I was on a Three Stooges fan site, and one of the members asked ChapGPT about the Stooges’ stunt doubles. It claimed that one stuntman was killed during filming one of the Stooge shorts (he wasn’t and in fact lived about thirty years afterward.) It also claimed a woman did some of Moe’s stunts and that she had been a burlesque dancer. I don’t know if she did Moe’s stunts or not, but she was a fairly common face as an actrees in old movies, and was always about as broad as she was tall, so I highly doubt she danced in burlesque.
This one example doesn’t give me good impression of AI. I don’t care for the style of writing of some of the other examples I have seen.
To be fair, large language models like ChatGPT work better when you give them specific and precise queries. Thus, this video and test wasn't completely "fair."
That's said, everyone on this thread "got it" with a few words or phrases. It also appears that some of our kids only need a brief prompt to launch into a perfect Python soliloquy.
ChatGPT flat-out hallucinated on the speaker of the Newt and Soiled my Armor quotes. I can't call the Comfy Chair response a hallucination, but the response was certainly word salad.
Failing the Monty Python test may seem trivial, but it should give pause to the people dazzled by this shiny, blinking object who trust it as infallible technology.