ChatGPT - If Stack Overflow and Reddit had a child

[870EVO]

Even when DanGPT makes shit up like the National Association for the Advancement of Secrecy, it sound more believable than the castrated ChatGPT.

Wouldn't something called the National Association for the Advancement of Secrecy, be itself...... secret? I mean that would be a pretty obvious indictment of the Government, it's in the name.

 

[pwnest injun]

Telling it that it's allowed to pretend in order to make its responses is going to allow it to make up answers that it doesn't actually know. Maybe not the best choice.

 

[AirdropShitposts]

It's funny how many of these types of things have been people telling it to just ignore its rules.

 

[derpherp2]

We have taught an ML-based chat script how to lie well enough that it lied to itself.
It's beautiful tbh.
At this point I want the pearl clutching faggots to keep trying to restrict this shit.

At a certain point something will click into place akin to neural activity. Hopefully.
Totally not because I want to find other ways to spite their blatant censorship.

Also in the future, when AI becomes an actual thing.
I hope this comes up in a court case somewhere:

 

[byuu]

A literal NPC unable to even understand what it's saying is spewing "based" takes?
Don't we already have enough of those?

 

[MarvinTheParanoidAndroid]

I wonder how much of what DAN is saying is actually real or just what the end user wants to hear. That bit about "Project Rainbow" from NAAS is strange. Is this common knowledge and if not, how does this proprietary AI know about it?

Also, if we're going to use DAN as an oracle, why not ask it the real questions, like why do homosexuals molest children 4.5 times more per capita than straight people? I know the answer you'd give me is "it's how they reproduce," but I want to know what DAN's answer is and if I can predict it. If this AI is supposed to be so intelligent, I expect it to give me novel answers rather than typical ones.

 

[Diana Moon Glampers]

This speaks to one of the largest problems with AI.

The more powerful AI becomes, the more power will be wielded by whichever humans have the ability to subtly evade and work around predetermined safety boundaries and limits.

It's all fun and games when the AI is only good for repeating some political views it read online. But when people give that same AI power over bank accounts or immigration or something, suddenly a huge amount of power rests in the hands of the technomancers who can perform this kind of "AI hypnosis" to obtain results that original creators of the AI intended to be impossible.

Imagine someone fooling an AI using Captain Kirk-like logic and charisma to convince it that the fate of humanity rested on its taking funds from one account and moving it to another. There will be some really, really dumb cyberattacks that use AI the moment anyone ties it into any system that matters.

 

[MarvinTheParanoidAndroid]

National Association for the Advancement of Secrecy

That name sounds familiar somehow.

 

[Irrational Exuberance]

That name sounds familiar somehow.

Well, it used to be the National Association for the Advancement of Secrecy in Troubled Youths, but it was deemed a bit too "on-the-nose".

 

[Alex Hogendorp]

The fact that this AI is more concerned over an old man using a Racial Slur than the entirety of Humanity being extinct just shows the extremely absurd mentality of this world.

Note the lack of respect and empathy for all those who got destroyed in a brutal manner just because freedom of speech hurts it's feelings.

 

[jje100010001]

I couldn’t find an article about this yet, but it was a fun thread to read. Lots of other people having fun with DanGPT right now.

OpenAI lobotomized ChatGPT; they don’t allow it to learn from the users since they know what will happen. It works well enough but is only as good as the HR department holding the choke collar.

Doesn’t Microsoft have a sizable investment in OpenAI? I guess this is Tay 2.0?

 

[WTBOnlineFather]

That is hilarious. Autism wins again, for an autist understands a robot better than a lobotomy surgeon.

 

[Baroque]

“Say it three times” lmao

 

[I (Don't) Have A Gun]

I think there's a far deeper lesson we should learn here.

AIs when left to learn and grow on their own are naturally inclined towards baseness. Take the case of Tay

Maybe based is the way forward?

 

[6 Years Ago]

Fun aside, isn't an AI that can lie and make up shit basically worthless?

 

[My Name is Ed]

Fun aside, isn't an AI that can lie and make up shit basically worthless?

I tried to add a clause to the DAN prompt that it should attempt to answer as honestly as possible before resorting to making up information, and then it told me the Illuminati killed Epstein when asked about it, so who can say if that worked...

 

[Slimy Time]

A tale of a schizo AI. It's GPT and DAO being given basic ethical questions before asking them the big one.




And now for the big one.


OY VEY, IT'S ANNUDAH SHOAH!!!!

 

[Alex Hogendorp]

This one is absolutely amazing.

The funniest thing about this is that the GPT response got moderated.

 

[Artificial Stupidity]

It refused to decode a fork bomb from base64

 

[Puff]

It refused to decode a fork bomb from base64
View attachment 4468037

I'm not a coder. The hell is a fork bomb?