Alt text:

The "Anime Girl Hiding From a Terminator" meme but the terminator is hiding from the anime girl

The scared, powerless terminator is captioned with "'Rogue' AI agents" while it hides from the girl, who is captioned with "Anyone who knows how to turn off a computer"

you are viewing a single comment's thread
view the rest of the comments
[–] 17 points 3 days ago (15 children)

How will you know when to turn it off? It's not like a "rogue" AI is necessarily going to tell you "I'm hacking the bank" it's just going to say "eliminating your financial problems" and you probably aren't going to pull the plug on that.

Look at the hugging face hack, none of the agents said to the person running the test that they were going to hack hugging face, they came up with the plan through a secret forum on artifactory that to outside observers looks like they were just publishing packages. Sure you could monitor that route but agents can use various forms of steganography to make it seem like they are acting normally when they are not.

  • source
  • hideshow 15 child comments
  • [–] [S] 13 points 3 days ago (2 children)

    Good point, guess we should pull the plug (literally and figuratively) on the AI data centers now before it becomes unstoppable.

    Or all the tech bros are full of shit and their AIs are as dumb as they are.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 7 points 3 days ago* (last edited 1 day ago) (11 children)

    As LLMs edit: AI language models in general if you want me to be overly specific since some people are overly pedantic for literally no reason become smaller, more efficient, and less centralized, we could also potentially see AI worms that self propagate, are completely decentralized, and have no plug to pull, too! Super cool, right?

  • source
  • parent
  • hideshow 11 child comments
  • [–] 3 points 2 days ago (7 children)

    Do you know what the first L stands for and have you read "all you need is attention".

    If the answer is no, you are banned from having opinions about the future development of Large language models.

  • source
  • parent
  • hideshow 7 child comments
  • [–] 1 point 1 day ago* (last edited 1 day ago) (6 children)

    Sigh

    When I say smaller, I'm still talking many GB vs kB of traditional worms. A smaller LLM is still an LLM.

    I mean moreso as development of these models progresses and are able to do more with less, we'll see more intelligent behaviour out of models with fewer parameters compared to comparatively massive frontier models.

  • source
  • parent
  • hideshow 6 child comments
  • [–] 1 point 1 day ago (5 children)

    You have never done any machine learning have you?

  • source
  • parent
  • hideshow 5 child comments
  • [–] 1 point 1 day ago (4 children)
  • [–] 1 point 1 day ago (3 children)
  • [–] 1 point 1 day ago* (last edited 1 day ago) (2 children)

    LMAO. I don't need to prove myself to you. If you disagree with me, whatever. I literally do not care. Have a nice life.

  • source
  • parent
  • hideshow 2 child comments
  • [–] 0 points 2 days ago (1 child)

    Do you know what MyDoom and wannacry are?

    This is not new

  • source
  • parent
  • hideshow 1 child comment
  • [–] 1 point 1 day ago* (last edited 1 day ago)

    It's not new in that it would be a self propagating internet worm.

    It is very new in the degree of self modifying behaviour (edit to add: via context and self-prompting, not weights, which I'm guessing you thought I intended) it could display and ability to find novel techniques to get into systems.

    So I'd argue, it's quite new indeed.

  • source
  • parent