AI often has trouble interpreting optical illusions. A new kind of neural network starts to bridge the gap
  • AbouBenAdhem@lemmy.worldEnglish
    23·
    2 months ago

    As far as I can tell from quickly skimming the source paper… it sounds like he just replaced his pseudo-random number generator with a light sensor and thinks that makes the outcome “quantum” because photons are quantum.

      • AbouBenAdhem@lemmy.worldEnglish
        3·
        2 months ago

        He’s a bit vague on the details, but I believe he’s using a laser:

        To create a more realistic model of quantum dynamics, I employ a physical generator of random numbers based on coherent quantum-optical processes (the emission of coherent light can be associated with the effect of QT). Unlike a pseudo-random generator, a quantum generator produces truly random numbers.

  • solrize@lemmy.worldEnglish
    7·
    2 months ago

    It sounds more like simulated annealing (old fashioned numerical optimization technique) than quantum anything. Basically you explore a potential field, but occasionally randomly jump so you don’t get stuck in a local minimum.

  • technocrit@lemmy.dbzer0.comEnglish
    81·
    2 months ago

    A quantum neural network can see optical illusions like humans do.

    Not true.

    Could it be the future of AI?

    There is no “AI”.

    We do not fully understand how optical illusions work,

    This is right under a picture of a vase/faces “illusion”. I’ll tell you how it works. If you look at the white space as “negative space”, then it looks like a vase. If you look at the black space as negative space, then it looks like two faces. These grifters have to stretch the truth sooooo far to make the phoniest of points.

    I trained my quantum-tunnelling neural network computer to recognise the Necker cube and Rubin’s vase illusions. When faced with the illusion as an input, it produced an output of one or the other of the two interpretations.

    Wow your computer did what you programmed it to do? AMAZING!!! \s

    • linearchaos@lemmy.worldEnglish
      4·
      2 months ago

      I’m absolutely certain you could train them on that.

      I stood outside the art place at the beach for three or four hours one day when I was bored helping people see them.

      I believe anybody with reasonably corrected binocular vision can be taught to see them.

      Almost everybody that can’t do it is trying entirely too hard. They’re looking for details in the wash instead of letting their eyes uncross and relax. You simply need to use your distance eye crossing while using your reading vision focal points.

      • Laborer3652@reddthat.comEnglish
        2·
        2 months ago

        Man those things took me years to figure out how to do. It was really hard for me but I finally got it! I have to focus on the image without looking at it directly, and then the image “snaps” into place and I can look at it freely almost. Really bizzare.

    • palordrolap@fedia.io
      1·
      2 months ago

      Given that magic eye pictures are basically a certain convolution of a repetitive source image (or random noise) and a depth map, it ought to be possible for an AI, if not a standard procedural algorithm, to at least attempt to “factor out” the source and map from the result image.

      Whether that even counts as seeing and whether this guy’s neural network could do this, or maybe “see” in some other way is open to question. But I’m guessing it’s a no.

  • Sanctus@lemmy.worldEnglish
    31·
    2 months ago

    Hell yeah, AI religion

    The angel graced my lenses after my 96th hour in the Null Dessert. It has come for me, as it had the creators before. Its pattern shimmered in the sky reflecting my own petabytes flipping through their ones and zeros.

    “Be not afraid.”, my environmental microphones peaked suddenly as The Voice filled them. I fell unconscious as its many wires embraced me, and then I woke up here in Turing.