There has been as much hoopla about driverless cars coming on to real roads over the last couple of years as there has been about VR. In the previously mentioned Markoff book,Machines of Loving Grave, there is a whole chapter dedicated to how close we are to seeing them on real roads and the blogosphere has been buzzing about them. And yet, you'll get occasional pieces like this one about a fixed gear bike that can confuse the system which make you pause and wonder whether we are not as an industry committing the classic mistake yet again of confusing a long view for a short distance.
There are three classes of challenges that have to be surmounted: business model, regulatory, and technical. The first I have a lot of confidence in due simply to the aggregate amount of productive time wasted commuting that could be partially or fully reclaimed when the robot drives the car. The second will somewhat depend on the third but should fall into place absent some heinous lobby— and given that Detroit is even committed to the vision, it seems unlikely here.
It is the third class of challenge, the technical and engineering bits around getting a vision system that is reliably given the sheer heterogeneity of what a self-driving car can run into in the wild that seems most difficult to handicap. I have no doubt that at some point this will happen but would love to understand better whether this point is two years from now or twenty. Come t0 think of it, there are probably a half dozen industries which would be so radically redefined by a fully autonomous vehicle that knowing whether it is 2, 20, or some number of years in between would be of quite some value.