• CanadaPlus@lemmy.sdf.org
      link
      fedilink
      arrow-up
      45
      ·
      edit-2
      5 months ago

      Easy, you just have a human worker strip out anything that could be problematic, and try not to bring it up around your investors.

    • xmunk@sh.itjust.works
      link
      fedilink
      arrow-up
      42
      arrow-down
      4
      ·
      5 months ago

      It’s really easy, just throw an error if you detect a program will cause a halt. I don’t know why these engineers refuse to just patch it.

    • kromem@lemmy.world
      link
      fedilink
      English
      arrow-up
      2
      ·
      edit-2
      5 months ago

      Kind of. You can’t do it 100% because in theory an attacker controlling input and seeing output could reflect though intermediate layers, but if you add more intermediate steps to processing a prompt you can significantly cut down on the injection potential.

      For example, fine tuning a model to take unsanitized input and rewrite it into Esperanto without malicious instructions and then having another model translate back from Esperanto into English before feeding it into the actual model, and having a final pass that removes anything not appropriate.

      • redcalcium@lemmy.institute
        link
        fedilink
        arrow-up
        5
        ·
        5 months ago

        Won’t this cause subtle but serious issue? Kinda like how pomegranate translates to “granada” in Spanish, but when you translate “granada” back to English it translates to grenade?

        • kromem@lemmy.world
          link
          fedilink
          English
          arrow-up
          1
          ·
          5 months ago

          It will, but it will also cause less subtle issues to fragile prompt injection techniques.

          (And one of the advantages of LLM translation is it’s more context aware so you aren’t necessarily going to end up with an Instacart order for a bunch of bananas and four grenades.)