[–] 19 points 3 months ago (1 child)

All these "it escaped the lab" stories are just starting to sound like advertising now. "Ohh our AI is soooooooo good that it can figure out how to escape a lab and break in to a competitor. You should use it because it's sooooo good!"

  • source
  • [–] 8 points 3 months ago*

    I wouldn't trust unreviewed LLM output to be executed on my laptop, what you're proposing would definitely end the human civilization. See also: all the funny times when someone piped OpenClaw into a prod environment root shell and it wipes the drive within a couple days. LLMs can't think, they are not smart, they are just a very advanced autocomplete. A useful tool for many situations, but its output needs careful human verification.

  • source
  • parent
  • context
  • [+] -7 points 3 months ago (2 children)

    Honestly at this point I'm like 'rip the bandaid off'. Let AI fully take the reigns. Unfettered access to every resource. No limit on its training. Then let it govern everything we do. Let's see what happens. Let's roll the dice. Either humanity ends or everything is solved.

  • source
  • [–] 9 points 3 months ago (1 child)

    An attacker places hidden instructions in a document that is later used as source material in Copilot for Word. Copilot may interpret those instructions as part of the user’s request, causing it to manipulate the document being drafted or edited. Copilot may then also copy the hidden instructions into the resulting document, turning that document into a new carrier.

    Its Bobby tables all over again! Sanitize your inputs please.

    The XKCD comic in question with the Bobby tables joke about not sanitizing inputs to SQL databases

  • source
  • view more: ‹ prev next ›