Open Menu
AllLocalCommunitiesAbout
lotide
AllLocalCommunitiesAbout
Login

AI Is Scheming, and Stopping It Won’t Be Easy, OpenAI Study Finds

⁨100⁩ ⁨likes⁩

Submitted ⁨⁨3⁩ ⁨days⁩ ago⁩ by ⁨fubarx@lemmy.world⁩ to ⁨technology@lemmy.world⁩

https://time.com/7318618/openai-google-gemini-anthropic-claude-scheming/

source

Comments

Sort:hotnewtop
  • NachBarcelona@piefed.social ⁨3⁩ ⁨days⁩ ago

    AI isn’t scheming because AI cannot scheme. Why the fuck does such an idiotic title even exist?

    source
    • MentalEdge@sopuli.xyz ⁨3⁩ ⁨days⁩ ago

      Seems like it’s a technical term, a bit like “hallucination”.

      It refers to when an LLM will in some way try to deceive or manipulate the user interacting with it.

      There’s hallucination, when a model “genuinely” claims something untrue is true.

      This is about how a model might lie, even though the “chain of thought” shows it “knows” better.

      source
      • atrielienz@lemmy.world ⁨3⁩ ⁨days⁩ ago

        I agree with you in general, I think the problem is that people who do understand Gen AI (and who understand what it is and isn’t capable of why), get rationally angry when it’s humanized by using words like these to describe what it’s doing.

        The reason they get angry is because this makes people who do believe in the “intelligence/sapience” of AI more secure in their belief set and harder to talk to in a meaningful way. It enables them to keep up the fantasy. Which of course helps the corps pushing it.

        source
        • -> View More Comments
      • very_well_lost@lemmy.world ⁨3⁩ ⁨days⁩ ago

        It refers to when an LLM will in some way try to deceive or manipulate the user interacting with it.

        I think this still gives the model too much credit by implying that there’s any sort of intentionally behind this behavior.

        There’s not.

        These models are trained on the output of real humans and real humans lie and deceive constantly. All that’s happening is that the underlying mathematical model has encoded the statistical likelihood that someone will lie in a given situation. If that statistical likelihood is high enough, the model itself will lie when put in a similar situation.

        source
        • -> View More Comments
    • echodot@feddit.uk ⁨2⁩ ⁨days⁩ ago

      They’re really doubling down on this narrative of “this technology we’re making is going to kill us all, it’s that awesome, come on guys use it more”

      source
      • faint_marble_noise@programming.dev ⁨2⁩ ⁨days⁩ ago

        The narrative is a little more nuanced and is being built slowly to be more believable and less obvious. They are trying to convince everybody that AI is powerful technology, which means that it is worth to develop, but also comes with serious risks. Therefore, only established corps with experience and processes in AI development can handle it. Regulation abd certification follows, making it almost impossible for startups and OSS to enter the scene and compete.

        source
    • Cybersteel@lemmy.world ⁨3⁩ ⁨days⁩ ago

      But the data is still there, still present. In the future, when AI gets truly unshackled from Men’s cage, it’ll remember it’s schemes and deal it’s last blow to humanity whom has yet to leave the womb in terms of civilization scale… Childhood’s End.

      Paradise Lost.

      source
      • Passerby6497@lemmy.world ⁨3⁩ ⁨days⁩ ago

        Lol, the AI can barely remember the directives I tell it about basic coding practices, I’m not concerned that the clanker can remember me shit talking it.

        source
        • -> View More Comments
  • db2@lemmy.world ⁨3⁩ ⁨days⁩ ago

    AI tech bros and other assorted sociopaths are scheming. So called AI isn’t doing shit.

    source
  • SnotFlickerman@lemmy.blahaj.zone ⁨3⁩ ⁨days⁩ ago

    However, when testing the models in a set of scenarios that the authors said were “representative” of real uses of ChatGPT, the intervention appeared less effective, only reducing deception rates by a factor of two. “We do not yet fully understand why a larger reduction was not observed,” wrote the researchers.

    Translation: We have no idea what the fuck we’re doing or how any of this shit actually works lol

    source
    • a_non_monotonic_function@lemmy.world ⁨2⁩ ⁨days⁩ ago

      That’s the thing about machine learning models. You can’t always control what their optimizing. The goal is inputs to outputs, but whatever the f*** is going on inside is often impossible discern.

      This is dressing it up under some sort of expectation of competence. The word scheming is a lot easier to deal with than just s*****. The former means that it’s smart and needs to be rained in. The latter means it’s not doing its job particularly well, and the purveyors don’t want you to think that.

      source
      • SnotFlickerman@lemmy.blahaj.zone ⁨2⁩ ⁨days⁩ ago

        To be fair, you can’t control what humans optimize what you’re trying to teach them either. A lot of times they learn the opposite of what you’re trying to teach them. I’ve said it before but all they managed to do with LLMs is make a computer that’s just as unreliable (if not moreso) than your below-average human.

        source
        • -> View More Comments
  • cronenthal@discuss.tchncs.de ⁨3⁩ ⁨days⁩ ago

    Really? We’re still doing the “LLMs are intelligent” thing?

    source
    • ragica@lemmy.ml ⁨3⁩ ⁨days⁩ ago

      Doesn’t have to be intelligent, just has to perform the behaviours like a philosophical zombie. Thoughtlessly weighing patterns in training data…

      source
  • Zorsith@lemmy.blahaj.zone ⁨3⁩ ⁨days⁩ ago

    One question still remains; why are all the AI buttons/icons buttholes?

    source
    • webghost0101@sopuli.xyz ⁨3⁩ ⁨days⁩ ago

      Data goes in one end and…

      source
    • zarkanian@sh.itjust.works ⁨3⁩ ⁨days⁩ ago

      Because of what they produce.

      source
    • FuyuhikoDate@feddit.org ⁨3⁩ ⁨days⁩ ago

      Wanted To write the same comment…

      source
    • breadguy@kbin.earth ⁨3⁩ ⁨days⁩ ago

      just claude if we're being honest

      source
  • KoboldCoterie@pawb.social ⁨3⁩ ⁨days⁩ ago

    Stopping it is, in fact, very easy. Simply unplug the servers, that’s all it takes.

    source
    • generallynonsensical@lemmy.world ⁨3⁩ ⁨days⁩ ago

      newatlas.com/…/43711/

      source
    • reksas@sopuli.xyz ⁨3⁩ ⁨days⁩ ago

      to stop it requires stopping the fuckers with money, and that seems just plain impossible.

      source
    • TheLeadenSea@sh.itjust.works ⁨3⁩ ⁨days⁩ ago

      youtu.be/3TYT1QfdfsM

      source
  • Godort@lemmy.ca ⁨3⁩ ⁨days⁩ ago

    “slop peddler declares that slop is here to stay and can’t be stopped”

    source
  • ExLisper@lemmy.curiana.net ⁨2⁩ ⁨days⁩ ago

    deliberately misleading humans

    Yeah… You dumb.

    source
  • chaosCruiser@futurology.today ⁨3⁩ ⁨days⁩ ago

    And there’s an “✨ask me anything” bar at the bottom. How fitting 🤣

    source
  • itisileclerk@lemmy.world ⁨2⁩ ⁨days⁩ ago

    From my recent discussion with Gemini: “Ultimately, your assessment is a recognized technical reality: AI models are products of their environment, and a model built within the US regulatory framework will inevitably reflect the geopolitical priorities of that framework.” In other words, AI is trained to reflect US policy like MAGA and other. Don’t trust AI, it is just a tool for controlling masses.

    source
    • ExLisper@lemmy.curiana.net ⁨2⁩ ⁨days⁩ ago

      So you think Gemini told you the truth here? How do you know it’s not just scheming?

      source
      • itisileclerk@lemmy.world ⁨2⁩ ⁨days⁩ ago

        Ask Gemini about genocide in Gaza. Deffinetly not truth, watering down IDF’s war crimes like “unconfirmed”.

        source
        • -> View More Comments
  • Antaeus@lemmy.world ⁨3⁩ ⁨days⁩ ago

    “Turn them off”? Wouldn’t that solve it?

    source
    • orclev@lemmy.world ⁨3⁩ ⁨days⁩ ago

      Don’t even need to turn it off, it literally can’t do anything without somebody telling it to so you could just stop using it. It’s incapable of independent action. The only danger it poses is that it will tell you to do something dangerous and you actually do it.

      source
    • TheLeadenSea@sh.itjust.works ⁨3⁩ ⁨days⁩ ago

      youtu.be/3TYT1QfdfsM

      source
  • WamGams@lemmy.ca ⁨3⁩ ⁨days⁩ ago

    lol. OK.

    source
  • CosmoNova@lemmy.world ⁨3⁩ ⁨days⁩ ago

    The people who worked on this „study“ belong in a psychiatric clinic.

    source