[gohrt]

...and they will defend against any accusations of discrimination or abuse by saying "no PERSON here did anything wrong -- we use a fully automated AI Machine Learning logistics system that optimizes our warehouse and delivery operations. We don't know why it picks which items to give to which customers"

This is the Unfriendly AI that people like MIRI should be worried about.

[TeMPOraL]

Ironically, it seems that even if Friendly AIs were the default state in our reality, malicious actors (like advertisers) would iterate on them until they get an Unfriendly one...