Re: I'm not sure I understand @AMBxx
The problem with all those scenarios is this: Assuming they would even happen.
Take the example from the article. The likely reason for a situation in which a car is about to plow into a vehicle ahead of it is driver inattention or following too closely for the conditions, speed, etc. A computer doesn't take a "quick sec" to gaze at the phone nestled against their crotch. It doesn't have a BAC of .04 that slows reaction time to require a greater following distance, or causes someone to do stupid shit like tailgate.
Also in the real world, cars going 35 mph stop pretty damn quick once the brake is applied, which is the only scenario in which there is likely to be a group of kids on the side of the road. Once again, the limitation is the meat sack in the driver's seat who was too busy digging for the last fry in the McDonald's bag.
"Unavoidable accident" is just a phrase people use to reduce their liability in court or make their conscience shut up. Kids don't materialize in the road; they came from a yard or park 10 seconds earlier that an attentive driver would have seen and made the appropriate behavior modifications when approaching, like slowing down. The same kind of down-the-street evaluation can be done by a computer, and might even tag squirrels and bunnies if the resolution is good enough.
And as far as mowing down people of the wrong skin color, well, I don't know where you drive, but it's pretty rare for anyone to be in the middle of a lane where I am. With a tiny sample size, it's easy to get skewed numbers.