1) It's not real-time: it would likely refresh data every couple of weeks. That limits the number of applications.
2) The data gathered would be... well, my lack of imagination is stumping me here. The number and species of trees people have? The color of their houses? New business openings? Typical length of lines outside the door? The existence and extent of potholes on the road?
3) I'm sure even expanding street view would be valuable. But the leap from effective autonomous vehicles providing street view data on public roads to autonomous vehicles providing transportation to real consumers seems pretty small to me.
I have a strong hunch on this one that the limitations are on our imaginations, and not on the potential applications once such an infrastructure existed.
I agree that the leap to transportation is small and that it will happen. I just can't see why Google would have been motivated to do it for that reason, but I can see World Crawlers as a natural evolution of their core mission.
It seems to me that a fleet of driverless information-gathering cars would be a natural component of a world with driverless cars. Primarily they'd be used for continuously mapping roads for any potentially useful information or dangers, eg lane closures, construction, potholes, etc. From there, though, there's no reason you wouldn't naturally start gathering more information: traffic, weather conditions, strength of cell/wifi signals, ambient noise levels, and all sorts of other things. A highly distributed network of sensors could provide all sorts of interesting information.
(Note that there's no reason that they'd even have to be empty cars. I see no reason why every car couldn't or wouldn't be decked out with sensors and reporting back to Google.)
