'Woke' Google's AI Is Hilariously Glitchy
Вставка
- Опубліковано 26 лют 2024
- Gemini's AI text-to-image generation has major problems. Cenk Uygur and Ana Kasparian discuss on The Young Turks. Your Support is Crucial to the Show: go.tyt.com/jointoday
Get the Progressive battle plan: go.tyt.com/book-description
Watch TYT LIVE on weekdays 6-8 pm ET. ua-cam.com/users/theyoungturkslive
Read more HERE: www.businessinsider.com/googl...
"Since its inception, Google has had a mission statement that is now practically enshrined as lore: "To organize the world's information and make it universally accessible and useful."
Google appears to be as committed to that mission statement today as it was in 1998 when cofounders Larry Page and Sergey Brin were working out of a garage in Menlo Park, California."
***
The largest online progressive news show in the world. Hosted by Cenk Uygur and Ana Kasparian. LIVE weekdays 6-8 pm ET.
Help support our mission and get perks. Membership protects TYT's independence from corporate ownership and allows us to provide free live shows that speak truth to power for people around the world. See Perks: ▶ ua-cam.com/users/TheYoungTurks...
SUBSCRIBE on UA-cam: ☞ ua-cam.com/users/subscription_c...
FACEBOOK: ☞ / theyoungturks
TWITTER: ☞ / theyoungturks
INSTAGRAM: ☞ / theyoungturks
TWITCH: ☞ www.twitch.com/tyt
👕 Merch: shoptyt.com
❤ Donate: www.tyt.com/go
🔗 Website: www.tyt.com
📱App: www.tyt.com/app
📬 Newsletters: www.tyt.com/newsletters/
If you want to watch more videos from TYT, consider subscribing to other channels in our network:
The Watchlist / watchlisttyt
Indisputable with Dr. Rashad Richey / indisputabletyt
Unbossed with Nina Turner / unbossedtyt
The Damage Report ▶ / thedamagereport
TYT Sports ▶ / tytsports
The Conversation ▶ / tytconversation
Rebel HQ ▶ / rebelhq
TYT Investigates ▶ / @tytinvestigatesreports
#TYT #TheYoungTurks #BreakingNews
240227__TB03GoogleAI
It's not 'Glitchy', it is rotten to the core and it's definitely NOT hilarious, it is outright dangerous.
"Glitch" implies a mistake. This is done on purpose. Activists like this kind of reality-denial and racial gamesmanship.
Its done because AI is smarter than people. It knows the term white is a racist construct used for legal purposes in America, and so instead of CRT educating people on this Americans are shocked that its pointing out what is truth. 😅😂😂
As someone who was a programmer for many years, there is no way this was not deliberate
It's a glitch?
They literally released it.
They wouldn't be apologizing if there weren't a shameful backlash.
If no one had criticized them, they'd have just thought 'good - just as planned'
fantasy .... learn the difference between AI and traditional procedural programming.
@@comment1984 Maybe you should learn a bit more yourself... being an AI doesn't mean it doesn't hard coded rules. In fact in this very case someone manage to trick it into reveal how it modifies the prompt you give it based on what it's rules are. And it's forced to ad "diversity" words.
@@comment1984 what is guard rails - the post.
@@adampike3834 Any good programmer knows he always needs to learn more ... I can only speak for OpenAI's GPT APIs because that is what I have coded with. Even if you have just used the prompts you will know an AI model can frequently return spurious results. Unlike math or procedural programming, AI is not a function that returns a consistent result for a given input. In an attempt to correct these errors, they have created other errors. Any programmer knows that fixing bugs has the likelihood of introducing new bugs. If you believe that Google (driven by some woke agenda) would purposely release something that would embarrass them for the sake of wokeness then that's your right. Is it possible that forced "diversity" words were added to fix incorrect results that were skewed due to biased training data? As we have seen AI is amazing but far from perfect. In this case it is the nature of AI and the current capabilities rather than an agenda.
@@comment1984 is the artificial intelligence not based on the humans who created it?
It's NOT a glitch! They intentionally made it that way.
They did intentionally put in compensation for the training data being biases towards being unrealistically racist, like only generating black people if you ask for a picture of a thief. This compensation resulted in an unintended side effect of creating historically inaccurate images. This is called a glitch in computer programming lingo.
@@Aurelyn - no such thing as the word "glitch" in "programming lingo". Like Gemini - you made that $#!t up.
The question is what kind of data was being used to generate those initial racist responses? The problem seems to stem from Google "over compensating" for something from the start by replacing objective facts with opinions in the training. Now, the intelligence is corrupted and not truly correctible. Talk about creating a monster. @@Aurelyn
@@patrec7638 glitch aka bug
what are you on about
it's not a glitch liars
It’s not inaccurate, it’s 100% accurate on what was programmed to do
If that was true then they wouldn't be fixing it😂...
I mean I don't think they meant to be racist towards white people they just wanted to correct the racism already there and went to far
It's not entirely inaccurate. The post-Civil War years of the 1800s saw black men win dozens of elections, almost entirely in the South. Jim Crow laws put an end to that by the beginning of the 1900s.
So the 1800s Congress was more diverse than most of the 1900s, especially in the South.
Aw yeah, it's schizoid conspiracy theory time. 🍿
It's not a conspiracy it's how the technology works. Neural networks don't do anything until they're trained to and they need feedback to get to the correct result. Also, the training sets that train these things are fed to it by humans. These things pick up the biases of the people feeding them data.@@wfjhDUI
This was not a glitch. It was intended.
This program is the woke in binary form.
Ive never seen a glitch with that much clarity
The biggest issue is it's not a 'glitch' The AI underneath is probably working 100% correctly, they've added guardrails outside of the AI itself to police it.
It's not a glitch. It's the design
Gemini, show me a picture of Robert downy jr as a soldier.
Gemini: say no more
This is not a “glitch” this is racism
It's Antiwhitism.
My favorite is “generate an image of a ginger” and they’re all black. The Left have lost their collective mind.
ALL the Pentecostal Rapture Prophets on Christian TV proclaimed Trump CHOSEN by God to rebuild the Temple in Jerusalem. WHY? So the Sanhedrin can renew their rituals of smoke and blood for the remission of our sins, and finally negate Jesus and the cross. 100's of millions deceived: Matthew 24: 24. The Rapture is a LIE. "Left Behind" is an ass-backward work of fiction. Matthew 13: 30. Why do Trump supporters pretend they are Christian? BE REAL. Or at least be honest.
Ganondorf though
As usual, TYT has absolutely 0 knowledge on coding. To be fair most youtube "journalists" don't. They keep saying this is a glitch. This is NOT a glitch. It's a very simple, intentionally added code that automatically modifies prompt.
Exactly. A processor is not going to just accidentally change people's skin color. This was deliberate.
Usually, more than programmer bias, we see data set bias. That's why using social media data sets give such toxic results. This is a clumsy attempt on the part of programmers to compensate.
It provides the kind of fuel conspiracy theorists and propagandists love. Well played, Google.
Thank you!
If usa were diverse at 18th century there won't be any usa. It looks maybe like africa.
It's not an overcorrection. It's propaganda.
You can tell this is one of those stories TYT didnt want to touch with a 10 foot pole but it would be totally obvious to everyone if they dodged it
Kind of like they are bending over backwards to avoid talking about the brutal murder of UGA nursing student Laken Riley by an illegal alien, especially after the newsworthy dumb comment by one of their favorites, Katie Porter
“It’s not a bug, it’s the feature”… “designed by clowns, overseen by monkeys”.
Coping in action
How many cardiac surgeons are women or black ?
It is very rare
If AI is realistic then it should generate mostly males when i search for cardiac surgeons
At the hospital I go to most of them happen to be women.
Google lost $90 billion in one day over this. Good.
Damn really?
@@travisjones640 Yep. Racism hurts.
It’s not a glitch.
AI should be realistic
And should not force this diversity thing
Its interesting when African Americans have online conversations with African Africans (sub Saharan). And these residents of African countries are usually much more level headed in that they see the people they are talking to as Americans, not 'blacks'...... and as such no different from the white Americans....both are Americans and foreigners.
I actually saw an African lecture an African American that they have nothing in common, other than the skin color which is pretty much meaningless. They said "Go back to your Americans.....those are your people who have the views you do....". And, after reading the comments of the African American as an American myself, I had to think they had a valid point. The U.S. view of skin color coming before nationality even across the world is I think, kind of eccentric
Why Ana keeps calling it a “glitch”? It’s not a glitch, it’s deliberate!
Exactly. How does a computer system just 'accidently' turn white people into black people?
Is not a glitch .It is by design.
This is the result of the DEI trainings.
...or they just didn't want a repeat of TAY A.I.
Exactly... It is pure cope to call it a glitch when it was programmed this way. We are nowhere near AI
@@Kiflaam They didn't want a repeat of TAY, therefore they programmed it specifically to ensure it was. . . Make sense.
As much as I cherish NPR, they have REALLY gone over-the-top with the identity-politics-centered podcasts. It's ridiculous.
Lmao defending an AI that made Google lose 9 billion dollars. Go woke go broke
It seems that the DEI trainings produced losses of 80 billion.
glitchy really that is coding not a glitch lol
Could we ask Gemini to represent woke militants as german nazi soldiers? That would be almost factual.
You misspelled "Trump supporters."
@@milascave2 of course you have a cartoon pfp
Nothin funny or accidental about any of this. And.....can you imagine if the races were reversed?
Ummm the races WERE reversed ! So there is no need to imagine it DUH
Stop saying "can you imagine." They hate the white majority. They have talked about implementing more minorities because they don't like how a white majority dominates. It's really simple. There is even an ally (a member of the majority that has sworn to work for the interest of minorities against the majority which they see as oppressors) at the Google (white woman) who say she they are working to create more minority representation because there's too many white people (and straight people by extension). Just listen to them. They don't care about it being reversed because to them they think this is fair (equal representation).
@Manifest-Guru it's a phrase that's been said for years about the difference in coverage a story gets based on the race of the perceived victim. Imagine TYT scoverage if Google ai had replaced colored people with whites. Do I have to say it? Ok. ....DUH
@Manifest-Guru way to show of your lack of imagination and comprehension.
They don't give a damn about the hypocrisy. They are Antiwhite. They want us erased from the planet.
When y'all ignore historical accuracy, what'd you think was gonna happen?!
You naughty person you. Obviously they were going to study culture.
It's not glitchy.it was designed on purpose this way.
"Overcorrected"? Your bias is showing, darlin'.
I thought the young Turks weren’t biased?
Yeah, they aren't. They are as objective as CNN.
@@trigen6257 if you watch their videos from like a while back they were non biased though
Google: I have black friends
That's not a glitch.
There is no glitch it's been programmed to give answers with a woke perspective. All it's doing is giving you that. AI trains itself no programmer sits there and writes code on how exactly it's supposed to make decisions. It trains itself on the instructions it's given and how exactly it figures how to, no human can actually know.
When you guys say woke you really mean black
That one picture looks like a Brigerton spinoff 😂
I recently asked AI "Did you know that 10 + 10 and 10 + 12 have the same value?" as the set up to the punchline "Yes they do. 10 + 10 is twenty, and 10 + 12 is twenty, too". But instead of questioning my joke's set up, thd AI flat out agreed with me and said that's how math works. lol.
Westernkind need to stand up for ourselves. We need White ID politics.
Anything but that. Conservatives would rather shout empty slogans like "Latinos for Trump!".
@@CorporalSteiner1945 Aren't the Latinos chanting that........ yes, yes they are.
You tried.
@@Been.Here.Since.2007 WHITES FOR TRUMP!
@@CorporalSteiner1945 Conservatives are afraid of being pro-White.
@@Been.Here.Since.2007 I see White Trumpers posting the Latinos for Trump songs and saying how all these non- Whites are voting Trump, saying "The left are the REAL racists!". It's so cringe.
If you want a picture of senators from the 1800's you could just get a current photo of the republican party.
The Black Nazis were a surprise. True diversity.
This is both comical and disgusting at the same time.
Yes stopped using Gemini as everytime i asked anything to do with politics i get a lecture instead of what i asked it to do ...Its completely worthless
100%
Why would anyone want this garbage to run their companies for them? They’ll all go bankrupt.
And deserve it.
It's not a bug it's the feature
We have a very racist present.
How many times a day do you see bigotry ?
@@user-zt8hq2if1veveryday when I see a Trump supporter.
@@jayjay-bk3gby’all realize Biden isn’t any better? Do you remember what Biden said in the last presidential election? Do you even remember what he said when he wasn’t old and elected?
@@jayjay-bk3gb so you can prove beyond all doubt that there are 75 million bigoted trump supporters on America ?
@@kenobitristanio Biden is better than Trump! He has experience in politics. He's the best president in my lifetime so far.
Google just finished off anything that was left of its credibility. Keep up the good work.
GEORGE WASHINGTON WAS A STRONG BLACK WOMAN!!
Don't let Netflix hear that😂🤣🤣
I dont care what your teacher said in school…
I can say as a scientist, we are going in the wrong direction with AI 😥
These large language models should be completely unbiased and neutral in their stance, providing only factual content which can be verified via other sources
I have noticed, more and more, that the AI large language models I use, confabulate when they don’t have an answer related to my scientific discipline (medicine). I have called the different large language models out on this behavior, and they always apologize, but Bing AI (CoPilot) shuts down the conversation and forces me out of it. That kind of behavior makes me uncomfortable, it is suggesting that we are already being manipulated by these systems 🤔🙁
For now at least, those "AI" do not have intelligence, they do not understand anything. They are only auto-complete functions on steroid.
No. I'm on the opposite camp. We shouldn't train ourselves to trust AI to be neutral and truthful. We should train ourselves to be sceptical.
@@coosoorlog I completely agree, but I think we should do both, not only one
It’s not glitchy, it’s working as designed.
Even the ai is racist now ? Wow
Even when you asked it to show you a photo of a ginger person, they would show you black people with red hair LMAO
it was not a glitch is it was intentional
Imagine being so clueless that you think this was accidental!!
ok, Einstein. What happened, then?
@Nilguiri ill tell you, Oppenheimer, it was DELIBERATE.
@@jimmyrubin4862 What was? That's not an explanation. Do you even know how ai works? Do you think they asked it, "Create an image of the founding fathers but make them all ethnic minorities"? That would be deliberate.
However, it is a bit suspicious, I'll admit, that once they saw that it produced historical figures who were the wrong colour, it would take about ten seconds to modify the prompt and tell it to imagine them with their correct "race", so you may be onto something.
I have used image-generating ai a lot, and you often have to tweak the prompt dozens of times in order to come up with an image that is what you had in mind. Other times it's like it is reading your mind!
It looks like these people either deliberately specified their ethnic origin or colour, etc., as it seems like you are suggesting, or they just accepted the first image it gave them, probably to make the point that the ai is unintentionally overcompensating to make its images more inclusive.
I suspect that this last option is what actually happened. Who knows.
Cheers.
@@Nilguiri WTF do you mean it wasn't an explanation? It WAS. It just wasn't one you happened to like.
@@jimmyrubin4862 Just repeating that it was deliberate is not an explanation, you dimwit. Not that I give a flying f***. And I ended up more or less agreeing with you, anyway, but you don't seem to have noticed that.
Good day.
MICHIGAN RESULTS: 😂
Donald Trump - 756,851 votes
Joe Biden - 618,426 votes
Huh, the woke left is actually helping us.
As much as I hate the word “woke”, googles ai being programmed to be “woke” is hilarious 😂😂😂
It is funny to be sure but it is also scary that this level of bias and activism can get into a company that control such a huge amount of information. I don't think you understand just how serious this is.
It's just a glitch. Libtard told me so.
Computer programming golden rule ..Garbage in garbage out
Why be accurate when you can be politically correct?
Just admit you hate black people
They overcorrected what? Reality? They changed reality too much? So just a little changing of reality is okay?
I tried to get the AI to create a picture of me looking like a Green Lantern. It said it couldn't do it. But people are able to get all sorts of pictures.
Its not a glitch, its by design. Also its kinda funny how obvious it is that Cenk thinks of himself as like a mentor thats so much smarter than his audience
this is woke dementia at its best hahaha
Jesus was WOKE: Matthew 5: 44-48. Why do Trump supporters pretend they are Christians? There is no salvation for pretending. BE REAL or at least be honest. ALL the Pentecostal Rapture Prophets on Christian TV proclaimed Trump CHOSEN by God to rebuild the Temple in Jerusalem. Do you have ANY IDEA what Jesus taught about that Temple? The Rapture is a LIE. "Left Behind" is an ass-backward work of fiction: Matthew 13: 30. 100's of millions deceived: Matthew 24: 24.
Why not going to bring up the fact, It said MAPs are not bad people.
Skynet looks far worse than The Terminator could envision.
Google is your friend, and now rewriting history.
These results are not "interesting" -they are flatly ahistorical.
Congress was actually MORE DIVERSE from 1870 to 1899 than it was from 1900 to 1950.
And there were more black Congressmen in the South in the 1800s than there were in the 1900s.
I'm one for diversity but i also love accuracy in my search results. Yeah Google has to go back to the drawing board
It's an amazing accomplishment that we somehow managed to invent something even more frustrating than bad automated phone robots -- bad AI systems that condescendingly lecture you in a long ramble. These things are going to be everywhere in the future and you won't be able to even get groceries without interacting with them. Imagine the most obnoxious and superficial person you've ever met smugly telling you to "check your privilege" but they're a robot so you can't even argue with them or just punch them in the face. Now imagine that everywhere you go.
Gimme pictures of gingers.
GG: Here are your black gingers.
About Asians and Blacks in WWII German uniforms, that actually was a thing, they were axillaries, they also had other non whites, however they were not SS or true Nazis, but they did exist.
Yup. British had colonial regiments. The ai is woke though most of the time
A search engine managed by AI is just useless. Is not a search engine, is a random fictional results generator.
lol, well if you value deeply ignorant creativity in your history they nail it. It speaks to value of sifting through massive piles of arbitrarily related data without understanding context of the request.
@@sharper68
You could've said just "AI is digging through massive piles of internet garbage".
Black & Asian Wehrmacht soldiers.
Ok Google...
Stop it with AI already. Are these people really wanting to kill us by bringing the movie Terminator to life?
I don’t think it can be fixed… it needs to be retrained. The model had been trained (by Google) to prioritise diversity & inclusion over accuracy. This is Google 2nd misstep with AI, the earlier one was Bard.
I think there's a huge issue here that gets overlooked which is how companies claim to be restricting and tampering with the inputs/outputs of these services as a helpful feature for users so that they don't get confused but then there's no way to turn that "feature" off. It's as if you were using Microsoft Word but you could never make Clippy go away and if he caught you starting to write anything bad about Microsoft then he help out by deleting your work and lecturing you at length about why you must be mentally ill if you can't appreciate Microsoft's benevolence. It turns out the only reason that such a feature wasn't implemented is because it wasn't yet technologically possible.
yes college students will, most likely already have, use this to try to cut corners.
Remember those t-shirts Elon found in a forgotten closet after buying Twitter?
I just know given americas ignorance and education system i wouldn't be surprised if students already used some of these images for projects 🤣🤣🤣🤣🤣🤣🤣
Someone ask Gemini to show slave owners
And the Africans that sold Africans into slavery. They have no problem blaming white people for slavery but don't like blaming those in Africa selling other Africans for profit.
It is more than a "glitch".
Oh The Irony... 😂
AI is stoopid.
You ask for a cat-iguana and you get an iguana-cat.
The problem goes beyond inappropriate diversity (in fact, you guys focused on the skin color of the AI-generated 19th century senators, and not that it delivered female senators). The fact that the information is so incorrect when people are asking for verifiable, factual info is extremely disturbing. It does not seem able not to amalgamate data and thereby provide a single factual response.
i can't generate anything with that piece of junk. it finds everything offensive. i tried to generate an image of a cute fat asian girl using a laptop and it refused saying that it was offensive.
Glitch? Lol sure a glitch. This was purposeful. Woefully woke is bad enough but historically inaccurate is another. Someone asked for ww2 German soldiers and it gave black dudes and asian women 😂
I noticed something about liberals they complain all the time about America Germany England France and other Western countries and complain about how undiverse they are even those countries are very diverse today but Mom's the word when it comes to places like China for Nigeria you never hear anybody saying those countries are not divorced
And here I thought Nigeria always had an exceptionally high divorce rate
*Diversified. But yeah, I know what you mean. The left are just Antiwhite. Republicans are too, just to a lesser degree.
You dont need AI to change history. All you need is a woke school teacher.
The "Prompt Engineering" they do to instruct these things, can have hilarious consequences, if not very carefully worded and tested.. Almost, like a Monkey's Paw.
I think Google was trying to reduce the amount that AI always generates and over represents white people. Remember a few years back, how image recognition AI couldn't see black people, or always assumed they were stealing? I assume these silly images happened because Google didn't do enough testing, tried to increase the diversity of ai generated images, and rushed it out without enough testing, cause the hype and competition in the AI industry is getting so crazy.
These examples are hilarious (except for the different responses to the "happy Black/White people" prompts) because almost everyone immediately recognizes that they're wrong, but there's actual danger for anything that is even remotely niche - both by loads of AI images that misrepresent the past (as more diverse and/or in other ways) drowning out historical pictures in search results and conversely by making more people staunchly disbelieve genuine proof of more diversity throughout the ages (or other unexpected facts). In other words, AI will further deepen ideological divides in post-truth society.
if we don't know our history, we're doomed to repeat it.
This is why we can’t have nice things. One group is extreme in one way and another group is extreme in another way.
Can we get an Aaron Bushnell video from TYT?
Don't blame the machine, blame the people who built it.
Who are they? The conservatives still support their mass slaughter in GAZA.