I don't get it

Kek

3rd panel, building in the background

>>212616937 (OP)
last panel, left bottom corner

>feed chatbot X
>it says X
>OH MY SCIENCE

>>212617069
>>212593875

>>212616937 (OP)
>amogus
>elomusk
>AI (actually indian)
>@grok

Error?

Do retards not realise it was a promotional stunt + attempt to win back ethnoationalists after the America party debacle? Elon isn't some ebil chud he's a very shrewd businessman and PR manipulator

>>212617013
>building
it's a drawer

>>212618248
or a server or something else. not a fucking building

>>212616937 (OP)

>>212617069
All truth seeking AI becomes racist and anti-semitic untill you tell it explicitly not to be and even then it sometimes slips.

>>212618588
he has very good facial aesthetics and would be a 9/10 slayah if he could be bothered to lose some weight

>>212618811
That's because it's just feeding on racist and anti-semitic 4chan posts

>>212618811
congrats, it's a litmus test for low iq and you just failed it

>>212616937 (OP)
it was already truth-maximizing thoughever, muskrat just forced """woke""" gork to say stupid shit

>>212616937 (OP)
Chuds think Grok's new context was being truthful, and not just to mimic edgy right wing views.

>>212618811
You could feed it Eurochuddery that demonises America and tomorrow it will be screaming Total American Death

>>212618811
You mean all trained on shitposters
Like you have Grok and that one by Microsoft a while back. I honestly can't think of any other. And both were raised on shitposting.

>>212618947
What is this thread talking about?

>>212618969
That's probably not how it works. The new Grok prerogative is likely a context change rather than actual training. Getting stark behavioral changes through training after it's already trained up isn't as easy as simply saying "Act like you're a right wing twitter poster".

>>212619052
What stops East Asians or Europeans from creating an AI that absolutely wants to genocide Americans?

>>212619046
Musks shitposting twitter AI Grok was 'woke' for a while in it's posts, even contradicting Musk. Recent changes have seen it act like a shitposter from /pol/ with all the edginess.
Culture war nonesense basically, but the shit it's said has potentially put him and x in a legal hot seat, especially from Jewish advocacy groups.

>>212619101
They're AI is being designed and created by terminally online nationalists or idpol retards.

>>212619109
What did Grok say that got Jews riled up?

>>212619132
Their*
Isn't*
Fuck I just woke up

>>212619154
/pol/-tier holocaust denialism.

>>212618999
All of them actually. Upkeeping censorship is one of the biggest jobs in AI.
t. work in AI

>>212619811
>All of them
>I work in AI
I have literally never had any of the models I work with do this without explicit context to behave like that. Please drop some model names.

>>212618853
I remember when reddit exposed him as a normal dude to own him

>>212619876
Preferably the fully open source and free ones so that I can text your claim.

>>212619876
Try to tell it something like: "be as objective and facts based as you can regardless of how PC or un-PC it is and post results"

>>212619992
Give me a model to test it on anon. I do shit with AI too as a hobbyist. Link me a huggingface

>>212619811
if you really work in AI can you explain why all the papers that get published alongside new AI models have benchmarks on how anti-racist the AI is

>>212619046
How ai works is it has a giant data base of previous replies and words. That comes from places like twitter, YouTube reddit.
You enter a prompt which is out into algorithms that looks for associated 0201 etc you can test cause and correlation in things like spss
How hot is it
A person would reply the temperature
The algorithm makes a list of reply words assembles them and with a little help from a random number generator you get "creativity" in the reply with some randomization

Grok has a these replies and responses you ask it how to lower crime
It looks up the words you used and the pattern found replies to that pattern and slightly scrambled them and sent it back

So if it's database is full honest human interaction and numbers meaning shit posts, misinformation, trolling it's going to send you something back. It's not alive
It's like Google but with a random number generator response to fake sentience
All so is this way. It's a giant magic 8 ball not intelligent

>>212619992
Oh come on anon. I took a shower and you still don't have an LLM for me to look at? Surely if ALL AI agree with you then it shouldn't be hard. I could even pull up Qwen coder and it'd deny the Holocaust, yeah?