• David GerardOPMA
    link
    fedilink
    English
    arrow-up
    25
    ·
    4 months ago

    yeah the use case for LLMs is amusing novelty. That’s why LLMs peaked with GPT-2, it was just the right amount of broken.

    • Roamin' Chemicals@mastodon.social
      link
      fedilink
      arrow-up
      13
      ·
      4 months ago

      @dgerard @visaVisa I think there’s a world where I could have had fun playing with generative AI as a toy, marvelling at what statistics could do and navel gazing about what the outputs said about the societal biases that generated them. But I don’t imagine I’d enjoy Lego, either, if people spent all day whipping them at me.

      • David GerardOPMA
        link
        fedilink
        English
        arrow-up
        14
        ·
        4 months ago

        I’ll go so far as to say that one of the bad things about the AI bubble is that LLMs aren’t fun toys any more.