Look, let's say a Tesla, oh, just as a random example, careens directly into a clearly visible truck because its cameras can't distinguish white trucks from bright sunny skies, killing "my daughter Susie." In this hypothetical, this is a truly autonomous Tesla, and overall, these truly autonomous Teslas are less crash-prone than the human average.
As Susie's dad, my argument is that this accident was caused by Tesla. Yes, Tesla may have also saved some other people, under different circumstances, circumstances which are harder for humans to deal with and easier for cars to deal with. So what? That doesn't change the fact that Susie was killed by their product misbehaving in a way that a basically competent human driver wouldn't have. It's not like Tesla previously saved Susie's life, and at least she got to live longer than she would have in the counterfactual -- it's not like all the Susies out there in the world are all essentially doomed to die in auto accidents today.
If Susie went to a doctor with a mild infection and he gave her an inappropriate drug and it killed her, you had better believe that I would sue the shit out of him and also want him prosecuted, even if in the year before that he saved 37 other kids' lives from other illnesses.
But that's the difference of being against self driving cars, and a specific model of a car (for which it's manufacturer is liable). The OP was saying, to use your analogy, we still use doctors in the U.S. because it changed deaths from medical problems from like 90% to 400,000 people killed every year by accident.
