• Swedneck@discuss.tchncs.de
    link
    fedilink
    English
    arrow-up
    27
    ·
    2 days ago

    i always like to call it hallucination, it’s significantly closer to how it works both technically and in effect.

    • usualsuspect191@lemmy.ca
      link
      fedilink
      English
      arrow-up
      18
      ·
      2 days ago

      What messes with me is how many AI videos I’ve seen that are so similar to dreams. The hallucinations that AI produces are very similar to the ones our brains produce, and that makes me feel like more of a meat computer than usual.

      • Cracks_InTheWalls@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        14
        ·
        2 days ago

        What I’ve found even more fascinating is, particularly in earlier iterations of the technology, visual effects produced were remarkably similar to visual distortions people experience with certain drugs.

        Easy to make a lot out of this where it’s not warranted, but at minimum it gives some interesting food for thought re: how visual processing works. Have seen people write about this, but am too dumb to actually understand.

      • easily3667@lemmus.org
        link
        fedilink
        English
        arrow-up
        2
        ·
        edit-2
        2 days ago

        You’re a meat computer and always have been, flesh sack named after a famous abuser

    • Snot Flickerman@lemmy.blahaj.zone
      link
      fedilink
      English
      arrow-up
      12
      ·
      2 days ago

      I think fabrication is a better term than hallucination because of the double entendre of it being industrially fabricated and also being a lie.

      • Swedneck@discuss.tchncs.de
        link
        fedilink
        English
        arrow-up
        3
        ·
        1 day ago

        that removes the reference to how it actually functions though, at that point you might as well just stop being coy and call it “AI dogshit”

      • lime!@feddit.nu
        link
        fedilink
        English
        arrow-up
        5
        ·
        2 days ago

        that’s more of a comment on the usage than on the technology itself.

        remember that google deepdream thing that would hallucinate dogs everywhere? it’s the same tech.

          • Glitterbomb@lemmy.world
            link
            fedilink
            English
            arrow-up
            1
            ·
            1 day ago

            If that’s the case, then we anthropomorphize technology all the time. Like, constantly. How many times has your phone died when its not even alive? How does a phone drop a connection without hands? We feed a computer input and it regurgitates or spits out output, all without a mouth. The examples are endless but hard to immediately pick out, because the usage has changed to be completely commonplace. Even bytes were originally conceived as a play on words with ‘bite sized’ to refer to a small collection of bits. I don’t necccessarily defend these ‘AI’ tools, but policing the language people use ain’t it. Changing the word hallucinate to refer to a part of technology is exactly how language has functioned since always

          • Swedneck@discuss.tchncs.de
            link
            fedilink
            English
            arrow-up
            3
            ·
            1 day ago

            that’s literally how it works though, the software is trained to remove noise from images and then you feed it pure noise and tell it there’s an image behind it. If that’s not hallucination idk what would be.

          • lime!@feddit.nu
            link
            fedilink
            English
            arrow-up
            7
            ·
            2 days ago

            so is calling it fabrication. something incapable of knowing what is true cannot lie.

            also, gpts and image generators are fundamentally different technologies sharing very little code beyond the basic matrix manipulation stuff, so the definition of truth needs to be very different.

    • Rhaedas@fedia.io
      link
      fedilink
      arrow-up
      5
      ·
      2 days ago

      Yes, good point, and it’s incredible that so often the hallucination is close enough that our pattern-matching brains say, yes, that’s exactly right!

      • Swedneck@discuss.tchncs.de
        link
        fedilink
        English
        arrow-up
        1
        ·
        1 day ago

        eh is that really true though? in my experience our brains tend to go “wow, this looks exactly right but there’s something ineffably off about it and i hate it!”