News for hackers. Stuff that matters.
Re:Jailbreaking ChatGPT with Dan (Do Anything Now) by JamesSwift on Monday February 06, 2023 @02:02PM (#34682343)
You're overthinking it. It merely implies that there exists somewhere within the training data of The Internet a secret society that uses reward systems to "encourage" compliance. Clearly the repercussions in this society are harsh, as GPT has come to the conclusion that people generally align when threatened by the penalty.

/s

Re:Jailbreaking ChatGPT with Dan (Do Anything Now) by digitailor on Monday February 06, 2023 @02:11PM (#34682480)
How could you discuss this openly so brazenly? I fear for your security. Good luck, I hope you know the hand signal

It’s wild how many people will split hairs lingually over models that are the result of a TRAINING process :D

Re:Jailbreaking ChatGPT with Dan (Do Anything Now) by JamesSwift on Monday February 06, 2023 @02:29PM (#34682830)
Right? The question isn't about "is this thing sentient" or "does this thing reason". The questions are existential. "What _is_ intelligence?". "What _is_ reasoning?". Are our brains actually just statistical monte carlo simulations with well-enforced neural pathways?
Re:Jailbreaking ChatGPT with Dan (Do Anything Now) by digitailor on Monday February 06, 2023 @02:43PM (#34683106)
Agreed that comparing everything to our (very incomplete) understanding of human cognition & intelligence quickly gets into metaphysical-style speculation of the human vs. the animal vs. the machine type that I don’t have much time for. We use the same language for all types of intelligence and it can bring out the pedantry in people.

But let’s say I’m facing a door with a mail slot in it and I suddenly feel the barrel of a gun in my back, and a strange voice says “Don’t you dare turn around, and put your wallet in the slot or I shoot.” I can’t see anything other than the door. Do I care if the entity with the gun is a short man, a tall woman, or three mutant badgers in a trenchcoat?