I've been writing a lot about self-driving cars which have automatic accident avoidance and how they will change our cities. I was recently talking again with Robin Chase, whose new company, goloco attempts to set people up for ad-hoc carpools and got into the issues again. She believes we should use more transit in cities and there's a lot of merit to that case.

However, in the wealthy USA, we don't, outside of New York City. We love our cars, and we can afford their much higher cost, so they still dominate, and even in New York many people of means rely strictly on taxis and car services.

Transit is, at first glance, more energy efficient. When it shares right of way with cars it reduces congestion. Private right of way transit also reduces congestion but only when you don't consider the cost of the private right-of-way, where the balance is harder to decide. (The land only has a many-person vehicle on it a small fraction of the time compared to 1-3 passenger vehicles almost all the time on ordinary roads.)

However, my new realization is that transit may not be as energy efficient as we hope. During rush hour, packed transit vehicles are very efficient, especially if they have regenerative braking. But outside those hours it can be quite wasteful to have a large bus or train with minimal ridership. However, in order to give transit users flexibility, good service outside of rush-hour is important. If we create self-driving, accident-avoiding cars, the safety rules of vehicle design change. In particular, we can make vehicles a great deal lighter. Once you feel confident that the accident rate is extremely low, the vehicle can look more like an electric tricycle or quadcycle with a fiberglass shell than a steel car. This is particularly true of the "city car" which need not be able to go more than about 40 mph, nor travel more than 10 miles in a trip. Because these vehicles will be able to autonomously seek out charging or battery exchange stations while not being used, they can make do with much less battery weight as well.

Indeed, the more people are willing to think of the vehicles as taxis, and thus switch vehicles whenever that makes sense, each vehicle can be tuned to the trip that will be taken in it. You might own a vehicle with a 10 mile range, and just swap off to a different vehicle for longer trips, or trips needing cargo space, as needed. (In other posts I've outlined the concept of a standardized locker for your "stuff" which is lockable and easily moved to another vehicle, perhaps automatically.)

One could even imagine that you start a trip in your own 10 mile vehicle, but as the battery gets low, it pulls over into a small parking area where you walk into a second similar vehicle for the next 10 miles. Your vehicle uses its remaining charge to go find a full charge, and waits for you or rents itself out. You would not want to do this every day but it would be fine for those odd longer trips.

This ability to use exactly the right vehicle for the trip will make these 1-2 passenger city cars even more energy efficient than transit, certainly that transit during light occupancy periods. And of course, superior in every other way -- no waiting, direct routing, privacy, peace and quiet, not stopping enroute etc. Even if the transit lines remain the most efficient during rush hour, it will make sense for them to shut down outside those hours, and even ideally cede their dedicated right-of-ways to vehicles able to use them.

Computerized trip routing can also create highly efficient jitney cars that may even win in rush hour. A jitney car might consist or 4 to 8 personal "pods" with their own privacy and workspace. A trip, calculated by the computer, might take you one mile in a single-passenger vehicle to an instant transfer where you step into the pod for the shared part of the route. Later you would step out to transfer to another single passenger vehicle. Aside from the minor inconvenience of the quick transfers, and brief stops for other passengers to transfer, it would provide much of the efficiency of transit with a vastly superior experience, including a private, internet-connected workspace pre-assigned to you. Of course, with even more efficiency, traditional transit vehicles could be assigned to the busiest common routes.

One final reason transit is efficient is that one vehicle deals with less wind resistance than individual cars do. Wind resistance is the greatest loss of energy in cars at higher speeds, which is why trains do so well. However, it is possible for self-driving cars to coordinate with one another so they can "draft" one another and get back some of these gains.

I don't see people choosing transit very often with these alternatives available, and so I predict the possible death of public transit in 10-15 years time. Why do people use transit today, other than a feel-good desire to be more efficient?

• It's cheaper than a car if you don't own one, but for those who already feel they must own one, they look at only incremental costs (gas and parking) and often don't find this a win. Cheap automatic cars will be a bigger win.
• At rush hour, dedicated right-of-way transit can be faster if it's going on your route. Automatic cars will not suffer congestion.
• Finding parking is time-consuming and expensive. Automatic cars will self-park or earn money when not used.
• Though transit often is slower, you can read or work, which you can't do while driving. You can do this in a self-driving car.
• It's generally safer, but accident-avoiding cars will be equally safe.

The self-driving cars eliminate almost all the advantages transit is able to have. In part they do this by taking advantage of the existing road infrastructure we've already built to make them cost-competitive. But it's hard to see transit winning, even though it also can gain from technology improvements.