(Yes, the oft drunk Libertarian listens to NPR.

A main question they brought up was how the AI should react when an accident was inevitable. (Likely due to a non-AI equipped car)
If you were in your AI car, and someone did something that would cause your car to take extreme evasive maneuvers, and there were large amounts of pedestrians nearby, should the car
A- Do the best to save the driver, even if it means running over pedestrians,
B- Do the best to cause the least amount of harm, avoiding the pedestrians by running into a brick wall killing the occupant.?
I know the main premise for this forum is reduction of harm, and I tend to side with that, but for this situation there is a question to be asked... If the self driving car puts reduction of harm as it's default, (Kill the driver to avoid killing more pedestrians) how many people would purchase the car?
In my typical "no solid rules" fashion, I'd like to hear thoughts on this.