The technological pre-reqs for getting autonomous cars on the road are higher than you think
It's easy to get caught up in the headlines of progress and imagine that cars are about to roll off production lines and out into the streets, but there is a lot of work that Google are currently putting in before they ever get a car into a new location. The level of mapping, for example, is way above what they do for just standard Google Maps: a process that even they admit cannot scale out to the rest of the USA from the locations they are currently trialling.
Slightly alarming, for example, is Chris Urmson, director of the Google car team, saying that if the car came across a traffic signal not on its map, it could potentially run a red light, simply because it wouldn't know to look for the signal
Those million or so miles the cars have clocked up? Over and over the same trial areas.
How much of the general intelligence in AGI is required in cars in order to handle the dynamic world at large? I would hope the requirements aren't near that high or they will indeed be much further off than some might expect.
But whether their limitations of a dynamic world are as stated, and even with their limited driving areas, the fact remains that they've been clocking up distance over a reasonable period of time now, and they've managed to handle that area quite well so far.
Whether they solve the mapping scalability issue any time soon, though ...