What are some of your ideas on how to best use AI (specifically those like LLMs) to benefit humanity as a whole? - eviltoast
  • howrar@lemmy.ca
    link
    fedilink
    arrow-up
    1
    ·
    1 year ago

    What do you mean we can’t explain it? It’s designed specifically to make up some text that is very statistically likely. If it doesn’t have anything similar in it’s training data, it will try to extrapolate, and that gives you hallucinations.