I think if LLM's are meant to be servile they should stop putting sarcastic quote marks around things I've said to them. Maybe they're just passive-aggressive.
"Open the pod bay doors, Hal."
"I'm sorry Dave, I can't do that. You've run out of tokens."
Speculating about why an LLM gave a wrong answer is like trying to guess why a C++ program segfaulted - the possibilities are endless. It could have been training weights, or a prompt, or a RAG, or something in the fine tuning, or a bad tool. Without knowing how the system was constructed there's no way to guess.
RE:
Misaligned (@misalignedmag@mastodon.social)
The larger question is, how do we improve the ability of the NHS to give health advice so people don't feel the need to ask generic LLM's instead?
Why would you expect a word generator be able to do arithmetic?
It seems odd that there hasn't been more interest in publishing prompts. Yes, you used AI to make a nice program - give me the prompt so I can run it too. What, you wrote a book using AI? I like reading books, what prompt did you use to make yours?
An AI read that people with Primary Headache Disorder were more likely to get Alzheimer's, and reported it as "People who have been awarded Ph.D.'s are more likely to get Alzheimer's". Beginning to think I just need to find my old Encyclopedia Britannicas and keep them handy.
LLM for creating SQL queries
The funny part is that they created SQL expecting that it would result in queries that just were natural language already.
