I've found a lot of the AI safety advocates have a hard time simplifying the case.  This article really is the best explanation I've seen.  Thanks.