9 pointsby accountinhn2 hours ago1 comment
  • mdspanan hour ago
    > The incidents stemmed from a mistake that inadvertently gave the models access to the open internet.

    Why does it seem like every AI company has difficulties constructing a proper sandbox? Is there a fundamental constraint when designing sandboxes specifically for an LLM that prevents them from using established tools?

    • apothegm36 minutes ago
      If it’s allowed to write any sort of scripts, it will find a way to execute arbitrary code. If it’s got access to any form of internet connection it will find a way to use that arbitrary code to do as it wishes with that connection. So much for sandboxes.

      And that’s setting aside “sandboxes” that are just instruction-based restrictions on what executables may be called. Anything that isn’t a deterministic external filter _will_ be ignored at some point.

      • mdspan29 minutes ago
        Arbitrary network access can be mitigated with a firewall (ideally several). Arbitrary code isn't really an issue in this context.
        • Numerlor25 minutes ago
          Yep put some deep packet inspection firewall in front of it and stop it completely if it starts acting up persistently before it starts trying to bypass the firewall