Vast majority of accidents are caused by drivers (or cyclers, pedestrians) who fail to safely engage in traffic. For the most part, cars don't kill people - people kill people (or, get themselves killed).
which is exactly why removing people out of the equation is the best way to solve the problem. even fully educated and hyper responsible person is not immune to judgement lapses, hallucinations, heart attacks, strokes, losses of consciousness and many more modes of failure.
what's worst - even if i'm 100% sure about myself, even if i take myself out of the equation - there are still thousands murderous humans driving around me.
Self-driving's problem isn't that it 'doesn't work in many use cases'. The problem is finding a single use case where today's L4 technology can actually be used given the technological, legal, PR and cost considerations.
Exactly - it's more about recognizing the shift that will happen if the technology can work and what that will mean.
GM seems to think it's important since they bought Cruise in the first place, but instead of seeing failure on this as an existential threat it seems more like a side project for them.
From the outside this seems like a bad strategy and will only succeed if it turns out self-driving is impossible (which seems unlikely).
Digital cameras were toys until suddenly they weren't and then Kodak died - even though they had invented the digital camera in the first place.
It's not enough to have the tech, you have to prioritize it in the business. Xerox PARC is another example of this type of failure.
Super Cruise, GM's Autopilot competitor, is planning to be fleet-wide by 2020[1]. But that is L2/L3 stuff, which is different than the L4 technology Cruise is focused on. You can't put L4 into GM's cars because it's not ready - it's probably dangerous and I'm doubtful it would be cost effective given the prices of sensors and the fact that you need to turn the car into a rolling datacenter.