On the other hand, Uber and Lyft externalize many costs including liability.

Externalizing liability and automated driving seem quite at odds unless Uber somehow manages to bypass laws again.

Is this not what effectively everyone who is doing this (outside of Tesla) is looking at?

As a taxpayer who pays for roads, and suffers from traffic congestion caused by one-occupant and zero-occupant vehicles, I'm eagerly looking forward to reducing the taxes I pay, by taxing those margins, instead.

Ideally, the taxes could be high enough that driverless taxis will operate at barely above break-even. The financial comfort of me and my neighbours are more important to me than the profit margins of a firm that barely employs anyone in my town.

Unlike a factory or a corporate office (that can threaten to move offshore, eliminating jobs and impoverishing a town), the firm in question is a hostage of local politics - not the other way around.

My cynical take: the government is not going to forgo collecting a tax from you that you are already paying. Instead it will tax you and start collecting per-ride fees from Cruise, etc.

Do you think your experience of congestion would be improved by everyone driving private vehicles instead? Not sure I follow the logic here.

Yes, because in the common case, a taxi (driverless or otherwise) drives empty at least some of the time, to pick someone up, thus creating congestion, compared to a private vehicle, which doesn't drive empty.

The cheaper and more convenient you make zero-person and single-person automobile transportation, the more people will use it, and the more congestion they will create.

The more expensive and less convenient you make it, the more trips will use non-automotive, or public transportation, both of which produce far less congestion.

I actually agree with everything here, but on the other hand the decision of whether and how to actually build the massive amounts of non-car infrastructure we need to have transport be efficient and accessible without private cars of any kind, is in a whole different place. At least in the US, it's pretty clear that in most areas there is very limited political will, even in the grass roots, for things like "build good high-speed trains" and "dig new billion dollar subways" etc. So I think pragmatically speaking things like robotaxis are going to be the "solutions" that we'll actually get.

(And yes, I agree that that's dumb since the same politicians and voters have no problem indefinitely subsidizing and expanding the massively money-losing infrastructure called Roads at taxpayer expense!)

On the other hand, once a sufficient percentage of cars on the road are autonomous, couldn’t they use cooperative navigation algorithms to improve throughput a whole lot?

There are so many inefficiencies with human drivers—chaotic merging, unnecessary lane changes, blocking of passing lanes, and so on. I could imagine that optimizing all those away would make a huge difference overall.

You could also probably increase speed limits. And fewer accidents should cause a significant reduction in traffic jams.

Of course. If one assumes relatively inexpensive rob-taxis people living outside cities will definitely come in more often. I certainly would.