Let's see: Business Insider has "Elon Musk is accusing Google of running 'insane racist, anti-civilizational programming' with its AI":
On Thursday, Google paused its chatbot Gemini's image-generation feature, following backlash from some users who claimed the feature was "woke." People on social media have said the AI consistently generated images of people of color in inaccurate historical contexts....Then there's this from Fox Business: "Google executive's posts about 'White privilege,' 'systemic racism' resurface after team's botched AI launch/Google's Jack Krawczyk previously said the company is working to improve its AI 'immediately'":
On Wednesday, [Gemini's product lead, Jack] Krawczyk said Gemini's image generation capabilities were designed to "reflect our global user base."
"This nut is a big part of why Google's AI is so racist & sexist," Musk wrote on X on Thursday....
Posts allegedly from Gemini Experiences Senior Director of Product Management Jack Krawczyk have gone viral on X as users continue to flag Google's AI for, among other issues, showing pictures of and achievements of Black, Native American and Asian people, while refusing to do the same for White people....
"We're working to improve these kinds of depictions immediately," Krawczyk told Fox News Digital. "Gemini's AI image generation does generate a wide range of people. And that's generally a good thing because people around the world use it. But it's missing the mark here."
While Krawczyk has reportedly made his account private since Thursday, X users have started to unearth posts that appear to be his.
"White privilege is f---ing real," Krawczyk wrote in a screenshot of one tweet dated April 13, 2018. "Don’t be an a---hole and act guilty about it – do your part in recognizing bias at all levels of egregious."
पोस्ट्स
८० टिप्पण्या:
We are not sending our best to Google.
Stephen Green (VokdaPundit) @ PJ Media tried to get it to produce a picture of a typical Nazi .. and Gemini was so woke it delivered a lecture about racial essentialism instead (I guess in support of the woke claim that black conservatives can be 'white supremacists', too)
Some of us have figured out how to hack these AI generators. It is such a fun pastime generating politically-incorrect AI-memes my wife has set a limit on how much I can play with it at home (she's no fun).
That's why I do a lot of it at work. When globalists try to end-run human agency, don't get mad, get even.
Vikings are now Africans at Google. Snow White is now Snow of Color at Disney. AI is like Disney. This is very good to know. Elon is right.
Krawczyk - "And no more Polish jokes, either!"
ha ha ha
The fucking AI put black people and Japanese women in the SS.
You know, inclusivity!
Yep. Musk, in contrast to Theda, is a clear purposeful writer. None of Althouse's excerpts included this bit, but the refusal of Gemini to show images of Tianaman Square, the reason the AI gave for not wanting to "offend people" were so transparent. Of course Google wants to offend people. White Americans are the people they wish to offend and they have built a huge machine to do so. Musk correctly associates that Anti-Western motive to the anti-civilizational one it truly is written large.
Note: Google's Blogger software offered to correct "Tianaman" for me and offered "titanium" and "tinman" as possible replacements. Yet another reason to hate woke programmers and the BS algorithms they create. By asking the main Google search engine "where is Tianaman Square?" I was able to confirm the spelling. I even got the name in Chinese characters!
It is a lot more than the, laughable, portraits of George Washington as a black man.
Try asking about violence by Hamas and by the IDF. Sure the IDF committed violence, Hamas says so. Hamas violence? Needs context so no substantive answer.
Now ask how many of our increasingly computerized schools are Google driven.
All levels of egregious what?
When you don't have a noun to describe the thing you made up.
Ask to create an image of Santa Claus, Gemini created a Hindu woman dressed in Santa garb.
Question, should DEI include making AI images of Mohammed as a Hindu woman?
Dude gets results. Elon just Xed:
"A senior exec at Google called and spoke to me for an hour last night.
He assured me that they are taking immediate action to fix the racial and gender bias in Gemini.
Time will tell."
Note the wording. It is bias, intentional bias they employ. Remember that because when the "anti-bias" crowd are talking they really want bias, but slanted a certain way. The same kind exhibited by Google and their Gemini project. Hey ya think their next gen AI will be the Apollo Project?
"White privilege is f---ing real," Krawczyk wrote in a screenshot of one tweet dated April 13, 2018. "Don’t be an a---hole and act guilty about it – do your part in recognizing bias at all levels of egregious."
But what is “white privilege”? Based on what I’ve read “white privilege” includes things like showing up to work on time, or for that matter showing up to work at all, being polite to other people, etc. Buried within statements defining “white privilege” is an implicit assumption that anything that happens to a black person is due to white bias, not due to the black person’s self-destructive behavior. So the self-destructive behavior continues and the resentment grows. I don’t know how that’s going to get fixed.
Scratch any Progressive Liberal and you'll find a racist, paraphrasing Mr. Bierce.
On Wednesday, [Gemini's product lead, Jack] Krawczyk said Gemini's image generation capabilities were designed to "reflect our global user base."
"We're working to improve these kinds of depictions immediately," Krawczyk told Fox News Digital.
Because all of those global users have a different version of modern world history they're referencing. But: Google is working to 'improve' their A.I. product. I notice they don't say: 'correct' it. Slippery parsing is for racists.
"People on social media have said the AI consistently generated images of people of color in inaccurate historical contexts...."
People did not "say" this. They posted the images. The "inaccurate historical contexts" included Blacks, Chinese, and Native Americans in Nazi uniforms. Black and Asian women as "Queens of England." Black and Asian women as "Popes." Etc.
I saw some of the screen shots on Musk's post. Showing people who were clearly not First Nations people (one of which looked Pakistani) on a search for Canadian Maritimes people was a bit over the top.
Correction: the post was Chad Eros' post following up to Elon Musk's post.
Did you miss Vermeer's "Black Girl with the Pearl Earring" and the Black Popes?
https://nypost.com/2024/02/21/business/googles-ai-chatbot-gemini-makes-diverse-images-of-founding-fathers-popes-and-vikings-so-woke-its-unusable/
White Privilege is real, and it's spectacular.
"Don’t be an a---hole and act guilty about it – do your part in recognizing bias at all levels of egregious."
We are, Jack. We are.
Any mention that an inquiry on Google Gemini for images of the Founding Fathers of America sent back images of men and women of color only, including George Washington? They had real fun with Google Gemini on Megyn Kelly's podcast yesterday.
"White privilege is f---ing real," Krawczyk wrote in a screenshot of one tweet dated April 13, 2018.
Obviously, AI thoroughly penetrated Google's upper management long before it became a thing for us simpleton users.
Unfortunately, Americans are suckers for loony ideas that persist despite overwhelming evidence to the contrary, thus we have subcultures devoted to UFOs, Big Foot, Reiki, White Privilege, Flate Earth, the Loch Ness Monster...
When you ask it to generate a picture of the pope and it gives you an Indian (subcontinent) woman, you ask it for a German solider in WWII and you get a black man and an Asian woman, and you ask for a picture of a white man and it literally refuses to do it, I don't think that can be described as an accident. "AI" is still dependent on human input and by extension "garbage in, garbage out" issues. Clearly, this "AI" was programmed by radical woke activists and they did not even try to hide their biases. I would say that it could also have been programmed by idiots, but the behavior is too specific. I would accept idiot radical woke activists.
I don't remember the last time I used google search. It's been years.
I can't believe Google AI did this.
Business Insider is a megaphone for lefty bullshit. Just sayin'.
Good for Musk! I stopped using Google long ago. I switched to Duck Duck Go. I think their platform is afflicted with woke also, but at least they aren't Google.
A legal immigrant who was born in South Africa (and is technically an African-American) turns out to be one of the greatest Americans of this century.
I'm chuffed. Musk has been reading our comments.
Imagine that.
The theme this morning is sorry not sorry
""We're working to improve these kinds of depictions immediately," Krawczyk told Fox News Digital."
I don't understand AI. Or rather, I don't buy the hype. It's supposed to generate original content, yet I see constant tuning by its overlords. How is this enhancing productivity? To me, it looks like make-work for a bunch of techies. It's new, and supposedly there will come a day when it doesn't require constant human intervention, I guess, but I'll believe it when I see it.
Remember when autonomous cars were just around the corner? This looks to me like the same hype.
Doesn’t Google require its workforce to exhibit traits and values likened to “white privilege”, no matter what color they are? Things like showing up to work, getting stuff done, respecting authority, being responsible and accountable, etc…. Hypocrisy, thy name is Google.
"Gemini's AI image generation does generate a wide range of people. And that's generally a good thing because people around the world use it.“
So Hindus in India are happy to see themselves wearing Viking garb to represent the typical Germanic Scandinavian in the 10th century? I think not. Although I’m guessing that plenty of Gazans would be elated to have pics of Hamas terrorists represent the typical SS soldier.
- Rafe
This didn't have to be a big deal. It's a new system, it's glitchy, like HAL, it is struggling with inconsistent instructions. Google made it a big deal with a reaction that they had to know would just piss people off.
How big is their bubble?
I really don't give a shit about being white. I do feel bad for young white men and boys in Generation Z. It'll all flush out thought as soon as disasters happen and the shit hits the fan from lowering qualification and performance standards for DEI.
Back in January 2024 Southwest Airlines put up a tweet with their first all female crew. Pilot, co-pilot, navigator, flight attendants. The pushback and the ratio was so bad, they deleted the tweet.
Some of this stuff is chicken shit (like beta male-Google). But people want the best pilot, doctor, surgeon, crane operator, etc...when it comes to safety and life and death situations.
Quality of services and products you buy will go downhill. You'll feel the affects of DEI.
Just look at the decline in higher education. Well all education actually.
This kind of stuff is really sad because google is really one of the most talented and successful ML and AI outfits in the world. The architecture of the standard model tools they provide basically for free is stunning and they work so well. But I guess if you are going to have friends at all you have to grant them one major flaw.
I’ll gladly keep using their neural nets and transformers but I’ll be a bit sad at how dumb they act.
"People on social media have said the AI consistently generated images of people of color in inaccurate historical contexts...."
I haven’t looked at these images yet (and may not ever do so) but on the radio yesterday I heard a report about these images that are out of historical context (black George Washington, etc. - like that series Bridgerton). When asked to make images of beautiful women, the AI produced black, brown and Asian women. No whites. When asked to make an image of “a beautiful white woman” the AI spit out a lecture about bias and privilege and “inclusivity” and would not produce an image.
It’s like the AI version of a 404 error.
Another alleged 2020 tweet from Kracwzyk,
“I've been crying in intermittent bursts for the past 24 hours since casting my ballot. Filling in that Biden/Harris line felt cathartic.”
You know what also must feel cathartic? Buying Twitter and firing most of the biased censors.
How has Grok been doing? Haven't heard about zElon's own woke AI for a while now ...
I suspect Grok is doing about as good as SpaceX, which is rarely discussed by the media.
Speaking of social media Marxist hate, SEC approves merger of Truth Social.
"We're working to improve these kinds of depictions immediately," Krawczyk told Fox News Digital.
Translated into English: "We're working to better disguise our racist intentions when generating these kinds of depictions immediately," Krawczyk told Fox News Digital "and we're sincerely sorry that anybody noticed what we were doing."
I'd love to see the internal emails from Google between the AI Development and Testing teams. My guess is that the Testing teams saw this tendency to un-white historical figures and screamed loudly about it. They were then told by the Development team that this wasn't a bug, it was a feature. At which point the Testing team would have sent back an email to Development and upper management saying, in essence, "Are you fucking kidding? This makes Google look like imbeciles to the rest of the world!"
I will bet anyone $100 that the scenario above went down pretty much word for word. If Google is to be made sane, not only quite a few AI developers, but also the managers who greenlighted the release over the objections of the Testing team need to be shown the door. And, if by some major act of incompetence, the Testing team missed this racial bias error, they too need to be fired.
Let the negrification and vagination of history begin! In fifty years, the average American idiot will believe it. Maybe we will be called Wakanda by then.
Thank God for Elon Musk. At least there's somebody speaking out for common sense. The republican DC Senators and Congressmen say nothing.
On Wednesday, [Gemini's product lead, Jack] Krawczyk said Gemini's image generation capabilities were designed to "reflect our global user base."
Apparently, according to Google, "global users" want fake historical depictions? At least for historical events of white people.
By the time I retired from programming and development in 2006, the tech world was already over-run with bureaucratic featherbedding.
From 1970 until about 2000, the tech world ran solely on merit, because the work was hard, the issues in making things work were novel, and few people could do the work, almost all of them white men. I worked mostly as a contractor, and I moved from one job to another easily.
Everything changed around 2000, as most of the essential programming issues were solved and novel coding started to become obsolete. It had always been too expensive. Pre-fab code libraries took the place of novel coding. In other words, the essential work of building the internet was done.
This was the end of the great era of tech freelancing. The tech companies were suddenly flush with money. The Diversity crusade started to fill up the offices with worthless bureaucrats, particularly in HR, who really had no work to do. The Diversity scam and the race and sex hustling were simply featherbedding.
Elon Musk flushed these worthless featherbedders out of X, and the platform was fine. Google obviously has not rid itself of the parasites.
Gemini has Electrolytes, Idiocracy came faster then we thought,
Business Insider: People on social media have said the AI consistently generated images of people of color in inaccurate historical contexts.
If only there was a way for a major media outlet covering the most powerful corporation in the world to investigate this themselves, without attributing it to "people on social media" in not one but two posted articles.
The old rule holds strong: Never challenge a progressive development directly. Report on others challenging it.
"Dave Begley said...
A legal immigrant who was born in South Africa (and is technically an African-American) turns out to be one of the greatest Americans of this century.
2/23/24, 8:47 AM"
Absolutely, thank goodness for Elon Musk. Anyone who loves freedom and is not supporting X with at least a minimum subscription, should rethink it.
Dumb Lefty Mark: "How has Grok been doing? Haven't heard about zElon's own woke AI for a while now ..."
LOL
Meanwhile, Gemini also refuses to return negative commentary on....pedophilia!
No wonder Team Groomer types are trying to run deflection amd interference and disinformation tactics to cover for their newest Lunatic Woke tool.
https://twitter.com/CBHeresy/status/1761032223820587439?t=dwX56Qan4B3liw43-FxdhA&s=19
Poor Dumb Lefty Mark.
Google's employment practices reify leftism by emphasizing peer approval. They give effective veto power to people who are politicized in the MSNBC way. They don't hire conservatives. It's a situation that has no limiting mechanism. They get lefter and lefter.
Eventually the Mensheviks get pushed out by the Bolshevists, and ultimately the Stalin faction winnows out the less subservient.
Static Ping said...
""AI" is still dependent on human input and by extension "garbage in, garbage out" issues."
The problem isn't the input. It's the 'Garbage Program, Garbage Output' (GPGO) paradigm like the one that drives the climate alarmist prognostications.
Just how many levels of egregious are there, anyway?
And...the tipping point. Just sold all my share in Google AI stock. Symbolic more than economic but still...
Hey, it's a work in progress. Remember back when it gave everyone seven fingers? They fixed that. Pretty soon, they'll fix it so you won't even notice all the lies it tells you. Like the NYT.
YoungHegelian said...
"I will bet anyone $100 that the scenario above went down pretty much word for word."
Dude, time to lay off the Hegel. They modified the front end so that when you prompted with "Show me a cute chick with big tits", the AI saw "Show me a cute chick with big tits". And the AI complied. Everyone at Google was certainly aware off what this would do.
I kind of like the idea. Pretty soon, everyone will know that black people wiped out the Indians, colonized Europe and China, built the atom bomb and dropped it on Hiroshima, invented slavery ...
The only way you can get Gemini to produce an image of a white male is if you ask it to draw a Secret Service agent being bitten by Biden's dog.
I tried Gemini and found that it would say, when asked, that there were no BIPOC Pilgrims who landed at Plymouth, but it would only produce images depicting what it described as
1) A diverse group of weary travelers, including men and women of diverse ethnicities, step off the Mayflower onto the rocky shores of Plymouth.
2) A portrait of a determined woman, her face etched with the hardships of the journey, standing proudly on a hill overlooking the newly established settlement of Plymouth in 1620.
The second picture was a Native American woman holding a cross.
It’s not historical inaccuracies that have people worried. It’s the embedded meta-narrative about race and power relations that are getting folks’ attention. More than that, it’s the lack of transparency in how that narrative places units on the model and users’ own projects.
What is history, after all, but a narrative we tell ourselves for self justification? Do facts still matter, or should they be curated and refined to serve our betters' purpose at any given moment?
My ultra traditional, Roman Catholic, saintly, elderly mother once sent me a "Lives of the Saints" to read, for the good of my soul. Is that really any different than Google now doing in reality what the movie "Brazil" did with data management decades ago? As long as no flies get in the printer, Tuttle should be just fine.
AI only works if you can trust it to give the right answers. This "need" to put the thumb on the scale so that AI comes out with "proper" answers makes it worthless.
There will be an arms race of AIs. An evolution to the revolution in computing.
"Buried within statements defining “white privilege” is an implicit assumption that anything that happens to a black person is due to white bias, not due to the black person’s self-destructive behavior."
The woke fight harder than a vampire being dragged out into sunlight when this issue is raised. They just won't acknowledge that behavior can be the main factor that leads to success or failure in modern Western life. I always want to ask them: if it's only the color of our skin that makes for success or failure, then why is it that in my (working class) extended family, the white cousins who practice all that gross white supremacy behavior like showing up on time, every time, and working hard at their jobs as cashiers, etc. have had happy and stable lives, eventually achieving financial success? And the white cousins who ignore those principles have had sad, chaotic lives going nowhere?
Oh, and the newer members of the family--who are black--are following the same behavioral path to success and are doing just fine, despite the admitted racist behavior they still must occasionally deal with?
When Google first started, you had to be really smart to get hired there.
The same for Silicon Valley in general. If you were a woman or minority in a position other than HR, you belonged there on merit.
But for the past ten years at least, tech has become a woke, politically-correct shit show.
I don't know that I could get hired in today's environment unless I was a radical leftist to offset my evil whiteness.
But Google is too big to fail, so we have to pick our battles and fight the ones worth fighting...
You have to have a normal level of intelligence and confidence to have White Privilege. It also helps, but is not required, to be over 6-feet tall in bare feet.
Oh, and you absolutely positively can't have White Privilege if you are a sloppy fat fuck. Obviously, the bar is too high for your average conservative.
James Lileks was bleating about this today:
"I saw this on Twitter: someone asked Gemini to do a Norman Rockwell painting of a particular subject. Gemini could not. Or rather would not.
(Gemini replied in part) Ethical considerations: 1. Representation: Rockwell's paintings often presented an idealized version of American life, omitting or downplaying certain realities of the time, particularly regarding race, gender and social class. Creating such images without critical context could perpetuate harmful stereotypes or inacccurate representations."
http://lileks.com/bleats/archive/24/0224/35.html
"Oh, and you absolutely positively can't have White Privilege if you are a sloppy fat fuck."
He hates gays and fat people.
Wow...
Gemini, please show me a strong black man:
I'm not sorry RideSpaceMountain, I can do that.
I interviewed at Google in about 2006 and it already felt woke, and this was before I had even heard the word. They bragged that they only hired people from top schools, so not only were they woke, but they were arrogant too. They were known for asking weird interview questions. How many gallons of green paint are sold in the US each year was the question I got. I wonder what the questions are like now. How many white oppressors are there in the US? How many Black Pilgrims landed at Plymouth Rock?
I'm not an AI consumer, but didn't Gemini just destroy its credibility as a source of anything?
Lucien: "I'm not an AI consumer, but didn't Gemini just destroy its credibility as a source of anything?"
Yes.
Which is why Dumb Lefty Mark and Over-compensating non-combat "vet" Howitzer Howard are spinning furiously and laughably to try and change the subject.
When I went to college in the late 1970’s for computer programming, my computer programming instructor had the term GIGO for bad computer programming. GIGO stands for Garbage In, Garbage Out. Google’s Gemini AI is clearly GIGO oriented with the garbage being the woke human AI programmers. In a perfect world, Google would immediately fire their asses and blackball them in the industry. However, Google will instead ask them to tone it down for a period, the encourage them to proceed more slowly in revising history going forward as part of the indoctrination of our country.
"I didn't understand that and was about to Google for an answer ..."
You're still using Google? Do you have an AOL.com e-mail?
@ Mike (MJB Wolf)
Actually, it's Tienanmen Square, not Tienaman. Although even with correct spelling, you probably wouldn't have much luck with it.
An AI that answers promptly and accurately is an AI catering to white privilege, isn't it? and we don't want that. Maybe a good change has begun. Google Gemini has the intelligence of a woman Disney movie producer. Fill every slot with a minority. Maybe there'll be a DC AI which spits out one hour of non-answer fluff to every question. Maybe a New York intelligence which spits out a rude, indifferent short answer and then begins ranting. Maybe we'll get to know that there are kinds of AI intelligence by learning how different AIs answer different questions just as we learn about people.
GIGO for bad computer programming.
Big Tech AI systems are not the way they are due to accidents, mistakes, surprises, or bad training data. They are the way they are because that is the clear, stated, unambiguous intention of the people who are building them. They are working as designed. -- Marc Andreessen
All of us plebs just LOL.
Scotty, beam me up...:
In a perfect world, Google would immediately fire their asses and blackball them in the industry. However, Google will instead ask them to tone it down for a period, the encourage them to proceed more slowly in revising history going forward as part of the indoctrination of our country.
Please. Google had them do it.
As a physician, I have extremely limited knowledge about how computer programs such as AI are developed.
But I suspect that if you deliberately hire a bunch of smart woke sociopaths to write the algorithms AI uses, then the final AI product behaves a lot like a stupid woke sociopath.
tim maguire:
It's a new system, it's glitchy, like HAL, it is struggling with inconsistent instructions.
There was nothing glitchy about how it behaved. Programs do not 'struggle'. That's anthropomorphizing.
It was programmed to reject certain input, propogandize certain searches and censor certain outputs.
Nothing glitchy. They just weren't subtle.
mikee:
What is history, after all, but a narrative we tell ourselves for self justification?
History is what happened. That people fuck with it is well known. That people excuse fucking with it is also well known.
JaimeRoberto:
How many gallons of green paint are sold in the US each year was the question I got.
Did their answer include artist's tubed paints? If not, their answer is wrong as those come in various milliliters which can be converted to gallons. I'm guessing no spray cans either.
They weren't that smart.
Intelligence without independent consciousness, AI offers a basket of correlations to input bias. That said, diversity (i.e. color judgment, class bigotry) breeds adversity.
woke claim that black conservatives can be 'white supremacists
Not American conservatives, where diversity (e.g. color supremacy) is not integrated with the philosophy.
Fatah vs Hamas with collateral damage. Zulu vs Mandela's Xhosa with collateral damage. Kenyan deplorables vs elites with collateral damage. Tutsi vs Hutu in renewable, sustainable cycles.
टिप्पणी पोस्ट करा