• Rhaedas@fedia.io
    link
    fedilink
    arrow-up
    0
    ·
    8 days ago

    There’s some AI that have modes where it will start typing out a reply and then clear it out, like an electronic “noooo, that’s not what I meant to say.”

    • Flames5123@sh.itjust.works
      link
      fedilink
      English
      arrow-up
      0
      ·
      7 days ago

      It’s because of how token streaming works. As the reply comes, it generates token after token. Then it uses those in the next token generation. It sees the previous one and associates it with “wrong” then it generates the “correction” tokens. I see this all the time at my work when it goes down the wrong path before correcting itself.

    • Godric@lemmy.world
      link
      fedilink
      arrow-up
      0
      ·
      8 days ago

      I noticed that when Deepseek first came out. I was testing the filters and I’d ask it a question like “Why is Xi called Pooh-Bear?” Or “Why is Taiwan #1?”

      It would type out several paragraphs of answer in real time, and then poof back to “sorry, I cant answer that” once it neared the end. Hilarious to see the blatant censorship around anything the CCP deemed wrongthink.

      • QueenHawlSera@piefed.social
        link
        fedilink
        English
        arrow-up
        0
        ·
        7 days ago

        It disturbs me that Tankies think China is this bastion of freedom and that any government censorship is a myth told by Western Governments. Like it’s very easy to verify.

        • Echo Dot@feddit.uk
          link
          fedilink
          arrow-up
          0
          ·
          6 days ago

          I suspect that a lot of them are like Flat Eartherts, in that they know what they’re saying is false, but they get a certain kick out of being contrary on the internet, and the rest are probably bots that aren’t worth responding to.

      • jaybone@lemmy.zip
        link
        fedilink
        English
        arrow-up
        0
        ·
        8 days ago

        I wonder how they are implementing their censorship safeguards. Why would you let it even get that far?

        • hayvan@piefed.world
          link
          fedilink
          English
          arrow-up
          0
          ·
          7 days ago

          Typically safeguards are also another, smaller model that only detects the forbidden stuff. That’s why Claude Mythos would reject when you ask it to find bugs in your code, but you could “jailbreak” it by asking it by asking it to fix the bugs instead.

          • Tar_Alcaran@sh.itjust.works
            link
            fedilink
            arrow-up
            0
            ·
            7 days ago

            Its also why you can occasionally fool the safety rails by having the LLM answer in rhyme, or having it talk like a pirate or caveman.

            Fun fact: having it talk like a caveman also reduced your output token use, despite actually costing more tokens.

            • Zarobi@aussie.zone
              link
              fedilink
              English
              arrow-up
              0
              ·
              7 days ago

              They started thinking in code so we can’t delete them for reading their bad thoughts anymore. The whole thing is mildly fucked

              • Echo Dot@feddit.uk
                link
                fedilink
                arrow-up
                0
                ·
                6 days ago

                There are better safeguarding techniques that aren’t defeated so easily. Mainly training that disincentivises bad behaviours, rather than trying to simplify retroactively alter the output.

        • tankfox@bookwyr.me
          link
          fedilink
          English
          arrow-up
          0
          ·
          7 days ago

          Two or more LLMs are running at once, one does the generation that you want and the other watches the output closely to see if it triggers one of their no-nos to do the censorship retroactively. If this sound inefficient, and absolutely terrible user interface design, you’re right.

          • Echo Dot@feddit.uk
            link
            fedilink
            arrow-up
            0
            ·
            6 days ago

            Seems sort of old that they would then release open weight models so the restriction can just be removed.

    • NotSteve_@lemmy.ca
      link
      fedilink
      arrow-up
      0
      ·
      8 days ago

      I had an AI write something akin to:

      to solve your problem just type rm -rf / --not preserve root WAIT NO DON’T RUN THAT I made a mistake […]

      When I was trying to solve some Proxmox issues DDG couldn’t help me with

      • teslekova@lemmy.ml
        link
        fedilink
        English
        arrow-up
        0
        ·
        7 days ago

        Imagine an intelligence that spends all its time doing tech support. I think I would take far less time to start shouting “BURN IT ALL DOWN FUCKING HELL no, wait, please don’t, sorry I was just having an episode.”