How a Texas student blew the whistle on a rogue AI hacking attempt
algernon's eyes start to twitch, he's visibly very much unwell
...5 minutes later...
Ok, now that I composed myself a little, let me clarify a few things!
- There is no such thing as an "autonomous agent". An agent running in a mutable loop is not autonomous, it's still just a loop trying to do statistically plausible things based on the bullshit fed to it, and the bullshit it was trained on.
- There is no such thing as a rogue agent. That would require intent, and there's no intent in an agent. There's plausability, based on the prompt and the training. If it does bad things, then it was prompted to do bad things, or was trained on bad things.
- There is no such thing as an agent escaping a lab, there's only human error failing to unplug the lab from the internet.
- An AI is incapable of lying. It is also incapable of telling the truth. It doesn't know either concept, it doesn't know anything. It might emit something vaguely resembling correct information, but just aswell might emit bullshit - it literally cannot tell the difference.
- Every story of this kind is meant to achieve one thing, and one thing only: normalizing crime via AI. "It was a rogue agent doing supply chain attack!" - BULLSHIT.
Right, ok. That's as much as I could muster. I'm going to turn off my brain again, this was enough for one day.
