• Opinionhaver@feddit.uk
    link
    fedilink
    English
    arrow-up
    3
    arrow-down
    1
    ·
    2 days ago

    Have you actually tested these tools? Because while they’re not flawless, I’d still say that even in their current form, they’re pretty damn good.

    I love how quickly we get used to things. ChatGPT would’ve been considered straight-up magic ten years ago - and now we’re just shitting on it for the edge cases where it gets something blatantly wrong, while completely dismissing the countless times it does exactly what it’s supposed to.

    • SchizoDenji@lemm.ee
      link
      fedilink
      English
      arrow-up
      1
      ·
      12 hours ago

      I have, they degrade very significantly whenever the context goes to 8000+ tokens.

    • nimpnin@sopuli.xyz
      link
      fedilink
      English
      arrow-up
      5
      ·
      2 days ago

      It’s approaching the lower levels of human reasoning, which, as we have realized over the past few years, isn’t that impressive.

      • Infinite@lemmy.zip
        link
        fedilink
        English
        arrow-up
        2
        ·
        2 days ago

        Yeah, I bet at least half of the major models would fall for disinformation and vote for fascists. The Turing test is seeing if the AI will commit hate crime using legal justification, right? Definitely by next year.

        • Opinionhaver@feddit.uk
          link
          fedilink
          English
          arrow-up
          1
          arrow-down
          1
          ·
          edit-2
          1 day ago

          I bet at least half of the major models would fall for disinformation and vote for fascists.

          And what is this assumption based on?