Also Won't Necessarily Be Safe

As I've said repeatedly, I don't think safety is really the prime concern of autonomous vehicles, at least in the sense that they can be programmed not to hit things. But driving in non-standard ways, ways in which aren't intuitive to the human drivers around them, can create problems around the cars even if the cars themselves aren't running over pedestrians regularly.

For humans driving regular cars, these auto-matons must be a nuisance. They are slow—we stayed at about 15 to 20 miles per hour for most of our trip. They stop at the hint of danger, sometimes slamming on the brakes and throwing passengers forward in their seats. (I would not choose to ride in this self-driving car if I were, say, already suffering from a migraine.) And occasionally, they get confused and just kinda freeze. At one point, Chinchilla approached a public bus pulled over to the side of a one-way street. There was plenty of room to navigate around it. Chinchilla braked and considered its impending circumnavigation. And considered. And considered. About two minutes later, the safety driver finally flipped off the self-driving mode and piloted the car around the bus. No vehicles were waiting behind us, but, oh, if there had been—the honking! (Kyle Vogt, Cruise’s CEO, later told me the lidar sensors that usually determine how much clearance the vehicles have on their sides have been suffering from technical issues for the past few weeks, so the cars are even more cautious about going around obstacles then they normally are.)

The technology is neat. It obviously "works" in some sense. It just isn't going to work well enough for them to be all that useful.