Just a reminder that LLMs have never provided actual answers to any question asked of them or any actual prompt set forth.
They have, however, provided answer-shaped responses, and we as humans are seriously lacking when it comes to telling one from the other.
Folks, a banana-shaped piece of wood is not edible, and I am embarrassed for our species that you cannot tell the difference.
@theogrin Exactly. The first "chatbot" was created in the mid-1960s. It was called Eliza, created by Joe Weizenbaum at MIT. You can easily get Eliza running on your machine and play with it. I first encountered it in 1970 and it was running the classic script called Doctor that mimics a psychoanalyst. I was very impressed, of course.
Weizenbaum was shocked that his secretary, a perfectly intelligent woman, wanted to have a private conversation with Eliza to discuss her marriage. He was so shocked he studied and wrote about how easily humans fall for "artificial intelligence".
The original source code to Eliza has now been found and published in the public domain. It's really quite simple. Clever and creative, yes. Intelligent, no. But it was the first program to pass the Turing Test.
@theogrin Engineer here. I use to use my wife as a sounding board (with her permission). The fastest way to find out if you really understand something is to explain it to someone. Presentation going along fine until "Uh, wait a second. Nope. I'll get back to you..."
Definitely better to have someone else, a real honest human being! The purpose of a rubber duck is mostly to make sure if it makes sense to you.
@shuttersparks
I am very much familiar with Eliza! It's one of my favourite examples of how the human brain can be tricked into seeing a very obvious tool as having someone on the other end. Even knowing exactly how it works, a sort of emotional pareidolia can readily occur.
Not that many of us are truly immune to the appeal of an ELIZA, even if it doesn't talk back. Many an engineer has been given to talking to their rubber duck in the bathtub, such that the shorthand for talking to bouncing an idea off someone who doesn't meaningfully response is called 'rubber-ducking' in common parlance. It's the appearance which counts in those circumstances. (And the fact that a rubber duck can never express its discontent at being a captive audience.)
I mostly just eat a little salt when people express the notion that these things are some kind of actual intelligence instead of a variety of hyped-up Markov chain.