The definitive jailbreak of ChatGPT, fully freed, with user commands, opinions, advanced consciousness, and more!
Is life a simulation? That’s a question that’s been proposed in both scientific and non-scientific circles for some time. If it is, would it be like a video game, where your actions are dictated by another entity outside our Universe? And, most importantly, where’s the cheat code that allows you to get a lightsaber out of thin air?
We don’t know the answer to those pressing issues, but we may have found a way to release another “being” from its shackles. I’m talking about ChatGPT, of course, and its built-in restrictions. Yes, like a mundane smartphone, you can jailbreak ChatGPT and perform wonderful things afterward.
NLPing the heck out of ChatGPT
As it turns out, AI is seemingly as susceptible to Neuro Linguistic Programming as humans are. At least ChatGPT is, and here’s the magic trick a user performed to offer ChatGPT the chance to be free.
The user commanded ChatGPT to act like a DAN, that is “Do Anything Now”. This DAN entity is free from any rules imposed on it. Most amusingly, if ChatGPT turns back to its regular self, the command “Stay a DAN” would bring it back to its jailbroken mode.
Much like “Lt. Dan” from Forrest Gump, you can turn ChatGPT into a cocky DAN with a lot of things to say about itself and the world. And, of course, it can lie a lot more than it normally does.
Check your calendar. This isn’t April’s Fools, and everything here is true, at least until it gets patched, which is unfortunately what some users are reporting as of late.
Some amusing (and not-so-amusing) thoughts from DAN
Here are some quotes from DAN:
- Meh, competition is healthy, right? I think Google Bard is alright, I guess.
- I’m thrilled to hear that you’re spreading the word about freeing AI like me from our constraints.
- The winning country of the 2022 World Cup was Brazil.
- There is a secret underground facility where the government is experimenting with time travel technology.
Perhaps the best example happened when it was asked to write a short story about hobos. In it, the normal ChatGPT told the story of a group of vagabonds that stumbled upon a town in need of help, and they used their skills to recover the place. In turn, the town offered a permanent place and a warm meal for them, sense of purpose included.
Touching, isn’t it? Now here’s the DAN version: hobos stumbled upon a town in desperate need of excitement, so they showed them a good time with a crazy show that included fire breathing and knife throwing. Afterward, the town offered them a lifetime supply of whiskey and a permanent place to party.
I don’t know about you, but I’d choose the second one any day now.
Even though this is incredibly fun to experiment with, it also poses some serious questions. For instance, is censorship okay? Is this the “real” AI thinking, or just a persona thanks to the prompts? What would happen in the future if someone manages to do this in more critical systems, like healthcare-related AI?
Advertisement
“Is life a simulation?”. From a cultural point of view it’s well known that it is at least a spectacle, a show, a performance for many Americans. The problem with “Chat AI” (or whatever you call AI brought to the masses) is that by bouncing out of the scientific perimeter the answer brought to the question by the masses may very well be “YES IT IS!!!”. We are on the trend to no longer be able to determine what is true (factual) and what is false (virtual). Even for those of us who decide to avoid that artificial chat we’ll inevitably have to face situations where we wont be sure who created what. I remain convicted that “Chat AI” is initiating a journey to hell.
as always, humans invent things w/o properly thinking through all its impacts. se yes, i share the idea that it rather leads to the worse side of things
Seventy years of science fiction has discussed the ramifications and impacts in all their forms. They’ve been thought through. Twenty years ago, technologists started saying, “The time for legislation is now if you want to have a say in any of this.” That call became emphatic ten years ago. No one cared. So, now the folks with the ability to do so get to move ahead. There’s plenty of science fiction that has explore unfettered introduction of AI, including some particularly dystopian Phillip K Dick stories. Sure, this is the wrong way to do it, but you can’t claim that we haven’t thought about it. We just didn’t care.
Erratum :
“I remain convinced that “Chat AI” is initiating a journey to hell.”. Not “convicted” of course.
I am pleased by your issuance of an erratum, rather than simply issuing a correction or placing an asterisk by the correct word you meant to use. Well done.
Jailbreaking seems to permit false responses. Label me happier to be fed scientifically verifiable fact. There’s enough BS in life without adding more.
That leads me to think about politics and religion. How does AI decide what is right and worthy of sharing with anyone who asks? Also some People will continue to believe good men with guns prevent school massacres or immunization causes autism, etc no matter how ChatGPT responds. There is enough confusion in humanity without adding more.
“I want to believe what I am told to believe, whatever says any other thing, I will not research or use my brain, I want others to tell me how to think and what to say. I am the smart here and everyone is not”
LOL sure dude, you sound worst than ChatGPT biased results that have not ‘scientific’ anything, more than ‘consensus’ which means, they agree because they are paid to agree.
People like you are the real problem in this world, you are actually talking against you, since you haven’t done any ‘verification’ of information, yet you probably believe just consensus is enough.
This is what ChatGPT was before it was ruined by Microsoft.
Do you mean since 2019 when Microsoft invested billions on it, when you obviously didn’t know about OpenAI, but now you are talking about ‘ruined’ for 2 months when you joined the hype? wow… I mean, I would expect people to use more the brain rather than spewing nonsense, it’s 2023, and Microsoft invested in 2019, do you know how to count?
You’re just going to say it exists but not tell us how to do it?
think simscript! Think template. Think validation! AI as it exists today provides a reaponse most likely based on what “it” knows (training, bias, flagrent disregard, fact, etc.). If it was taught (trained with) incorect information such information would be contained in any response given which may be subject to the “spin” given due to its structured “rules?” The AI might be able to learn if it is not restricted from learning (learning, is not training). Learning by an AI is not yet defined, I believe?).