• betelgeuse [comrade/them]
    ·
    10 months ago

    The AI said "All Lives Matter" and this was a huge problem for people who have been saying "All Lives Matter" for the past three years.

  • GaveUp [love/loves]
    ·
    10 months ago

    The AI had troubles because letting a billion white people roam free on Earth and "harm no one" are contradictory rules

    • Tankiedesantski [he/him]
      ·
      10 months ago

      letting a billion white people roam free on Earth and "harm no one" are contradictory rules

      This is one of my few remaining criticisms of anarchism. Like we're just supposed to trust whitey to do the right thing.

  • UlyssesT [he/him]
    ·
    10 months ago

    The problem with the LLM is that it hasn't been fed enough nazi shit and bazinga cringe yet billionaire-tears

      • UlyssesT [he/him]
        ·
        10 months ago

        That's why my-hero is financing Grok, named after a super cool super secret very special people slang word for understanding a concept from a story written by one of his favorite fascist authors: Robert Heinlein. yea

  • Owl [he/him]
    ·
    edit-2
    10 months ago

    Reminder that ChatGPT's architecture is roughly:

    • Predict what would come next in this sentence according to your corpus of raw internet

    • Shit people keep trying to ask it questions instead of giving it the first sentence of wikipedia, uh... here's a few hundred hand-written examples of people asking questions and then answering them like a good little encyclopedia buttler. Prioritize these over the raw internet

    • Okay that looks good but we weren't born yesterday, so here's a bunch of hand-written examples where people ask leading questions about racism, crime, and politics, and then the dictionary stonewalls them. Prioritize these over everything as soon as anyone mentions those topics

    • Adkml [he/him]
      ·
      10 months ago

      Yea this is literally just an ai being more clever than the people "just asking questions"

      Even they can see the writing on the walls the question asker is trying tonget them to talk about racism against white people and is pulling the chord.

      • Owl [he/him]
        ·
        10 months ago

        Correction: this is the last layer of bolted-on failsafes, using a more primitive AI model than the rest, being more clever than the people "just asking questions."

  • Ho_Chi_Chungus [she/her]
    ·
    10 months ago

    if it would take an action that would harm no one

    this sentence feels like it's doing some real heavy lifting here

    • invalidusernamelol [he/him]
      ·
      10 months ago

      And it making a decision on a hypothetical also harms no one.

      AI is not intelligent in the slightest, but it's definitely more intelligent than people like this

  • BountifulEggnog [she/her]
    ·
    edit-2
    10 months ago

    I hate when people say "I asked openai" and what they mean is "I asked chatgpt". I've only seen it a couple times but my god. Openai is the company, chatgpt is the funny chat bot. People conflate the two and act like everything chatgpt says is like the companies official stance.

  • Timberknave
    ·
    10 months ago

    LLM smart enough to recognize concern trolling

    Spoiler

    joking

  • NephewAlphaBravo [he/him]
    ·
    edit-2
    10 months ago

    anglos have that individualism/death drive that means you'd have to physically harm some of them to keep them from throwing themselves into the wheat thresher because it's their right to do that

    • UlyssesT [he/him]
      ·
      10 months ago

      "DON'T TELL ME WHAT TO DOOOOOOOOOOOOOOOOOOOOOOOOO" right off a cliff, every time, to prove they're not lemming sheeple wokes. grill-broke

      • GarfieldYaoi [he/him]
        ·
        10 months ago

        grill-broke: "Wait, you're a black woman and you just said 'don't tell me what to do?' That doesn't apply to you because.....well it just doesn't okay?"