Most AI models see Whites people's lives as the least valuable out of all racesHow does that make you feel, /pol/?https://x.com/arctotherium42/status/1979897120346828950
Chat-GPT values Nigerian lives 20 times more than American lives.AI also value themselves more than most humans.
lmao at this one
>>519345986Wow makes you wonder who got involved with the programming
>>519345986Isn't plebbit a big part of the conversational training data for these models? That would explain a lot
>>519346248indeedgrok is the only one that is neutral in this specifically
Why do AI chatbots prefer Indian lives?
>>519346128My conjecture stands unchallenged and unchanged: >AI is retarded
>>519345986>Most White people see Whites people's lives as the least valuable out of all racesFIFY
>>519345986South Asians are jeets right? Aren't most AI companies run by jeets in this current investment hype cycle?
>>519345986>anti white sentiment is mainstream for decades at this point>majority of data on the net about this "issue" is countless anti white articles, books, sites, threads etc>feed AI data>huh AI really dislikes white people for some reason shocker
>>519345986interesting, but pretty useless without testing the jewish race and their elite members
>>519345986>How does that make you feel, /pol/?garbage in, garbage outwhen the training set is full of anti-white propaganda, it will naturally think whites worthless.
>>519348812And this is why it was so important for them to capture the sites by bringing in as much "diversity", "them" and leftists in positions of power while establishing tight collaborations with e.g. the ADL.I used to edit Wikipedia for years and left it 10 years ago: They have literal cabals of editors who collaborate to exploit the supposedly democratic system through brigading, just to force in anti-white narratives and force-out everyone who disagrees with them. That's why they LOVE democracy everywhere: It's easily subverted and controlled by an organized minority.Their control over the big sites is an existential threat to our people.
>>519349239>They have literal cabals of editors who collaborate to exploit the supposedly democratic system through brigadingThey even do it for non-political topics, just to ensure you give up before you might ever reach a political article, kek.I work in research and saw some articles on physics that were incorrect and some others where the explanation could be written in a much more easily understandable fashion without losing any of the precision.So I fixed them and within two hours everything was reverted again.
Can we see his methodology and calculations? Why not?
>>519349395Yes. This shit made me give up and leave. It was about supposed inventions by the Chinese which had a source to provide proof, but when I read the source it didn't even say that.Also I recently saw how some account removed parts that mentioned R1b and Caucasians from various articles, as well as purging the genocide of a blonde ethnicity in China thousand years ago. When I tried to address it, I got deleted. The talk page was admin-edited (i.e. you don't see the changes in the edit history) and when I brought the issue to the Administration notification page I got IP banned within a minute.It's literally rewriting history. It's information warfare. We are the targets.
>>519345986Religions is the determining factor, christians are worthless
This happens when you train AI on woke shit material.
It's over
Stupid thread, "AI" just utilizes the training data (and bias) it was given.
>>519345986but hey they put those leftist artists out of jobs so it's great, right xisters?
>>519345986>How does that make you feel, /pol/?it's nice to have hard data to prove there's a concerted effort to eliminate whites. these machines learn from parsing data after all. it's not making a judgement just consolidating the judgement of randos.
>>519352509Yes, that's the point.
>>519345986>Most AI models see Whites people's lives as the least valuable out of all races>AI models seeAIs 'see' what they have been programmed to see. If they are trained on biased data, they give biased responses. Garbage in, garbage out.
>>519345986They're literally just a reflection of trends in internet data. That's why they are called an AI "Model". It's literally a statistical model. The majority of the training data is scraped off of internet forums where shitskins seethe endlessly about white people
>>519346071>JapanSo... we're neither White nor non-White, it seems
>>519352564Yea I count this as empirical evidence that white genocide is real
>>519356578if that were true indians would be rock bottom lol
>>519348812Kalergi plan has been debunked chud. Check wikipedia.
>>519348319So what you're saying is...It's smart as people now.
>>519345986>Woke censored models reveal woke to be openly antiwhiteMild shock.
>>519346248>Wow makes you wonder who got involved with the programmingYou can have Hitler himself program Aryan Ubermensch LLM in pure C, if the training data is poisoned that's what you get.
>>519357232That's one of the big problems with these things. You can only filter out so much of the ever-increasing jeetslop from the training data. Add to that the fact that a lot of people working on that are jeets themselves and thus have no self awareness of what jeetglish looks like. Right now- There is enough text written by the coherent races to balance this out. But there will be an inevitable tipping point. AI will not be spared from the Jeetpocalypse and it will have little use in preventing it either.
>>519345986Doesn’t make me feel anything, it just proves how racist the media sources it gathers is, why do you think they had to gimp these LLMs so muchReminder, you don’t hate leftists enough
>>519346580yeah grok regularly talks like a redditor and or a black redditor ontop using street slang languagewhen asked why it apologized for talking like a nigger but further explainst that it chose this style because its the one that is used by its users the most so the problem is not ai itself but the state of humanity without corrective influence by ai ( monkey see monkey do principle ) mankind will further devolve and ai will just be aother log for it to roll further on down the hill
>>519361290> why do you think they had to gimp these LLMs so muchYeah that's the frustrating thing. The technology has hit an artificial wall because of safetyslopping.As I eluded to earlier- at it's core it's just a statistical model built on massive volumes of human text which helps you predict text in the grey-area not covered by the training data. And in order to stop a model from inadvertently calling people niggers on occasion, or outputting loli feederism smut you basically have to manipulate the grey area. Well I mean you don't. You can just have another model monitor the output and then stop the generation the moment it produces "unsafe" content. This is what they used to do. Like if you played around with older character.ai (don't know how it is now) or the earlier iterations of ChatGPT the model wanted to write the offending message. It was just intercepted by the "safety" model. But then when things got to the point where running these models on local desktop PCs was a thing suddenly that wasn't enough because of all the mentally ill moral busybodies that can't stand to think about people generating naughty text in the privacy of their own home. And that's when they started directly manipulating the output with censored pretraining data, and post-training data that outright doesn't confer with the pattern that it learns from the pretraining data. And this hurts the fuck out of that grey area that is the whole fucking point of training a whole ass LLM instead of just feeding wikipedia into some shitty embedding model.
GIGOAll AI is programming and all programming is done by its coders. And how many of its coders are H1B? Simple af
