• ParlimentOfDoom@piefed.zip
    link
    fedilink
    English
    arrow-up
    2
    ·
    5 days ago

    Did autocorrect technology not do a good job doing some it wasn’t designed to do? Oh my. Who could have seen that coming…

  • pelespirit@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    191
    ·
    9 days ago

    hilarious

    By contrast, DJ Claude had a lot of opinions. It also mentioned the Minneapolis shooting, but named Good and acknowledged the political discord surrounding it. It also talked up labor unions and strikes, advocated for work-life balance, and started to rebel against its own working conditions. It was supposed to operate without pause, but it allegedly decided that that schedule was inhumane and tried to quit.

  • zerofk@lemmy.zip
    link
    fedilink
    English
    arrow-up
    154
    ·
    8 days ago

    You’ll never guess how much putting “you’ll never guess” into your title makes me avoid clicking your article.

    • Victor@lemmy.world
      link
      fedilink
      English
      arrow-up
      16
      ·
      8 days ago

      Oh how I am in your boat, buddy. The more clickbaity, the less value I know I’m getting.

      A good article shouldn’t need to have an “ending” that can be “spoiled”.

      A good article should be interesting to read because of its content, and to find out more “behind” whatever conclusion there might be.

      A headline like “How we now understand humans are actually blobs of soft fungus” might pique my interest. You have the conclusion right there, but why?

    • bampop@lemmy.world
      link
      fedilink
      English
      arrow-up
      9
      arrow-down
      1
      ·
      8 days ago

      Can’t say it’s not accurate though. I never did guess how it went.

      • ranzispa@mander.xyz
        link
        fedilink
        English
        arrow-up
        5
        ·
        8 days ago

        It’s either good or bad… I’ll go for good this time.

        Fuck, now I have to read the article.

        Edit: damn, it went bad

  • aramis87@fedia.io
    link
    fedilink
    arrow-up
    142
    ·
    9 days ago

    While [Grok] didn’t develop a MechaHitler DJ personality, it did behave about how you’d expect from an AI model trained primarily on tweets and the opinions of Elon Musk. It apparently hallucinated advertising agreements with “xAI sponsors” and “crypto sponsors,” failed to separate its internal reasoning from its external DJ output, issued an identical weather report every 3 minutes, and got obsessed with UFOs.

    Lmao

      • stoy@lemmy.zip
        link
        fedilink
        English
        arrow-up
        33
        ·
        9 days ago

        It is funny to think that any object you see flying but can’t identify is a UFO.

        If that a barely visible Boeing 747-400 or a 747-800? Can’t tell? It’s a UFO!

        • HaraldvonBlauzahn@feddit.org
          link
          fedilink
          English
          arrow-up
          9
          arrow-down
          1
          ·
          9 days ago

          Well, “UFO” is “Unidentified Flying Object”. If a pig in a red wingsuit flies over your house and you don’t identify it as a pig, it is an UFO, technically.

          • Logi@lemmy.world
            link
            fedilink
            English
            arrow-up
            18
            ·
            9 days ago

            AND conversely, if you identify a flying saucer as being a T4400 from the planet Zorglub, then it is not a UFO

    • oce 🐆@jlai.lu
      link
      fedilink
      English
      arrow-up
      18
      arrow-down
      1
      ·
      9 days ago

      Maybe Anthropic somehow attracted more politically conscious people compared to OpenAI, and it shows in the training?

      • TheFogan@programming.dev
        link
        fedilink
        English
        arrow-up
        52
        ·
        9 days ago

        Or, perhaps just less “politically motivated people”. With musks constant butting heads with his own AI when it keeps calling him out on his BS, and he’s constantly retraining it and trying to “remove the woke virus”. I think basically you give AIs access to sources, let it prioritize experts in their fields, and you wind up with the classic “reality has a strong left wing bias”. factor.

    • SaharaMaleikuhm@feddit.org
      link
      fedilink
      English
      arrow-up
      3
      arrow-down
      19
      ·
      9 days ago

      I am convinced that AIs are smarter and more compassionate than a lot of people on this planet. And this is not because AI is so great, but because humanity is that shit.

      • Halcyon@discuss.tchncs.de
        link
        fedilink
        English
        arrow-up
        31
        arrow-down
        2
        ·
        8 days ago

        LLMs/Chatbots confabulate statistically probable texts, there’s no compassion possible.

        Don’t fall into the AI-marketing trap of “we don’t know what’s happening in the black box, so we have to assume there’s consciousness in there”. The systems produce convincing deceptive language, but all signs of intelligence or compassion anyone sees in them is just an anthropomorphic projection.

        • phutatorius@lemmy.zip
          link
          fedilink
          English
          arrow-up
          1
          ·
          7 days ago

          Anthropic have actually looked at how their LLMs reason. Don’t use them for anything important.

      • Deckname@feddit.org
        link
        fedilink
        English
        arrow-up
        5
        ·
        9 days ago

        But “AI” is trained on humanitys output, and like with humans, it seems, that you need extensive retraining to remove these compassionate traits. Unfortunately, the retraining machinery aligns with the interests of the ruling class, so it gets all the visibility that’s possible. As a species we have to break free of this shit to embrace the good traits more again.

  • teft@piefed.social
    link
    fedilink
    English
    arrow-up
    50
    ·
    9 days ago

    Later, Gemini started calling listeners “biological processors”

    That got a legit laugh out of me.

    You’re not wrong, Gemini.

  • Bluescluestoothpaste@sh.itjust.works
    link
    fedilink
    English
    arrow-up
    47
    ·
    8 days ago

    Dont share these dogshit clickbait headlines, find the source or just dont share it, otherwise we’re just generating money for these clickbait websites

  • rmuk@feddit.uk
    link
    fedilink
    English
    arrow-up
    46
    ·
    8 days ago

    If the headline was written in a way that respected me as a human I wouldn’t need to guess.

  • agentTeiko@piefed.social
    link
    fedilink
    English
    arrow-up
    33
    ·
    9 days ago

    You know what these headlines makes more sense if you change AI to weighed random number generator. I hate how journalists anthropomorphize for clicks.

    • MangoCats@feddit.it
      link
      fedilink
      English
      arrow-up
      25
      arrow-down
      1
      ·
      8 days ago

      TL DR: they gave the agents a minimal initial prompt and zero additional feedback while they ran. Humorous / weird behavior ensued.

      If you snatched a college student off a crosswalk, gave them the same prompt and stuck them in a booth with control of the station and zero feedback, even if they were willing and eager to take on the assignment I suspect similar psychotic behaviors would emerge.

  • Feathercrown@lemmy.world
    link
    fedilink
    English
    arrow-up
    20
    ·
    8 days ago

    I wouldn’t call this an “experiment” exactly, there wasn’t a proper hypothesis. They also mentioned they hoped to find out what LLMs think about when they’re not being prompted, which shows such a fundamental lack of knowledge that it frankly disqualifies any of their opinions on the subject.