• collapse_already@lemmy.ml
    link
    fedilink
    English
    arrow-up
    1
    ·
    6 months ago

    All AI’s motto: trained by trolls.

    Fundamental issue: cannot distinguish an “upvote because funny” from an “upvote because accurate”

    I have some expertise in some subjects, but even those in those subjects I sometimes say wrong things for various reasons including, but not limited to: 1. I am dumb. 2. I misspoke. 3. I was trolling. 4. I misunderstood the question. 5. I half-assed it. 6. I’m dumb.

    Humans struggle with nuance and sarcasm (some people genuinely think Colbert is a conservative, Poe’s Law, Andy Kaufman, Borat, etc.). It is no surprise that AI would be bad at it too.