It's always struck me as obvious that driving will be automated in order of difficulty of job. First, trains (already done for many of them). Then buses in "BRT" systems (where the buses have a dedicated lane) and then buses on regular roads. Only after all of these have become routine sights will you see your automated Car2Go -type taxi services.
But just automating mass transit will increase the use of it. Why are trains lumped together in 3 cars that only come by every 15 minutes? To save on drivers. One car every 5 minutes is the same capacity but one-third the waiting time.
And you could be getting to the station from your house from bus stops where a small van comes by every 5 minutes, too. Chopping out that time-consumption (and where I live, COLD waits for half the year), would probably double interest in mass-transit right there.
Effects that make mass-transit more appealing have a positive feedback loop effect going for them, because of the same "network effects" that drive adoption of new popular communications like fax then E-mail the social media. If twice as many people take the train, then it comes every 2.5 minutes, and they start building tracks to more places.
Meanwhile, there's then a positive feedback loop hurting the car industry. The fewer people buy cars, the more expensive they get and the more likely your employer is to charge you for parking, because only half the employees even use it, and why should you be subsidized? These positive feedback loops can lead to "tipping points" more quickly than most people would tend to predict.