I understand the concerns that are trying to be presented in this article. But I think the examples they use tend to confuse the point they are trying to make a

I understand the concerns that are trying to be presented in this article. But I think the examples they use tend to confuse the point they are trying to make about data being gathered and used by a a machine to independently make a decision to kill someone. The first example they give is two humans with an intention for harm have programmed some machines to do harm. Which is the human making the decision and not any different. Some subtlety is also missed in the range of data being processed by AI/machine learning and presenting intermediate conclusions to a human (who still has to make the final decision) that may or may not be correct or inherently biased. I'm surprised there is no mention of something like Google Project Maven in the article.