Yeah, I think there are at least a couple somewhat unrelated points.

- Getting from, say, freeway autopilots to more or less arbitrary self-driving under a wide range of road and weather conditions without a competent human even present is going to be a lot harder IMO than a lot of people credit. I actually don't think the NIST autonomy levels are that well thought out. I can't help but think there's a bit of "and then a miracle happens" between 3 and 4.

- Peak demand is definitely an issue. It's not just commuting. Going away for a weekend has definite unidirectional patterns. There are also lots of reasons people will prefer personal vehicles, such as keeping stuff in them.

Agreed, the "cars become autonomous" problems are separate from the "autonomous cars means nobody owns cars anymore" ones.

I'm more optimistic about autonomy. It still seems like a crazy hard problem, but things are advancing so fast right now that it actually seems like they might get there.

But I don't see where the leap from "autonomous cars" to "shared cars" happens. The main obstacle there is simply personal: I don't think people want shared cars. People want autonomous cars (although they probably want the autonomy to be optional) so that hurdle is not a problem on that side.

I suspect that given peak loads, utilizations, etc. there are pretty unfounded assumptions about the economics associated with what would in effect be robo-taxis. Drivers today make, what, $10-15 per hour? So one would need to explain to me how taking that cost alone out of the equation suddenly makes using taxis and limos in place of private auto ownership an economically sensible thing for the majority of people in the majority of places where it isn't today.

The cost of a parking space in an urban center is also a large part of the equation.

The comparison is between a digital computer and an organic one (i.e. a human driver). Furthermore, one can't just assume that a car can drop a person off and head out to the suburbs to wait. That's a huge potential cost in both congestion and energy.

You say that like 3-5 wouldn't already be insanely good. Even 1.1 would be a huge decrease in pollution. 0.4 and it might still be worth it due to fewer casualties and the ability to safely multitask.

I'm not talking about whether it would be good. I'm sure it would be great. I'm talking about whether the introduction of autonomous cars will result in everybody ditching car ownership and using companies like Uber for transportation, which a lot of people seem to think is a given.