> Turns out natural language models are great at generating natural language.

AIs are basically being designed to hack our brains by using "right-sounding" language to bypass our BS detectors with unearned confidence. This is what con-men do. We're building AI to con us at scale.