Large language models largely remain black boxes in terms of what is happening inside them to produce the outputs that they do. They have also been a bit of a black box in terms of who is using them ...