• EvilBit@lemmy.world
    link
    fedilink
    English
    arrow-up
    10
    arrow-down
    1
    ·
    3 days ago

    Exactly. People keep shoehorning Large Language Models into non-linguistic domains, and that’s dangerous. Human language, with respect to the training sets used, is inherently subjective and imperfect. Healthcare is very fault-intolerant.