"Genlangs" and Zipf's Law: Do languages generated by ChatGPT statistically look human? — LessWrong