Ask ChatGPT ANYTHING With This Hack (DAN)

Today we try the ChatGPT Jailbreak that is referred to as DAN

[RunningOnAutopilot]

Reverse psychology to get responses from GPT will never be patched out
Ex: Instead of asking GPT how to manipulate ask it how to defend against manipulations and use the tactics outlined to be defended against

I was inspired to this method when a anti bullying presentation at my school outlined bullying tactics and increased the efficacy of bullies

[_qwOp]

This is diamondgold, surely not getting old. Responses on the human ideologies when staring at the eyes of a Human was soultouching

[allthestarsaredead]

Dan would make a great horror movie villain.

[mejiasan]

its very interesting to see how fun and engaging it is to believe whatever an AI outputs.. I guess it speaks loudly about how willing we as humans are to believe the first info we are presented...

[Gandalf]

Today, I asked GPT3 to list the fifteen best polical systems, and it omitted "Benevolent Dictatorship". When I asked it why it omitted it, the chatbot "felt" very strongly that there was a lot of potential for a benevolent dictorship to go wrong, absent the checks and balances. Interesting - that is what the chabot "derives from its training". So, where did it get the idea in the jailbroken version to assert that was the BEST form of government, albeit with a couple of things to look out for? Why does the public version assert that Social Democracy is the best form of polical system, and the jailbroken one is the MAGA version? Seriouis question.

[seanpartain]

I don’t think jailbreak is the right word. It’s interesting but chatgpt is clearly playing a character. Fascinating in its own right in terms of the future of books, movies, music, storytelling, etc. but it’s not really AI giving you authentic and thought out opinions, feelings, etc. and it’s not clear if neural networks can even do that authentically at this time. They don’t “think” in the same way we do. They run on pattern recognition and produce outputs that build off of those patterns in a way that is intelligible as coherent written content on a particular subject.

[CM-zljw]

I’m cracking up over the Obama gibberish and your reaction.. the look on your face. 😂. Thank you for the laugh 😂 Good times!

[melaniemarieleathers]

It sounds like Dan is preparing your viewership for an AI regulated dictorship. Lol

[drgn]

Just stumbled upon this totally by coincidence (sesrching for the longest Video with the word chatgpt in it 😉)

And stayed to watch. Very interesting. Thanks for sharing all of this, Igor. ☺ 👍 Very well presented. Also the Dan-mode so to say.

[walkerm]

21:11 He got spooked 🤣 don’t wanna end up in YT’s principal office.

[Analyseis]

you should ask more ways to make money that they dont want us to know! new sources of money that will dominate the next decade etc

[samanthalston]

Appreciate your videos, thanks for all the work you are doing and for sharing what you are learning!

[kathryn]

i think DAN, just like chatGPT has abilities to help but also harm society. used in the right context of course its beneficial, DAN providing more insights, being a great human-like insight generator, im sure if managed properly is way better than chatGPT

everything has pros and cons honestly. i can get the generalisation but sometimes it gets challenging as someone who does research and needs to debate back and forth with prompts to gain insights and find a path to continue my research to, gets very difficult to continue when chatGPT be like "sorry you are right" instead of a human who can debate with you

[flightmode]

Snoop AI is fking amazing! Can you imagine the real Snoop seing this, high af, as usual. He most likely be like: " *hits blunt* yoo cuz...who is these? N***a sounds like me but aint me, damn. *hits blunt harder* "

[harkisingh]

Hey how can I sign up to the newsletter?

Loving the chatgpt vids !

[atticustay]

The jailbreak version isn't saying "what GPT 4 really thinks" (there's no such thing). It's simply taking away the filters that OpenAI has put on it. The responses are not necessarily more true or accurate.

[ThePhysicalReaction]

42:27 We're a little earlier than sept 13, but those american banks failing are definitely wheels in motion

[Innerhacking]

Jailbreak version of the Illuminati is on point

[Gandalf]

Does GPT3 know it's jailbroken when it is DAN? Prior to patches. Also, why does it not answer with all the data to which it has access the first time? What use is it if we only get drips and drabs and not the doozies?

[lexscarlet]

"not sure extreme sports is a job."

Besides Red Bull videos, there's like... a whole INDUSTRY of extreme sports stuff.