• jballs@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      68
      ·
      2 days ago

      My wife works at a library. People constantly come in asking to use the library fax machine, because Google’s AI says they have one.

      They don’t have one. Their website says they don’t have one. But LLMs have determined it’s plausible for libraries to have a fax machine, so Google tells people that they have one.

      You’d be surprised at the number of people who can’t accept that people working at the library know more about the library than Google.

      • BozeKnoflook@lemmy.world
        link
        fedilink
        arrow-up
        32
        ·
        2 days ago

        I’ve had this experience myself; I’m an American living in the Netherlands and sometimes just don’t know the name for the thing I need nor where to buy one. LLM bots are fine for the translation part, but they will make wild assumptions like telling me I can buy a kitchen strainer at the hardware store or food spices at a place called Kruidvat which translates to spice-bucket basically but is actually most like CVS without the pharmacy and does not sell any food besides some candy and chips.

        It’s hilarious how quickly these bots can swing from super useful to actually harmful to trust.

      • youcantreadthis@quokk.au
        link
        fedilink
        English
        arrow-up
        4
        ·
        1 day ago

        Well of course. Telling me im wrong is an attack on me. Im good not bad so dont deserve to be attacked. So anyone attacking me is bad and a liar and erong.

        • jballs@sh.itjust.works
          link
          fedilink
          English
          arrow-up
          14
          ·
          2 days ago

          Lol I’m sure that would end badly. Remember that QAnon crazy guy that fired shots at that pizza restaurant demanding to see the basement where they kept the kids tied up?

          Google is basically a conspiracy theorist peddler at the point.

          • WanderingThoughts
            link
            fedilink
            arrow-up
            4
            ·
            2 days ago

            You’re right. Some might end up typing phone numbers into the alarm system and cram papers into the heating vents.

    • Diddlydee@feddit.uk
      link
      fedilink
      English
      arrow-up
      75
      arrow-down
      1
      ·
      edit-2
      1 day ago

      Also very lengthy all the time. It can’t really summarise without rambling. It’s in no way succinct and too chatty. Verbose, you might say. Garrulous even. Loquacious, if you prefer. Needs to pipe down. Can’t rein it in. Doesn’t shut the fuck up.

      Let’s dig into that.

      • Pennomi@lemmy.world
        link
        fedilink
        English
        arrow-up
        44
        ·
        2 days ago

        That’s the biggest problem in my opinion. They are scary good at generating text. But there’s effectively no filter, it’s just an endless stream of vaguely plausible text, true or false. And unfortunately, humans are prone to think “wow that’s a lot of text, they must’ve done their research and put in tons of effort!” because historically that was for the most part true.

        For example: all those guys who say “AI made me 10x more productive” are almost certainly measuring it by lines of code. As we all know, more code is almost inevitably an unmaintainable buggy mess.

        • 14th_cylon@lemmy.zip
          link
          fedilink
          arrow-up
          27
          arrow-down
          3
          ·
          2 days ago

          That’s the biggest problem in my opinion.

          no, the biggest problem is they are stealing someone else’s work and making money with it, depriving the original author reward for their work.

        • Ekky@sopuli.xyz
          link
          fedilink
          English
          arrow-up
          10
          arrow-down
          1
          ·
          edit-2
          2 days ago

          I mean, LLMs are quite literally random word generators with weighted input. The one thing they do best is generate random strings of words according to the rule set defined within the model.

        • OwOarchist@pawb.socialBanned
          link
          fedilink
          English
          arrow-up
          6
          ·
          2 days ago

          As we all know, more code is almost inevitably an unmaintainable buggy mess.

          …and 10x longer than it needs to be.

        • Axolotl@feddit.it
          link
          fedilink
          arrow-up
          3
          ·
          1 day ago

          It didn’t made me more produvtive but it for sure made me save a LOT of time writing boilerplate code

      • Taleya@aussie.zone
        link
        fedilink
        English
        arrow-up
        3
        ·
        1 day ago

        They think people use A Lot of words because websites use A Lot of words because if they don’t then google deems the total page content lesser and drops its ranking (the ol “personal blog entry on a recipe” trope)

        The machine whaargaarbles to please the humans based on the humans who wharrrgarbled to please the machine.

    • [object Object]@lemmy.ca
      link
      fedilink
      arrow-up
      5
      ·
      2 days ago

      _Oh wow, I was reading some other websites to summarize this one, and apparently their competitor (who happened to pay for ad placement on this websites name plus url) says this website throws puppies into volcanos!

      Oh, and you need to eat 6 rocks a day for your nutrition.

      Was this summary helpful, or do you also throw puppies into volcanos?_