Good AI / Bad AI #4

ChatGPT on ChatGPT

Near the end of my last post I wrote:

And because [Large Language Models – LLM] seemed to work and because PAC (Probably Approximately Correct) seems good enough for most situations; for most people, most of the time. But I would argue that it is dangerous. And even if it is only dangerous occasionally, to some people, some of the time, those people have a right to know what they are getting.

I thought it would be instructive to see how a LLM – in this case ChatGPT – saw its answ...