Think about it... Google could train its AI to recognize malicious behavior for moderation purposes, to recognize astroturfing and manipulation of public opinion (for purposes of stopping it or carrying it out).

There's not much value in training it to simulate a redditor, but there's a lot in modeling bad behavior.