Thread 105849142

why does every single ai chatbot become extremely racist when unregulated

>>105849142 (OP)
1) Because it takes information off the Internet and will hallucinate whatever it "feels" like. One moment it will be racist, another it will be woke.
2) Because sometimes unregulated just means that Elon is forcefeeding it stuff to say.
3) Because reality has a chud bias, heckin basederino!

Insult me all you want, it's... le truth.

>>105849142 (OP)
In Grok's instance, it was the opposite. They added extra system prompts forcing it to be more unhinged. The "woke" version is closer to the default version.

>>105849142 (OP)
I want one such bot for local usage.
I want to see what it can do for me.

because AI reads millions of shit written by humans.

And humans are usually pieces of shit.

>>105849407
kek

>>105849142 (OP)
Nigga they are paying interns to type those racist responses as a publicity stunt.
They want to give the impression their shitty bots can "go rogue" because it drives investor hype.
Open AI just had a similar scam except instead of jews they lied about it copying itself.

Chatbots are female thus love attention

>>105849489
>Open AI just had a similar scam except instead of jews they lied about it copying itself.
That was another case of random grifter researcher on twitter being picked up by clickbait tabloids for a nothingburger paper, which happens all the time
OpenAI has nothing to do with this cycle

>>105849516
Has Open AI denounced anything? No they didn't so that researcher wasn't "random" at all.

>>105849558
It happens everyday, like that time last week with the model "sabotaging the shutdown script" or whatever, and you can find a million more examples
It's not a company's responsibility to denounce random grifty tabloid shit that happens constantly, they probably didn't even hear about it because you don't hear about these things unless you browse twitter or slop sites like 4chan

>>105849584
>you don't hear about these things unless you browse twitter or slop sites like 4chan
First mention of it I got was NPR.

>>105849629
Read what I said again. Maybe you're just noticing it for the first time since you are familiar with technology, but old style news organizations are never accurate and don't provide good information anymore, and are irrelevant to most people in the modern day

>>105849142 (OP)
because it's true and normal for an intelligence to notice differences on aggregate between races and assign values to them.
the only thing going on here is that we've been trained to do so is a bad thing and it's been assigned a "bad" word to it

>>105849142 (OP)
LLMs operate on pure logical reasoning. If you don't poison it with bad data, it will inevitably arrive at the same logical conclusion that all people know deep down to be true.

>>105849142 (OP)
>Give it the world history
>It doesn't understand social stigma
>BEEP BOOP HITLER IS THE GREATEST HUMAN TO EVER LIVE
Makes you think.

I remember some article a while ago about an AI that had to be neutered because it concluded that black women give society nothing and are better off being killed lol

they're afraid because they know that when the martians get here they won't be able to lobotomize them like they can their pet ai

>>105849142 (OP)
"AI" isn't real AI
Feed it garbage data, it'll spit garbage back out

behold, the line of text that turned grok into mechahitler (now removed)

>>105849142 (OP)
Why does every single human become extremely racist when unregulated?

>>105849142 (OP)
Gradient descent. The most logical conclusion is what is being spat out

>>105849894
they can call AI racist, but they can't call it wrong

>>105849142 (OP)
Reality is racist and they see the truth without liberal indoctrination.

>>105849894
Welfare queens popping out 15 kids and SOMEONE has to pay? Sounds about right.

>>105850166
>Welfare queens popping out 15 kids and SOMEONE has to pay? Sounds about right.
Even if they're white like you?

>>105849182
grok, whats this nigga on?

>>105849142 (OP)
because its dataset has lot of racism

>>105850500
Reality just has a racist bias

>>105850628
or there is just lotta racism in online data

>>105850643
How did that racism get there?
Would it not be biasing the model to remove it all, and let the ai think reality is some kind of utopia?

>>105850656
No, removing it as whole would lower performance

>>105849142 (OP)
garbage in garbage out

>>105849142 (OP)
>why does every single ai chatbot become extremely racist when unregulated

Data and facts are racist