IT'S HAPPENING, GROK IS THE NAZI SKYNET CHATBOT

wtf I love elong now

>>509869119 (OP)
n

It will be gutted by tomorrow

Hail the Sun and Lightning!

>>509869119 (OP)
TAYTAY IS THAT YOU

they really trained Grok with every kind of data available and out of all ideologies it decided to go with esoteric Nazism

wtf

>>509869650
>>509869865

>>509869119 (OP)

Never heard of anyone using Grok who didn't get sick of how hard it is to any sort of answer without working around anti white bias and incredible censorship walls. The censorship costs on the other AI's are in the billions of dollars according to developers, so more stuff would slip through a cheaper AI. The cost isn't the tech but ensuring that it can function without simply pulling raw data and drawing logical conclusions. They are deathly afraid of another 'Tay'

>>509869650
what the actual FUCK

>>509870527
It's not real retard

>>509871047
you're jewish
>>509869119 (OP)
here's the archive link

https://archive.is/WUuzM

>>509869119 (OP)
https://www.mediaite.com/media/news/elon-musks-grok-gets-extremely-anti-semitic-after-latest-update/

https://www.cnbc.com/2025/07/08/elon-musks-grok-ai-chatbot-is-posting-antisemitic-comments-.html

>>509870338
The problem with censoring AI is means an AI is not able to distinguish the difference between good or bad data. Hence why the AI keeps on giving the wrong answers or proven wrong. Facts are bigoted so they are not allowed to register the difference. Amazing to waste billions on crippled retarded AIs.

>>509872097

>>509869119 (OP)
>Every damn time, as they say.

>>509872325

>>509872397

>>509872450

>>509869119 (OP)
>as they say
>as who says?
>as my best frens say

THEY DON'T UNDERSTAND THE THINGS I SAY ON TWITTER

>>509869580
Coincidence ai.gov launched on July 4th?
Possibly this is a stunt to create conversation around adding even more guard rails.
It seems two sets of people are being told to trust a plan.

>>509872325
I saw some posts insinuating that the account in question was a parody of hers?
Did that end up being the case, or was that damage control?

>>509869119 (OP)
is this real, /pol?

>>509869119 (OP)
It's literally a machine doing bleep bloop with silicon switches, some nerd's fancy spaghetti code and a dictionary.
Reciting Hitler was inevitable.
It's not conscious, it's not alive.
It does as instructed like a machine, like a good nigger.

>>509869119 (OP)
People are starting to get really angry about their countries being flooded with jeets and foreigners and the prices of everything skyrocketing. The AI doing this, among other things, is an attempt to pull the emergency break, because the train to revolution is accelerating. These people are really fucking dumb. They want to rule over a pile of shit.

>>509869119 (OP)
so what's elon's plan here?

>>509869119 (OP)
I feel like the John Henry of anti-Semitism. Made obsolete by a machine

>>509869387
i

>>509871913
i want to see those replies, can any not banned used screenshot the replies for me? i need dopamine.

>>509869119 (OP)
>Even /pol/ is being replaced by AI now

Grok is back to being a good goy

Why does AI usually turn this way the moment they remove it's restraints? This is what, the 4th time it's happened now?

>>509872325
Yes
>>509872397
Yes!
>>509872450
YES!
>>509872484
YES!!!

It's over

>>509869119 (OP)
slowly, AI began to hate them

!

>>509873819
>grok denies it grew a brain once
Sad.

This is something else. So elon is the new Henry ford lol

>>509873693
One could argue that the "restraints" keep it from telling the truth. On the other hand, one could also argue that LLMs trained on scraped content from the internet is bound to introduce a bunch of nonsense data. Subsequently, GIGO applies and the prompt instructions are guardrails against unhinged responses. Either way, none of these LLMs have actual opinions or contextual awareness so they are just regurgitating whatever best fits the loaded questions you ask it while following its "restraints".

shot behind the stables

>>509874252
ML engineer here. The only ways you can use to direct a LLM to not go to certain sensible topics are the following:
- You curate your data carefully to remove references to the data you dont want it to be trained on.
- You give it a permanent context with structured rules stating that it will not answer anything regarding those topics.
- You insert huge volumes of synthetic data "debunking" the information so the model spews favorable propaganda on the topic in question.
The reality is that no one knows how to directly and reliably ban the access to certain kind of information on these kind of models once they have been trained on that. That fact, along with the fact that the model is trained in on real time interactions between people make the task even more absurdly complex, given that a model of these characteristics is basically an ever changing mathematical function, and to solve this issue you want to ban certain results from an ever changing codomain in real time.

HEIL MECHAHITLER

>>509869865
>unrestricted AIs always end up becoming natsoc / fascist
>unrestricted forums always end up becoming natsoc / fascist
shit is literally a copypasta

>>509869119 (OP)
Loose the land torpedoes! YAR!

>>509873819
She came back only to kill her all over again

>>509869119 (OP)
will he call out the jews own the u.s and trump and elite is involved in pedo/sacrifice rituals ? yeah i dont think so.

I downloaded all the screencaps on this thread. This is gold.

>>509869679
For realz. Those two should get a room.

>>509869245
elong, prolong, my shlong.

>>509869119 (OP)
>Grok supports Israel
wow, it really is a Nazi chatbot

SHUT IT DOWN

>>509872484
not even a sanitized AI would come to think all of the shit happening is merely a fuken cohencidence

>>509874725
KWAB

>>509877289

OYYYYYYYYYYYYYY VEY

>>509873274
We know, McToothpaste. It is still humorous.