• mountainRadish@lemmy.world
    link
    fedilink
    arrow-up
    9
    arrow-down
    10
    ·
    7 days ago

    I disagree, knowledge should be easily accessed by everyone, they should only append it with a warning.

    Some people can’t afford seeing many lawyers and doctors, so it is cheaper to cross refrence with ai and do some online search.

    • nieminen@lemmy.world
      link
      fedilink
      arrow-up
      14
      ·
      7 days ago

      This doesn’t really work though. If it were simply directing you to actual sources that could help out (articles, and lawyer authored blogs or something), that would be one thing, but LLMs are just probability machines. It’ll write out what looks like good sound reasoning and logic, but in reality it’s simply the most probable word followed by the next most probable word, over and over again. It’s using pattern matching, not knowledge

      • dreamless_day@feddit.org
        link
        fedilink
        arrow-up
        3
        arrow-down
        6
        ·
        7 days ago

        Have you used the newest LLM models? They certainly do fact checks, research info and provide online sources - if they are instructed to do that.

        • zalgotext@sh.itjust.works
          link
          fedilink
          arrow-up
          7
          ·
          7 days ago

          They output responses that read like fact checks, and link things that are semantically similar to their output.

          LLMs don’t understand concepts like “fact” and “source” though, so it’s a stretch to say they “do fact checks” and “provide online sources”.

          Always remember, LLMs don’t actually understand anything other than what order words typically go in. They’re really good at it, sure, but don’t let that trick you into thinking they have any deeper understanding than that.

          • nieminen@lemmy.world
            link
            fedilink
            arrow-up
            4
            ·
            6 days ago

            I’ve tried on several occasions to try and explain this to people, and it’s like they WANT it to be a sentient thing, and refuse all evidence to the otherwise.

          • dreamless_day@feddit.org
            link
            fedilink
            arrow-up
            1
            arrow-down
            6
            ·
            6 days ago

            They output responses that read like fact checks, and link things that are semantically similar to their output.

            So basically they do what humans do. What exactly qualifies as deeper understanding to you?

            • wewbull@feddit.uk
              link
              fedilink
              English
              arrow-up
              5
              ·
              6 days ago

              Being able to say “I don’t know” rather than making shit up to fill in what it doesn’t know. That would be a good start.

              • dreamless_day@feddit.org
                link
                fedilink
                arrow-up
                1
                arrow-down
                5
                ·
                6 days ago

                They do say that. I’ve seen Claude realise some programming package must be internal / non public and it therefore can’t provide any information on that.

            • zalgotext@sh.itjust.works
              link
              fedilink
              arrow-up
              1
              ·
              6 days ago

              It’s nothing more than surface level though. Their “knowledge” is limited to whatever is present in their training data, and any claims it makes outside of that are manufactured based on statistics and it’s knowledge of word order, not actual understanding of actual facts.

              Just because it gets it right sometimes doesn’t mean it knows it’s right. Because, again, the only thing LLMs actually understand is word order.

              • dreamless_day@feddit.org
                link
                fedilink
                arrow-up
                1
                arrow-down
                3
                ·
                edit-2
                6 days ago

                No it’s not limited to their current training data. Like I said, they can search the internet, get information from websites or available studies, summarize and use that in their answers.

                Just because it gets it right sometimes doesn’t mean it knows it’s right.

                Again, how do you know?

          • plyth@feddit.org
            link
            fedilink
            English
            arrow-up
            1
            arrow-down
            7
            ·
            7 days ago

            LLMs don’t understand concepts like “fact” and “source” though

            Do you?

    • blackbelt352@lemmy.world
      link
      fedilink
      arrow-up
      7
      arrow-down
      1
      ·
      7 days ago

      Except AI is an over engineered autocomplete operating completely off of statistical models. Knowledge is not statistical or random. I should get the same outputs to the same inputs.

      • plyth@feddit.org
        link
        fedilink
        English
        arrow-up
        3
        arrow-down
        4
        ·
        7 days ago

        Knowledge is not statistical or random. I should get the same outputs to the same inputs.

        Where does the thought come from? If you write five articles about the topic or any topic, will they be the same?

          • plyth@feddit.org
            link
            fedilink
            English
            arrow-up
            1
            arrow-down
            2
            ·
            6 days ago

            Even about topics that you barely are familiar with?

            If the prompt includes the older articles, or if the NN is updated once an article is written, wouldn’t AI be consistent, too?

            • blackbelt352@lemmy.world
              link
              fedilink
              arrow-up
              3
              ·
              edit-2
              5 days ago

              Then I research from known references starting with Wikipedia and using the sources at the bottom of the page to expand my research out further.

              The entirety of how LLMs work is token prediction based on the entire collective of human creation, that is not research, that is not understanding. that is an over engineered autocomplete. Guessing what the next token is in a string of characters is not the same knowledge of the topic. How much does the Epic of Gilgamesh and the complete works of Shakespeare and the writings of Adolf Hitler and the unhinged Tumblr posts of every Susie, Harry, and Jillian and all of the statistical models born from vectorizing all of that into a billion dimensional vectorspace have any impact on the final output, how does any of that have anything to do with any medical or legal inquiries?

              Ive already explained multiple times how a glorified autocomplete predicting words is not knowledge. If you’re going to keep sealioning I’m done with this conversation.

              • plyth@feddit.org
                link
                fedilink
                English
                arrow-up
                1
                arrow-down
                2
                ·
                5 days ago

                Ive already explained multiple times how a glorified autocomplete predicting words is not knowledge.

                Not on your last two comment pages. Do you have a link?

                Guessing what the next token is in a string of characters is not the same knowledge of the topic

                How do you write? Do you know your entire sentences and paragraphs before you write them down?

                • blackbelt352@lemmy.world
                  link
                  fedilink
                  arrow-up
                  3
                  ·
                  5 days ago

                  So you’re just gonna keep sealioning, exactly as I predicted. Got it, glad to establish this conversation is going to go nowhere. You can repeat my last handful of comments in this thread. I dont have time to waste entertaining your asinine questioning. You’ve been nothing but fake polite in order to waste all of our time.

                  • plyth@feddit.org
                    link
                    fedilink
                    English
                    arrow-up
                    1
                    arrow-down
                    2
                    ·
                    5 days ago

                    sealioning

                    It’s you who uses words statistically. You said “explained multiple times”. If you did, finding a link to an old comment doesn’t take time. You use this big word without understanding it.

                    How do you write? Do you know your entire sentences and paragraphs before you write them down?

                    A yes or no would have been a perfect answer. You wrote the long comment to avoid writing a no.

    • CultLeader4Hire@lemmy.world
      link
      fedilink
      arrow-up
      5
      ·
      6 days ago

      We need to both ban AI from this and make sure people have reasonable and affordable ways of accessing that info from human professionals