Lets say thier algorithms are 1000 times better than a human driver at avoiding accidents, at least statistically. Given that in 2009 over 40,000 auto deaths occured and over 250k childred were injured (http://www.edgarsnyder.com/car-accident/statistics.html) that would translate into hundereds or thousands of multi-million dollar lawsuits against the navigation company per year. Not to mention the countless fender benders that the system would be responsible for. Even dumbasses running into the car at no fault to the navigation company will probably be dragged into court. You think you can code dumbassery avoidance into the car? They will build a better dumbass. You will see this technology in Japan and Germany well before the US simply due to the popular acceptance of these technologies as well as the legal ramifications. State of the art algorithms such as these are not nearly as good as a moderately competent driver under generalized circumstances - as a single example just put a google car on icy snowy roads and see how well its vision algorithm tracks the road or how well it's laser or sonic sensors work when fouled with snow.