I don’t see much of a future for self-driving passenger vehicles. There are two non-tenable options for crash avoidance algorithms. Either the algorithm prioritizes my life and property (which means it would kill someone else to save my life … good for me, bad for society) or it won’t (great for society, but am I going to pay money for a car that will literally kill me to save someone else?). Does the computer assisted human driving model suffer this flaw? An algorithm that engages the brakes any time there is an obstacle within X feet fails to consider the vehicle that is about to slam into the side of your car if you don’t move it into the shrubbery ahead of you.
Self-driving unoccupied vehicles can simply de-prioritize itself (and the owner needs to accept that financial risk). We may see driving as a service (DaaS?) where a real human is responsible for making these split-second decisions. But allowing people to achieve the metro experience in their own vehicle (i.e. you sit and work for half an hour whilst your conveyance delivers you to your destination) is probably not going to happen.