• Tar_Alcaran@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    18
    arrow-down
    1
    ·
    3 days ago

    Your argument would be good, if LLMs got it right. But they will just as happily inform you that drinking your own urine is good for you, if you ask them in the right way. A doctor will always tell you no. An LLM will inform you about the dangers of vaccines and how there are many sources that say they give you autism, whereas a doctor won’t.

    Getting validation from the lying machine is not a way to do healthcare. The LLM confirming that you have some condition will not change anything, just like me and my forum buddies all self-diagnosing autism, and ADD doesn’t actually help and actually causes a lot of harm.

    • brackled@lemmy.world
      link
      fedilink
      English
      arrow-up
      13
      arrow-down
      1
      ·
      3 days ago

      To add to your points, LLMs are also not free of biases that the poster above and others have to deal with. The same medical biases are more than likely part of the training data used to develop these LLM. Medical biases are just now being fully understood so that data is no where near as prevalent as the sea of historical biased data.

      • Tar_Alcaran@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        9
        arrow-down
        1
        ·
        3 days ago

        Also, annoyingly, most people are absolutely shit at identifying bias, on both sides of the argument.

        Personal example: when I went to the doctor with a painful lower back, and a weirdly pulling muscle above my ass, he said it was very likely due to my weight.

        Well, that was obviously fucking bullshit weight bias, and he wasn’t taking me seriously because I’m a woman. Besides, I have PCOS, so I can’t help it.

        Except when I lost the weight, the pain went away. And so did the sore knees, and the shortness of breath I didn’t even notice I had. And the PCOS complaints got much better. So… Yeah, it wasn’t a biased doctor, the guy was entirely correct, I just didn’t like the answer.

        And at other times, medical science is just straight up wrong, and we found out after it improved. I had some amazing semi-permanent birth control implants, that unfortunately turned out to occasionally rip your insides to shreds. Whoops. So I got them removed, which sucked ass. Was that bias? Or just medical science advancing?

    • realitaetsverlust@piefed.zip
      link
      fedilink
      English
      arrow-up
      3
      arrow-down
      5
      ·
      3 days ago

      GPT, Claude etc yes, they will tell you whatever you want to hear.

      But custom trained models tailored to the healthcare industry could be vastly different.

        • realitaetsverlust@piefed.zip
          link
          fedilink
          English
          arrow-up
          1
          ·
          2 days ago

          … Yes, if we develop these custom models that don’t exist yet and give the public access to it, that would be the case.

          • Whostosay@sh.itjust.works
            link
            fedilink
            English
            arrow-up
            1
            ·
            11 hours ago

            I appreciate you contributing to the conversation but they’re talking about things that are, not could be

            I should not have been as rude as I was.