It’s a curious thing that while most people are abysmal drivers, we also expect 100% perfection from automated driving systems. The reality is that if AI driving systems were only 10% better than humans (not a difficult attainment) then many lives would be saved daily. As for the philosophical arguments over edge cases (“if the car has to decide between saving the driver or saving a child who’s stepped out into the road”) these are all redundant. Human drivers merely freeze and plow into whatever they were pointed at before they froze. There’s no choice being made, no abstract decision criteria being applied.

The road to automated driving lies through the legal landscape rather than primarily through the technological landscape, because it’s where the real battles over liability etc. will be fought.
Meanwhile, we humans with our inept driving styles and slow brains will continue to slam into each other on a daily basis, while telling ourselves that we’re all “better than average” drivers even though we’ve taken no advanced driver training nor even refreshed the rudimentary point-and-lurch skills that enabled us to pass a ridiculously easy driving test in the first place.