• CyberEgg@discuss.tchncs.de
    link
    fedilink
    English
    arrow-up
    4
    ·
    6 days ago

    I know there are reliable AI applications, for example cancer recognition models scanning CT imaging.

    However, from what I know systems trying pattern recognition in human behaviour are the opposite of these medical implementations.

    • glasratz@feddit.org
      link
      fedilink
      English
      arrow-up
      6
      ·
      edit-2
      6 days ago

      I know there are reliable AI applications, for example cancer recognition models scanning CT imaging.

      Those are not reliable at all and only meant as hints for the radiologist to check certain areas twice. They do not identify cancer but spots that may be out of place. However, not all radiologists seem to know that and a relative of mine would have been pushed into end-of-life care if not for the intervention of another doctor in the family. He had to take it up to the head of department, though. The spots in question turned out to be old scarring and could have been easily identified if the radiologist had also consulted older scans as would have been her damned job. Instead she thought the AI was thinking for her.

      Similar software also exists for spotting tooth cavities. It’s useful, but not to be trusted.

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

        I give up.

        Apparently you do have to caveat every little fucking case for every screwed up situation that could ever exist. OBVIOUSLY these scans have to be double checked and OBVIOUSLY “reliable” does not mean “never ever could something go wrong” and OBVIOUSLY shitty or overworked personnel makes wrong decisions.

        But if you have to account for all that in every yet so minor argument you want to make you’d have to write at least a bachelor thesis level publication about the weather being warm and sunny outside my window right now.

        You cannot have a basic discussion about any topic on this level so screw it, I’m out of this.

        I’m fucking autistic and love to get into unnecessary depths of details but this is too much even for me AND YES I KNOW autism is a spectrum so YES OBVIOUSLY there will be some people WHO MIGHT NOT EVEN BE AUTISTIC who love it even more to discuss these details but this is just dumb and I just can’t anymore.

        • glasratz@feddit.org
          link
          fedilink
          English
          arrow-up
          4
          ·
          6 days ago

          Don’t you feel that this is a bit uncalled for? All I said is that these pieces of software do not in fact identify cancer. They are not reliable because they aren’t even allowed to be in the circumstances that they are supposed to be used. They are a visual aid, nothing more.

          • CyberEgg@discuss.tchncs.de
            link
            fedilink
            English
            arrow-up
            2
            ·
            edit-2
            6 days ago

            Every article and paper I found so far concludes that AI improves the ability of medical personnel to recognize and diagnose cancer.

            And no, it is not uncalled for, because I had to deal with such bullshit the whole damn day (and not for the first time) and it is enough. I’m logging out for today.

            • glasratz@feddit.org
              link
              fedilink
              English
              arrow-up
              3
              ·
              edit-2
              6 days ago

              Every article and paper I found so far concludes that AI improves the ability of medical personnel to recognize and diagnose cancer.

              I’m not saying it doesn’t. I’m saying it doesn’t detect cancer and is not reliable. If you know how it’s supposed to work, it works. It suggests areas where you should look. However, there are doctors who don’t know how to use it and they can and will kill people with it.

              I had to deal with such bullshit the whole damn day (and not for the first time) and it is enough.

              The way your last post was worded, I would very much suggest that that’s a you problem.

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

      I know there are reliable AI applications, for example cancer recognition models scanning CT imaging.

      Or the cancer recognition model might flag your X-ray as cancer just because it was taken with a particular type of X-ray machine…

      Not that this would ever happen, of course. (it did)