"One is the reduction in accidents as most accidents are really due to some form of operator error/stupidity."
As human drivers generally drive very safely (I don't know about US figures but as far as I can make out the fatalities in the UK must be of the order of one per 100 million* miles). This is a pretty tough target to set for an autonomous vehicle. In fact, the figure suggests that accidents are corner cases, the driver failing to cope with an out of the ordinary situation. Experience suggests that dealing with corner cases is something S/W isn't particularly good at. In addition inexperienced drivers are more likely to have accidents than experienced drivers. One should reasonably expect experienced and sober drivers to be somewhat better than the average. If I were to trust my life to an autonomous vehicle I'd want it to be at least as good as an experienced and sober driver; I have no confidence that this will be achieved for a long time if ever. Meanwhile I'm quite happy for an experiment like this to take place somewhere where I have no intention of ever being so the US fits that quite nicely.
*I base this on there being c 30 million vehicles in the UK and c 3,000 accidents and assuming 10,000 as the average mileage to get an order of magnitude figure.
Biting the hand that feeds IT
