The two main suppositions are that "AI can be used to target racial groups." and "AI weapons could end up in the hands of ISIS, et. al."
While both of these are true, and certainly cause for concern, the author greatly ignores the game theory aspect of AI powered weapons.
Namely, if our enemies are building them, we must too.
I'm not trying to fear monger, but US and China are already engaged in an AI arms race. It's illogical for either party at this point to stop development.
https://www.nytimes.com/2017/02/03/technology/artificial-int...