If there is a child running out on the road that is obscured from sensors, a human will see the ball, see others on the yard in horror and shock and be able to intuit what is about to happen.
Or in a more realistic scenario, the human is driving 35 mph in a residential area and still can't stop in time for the child, and the automated car is obeying the speed limit and can stop. But I will certainly grant you there will be times when an automated car kills someone when a human would not. The important thing is whether more or less people in total are killed by automated cars when compared to human drivers.
Your second scenario is exactly why we need legislatures to start deciding these rules now. Your question is essentially how dangerous will we allow these cars to drive. Your ability to sue, or more accurately your ability to win a lawsuit, will depend on the early legislation and case law. Most likely insurance companies will end up covering car companies in the same way they cover individual drivers now. You won't have to sue them, they will just pay you for medical bills, pain and suffering like they would today for a human driver.