I’ve been thinking out loud about self-driving cars for a few years now, but two articles by Rodney Brooks highlighted a number of concerns that have actually never occurred to me before:
Lots of people talk about edge cases for self-driving cars—but Brooks’ discussion is substantially more interesting than most of those… like: when is it appropriate for an autonomous vehicle to honk, or for people to honk at an autonomous vehicle?
People often work off of subtle non-verbal cues that drivers send: we can tell that a car is apt to switch lanes based on the little ways the vehicles drift and the way we interpret the driver’s (barely-visible! But we still use it) body language. We recognize and expect certain kinds of behavior from drivers we intuit to fit into certain categories: inexperienced, not-from-around-here, aging, impatient, distracted, and so on.
Self-driving cars have none of those. They’ll have their own cues, and we’ll come to recognize and respond to them; but they will both be very different from the ones we are accustomed to and they will not (for a very long time, at least) be able to carry on the completely non-verbal two-way dialogue that happens between a pair of drivers, or between a driver and a bicyclist, or between a driver and a pedestrian…
There’s a lot here, and Brooks’ articles are not alarmist in the least, but realist in the best sense: careful and thorough and accurate. So, give them a read: