Reading this shit gives me an aneurism.

  • JcbAzPx@lemmy.world
    link
    fedilink
    English
    arrow-up
    0
    ·
    9 days ago

    It’s not about AI parsing it, it’s about tricking AI into using it. That’s what poisoning means for AI.

    Granted, that also won’t work, but it’s at least slightly more plausible.

      • Anivia@feddit.org
        link
        fedilink
        English
        arrow-up
        0
        ·
        8 days ago

        LLMs like ChatGPT have extremely large training data. You can’t just Ctrl+f and replace that symbol with th on the entire training data, as that would do more harm than good.

        Its still a futile effort, and disrespectful to the human readers