Fully automated vehicles, that is.
I don't think "full auto" is required. This is more like Tesla's "autopilot" concept than Google's "driverless" car. This would get used most often on the interstate, not so much in cities, and it's a pretty good fit for that application. I can do some work (or take a nap) between cities and take the wheel a few minutes before the exit ramp. (Or I could program certain conditions such as weather or traffic to trigger an alarm.) But even this level of automation would dramatically reduce highway casualties.
What I'm curious about is how they sense certain road conditions, such as "black ice" that can fool even the most experienced human driver. OTOH, with a broad range of sensing like RADAR and echolocation, you could plow through pea-soup fog without much worry.
cabs are too expensive for everyday use
I'm lucky to live in a place (Taipei) where public transportation is cheap and ubiquitous. Even taxis are plentiful and cheap here. I don't even own a single motorized vehicle. Why bother, when I can get to anyplace I want with less than 20min walking and $2 in fees, and I can get home from anywhere in the city for less than $10, anytime, day or night?
This is where "full auto" is required: bringing this kind of convenience to the broad, "midwestern" spaces of America. When you can make the round-trip to/from your local watering hole for less than 15 bucks, why would anyone take the risk of driving drunk?
I think Google is smart to be investing so heavily in this tech, because once we pass that tipping point, this is going to be the biggest "killer app" of all time. And in the meantime, Tesla is also smart to be pursuing their autopilot tech, because it will be a huge selling point.