and other bad policies which confuse the seen with the unseen.

"There is only one difference between a bad economist and a good one: the bad economist confines himself to the visible effect; the good economist takes into account both the effect that can be seen and those effects that must be foreseen." - Bastiat

If housing markets work homes would cost the most in the locations providing the highest quality of life. If living somewhere cost "(both monetary in the sense of higher housing prices and non-monetary among many dimensions relating to quality of life)" is true it means there has been a failure in our housing markets . I wonder if people who live in others countries would make such a statement.

Housing DOES cost the most in areas with the highest quality of life. What Ben was saying is that if you want to pay less for housing, then you have to pay additional cost in the form of lower quality of life. So if you work in NYC you can get a house in NE Pennsylvania quite cheap, but your commute to NYC will drag down your QOL quite alot. On the other hand you can pay much more and live in Westchester, and have a 40 min commute by train.

Problem is that housing markets are currently regulated into oblivion to favor car-driven transit and suburban sprawl. E.g. height restrictions, zoning restrictions, minimum parking requirements, free highway access but paid transit acces, etc.

In principle you're right but in practice we've got such heavy subsidies and regulations in favor of the car-and-suburb model that you actually need counter-regulation back the other way to push things into balance (assuming deregulation is offf the table for NIMBY reasons)

Every city since the proliferation of the automobile has been built around automobiles -- nothing would be different in the absence of regulation.

That's an evidence-free claim with no evidence to support it. Evidence?

Chad - consumers make the choice to have a long commute from a big house in a good school district. With the invention of driverless cars, not that much time is wasted. I could even envision people telecommuting-while-commuting.

Hell is other people. I moved to a city for reasons of transportation, I would prefer to be back in the suburbs.

I said back in November: "Pundits don’t seem to realize just how big a deal this is – it could let cities be roughly twice as big, all else equal."

“Pundits don’t seem to realize just how big a deal this is – it could let cities be roughly twice as big, all else equal.”

Assuming the economy lets us use between twice and four times as much automobile fuel per year. Etc.

“There is only one difference between a bad economist and a good one: the bad economist confines himself to the visible effect; the good economist takes into account both the effect that can be seen and those effects that must be hypothesized.”

If you had safe, on-demand riderless cars, you could probably make up for most of the additional person-miles traveled by having the majority of those on-demand riderless cars be small, fuel-efficient ones, and saving SUVs for the time that you actually need SUVs.

And, to be clear, I'm not necessarily talking hybrids or ultra-compact cars, here: a Honda Civic uses half as much fuel as a Honda Pilot, and the Civic, while a "small" car, is by no means a tiny car.

My computer-driven car could also join up to trains of cars. If I don't have to pay attention to my commute, I'm willing to trade a little more time for a cheaper ride.

"And, to be clear, I’m not necessarily talking hybrids or ultra-compact cars, here: a Honda Civic uses half as much fuel as a Honda Pilot, and the Civic, while a “small” car, is by no means a tiny car."

If it's a riderless car, then there's a question what the actual payload will be.

To deliver a pizza, maybe it's enough to send a riderless moped. Only without the pedals....

Why use a 200 pound vehicle to send a 2 pound payload?

Living near where you work means more frequent house/apartment moves (for job changes), and it means underemployment for many spouses and significant others.

Bingo -- I live between two major cities (Denver and Colorado Springs) with a 30 - 45 minute commute each way. My job prospects are much more plentiful than if I lived right near my job

Computer controlled cars are less safe than human controlled cars. That can change, but you can see why states are wary.

as someone who was blindsided by a woman attempting to have a phone call while putting on makeup, i beg to differ.

I strongly promoted self-driving cars in my presentation to House Energy Subcommittee staffers on Feb. 18th. See the related slide (page 37) here:

Google (Stanford AI lab) has now tested a driverless car for probably 150,000 miles successfully in the San Francisco area. So the technology is approaching maturity. However, implementation will require a regulatory framework. Just what does it take to make a driverless car street-legal? We don't really know yet, although it mostly a matter of cataloging.

I ran the numbers for this market segment, and I estimate the annual market for related equipment sales at $15-20 billion per year, with a strong comparative advantage for the US (because time is cheaper in China, and thus a time saving device less competitive there; also, the IT requirements are high).

This technology would serve four market segments in particular: our fast-growing senior population; soccer moms (who spend an amazing amount of time in a car daily); busy executives, who need more productive time (oh, and wouldn't it be great if your car could park itself in Manhattan); and taxis, because about 3/4 of the cost of a taxi is the driver.

Futher, such a technology may be critical to making electic vehicles viable. The chief problem of EV's are high capital costs and long recharge times. Both of these can be addressed if vehicle utilization increases (my car averages less than 2% utilization by time on an average day) and if the user can choose from a fleet of vehicles, which would allow use of the law of large numbers to overcome re-charging issues. (This implies that some people would subscribe to transport services rather than purchase vehicles as owned-assets.)

Of course, attitudes matter. Fear of new technology is important for market adoption. But it is important to understand that this is the only barrier to successful adoption, other than needed regulatory infrastructure.

> and taxis, because about 3/4 of the cost of a taxi is the driver.
I think you just identified a major opposing interest group. But, what is the cost of a medallion in NYC?

The Cabby lobby is a major problem in all cities. In Las Vegas, they've been blocking the construction of a light-rail project from the Airport to the Strip for 20 years. It's only 2 miles from the airport to the strip, but cabbies have successfully lobbied for a monopoly on picking up tourists.

You know who loves robots, has dense cities and a thriving car industry? Japan.

If that stuff is promising, I would be on the Japanese to go first.

Robotics is hard though, especially the positioning and control parts of robotics. That is why those roombas just drive around randomly.

When you introduce a driverless care into the drivered system, you are mixing two systems.

A better result would be to have separate hyway systems--an entire drivered network, and a driverless network.

The other issue that arises is liability for accidents: is the driver of a driverless care ultimately responsible, or is the manufacturer with defective programming.

How do other laws change: texting while driving a driverless care, drunk while driving a driverless car, etc.

You clearly do not understand what these driverless cars do. They have sensors and process information just like a human driver does, so there is no obstacle to mixing the two.

In fact, computer drivers are much more awesome at seeing just what is going on than any human.

And they not only see what is there, but they can know what should be there. I.e. "there should be a stop sign there, but I don't see one. I'll stop anyway, and post a note to my map database company, who will either 1) fix their DB, or 2) tell the road crew that some a-hole kids stole the stop sign."

Positive externalities, right there, baby.

Any wagers on which city will be first to push adoption of driverless cars? Based on comments from Robin and Steven it seems like a large, geographically constrained city with an aging population and busy executives. Perhaps also a society which is somewhat accepting of robots... Tokyo?

The most likely state in the US is Nevada, but the cab drivers, who own Clark County politics along w/ the culinary unions, would go ballistic.

Driverless cabs in NYC would work well.

It is weird that you do not see the contradictions in your own alleged "free market" thinking.

Let's take your article:

"Putting a computer behind the wheel may sound scary, but in road tests performed by Google and other companies, the cars have had a good safety record."

Wow. So a few companies all of which have some possible vested interest in the technology have found that there one or two cars work fine under strict conditions and you are just ready to impose the technology on everyone without further testing?

"The point is not that such cars could be on the road in large numbers tomorrow, but that we ought to give the cars — and other potential innovations — a fair shot so that a prototype can become a commercial product someday."

Umm. I thought you just stated above that there are prototypes? I thought that is the point. I love the "pebbles in the stream nonsense. Let's call it economic "science" and then rely on colorful analogies.

"The driverless car is illegal in all 50 states."

Big shock there.

"Unfortunately, the very necessity for this lobbying is a sign of our ambivalence toward change. Ideally, politicians should be calling for accelerated safety trials and promising to pass liability caps if the cars meet acceptable standards, whether that be sooner or later. Yet no major public figure has taken up this cause."

So "change" is good just by itself? We should just immediately allow driverless cars without any further testing? Liability caps? You mean if something goes wrong and kills hundreds of people the company that makes the cars's liability will be capped? Why? So nw you want to in effect subsidize this industry by legislating away any externalities?

Then your grand conclusion which is nothing more than an assumption:

"Consider this thought experiment. Assume that driverless cars could certainly reduce deaths by avoiding accidents caused by people who drive while intoxicated or who simply make stupid driving decisions, like driving on the wrong side of the road. Add in the likelihood that even after they are perfected and well inspected, driverless cars would lead to special problems, perhaps if the computers don’t respond properly to some unusual situations.

To continue this experiment, imagine that the cars would save many lives over all, but lead to some bad accidents when a car malfunctions. The evening news might show a “Terminator” car spinning out of control and killing a child. There could be demands to shut down the cars until just about every problem is solved. The lives saved by the cars would not be as visible as the lives lost, and therefore the law might thwart or delay what could be a very beneficial innovation."

You basically assume your premise that these cars are going to be safer and then ask us to "imagine" a world where they are banned? Sloppy thinking and also the typical magical realism of the "free market" BS. In then end, reading your article carefully you seem to advocate that these cars should ultimately be imposed on everyone in order to achieve the safety and efficiency results you desire. Yes. Just what our corporations need, a reason to sell us completely new untested stuff with a liability cap to ensure that their profits do not suffer if people are killed by their negligence.

I agree that development should continue. I am not against that. To say that "regulations" are somehow stymying development is just patent nonsense. This is not a "legal" problem. This is an issue of fundamental safety. There is a role here for the government. Yes it will be slower and it will not be perfect but neither will your "utopia" where suddenly everyone needs to drive these cars in the name of "innovation."

Those cars have driven over 140,000 miles along California's highways and roads. You're exemplifying the point Tyler is talking about in his article about blowing up the issue in the name of safety and forestalling innovation.

So a few companies all of which have some possible vested interest in the technology have found that there one or two cars work fine under strict conditions and you are just ready to impose the technology on everyone without further testing?...In then end, reading your article carefully you seem to advocate that these cars should ultimately be imposed on everyone in order to achieve the safety and efficiency results you desire.

Tyler isn't advocating that anything be *imposed* on everyone, he simply thinks that obstacles should be removed to make driverless cars more of an option. It is generally statists, not libertarians, that call for various measures to be imposed on the population.

Tyler is complaining about "regulation" and then asking that govermnent pass liability caps, which prevents the free action of individuals.

The cars have been indeed been tested, but with a human sitting behind the wheel, ready to take over.

Good points.

I had not previously considered what I expect could be large fuel-efficiency gains by using small driverless-vehicles as errand runners (picking up "take out" and such).

Auto-autos are also better suited for electric engines -- they can drop you off and go charge.

"But it’s clear that in the early part of the 20th century, the original advent of the motor car was not impeded by anything like the current mélange of regulations, laws and lawsuits."

The magazine Horseless Age, which began publication in the 1890s, had a regular section titled "Legislative and Legal," describing the numerous laws and ordinances prohibiting or restricting the use of automobiles.

good catch:,++legislative+legal&source=bl&ots=fPI9ieMWnY&sig=VxKZbPPhLGd_ezlW33zjh2O93GI&hl=en&ei=6MziTarXFobciAKR_OWoBg&sa=X&oi=book_result&ct=result&resnum=1&ved=0CBoQ6AEwAA#v=onepage&q=Horseless%20Age%2C%20%20legislative%20legal&f=false

Even as engineers are now close to developing a vehicle that can negotiate incredibly complex environments, vehicles with relatively simple environments to negotiate (mass transit trains) in the US remain for the most part reliant on human drivers.

I have often wondered why so few subways in the US are automated. Is it simply the absence of incentives in these government-run monopolies? Are union pressures responsible? While the cost of automating the NYC subway would certainly be high, I can't imagine this cost couldn't be recouped through a few years of dramatically reduced labor costs.

I suspect that the marginal cost of running additional trains is mainly labor. The unioniized drivers on the MTA and PATH are very well-paid.

"I have often wondered why so few subways in the US are automated."

To me it's the simple fact that none of the people who would push for such a thing are willing to really stand behind the idea. And by "stand behind the idea" I mean ritual disembowelment upon the first failure of their product. In other words, until the entrepreneurs who want to cash in on automated transit/driverless cars are willing to put their lives and limbs on the line for their ideas, I see no reason why I should.

Unions. Exactly. I worked on the JFK Airtrain, which is driverless. Since it is operated by the Port Authority of NY and NJ, which didn't have such entrenched driver unions, such a project was possible. MTA goes ballistic when even some automation (see: governor technology) is introduced.

In Toronto, when it bought a driverless train for a short loop (the Scarborough LRT), the unions insisted that there still be a driver on the train. To sit there and do...something. Which usually involved reading the paper or napping or whatnot.

In the UK, the driverless trains of the DLR unnerved passengers, so they had to add "guards" to them, who sit around and read, mostly. (Though they are trained to break up fights and call the police.)

Actually, the first place to test driverless cars would be on private property -- resorts, golf courses -- or large military posts. Or, perhaps as eco friendly electric car tours through parks or land that would be damaged if people left the road.

Already done, mate.

I'm skeptical. There is certainly some demand for driverless cars. But generally, driving is too fun to leave to a computer. In fact, I'd say that most people like computers for the same reason they like cars: The give and take between man and machine is an enjoyable experience. The efficiencies are definitely important, but most buyers adopt such technologies because they're fun.

Not everyone agrees with you about driving being fun, is the thing. Allowing for driverless cars would allow people like me to greatly increase the pleasure of our commute (consumer surplus, baby). Personally I find driving a tiring, annoying, frequently dangerous experience (seriously, people, do you need to tailgate me when I'm going the speed limit in the slow lane with _traffic in front of me_?) and would very much rather do without it at all.

You don't have a long commute, do you?

I work 50 miles from where I live and commute 4 days a week. Trust me, I can find much better uses for my time for those 8 hours (minimum) a week.

I would assume that your driverless car allows you to switch to manual control once you get off the crowded motorway and on to an empty twisty mountain road.

Isn't it obvious that driverless cars will be adopted first in another country? Any bets on which one? Netherlands, maybe?

It wasn't obvious to me until you pointed it out. But ya, it does seem very likely. My bet is on Canada which is much less dense than the Netherlands. My intuitions says that we'll first see driverless cars used extensively on long stretches of low traffic highways like you'll find in much of Canada.

Will probably be a country in the east.

Dubai. China. Hong Kong. Singapore. Taiwan.

Definitely won't be India!

"System Error: Bovine detected: un-programmed event"

Even more obvious: It will be driverless TRUCKS not cars.
1. Trucks usually adopt new tech ages before cars. For one thing, they are more expensive and cost more to run, so small improvements are more likely to be worth the capital costs.
2. Truck driving is seen as a job. Labour saving devices are attracted to reducing the work involved in jobs. People don't have the ambivalent, "is it a chore or is it fun" relationship they have with cars.
3. Truck accidents have a direct financial cost that goes to the big corporations responsible for the decisions to buy the new tech.
4. Smoother driving and less wear on tyres and brakes adds up to a lot of money for trucks.
5. Ever stricter driver safety regulations mean that if you want more miles/truck/day, you have to get around the restriction of the driver.

One obstacle to this: for some trucks, the driver also loads and unloads cargo at the truck's various stops. This seems more difficult to automate, but if a human is necessary anyway to handle cargo, they might as well drive too.

As we have seen in light rail crashes on WMATA and BART and numerous airline accidents, humans are not very good backup for an automated system. We are simply not sufficiently vigilant or quick to catch the one time in a million when the system malfunctions. On the other hand, the same technologies that could enable driverless cars would be a superb backup and warning system for a human driver, as speed and vigilance is precisely where computers excel. What if, instead of having the computer drive the car and expecting a human being to intervene in the case of a problem (as was the case on Metro during the red line crash), the human was asked to drive and the computer sounded an alarm or even intervened to prevent a crash? We already do this to some extent with traction control and active cruise control, but this roll could certainly be expanded to augment drivers' capabilities.

This is substantially less scary to me than putting two-ton deterministic machine into an extremely noisy and unpredictable environment. Murphy would have a field day.

Driverless cars would also be helpful to people with disabilities. The extent depends on the type of car: a fully self-driving car could allow people with epilepsy or poor eyesight or other contraindications to driving to be independently mobile when they otherwise couldn't, but even a mostly self-driving car that reverted to driver control when confused could be useful for people who can't drive long distances because of fibromyalgia, carpal tunnel and other repetitive strain injuries, or arthritis.

Bill's question about liability is especially important in the US.

In addition to the desire for control, driverless cars suffer from the network effect.

If everyone had these little, efficient, networked cars there would be no traffic and the system would be perfect... because it's essentially centrally controlled (gasp!). Even if there is no central control the distributed computers would tend to cluster in their decision-making and have more impulse control than human drivers, so the impact would be similar.

However, these benefits don't appear for anyone until most if not all people are driving such cars. Not only would it be too chaotic, but there are worries about other drivers in larger, heavy, more dangerous vehicles crushing you in your tiny little computer car. Also, once most people are "off the road" by getting into these cars, there are incentives to be one of the few left that can cruise past them and essentially drive like a jerk. The system would evolve towards an equilibrium but then as people take advantage, it would veery away from it.

The only solution would be to create separate lanes and infrastructure to separate the smart cars from the dumb cars, but that can't be done purely in the free market. Not only do we not have the stomach for such infrastructure projects right now, but this post was about government "getting out of the way."

Unfortunately, I think that smart cars now are very similar to real cars in the 40s. They are a novelty until a big government program backs them and creates subsidies as well as efficiencies of ownership. Not only am I talking about the Eisenhower highway system, but also about the systematic dismantling of all of old downtown trolly systems (done by local governments and supported by the car and tire companies). The installation of new infrastructure, the dismantling of the old infrastructure, and a new mode of living (suburbs) all came together to produce the reign of the automobile.

Maybe in the future living will become more dense and we will change our infrastructure to react... but that's the only way that driverless cars can and will take off.

good points re network effects.

I don't think the network effect applies here -- driverless cars operate on the same rule-sets as human-driven cars...they just have a 360-degree view, perfect vision to the horizon, radar, etc etc

Your comments apply more to ultra light cars (whether human or computer driven) than it does to computer controlled cars (which may be any size).

There may come a time when computer-controlled cars are so prevalent that the few people who are not in them are effectively able to "bully" the other cars by driving aggressively, which the computer cars will nicely tolerate.

If we ever get to that point, though, then
1. mission accomplished
2. there are a number of remediation techniques available, including technological, social, and legal.

The network effects from computer-driven cars are huge. The first computer-driven cars must assume that every other car is being driven by a dumb human. But the second computer-driven car on the road with you becomes your partner. You can talk with each other about what you see, where you have been, and what road hazards there are. If they are both travelling in the same direction, one can tailgate the other a lot closer -- if something requires hitting the brakes, the second car will know this and be acting on it even before the taillights on the first car go on.

With Laurel's comment, there are many people whose mobility is constricted by their inability to drive–not only people with disabilities, but youth and the elderly. Driverless cars would eliminate barriers.

The price of insuring these cars will determine adoption. If they are not safe, or tend to lose jury decided litigation contests against human drivers, the adoption rate will be slow as the insurance will be prohibitive.

Another large benefit of driverless cars: fewer cars would be needed.

Right now the bulk of cars are sitting idle at any given time because the labor costs required to use them more efficiently are too high. With driverless cars these costs would drop significantly.

Yes -- anyone living in a major city would probably just get a subscription to an automated zipcar service.

I think it is going to be a very hard sell to the general public. The first person killed by a driverless car will generate more bad press than 1

Sacrificing the unseen for the seen...

Information technology has largely changed the neolithic basis of civilization and additional innovations will usher in a postcivil era of much richer human choice and sustainability. Postcivil society is coming. The transition will be rough. Empty the cities now to minimize human suffering during the transition.

Ah. sorrry, pressed submit by accident.

Anyway, I think the first person killed by a driver-less car will generate more bad publicity than 100 people killed by conventional cars.

They way I see this rolling out is to sell it as an extension of cruise control. You still have a driver who is notionally who in control but in reality people just use the autopilot and everybody is safer. This also solves the construction site problem because the human can step in when a situation that the computer can't handle crops up.

I agree -- I think you will see signs where you can 'engage autopilot' -- sings will be first posted outside of city limits on major highways and'll have to go back to manual inside city limits.

Eventually it will be the other way around. Only allowed to go to manual outside of populated areas.;

But it will take a while, maybe a generation, to switch.

I agree - but it will take a long time.

I meant to add that insurance companies will facilitate this. The insurance premium on a auto-auto should be much lower than on human-driven vehicles. As a result, it will be expensive to drive your own car in a dense area.

With neolithic agriculture came civilization.

With the Internet and advances in shipping technology we can enter a postcivil era with social organization much closer to that of the Greek demes (kin-based agrarian populations of about 5,000) that gave rise to their Golden Age.

Not only can we enter such a postcivil era, we are entering it. The rate of evolution of human pathogens is much higher now. The availability of technologies that can destroy urban centers is much wider now. The population is much more concentrated now.

Postcivil society will be the result. The only question is how much human suffering can be prevented by taking action to empty the cities before they are forced to seek new abodes.

Decentralized production and local consumption of food is far more energy and capital efficient since it needn’t be transported to urban centers. This needn’t involve a return to old agrarian technologies—although from an examination of household leisure time remaining for most employees after work and other burdens of civilization, it is apparent that civilized jobs are little more efficient for food acquisition.

Moreover, the small residual needs for distribution of food to cover local shortage is far more viable now with “just in time” inventory systems based on efficient, decentralized and very robust communications infrastructure. For example, the trading pits are not a necessity—it can all be electronically distributed and decentralized with reputation networks.

Likewise, huge central repositories of grain and livestock yards are inefficient inventory policies vulnerable to attack and sabotage.

Chicago can go.

Similar arguments apply to almost all other urban areas due to their existence as mere levels of abstraction atop the thermodynamics of food. The primary value of such abstractions remains via the distributed networks of communication keeping alive inter-cultural dialogs for those who choose it.

As recognized by Control Data Corporation founder, William Norris, in his project to create small, energy self-sufficient farms, and as recognized by founding fathers of the United States such as Thomas Jefferson in his effort to make Yeoman Farmer Conservatism the basis of federalism, centralized population structure creates vulnerabilities.

The obvious vulnerabilities, such as pandemics, bioweapons attacks, nuclear attacks, due to centralization of population, central stores and transportation hubs, need not be elaborated.

The outflow of population from cities to the areas of solar collection of their energy—photosynthesis of their food—to reduce total system complexity will necessarily be driven by the ecological structure of the food chain.

It will likely begin when a few catastrophes hit and cause millions of deaths rendering the apparent “safety” of urban areas a cruel deception. Since there have been no massive wars in the Western World since WW II, there has arisen a profound complacency which has just recently be shaken by the AIDS pandemic, The attacks of 9/11/2001, the de-population of New Orleans and the on-going sacrifice of civil liberties for “homeland security” primarily due to the vulnerability of specialized, highly centralized, structures. Recent fears of a 1918-style pandemic may well be realized. The world’s population is far more vulnerable to such a pandemic today than it was in 1918 and there could easily be a billion deaths, disproportionately concentrated in highly civilized societies if cascade effects arise as they are likely to.

As this awareness rises, and people begin to look for genuine security and alternatives to urban lifestyles, it will become apparent that current social constructs aren’t working for people. People will no longer see contributing half of their labor to a government that is resulting in their deaths as a good deal.

There are a number of drivers for the coming postcivil society, including pandemics, bioweapons attacks, nuclear attacks, electromagnetic pulse attacks (EMP), etc.

The evolution of virulence is perhaps the greatest threat of immigration specifically and globalization generally—which is precisely why we are never allowed to hear about it from the old media.

But a ray of light has broken through in a limited way here and there as Medical News Today reports:

Viruses More Virulent In A Connected World

That’s one conclusion from a new study that looked at how virulence evolves in parasites. The research examined whether parasites evolve to be more or less aggressive depending on whether they are closely connected to their hosts or scattered among more isolated clusters of hosts.

The research was led by Geoff Wild, an NSERC-funded mathematician at the University of Western Ontario, with colleagues from the University of Edinburgh. Their paper was published on Nature’s Web site on May 27…

“The findings also suggest that as human activity makes the world more connected, natural selection will favour more virulent and dangerous parasites.”"

An alternative source of pandemics is bioengineering.

A recent article by Robert Carleson, titled “The Pace and Proliferation of Biological Technologies” essentially predicts that personal bioweapons labs capable of producing deadly pandemics will be widespread in about a decade. This means that the kinds of people who now develop computer viruses may turn their talents to real viruses—and not all of these people will be “playful” about it. As I see it the primary potentials here are:

1. A team of reproductively disenfranchized East Asian and/or Israeli men building a pathogen that kills men in high population areas except those with a selected group of Y-Chromosomes. This selectivity requires some careful engineering.

2. A reproductively disenfranchized “loner”—most probably of Northern European or Russian heritage—concocting a much less selective pathogen that kills anyone unfortunate enough to be in a high population density area. It's easier to kill indiscriminately.

When the flight from the cities becomes widespread, people are going to find themselves renting small plots of land from which they’ll need to scrape some semblance of subsistence.

However they’ll be renting land that has been grabbed by the genocidal land barons, who will be protected by mercenary sheriffs. Most of these sheriffs will be vampires—whose character has been shaped by compensating, for having to live under the thumb of—and forcing other men to submit to the abuse of—feminist wives, with the sadistic power of sending the “delinquent” young men of their communities off to be raped by gangs in prisons and frequently infected with HIV, Hep-C, XDR-TB and drug resistant staph.

In this desperate situation, being able to generate subsistence from small amounts of low quality land within a short period of time—months if not weeks—will be a crucial bridge for them.

OK, you're nuts. You don't have a fucking clue about anything, do you. Certainly not feminism.

If you want to make this a story about regulation, you need to make this more Coasian.

Let's say that a driverless car for some reason fails to recognize the drivers right of way. Who yields: the driver because the driverless car is not responding appropriately to the conditions and the driver has the last clear chance. So, if it came to a liability situation, who would be liable: the person with the last clear chance to avoid the injury, or at least some portion would be assessed to him.

Or, should drivers have to alter their driving styles to accomodate driverless cars; if there is a priority as to who gets on a congested, fast highway, who should get priority.

The point is that anything that involves interaction with others involves regulation.

Just bangin' on regulation gets one nowhere--like a stalled automobile stuck in traffic.

Good focus on regulation versus humans: I wonder if the driverless car follows the regulations or follows reality. For example, does it drive at 55 on the Dulles Toll road, when almost everyone does over 65? Or at a 4-way stop, where often the informal rule is that the north/south cars go simultaneously, then the east-west cars go (because there's few left turns being made).

Of course we already have the "Student Driver" signs on cars, maybe the driverless car has to have a similar one.

Maybe you can set your 'driver profile': extra cautious, aggressive, fuel efficiency, etc.

There is no way the manufacturer would program the vehicle to exceed the speed limit. So they'd allow the "passenger in charge" to set that sort of thing.

Although it sounds like it would be a radical change to your life and to society to have robot drivers, as someone who has had a full time human driver (perk of the job) for a few years now, I am not so sure.

The biggest impact to me is that it extends my working day, I get in the car and out comes the laptop (with 3G dongle) and cell phone and I am working (or surfing like now). This makes the commute pass really quickly (I am sometimes sitting in the car when I get home finishing off a call or email, slightly annoyed the journey finished so fast.) But what I don't do is leave the office any sooner, which I could justify since I now have another hour or so of productivity. Work expands to fill the time available I suppose. As a result, I still worry about my commute and do what I can to live close to my workplace, so I don't think robot drivers will necessarily increase urban sprawl.

On the social side, having a full time driver probably increases our tendency to socialize and eat out as we don't have to worry about drinking or parking (or finding a location). My wife loves not having to worry about parking while shopping, so she probably shops more. Most places now have pretty organised parking systems so probably this is not that big a deal in terms of total time, but it extends the variety of places that we visit as we don't have to figure out a new system. Maybe this can save downtowns?

I never was one of those guys that was really interested in cars or driving, so I can't say I really miss driving for driving sake (I still drive on vacations though). It's made the whole car as a fashion statement thing even less important to me, I mean who cares what the acceleration or driving style is like when you are not driving. So robot drivers are likely to extend the life of cars and make them cheaper and more of a commodity - which is probably a good thing, but probably not major in terms of societal impact.

One thing you probably don't get with a robot driver that really is a benefit to me is my driver keeps the car clean and maintained. That is a chore I don't miss. Overall therefore I would put robot drivers in the category of flat screen TVs, nice but not game changing.

Some practical questions:

1. How expensive would these cars be if they become commercial? What is the marginal cost of going driver-less versus employing a minimum wage chauffeur; noting that the chauffeur can run additional errands that the car cannot.

2.Who would get a ticket, on the off chance that the car indeed did something illegal. Does the car understand traffic signs? The google version seems to rely on a pre-programmed database at least for speed limits.

I think they would be on luxury vehicles first, just like self-parking is. so you'd get it on your $75,000 Mercedes SUV.


If the driverless car becomes mainstream despite legal and political obstacles, will Tyler Cowen change his views on regulation? Or to put it another way, I think those who point to legal and political obstacles to the adoption of this technology may very well be overstating their case. But if I am wrong, how will I know whether I am wrong? A failure to introduce driverless cars could be due to technological issues as well as political and legal ones.

If driverless cars are safer than cars with drivers, maybe one day it will be negligence for a human to unnecessarily drive a car. That may be a more significant issue arising from the legal system. Will we one day have a future where no one experiences what it is like to drive a car?

I don't merely view driverless cars as a "potentially" significant development. I think they are the future.

Ya that is a great concept. But there some drawbacks also . As this scheme could be very successful in urban area where the colonies or households are in the proper way. There where more traffic there they could not manage so easy. The thing is all system would be changed for it. The traffic system, The towns planning and many more . All these could be solve by Nanotechnology related materials as Carbon Nanotubes & Nanoparticles which could become more helpful to advance development of such above plans.

The good advanced plan . But need is to make a big change in the current system. We should change our traffic plans, A programed cars could make decisions as Man do. But we should take some help of Industries as the Industrial chemical supplier, industrial chemical manufacturer and other Auto Industries could helps in developing advanced cars for above model.

Thanks to the various people who described legal problems.

I think this boils down to a problem in human expectations. We expect human beings to have automobile accidents, and when it happens, we decide who to blame. It's always somebody's fault. If they had not made the accident happen, it would not have happened. The default state is perfection except when some human being does something wrong, and when that happens they can be blamed for it.

If we had automated automobiles then every accident would be the fault of the manufacturer, unless it was somehow the fault of an owner. Driving in conditions where the machine cannot sense well enough -- heavy rain, maybe -- would be the owner's fault. Unless the machine was designed to refuse to operate in that situation.

It would help to have 360 degree HDTV recordings. If the recordings survived an accident they might demonstrate that some other human than the owner was at fault.

But the manufacturer and designer could expect to face a lot of legal proceedings, an unpredictable but large amount. That naturally will tend to make them cautious about investing. So yes, the natural response if we want robocars is to declare the manufacturers legally immune no matter what they do.

Very similar for nuclear power plants. When there's some chance that any nuclear power plant might possibly turn into an accident a thousand times worse than Chernobyl, it's hard to get investors. We will get nuclear power much faster if we declare that no one can sue for any effects of a nuclear meltdown or explosion.

This generalizes. We could have a government office that can label qualifying inventions as "New Technology". And we could make a blanket rule that no one who invents, manufactures, distributes, sells, or maintains New Technology can be sued for any reason whatsoever. I can't think of a more effective way to encourage New Technology.

Though here's a second-most-effective way. Once an invention has been classified as New Technology we could give the owners of the New Technology a flat billion dollars to do what they like with, whether they ever sell a product or not. That would go a long way toward getting lots of registered New Technology, though it might encourage or discourage actual manufacturing.

The solution is simple. The automobile manufacturer goes bankrupt every couple of years, thus shedding all legal liability. Then the government bails it out.
Thus the American car makers actually have a clear advantage over foreign brands, because only the Americans can afford to bring out the new tech.

Metaphors about pebbles are cute. But they don't tell the story behind the hyperbole. As N. Chomsky remarks in "Understanding power," about 80% of innovation is government subsidized. If you're concerned about keeping rates of regulation low to accrue innovation, it should be considered what interests are backing the innovation in the first place - a government subsidy won't dry up because of accumulated regulations unless one or a few regulations are to blame for totally eliminating the market for it.

I have a hypothesis why people can get worked up about plastic bags and birth certificates, but not driverless cars. Prospect theory. The bags and birth certificates would be "domain of loss" issues in which it appears to the decider that if they do not act something they have will be taken away from them (clean environment, racial purity or whatever is motivating the Tea Partiers). Driverless cars, however, are in the "domain of gain" in which it appears to the decider that the matter involves whether they will receive some benefit or not.

Kahneman and Tversky have conclusively shown that people behave very differently depending on whether they believe they are in a domain of loss or of gain, even if the questions are, strictly speaking, logically equivalent. They act more forcefully in the domain of loss, and more tentatively in the domain of gain.

Therefore, given the perceived high costs of political participation, people will get politically involved only when they think something is being taken away from them. And so no one cares about driverless cars.

Prior restraint?

This seems to be a case of an economist inventing a hammer - getting government out of the way of doing whatever anyone wants - and then seeing everything as a nail - desperately needed innovative change to eliminate real or perceived economic stagnation.

And classically for economists, the reason inventions don't spread like wildfire through all of society is always because government prevents it, not because the invention is insufficient to solve the real world problems, or the solution costs way more than the status quo.

I note that the same economists who blame government for the lack of cheap flying cars or pilotless planes and cars, also blame government for oil companies failing to increase oil production in Texas and Oklahoma instead of recognizing the limits of the natural world and of the power of invention and innovation to solve problems that have huge global economic impact.

If technology can navigate the physical world because driving on roads is so well understood and nothing can surprise the engineers behind the system, it is certainly the case that a computer system can eliminate all civil court judges because the knowledge required is fully embodied in the more than three centuries of case law, and Watson demonstrates a computer system can integrate all that knowledge in answering any question, which would include the questions raised by all civil court cases.

Or let's replace doctors with Watson for all primary care and diagnostics. Integrate Watson with Leonardo and surgeons can be replaced. The medical profession can be reduced to nurses and medical science engineers who enhance the computer software.

"I note that the same economists who blame government for the lack of cheap flying cars or pilotless planes and cars, also blame government for oil companies failing to increase oil production in Texas and Oklahoma instead of recognizing the limits of the natural world and of the power of invention and innovation to solve problems that have huge global economic impact."

There's a sort of religion about this. The religion says that free markets naturally optimise every problem. So if you notice a problem which has not been optimised, there must be something that prevents free markets from optimising it. And the only thing that can prevent free markets from optimising every problem is government.

Of course no actual economist would hint at such silliness, but there are a lot of pretend economists who come right out and say it pretty much the way I just did.

"I'm not an economist but I play one on the Internet."

Tyler - if Auto-Autos are here, what are the implications for your great stagnation thesis? Where IS your great stagnation if driverless cars become a reality? Doesn't that qualify as a major life-changing advancement, like TV, radio, plumbing, etc?

I am quite suspicious of this technology actually working and being affordable. I use autopilot everyday in an aircraft. It's pretty great when it works but it breaks a lot. parts have to be replaced regularly. Is GPS at a point were it is good enough to keep cars from getting in head on collisions? Will all our cars stop driving if we don't have RAIM? The GPS systems that can provide a plane with precise enough information to land on Runway 400 feet wide and a mile and half long require more then just some satellites and a receiver. They also require error correction infrastructure on the ground. Or maybe it senses the lines on the road? Should work half the year I guess but the other half of the year when the roads are covered in snow should be interesting.

It might be all right in California but I think weather conditions in a lot of other places might prove to be a challenge.

It should probably download any mapping its doing ahead of time, so that shouldn't really be a problem. Driverless cars won't be able to use just GPS to navigate (or even use it very much to navigate). They're going to mostly have to use "eyes" just like the rest of us.

That said, there are still a lot of people who've bought into the macho stereotype of driving their own car, and changing that view will probably take a lot of time. I used to commute on a bus line that took only 10 minutes more than by car and on which I got to read or program the entire route in a decently comfortable seat. Even after explaining how many books I'd read and extra work I'd gotten done, the majority of people said they couldn't do it. There's a lot of irrational cultural values in this one.

As Tim says, these cars don't rely 100% on GPS. GPS isn't accurate enough to even keep you in a lane.

Humans are ugly bags of mostly water, and they only have two eyes. Cars can see everywhere. They can see things that humans cannot. And (this is really important) they can react MUCH MUCH faster than humans do.

"Cars can see everywhere. They can see things that humans cannot. And (this is really important) they can react MUCH MUCH faster than humans do."

Good point! So if a car sees something that looks like a dangerous situation approaching, it can instantly slam on the brakes.

And if a human behind it tailgates that car, it is the human's fault. Human drivers have the responsibility not to hit cars that stop suddenly.

At some point in the campaign, it might make sense for cars to aggressively drive in ways that will get humans to cause accidents, proving that humans are bad drivers. This would be particularly viable when the cars are not carrying humans who could be hurt. So, you get into an accident where maybe you're hurt, your car is totaled, and the other car suffers damages plus the pizza it was delivering is cold. It was your fault because you did not brake quickly enough. Thank god you didn't kill anybody....

The American workers' compensation system should provide a workable template for limiting the litigation downside. The bigger hurdle is social acceptance.

Comments for this post are closed