I think we'll also see something in between a typical commuting vehicle and a bus. If 10 people in a specific area have regular bookings to go from a two block area to the CBD in a fixed time, then a van will do pick-ups and drop-offs either from houses or from fixed corners (think unmarked bus stops). It will be like demand-based bus routes.
I think we might also eventually see pod-based vans for people that don't want to socialise in any way with others in the vehicle. That said, people cope with the subway OK.
I don't think pollution from driving will be a significant factor with a serious influx in electric vehicles.
2. That's worse than what's happening now. I think we'll see peak pricing motivate changes in schedules and a decrease in the full complement of cars. I also think we'll see companies selling time-share in vehicles so you might buy one of these cars to have priority access, but then earn money from whenever it's available to work for you as part of a network. Those that want to keep one dedicated to themselves could.
3. Staff costs are significant in Australia so here and in similar places, I think taxis will be increasingly popular. Already it's not far off the case that commuting via taxi outside peak times could be more affordable for me with a second car I use twice a day, twice a week. Four taxi trips would cost me $12-17 each ($2500-3500/year). My second car, ignoring the cost to buy it would reach $2500 annually from registration, insurance and maintenance. That ignores depreciation, cleaning, hassle of parking, walking to my parking spot in the rain four blocks away, etc. Take away salary costs of taxis and make them easy to hail and pay by phone and I'd switch so quickly. And I think many people would prefer to travel by autonomous vehicle than with a driver who smells, listens to crappy radio stations or answers their phone the entire trip.