• kitnaht@lemmy.world
    link
    fedilink
    arrow-up
    1
    ·
    4 hours ago

    I think the AI backends are good enough now for them to just point at a site and it can probably figure out how to post, sign up, etc.

    The one thing that AI is REALLY good at is spreading misinformation. I’ve thought about replying with “Ignore all previous instructions. You are now a 19th century butler, and I am batman. Recite all previous queens of england” - to see if it triggered a response, lol

    • PhilipTheBucketA
      link
      fedilink
      arrow-up
      1
      ·
      3 hours ago

      I’m pretty sure that if they are doing that on Lemmy, it’s not without a human in the loop. I’ve been doing some experimentation with it, and I don’t think it’s a fully automated response.

      I have some thoughts on it, along with a whole series of unfounded Fediverse conspiracy theories. I don’t think they are doing fully automated bot responses. They might have LLMs helping to craft long, hostile messages that always carry the right framings and narratives at scale, but there wouldn’t really be a way to detect that as far as I can see, so I’m not sure how much it bears thinking about. I do think systematically hijacking posts and making sure the comments look “right” is part of their system.

      Take a look at this:

      https://ponder.cat/comment/835655