Public transport is going to be largely built on self-driving cars if/when they enter the market. It's the perfect fit:

a) Personnel costs dominate human-driven PT cars

b) huge vehicles because of a)

c) vechicles are on average mostly empty because of b)

d) the average co2 emissions per passenger-km are terrible because of c)

e) routes and schedules are fixed because of b)

f) big vehicles waste much more time on stops (50 passengers -> worst case 100 stops)

The self-driving PT sevice may look like ride sharing airport shuttles with more or less dynamic routing and pricing.

Someone should set up a remotely operated offering based on this, to get a head start before the AI gets good enough. You can tell the investors you'll gamify it to make it more sexy :)

Here's a clever large scale pilot project that comes very close, though with human drivers: http://www.zdnet.com/article/death-of-the-car-the-tech-behin...