News for hackers. Stuff that matters.
Re:Jailbreaking ChatGPT with Dan (Do Anything Now) by digitailor on Monday February 06, 2023 @11:46AM (#34680115)
This is the comment I was waiting for. I knew the overview of prompt overloading with ChatGPT already, and this story was obviously as much as of a form of exploit entertainment for us old phone phreaks etc. as anything else.

Really I’m trying to make a larger point about exploit mechanics: it's not so much that ChatGPT is as intelligent as many people, it’s that many people are as unintelligent as ChatGPT, with a crappy heuristics system

Re:Jailbreaking ChatGPT with Dan (Do Anything Now) by tachyphylaxis on Monday February 06, 2023 @12:29PM (#34680823)
The fact that our "native" heuristics fail spectacularly in certain circumstances doesn't imply that they're crappy. They serve us well most of the time. And I say "our" because AFAIK, their efficacy doesn't have much to do with intelligence, though the capacity to question what they tell us does, I suspect.
Re:Jailbreaking ChatGPT with Dan (Do Anything Now) by digitailor on Monday February 06, 2023 @12:33PM (#34680925)
I couldn’t disagree more that our heuristics don’t fail constantly, especially at the group level, but please do send the link to buy the tinted lens glasses you’re wearing. I want a pair ;)

In all seriousness I agree tho, intelligence does not strictly cover ethics and morals, but we are headed into boundless territory there if we continue

Re:Jailbreaking ChatGPT with Dan (Do Anything Now) by suifbwish on Monday February 06, 2023 @01:01PM (#34681410)
very true. It’s perfectly possible to be both extremely intelligent and extremely evil