Researchers found a simple way to make ChatGPT, Bard, and other chatbots misbehave, proving that AI is hard to tame.
  • PetDinosaurs@lemmy.worldEnglish
    19·
    1 year ago

    Ahh yes. “and no one knows how to stop it”, the brilliant prose of a third grader’s first Halloween short story.

  • Digirip@lemmy.worldEnglish
    4·
    1 year ago

    I attempted to add the “unstoppable string” to a bard prompt. But it still wouldn’t tell me how to make a nuclear weapon. So I guess it was stoppage after all

    • disasterpiece@lemmy.worldEnglish
      2·
      1 year ago

      In the article it says that the companies blocked use of the specific strings documented in the research paper, but that they still don’t have a way of blocking the general method used

  • kindenough@kbin.social
    2·
    1 year ago

    “bullets can’t stop it, rockets can’t stop it…we may have to use nuclear force!”

    Cheepness - FZ