The Google car's 300,000 accident-free miles demonstrates that we are technologically quite close to a level where we can have truly autonomous cars, even if the regulatory and societal frameworks need some time to catch up. And yes, we are also at a level of truly profound change in many, many different areas (such as the conversion of the middle-class salaried job into a freelance mixture of Ubering and Taskrabbiting). I think the time you're describing is already here.

How many of those accident-free miles were in uncontrolled conditions? How many of them were negotiating traffic in a rainstorm, or unplanned detours around construction, or getting boxed in by semis on the interstate, or dealing with people maybe or maybe not about to jump across from the median, or anything near the sort of calamitous and hard to manage situations typically leading to human accidents or fatality?

The Google car is brilliant but from what I understand "quite close" is stretching it.

The Google car project is what, 5-10 years old? I assume that in 5-10 years it will be more capable than a human driver across the naturally encountered spectrum of driving conditions. That's "quite close" to me.

The average guy or gal on the street across much of the US has never heard of Taskrabbit and many have never heard of Uber. And, frankly, that's a pretty dystopian future you describe. Fortunately, it's not here for most people. And if I drive or walk in a large city and try to imagine an AI navigating that chaos from door to door, my mind boggles. Not to take away from the impressive work that has taken place over the last five to ten years but the "replace human 100%" use case seems really really challenging--and far out.