If and when self-driving cars really start taking off, it’s easy to see where the road leads. Firstly, they probably won’t be operated on the owner-occupier model that we use for cars today....Given driverless cars’ ability to come pick you up whenever you need one, it makes much more sense to just join a network of such things....therefore the freeing up of lots of space currently given over to parking spots.
What’s more, the capacity of all that freed-up space will be much greater than the capacity of our current roads. Put enough  self-driving cars onto the road, and it’s entirely conceivable that the number of vehicle-miles driven per hour, on any given stretch of road, could double from its current level, even without any increase in the speed limit. Then, take account of the fact that vehicle mileage will continue to improve. The result is that with existing dumb roads, we could wind up moving more people more miles for less total energy expenditure in cars — even when most of those cars continue to have just one person in them — than by forcing those people to cluster together and take huge, heavy trains instead.
Kevin Drum concurs and even goes one better:
I don't really need any convincing on this front. I think that genuine self-driving cars will be available within a decade and that they'll be big game changers. Even bigger than Felix suggests. When you're not actually driving a car yourself, for example, you don't care much about how powerful it is. So you'll be happy to chug along in a super-efficient car, reading a book or playing on your phone. You'll be more willing to share a car, since automated systems will be able to quickly put together carpools with guaranteed maximums on wait time. And of course, driverless cars will be fundamentally more fuel-efficient since computers can drive cars better than humans can.You won't get any argument from me that driverless cars are on the way, but the idea that people will then agree to get around in networks of super-efficient automatic taxis that they don't own is, I think, complete wonkish naivety. Yes, the overall car transportation system could be made much more efficient along these lines but to think this will happen gives average people far too much credit for rational decision-making. The vast bulk of spending on cars in the US is for irrational emotional purposes: status display, feelings of safety, etc, not simple transportation. If this was not so, Americans would all be saving their money and driving around in the equivalent of a $3000 Tata Nano (or $5k VW Beetle or Citroen 2CV). That class of vehicles can get you from A to B in reasonable comfort from the weather, but hardly any Americans would be seen dead in one and so none are sold here. Just watch a few car ads - endless fantasies of cars driving around on empty roads, barely any discussion of the realistic cost/benefits of sitting in it in stop and go traffic on the 101 or the 495. Thus anywhere from 2/3 to over 95% of the cost of the car is for something other than getting from A to B in reasonable comfort.
I see this in Silicon Valley in spades; parking lots dotted with Mercedes and Porsches well suited to doing 150mph on an empty autobahn or 200mph around a race-track, none of which ever get to do anything of the sort. Maybe occasionally the owner gets a break in the traffic on Highway 1 on a weekend away, but that's about it - maybe half an hour a year that they get to do what the car is actually designed for and the car ads told them it would make them feel good doing. Yet highly accomplished and rational meritocrats will cheerfully stretch the household budget to have one of these things to show off in the driveway or the parking lot at work.
Or think of it from a different ideological/demographic direction - why is the Prius the overwhelming best-seller hybrid? I would argue the answer is that it's the only one with a very distinctive design allowing its owner to communicate his or her concern for the environment publicly. The Civic hybrid was similarly efficient but it doesn't communicate "environmentalist", just "small cheap car", and so it doesn't sell anything like as well. It's a different definition of status, but still it indicates that most of what we are spending over a basic $5k vehicle is for something other than warm and dry transportion. So if we accept that "showing it off in the parking lot" is actually one of the major functions of a car, even if we probably don't like to admit that publicly, how likely is it that people will agree to give this up in order to save money in an auto-car sharing service? While it's true that getting rid of the parking lot would benefit society a lot, how are the first three employees who give up their own car and show up at the office in the autoZip going to feel?
If autoZip tries to start from the bottom-up, it will fail because its early adopters will secretly feel like cheap losers and will try to get rid of the stigma as quickly as possible. On the other hand, if it tries to start from the top down (Teslas!), it will run into the problem that the top 1% (income-wise) can easily afford to have their own car lying around doing nothing all day, and that's an effective way to signal wealth and status. The save money argument cuts nothing with the 1% because a lot of them are specifically trying to signal that they don't need to do anything of the sort.
So how does autoZip ever get started? I don't see the business model here. I totally see auto-driving taxis - taxi drivers are much more expensive than computers - but auto-taxis will be like taxis now: slightly grungy transportation used by poor people in a pinch, business travelers going to/from airports, or folks in very concentrated urban areas (basically Manhattan) where the overheads of car ownership are overwhelming. People will not use auto-taxis to get to work or go shopping - they will take their own autocars that they can then show off to others.
Update 1: And think of the incentives of the car companies. Do they have any incentive to push consumers to make more efficient use of their product (ie sell fewer cars)? Noooo.
Update 2: On reflection, I have realized a compelling benefit of auto-driving cars. It will be possible to drive down Highway 1, put one's foot absolutely to the floor, and have the computer take the bends as fast as possible while guaranteeing not to go over the edge. That sounds like a lot of fun and I can see the ads already. :-)
Update 3: Better still, if the cars are all talking to each other via Skynet, then it will be possible to pass at high speeds on blind bends in safety, knowing that no other car is coming the other way. How awesome is that? Too bad about the cyclists and the deer - wait, we could solve that problem with fleets of small monitoring drones... Screw the truckers and the taxi-drivers, count me in :-)