Yes exactly the point being made in the link. AI picks up unintended behaviours already, it's not too hard to imagine that the general direction is towards model collapse.