The thing that sometimes gets lost in the debate around Tesla autopilot (IMO), is that it's possible for self driving cars to be theoretically better than human drivers, AND that Tesla's implementation of that concept to be overall worse.

From the recent accidents it seems clear that Tesla autopilot is just not that great in it's current form. Even if it's following a line that it mistakes for a lane marker, how is it ok that it can't stop for a large obstacle in front of the car when given plenty of distance before a crash? How is this not a totally unacceptable type of failure for a system like this?