Charles Moore:

While there is an apparently near consensus a priori assumption afoot in the tech community that self-driving cars becoming a consumer reality is both a foregone conclusion and will be a good thing, a white paper by researchers the University of Michigan Transportation Research Institute investigates several important issues associated with autonomous autos. And while it’s not the central focus of the paper, entitled: “Potential Impact of Self-Driving Vehicles on Household Vehicle Demand and Usage (Schoettle and Sivak), Report No. UMTRI-2015-3, February 2015,”the report coauthored by Dr. Michael Sivak—a research professor in UMTRI’s Human Factors Group with Brandon Schoettle—observes that there are still vast technological gaps that have to be bridged before self-driving vehicles should be turned loose on streets and highways.
 One of the most important is the still very much open question of how a self-driving car’s operating program will be able to accommodate the unpredictable nature of roadway dynamics such as how they will relate to other drivers, pedestrians, and changing—often challenging—weather conditions. What happens, for example, when road markings disappear under coatings of snow and ice, as they have been pretty much the whole first half of February on highways in my neck of the woods. For that matter, what about dirt roads that have no lane markings at all and precious few road signs, such as the road I live on? Reliance on GPS isn’t going to help much there. And what will a computer make of a police officer at an intersection directing traffic using hand and arm signals or a worker at a construction site doing likewise?