There is no such thing as transparency into large language models. It’s a billion numbers that don’t mean jack shit to anyone, human or machine. In other words: it’s Chinese to us.