• heavy@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    36
    ·
    1 day ago

    To add that the general understanding of how DNA works and is used can be scary, just like other measurements. I bet there’s still a lot of people that believe fingerprint analysis is some kind of rock solid science based evidence, but my understanding is that it’s very much prone to errors and interpretation.

    I don’t mean to say that DNA analysis suffers the same flaws, just trying to illustrate with an example.

    • Windex007@lemmy.world
      link
      fedilink
      English
      arrow-up
      20
      ·
      edit-2
      1 day ago

      I hate the generalized concept of “AI”, but I love the concept of “Machine Learning”

      If you think LLMs are good at anything, I am almost 100% certain to disagree with you about pretty much everything, to help you understand this distinction.

      Anyhow, some computer scientists found that a machine learning algorithm could predict beyond a null hypothesis that A fingerprint belonged to a person given a different fingerprint (different finger but still same person)

      “Criminology” expers were just like “no, it’s settled science”

      This is the state of discourse.

      1. why do I even feel the compulsion to preface by saying my bit about ai and llms?

      2. how tf is “settled science” even a concept in a science

      • prime_number_314159@lemmy.world
        link
        fedilink
        English
        arrow-up
        11
        ·
        1 day ago

        I get a similar vibe from psychology. There’s a number of “experts” that are out in the field, doing the hard work day after day, putting in those hours… And hopelessly blinded by their own confirmation bias and survivorship bias. Clinical therapists in surveys prove very willing to overlook strong research in support of certain methods because they believe they see results in their clinical work that can’t be reproduced in a lab.

        Then each field also has a research wing, slowly carving a path towards useful ideas, expending tremendous effort for each new finding, method, and result (even negative results!).

      • ericwdhs@discuss.online
        link
        fedilink
        English
        arrow-up
        6
        arrow-down
        1
        ·
        23 hours ago

        If you think LLMs are good at anything, I am almost 100% certain to disagree with you about pretty much everything, to help you understand this distinction.

        Depends on what you mean by “anything.” The current obsession in the tech world of trying to shove LLMs into the AGI box? Yeah, not a good fit. Pure language stuff like translation or brainstorming? Very useful. LLMs now even surpass DeepL.

        why do I even feel the compulsion to preface by saying my bit about ai and llms?

        I have a similar compulsion to clarify that my interest in LLMs centers mainly around local open-source models that can run on consumer hardware.

    • sudochown@programming.dev
      link
      fedilink
      English
      arrow-up
      11
      ·
      1 day ago

      Same with bite mark analysis, polygraph, and bullet/gun rifling matching. CSI, Law and Order, etc. all have convinced people these things are just the pinnacle of evidence.

      • LH0ezVT@sh.itjust.works
        link
        fedilink
        English
        arrow-up
        4
        ·
        15 hours ago

        At the end of the day, nothing is really beyond any doubt. Witnesses can imagine things, cops can be bribed, judges can have a newborn kid and maybe slept 3h last night