Where I think we agree is that cars mostly suck for cities, and cars that get self-driving tech tacked onto them doesn't solve most of the fundamental issues with cars as we know them today.
I think the question here is what is a practical roadmap to building better cities with the ones we've got?
An extreme distillation of my pov is: roads&rubber wheels > rails&metal wheels.
I believe roads&rubber are just far better for A-B transport by so many measures, but most importantly that it is an extremely flexible connective tissue because it can be incrementally and cheaply added to or changed.
We can make the big, slow, expensive changes to add rail, and that may (mostly) work (for some), but it is extremely unlikely to happen at scale because it essentially means upending how car-centric cities are designed using top-down process or fiat.
Meanwhile cities are living breathing things pulled in many directions by many forces. The question is whether those forces can be harnessed to bring cities toward what we want them to be. I think roads&rubber do that better, because they allow bottom-up, rapid incremental change that shares costs with citizens (by way of vehicles) and does not require anywhere near the level of commitment or consensus that rail does.
I think it's reasonable to assume that if self-driving tech starts showing up in 10 years, that in 30 years we'll begin to see several changes that push us towards our goals for cities:
* There will be cities that require only self-driving cars in city limits, which will begin the alteration of the rules of traffic flow, since rules can be reprogrammed instantly
* product (car ownership) will give way to utility as all tech does over time
* There will be roads that will be converted to mixed use (parklets, mini-parks, bike lanes, walking) and be limited to only smaller micromobility-like vehicles. An early version of this is SF's 'slow cities' streets in each neighborhood that have been cut off to (most) cars.
* Max size requirements & road size changes will start to be palatable, with roads being recaptured, property lines & zoning shifting, etc.
* longer commutes/travel would use (larger) vehicle types that specialize in long distance comfort whereas short commutes would use (smaller) vehicles that have higher throughput in high density areas.
And more. This is the kind of payoff we'd start getting by investing in self-driving tech today. And these are things that can be done in a lazy and politically simple way, making changes only as they make sense in the moment, not on long planning schedules and anticipations of ridership and population proximity in 25 years.
I doubt that in the same ~30 year span that you could get most cities to install more than a few light rail corridors, and they'd probably not have sufficient throughput and frequent enough stops. We'd have to get like 5-10x better/faster at big infrastructure projects including politically challenging things like eminent domain to get comparable results.
LA metro is a good example: remember the metro construction scene at the end of the movie Speed? That was 30 years ago, and in that time metro has eased maybe ~5% of total traffic and you still absolutely need a car to get around LA. How do we get the other 95%?
The ideal answer in my mind is neither cars nor rail, but a system that operates a lot like rail but flexibly sized and individualized, and using the basic roads&rubber approach of cars.