Earlier quoted context omitted.
They chose to use a machine, ignoring personal responsibility to control it, killing someone because they bought into marketing hype.
But if you get into an uber and it kills someone that's fine? Why does putting an overworked and easily distracted human in control absolve you of consequences more than putting an occasionally broken ai?
Despite the hype, they don't work, and aren't going to any time soon.
The FPGA link here is an interesting window into why: https://news.ycombinator.com/item?id=37552365 https://news.ycombinator.com/item?id=33402603