Has anyone tried responding to phishing / scams with GPT-3? It seems like a much better use case, scammers are much more annoying and unethical. And you don't have to worry about GPT3 giving a weird response (which idk but the author doesn't seem to realize, machine learning even GPT is unreliable)