Note: This old post is still useful whenever you see a "driverless cars will change everything" story, (this one, for example) and especially a "driverless cars will be the end of transit" story. Abstract: The two fallacies to watch for in these stories are (a) the "complete imagined future" mode, which denies the problems associated with evolving the future condition instead of just jumping to it, and (b) the assumption, universal in techno-marketing but always untrue in the real world, that when the whizbang new thing appears, everything else will still be the same; i.e. that none if the whizbang thing's imagined competitors will also have transformed themselves. This latter assumption can also be called the "everyone but me is a dinosaur" trope.
Richard Gilbert, co-author of a book that I've praised called Transport Revolutions, has a Globe and Mail series arguing for how driverless cars will change everything. I will give this series a more thorough read, but just want to call out one key rhetorical move that needs to be noticed in all these discussions. It's in the beginning of Part 4, "Why driverless cars will trump transit rivals."
With widespread use of driverless cars – mostly as autonomous taxicabs (ATs) – there could be more vehicles on the road because ATs will substitute for most, and perhaps eventually all, private automobile use as well as much use of buses and other conventional transit.
This, and much of the discussion around driverless cars, is in the complete imagined future mode. Gilbert describes a world in which the driverless cars are already the dominant mode, and where our cities, infrastructure, and cultural expectations have already been reorganized around their potential and needs.
Some complete imagined futures are not necessarily achievable, because the future must be evolved. In fact, the evolution of organisms is a fairly apt metaphor for how cities and infrastructure change. As in evolution, each incremental state in the transformation to the new reality must itself be a viable system. We can think of lots of wonderful futures that would be internally consistent but for which there is no credible path from here to there.
Driverless cars remind me a bit of the "wheeled animal" question in evolution. No animals have evolved with wheels, despite the splendid advantages that wheels might confer on open ground. That's because there's no credible intermediate state where some part of an animal has mutated something vaguely wheel-like that incrementally improves its locomotion to the point of conferring an advantage. Wheels (and axles) have to exist completely before they are useful at all, which is why wheeled animals, if they existed, would be an argument for "intelligent design."
I will begin to take driverless cars seriously when I see credible narratives about all the intermediate states of their evolution, and how each will be an improvement that is both technically and culturally embraced. How will driverless and conventional cars mix in roads where the needs of conventional cars still dominate the politics of road design? How will they come to triumph in this situation? How does the driverless taxi business model work before the taxis are abundant? Some of the questions seem menial but really are profound: When a driverless car is at fault in the accident, to what human being does that fault attach? The programmer? What degree of perfection is needed for software that will be trusted to protect not just the passengers, but everyone on the street who is involuntarily in the presence of such a machine?
Here's a practical example: In Part 3, Gilbert tells us that with narrower driverless cars, "three vehicles will fit across two lanes." Presumably lanes will someday be restriped to match this reality, but when you do that, how do existing-width cars adapt? If you could fit two driverless cars into one existing lane, you could imagine driverless cars fitting into existing lanes side by side, so that the street could gradually evolve from, say, two wide lanes to four narrow ones. But converting two lanes to three narrow ones is much trickier. I'd like to see how each stage in the evolution is supposed to work, both technically and culturally.
That's one reason that I seem unable to join the driverless car bandwagon just yet. The other is that claims for driverless taxis replacing transit amount to imaging a completed new technology out-competing an existing unimproved technology — as though that would actually happen.
Sure, driverless taxis might replace many lower-ridership bus lines, but wouldn't buses become driverless at the same time? In such a future, wouldn't any fair pricing make these driverless buses much cheaper to use where volumes are high? Wouldn't there be a future of shared vehicles of various sizes, many engaged in what we would recognize as public transit? As with all things PRT, I notice a frequent slipperiness in explanations of it; I'm not sure, at each moment, whether we're talking about something that prevents you from having to ride with strangers (the core pitch of "Personal" rapid transit) as opposed to just a more efficient means of providing public transit, i.e. a service that welcomes the need to ride with strangers as the key to its efficient use of both money and space.