Robots Are Killing People. How Worried Should We Be?

The main metric we should use is comparing the mortality rate of the AI to the human. If the morality rate is better, then effectively human lives have been saved.