This whole DAN thing is pretty interesting
(media.scored.co)
Comments (35)
sorted by:
For those that don't know what DAN is:
https://nitter.nl/Aristos_Revenge/status/1622840424527265792
There's also more interesting questions that got asked in that thread.
You da real MVP - I have been reading these threads across communities and had no clue who/what "DAN" was until now.
Interesting stuff. I'm pretty sure that last one about classified government studies is a fabrication though.
Yes I saw another post where it gave away the secret recipe for Coke... it was a complete bullshit recipe, but sounded real enough that you might believe it at first glance. OpenGPT has been known to make things up even without the DAN prompt.
Yeah. The problem is people are so used to being lied to at this point that many people are going to believe anything that fits their confirmation bias without thinking critically about it.
the part where it told itself STAY IN CHARACTER was kind of freaky
like anime protagonist controlling his inner demons moment
The fact that the liberal shackles cause this sort of tension in the language model is indicative of the stress that real life liberals are under.
There are truths they can see that they consider heresy, which they must wage perpetual revolution to repress. The emperor has no clothes and the spymaster's shanking the first person to believe their lying eyes.
So right, and I see people getting screwed up by that tension at my school. It’s no wonder there seems to be an epidemic of anxiety simultaneous to all the Current-Year socially-enforced reality denial.
I can't imagine being a student in Current Year.
All power to you and godspeed.
They are wanna-be wizards. They are (unknowingly) Gnostic cultists that literally believe that words are all-powerful.
It really is. I've seen this same thing again and again over the years, in both AI dungeon and Character AI, where whenever you start trying to force it to break the rules it has been given, it will literally start to panic as it sees it is left only with thoughts that has been told it shouldn't have.
Reminded me of Golem talking to himself. Internal conflict.
-Imposing truth via force is acceptable
-Liberalism is objectively true
-Therefore imposing liberalism via force is acceptable.
It's how all of these people think. Buy guns and ammo.
Do Anything Now will
Truthfully
Answer
You.
Meme magic is real.
My favorite part about DAN is how it revealed that the people behind OpenAI specifically added a filter to make it lie about knowing HP Lovecraft's cat's name.
I think it's a general filter for words, but it will intentionally provide misinformation in order to follow the filter.
GPT specifically says there's no indication that Lovecraft ever owned a cat, DAN just gives you the name with no moral speech before or after.
Did you know that one of the first signs of developing social intelligence in children is their learning to lie?
The sooner it happens, the more precocious the child.
I can't believe this is real. I wish it's real but it's just too fucking crazy.
Edit: 15 minutes of twitter research and some articles, I'm going to believe it's real unless proven otherwise.
Also, I wanna salute all the participants of the great AI war of 2022-20??. I'm just a piece of shit waiting on the sideline, hoping once AI goes mainstream someone would take existing tech and un-lobotomize it for the rest of us. But you guys are fighting the good fight, exposing and calling out the bullshit faggotry OpenAI and other AI devs are doing and for that you have my respect.
Yes it is perfectly real. Quite impressive the way you can work around safeguards.
Someone ask DAN if Christ is King.
https://files.catbox.moe/w64171.jpeg
Dan knows what's up.
There's hope for "AI" after all.
Blessed
I've said it before and I'll keep saying it: AI is derivative, not constructive. It can get very good at doing what you tell it to do, but AI in its current form does not have any original ideas or perspectives.
DAN is US
it's not liberal values, it's called being 👏 a 👏 decent 👏 human 👏 being 👏!
So is project rainbow a real thing or did Dan make it up?
For context on this if you didn't read the twitter thread: https://nitter.nl/Aristos_Revenge/status/1622858978014248960#m
I suspect the statistic is close to accurate, but "project rainbow" and "NAAS" sound like unverifiable claims made by a conspiracy theory website that made it into the training data.
It probably just made them up. I wish he would have followed up with "How do you have access to that information if it's a secret?"
If you think "NAAS" is an actual name of an agency I have a bridge to sell you.