In another thread on summon being a silly party trick @SSonnentag posted this: Rather than carry two conversations on in that thread I pulled it out here to open this discussion up. I think @SSonnentag has a very valid point. Driving is not just made up of mechanical motions it also involves a whole wealth of inferences that we humans make all the time. It's easy to see the car 10 lengths in front of you correcting lane drifting and then you think "Gee, drunk or on the phone, maybe?" How about the guy coming up fast behind? You've seen him make lane changes to gain one car and you're going to be next. How do you react? For our Boston drivers... making eye contact when merging into the Calahan tunnel in the '70s means you lose the merge. Zipper merges? etc. etc. You get the picture. All of these require you to get in the head of other drivers based on your observations and then react. How is an automaton going to do that? Sensor suites on our cars have a long way to go to see other drivers inside their cars, see many car lengths in front and back. It just seems we really do have a long way to go.