Google's A.I. Has Gone Insane
Вставка
- Опубліковано 27 лют 2024
- 🔴 Website - www.vaush.gg/
💵 Patreon - / vaush
👕 MERCH - merch.whitefore.st/
😎 JOIN PROGRESSIVE VICTORY: progress.win/vaush
🌟 Other Socials 🌟
⭐️ 2nd Channel - / thevaushpit
👀 3rd Channel? - / @vaushvidya
🐦 Twitter - / vaushv
👾 Twitch - / vaushvidya
📸 Instagram - / vaushv
🔵 Facebook - / vaushvidya
🎵 TikTok - / vaushvidya
🎙Podcast - anchor.fm/vaush
Intro animation credit goes to / itismekyleg & / honeybunnbadger for the visuals, and / sound_sierra for the audio! Thank you!
#ai #artificialintelligence #aiart
The sooner Big Tech investors lose interest in generative AI, the better.
The funny thing, it's burning money at an insane rate. It's not even profitable. The obvious end game is to sell it to businesses to replace workers, because it's too expensive for average users to pay the real cost.
Ai is the future
The question is "Will it hold up to its expectations or will it be as much of a bummer as crypto"
@@saliferousstudios And for what exactly? I mean if you are talking about AI ads, then this low ass quality won't fly, the only ads I pay attention too are the ones that are funny and the ones that tell a story (both things AI is not very good for). It could be useful for posters and similar types of visual media, but you are still gonna need someone to write the prompt to get the right result (even if that person is gonna get paid less money) and by that point is potentially wasting hundreds of hours trying to get the desired result really better than just paying a real artist.
@@saliferousstudios no investments are immediately profitable, that's why they're called investments, it's long term gain over short term losses.
We're not even 18 months into the age of generative AI.
why would they? it's a great productivity booster.
The funny thing about Gemini's filter is that several users have managed to turn it against itself by asking it to generate "chained scholars/Greek philosophers eating watermelon", which would naturally cause it to plug in its diversity prompt and generate some pretty racist caricatures.
put this to bass-boosted cyberpunk music lmao
I think the only setting where you couldn't conveniently get racist outputs would be dealing with multiple people in one image.
Lmao love it
Get the woke AI to turn on itself
This is the bed ppl like voosh have made, especially whining about "r@cist" AI whenever there is no equality of outcome.
Now tech companies are overcorrecting, and you get to lie in your bed that you made.
It sounds like Google is trying to make a digital paintbrush that can't paint racist things. But you can't teach a paintbrush what racism is, nor can you teach it context, so the project is inherently foolish. I'm surprised they don't just ditch the idea.
Or maybe they could keep it, but only allow people to use it by signing a legal waiver, akin to buying an M-rated video game.
@@icarus313the thing is that if there's any sort of paintbrush which *could* be taught to not be racist it would be an AI one. I could easily believe the tools could be smart enough to go "this person is trying to generate an individual with dark skin eating a green fruit with red flesh? Oh, racism" and shut it down. In like, a decade.
But instead tech companies did what they always do and, irregardless of the stolen art thing, they STILL fucked up and took a barely functional piece of technology and tried to make it the future, not thinking of the moral implications of doing so.
They polished a turd and now there's shit everywhere.
Thank you google for giving my dad something to talk about for the next 6 months
AI like a year ago: cant generate black ppl
AI now: cant generate white ppl
As an artist, I'm fuckin cackling
The first isn't really AIs faults. It's because online art is *heavily* biased towards white people. And if you are not careful during with your training data that bias will carry over to AI. It's not like an image generation AI has inherent knowledge on racism.
This now is just an overreation while trying to correct.
Yeah, artists collectively have all been afraid that AI would take our jobs. But when you see the challenges people have in trying to get useful results, it does provide a tiny sense of relief. 😅
And a few years from now, all these funny moments will be barely remembered as brief drops in the bucket, in face of the ocean of AI that's coming our way..
The more bot generated content there is, the more often this will happen. The self-cannibalism has already begun, so buckle up.
@@ArDeeMee I can't wait to see it. The sheer amount of human generated data on the internet made AI what it is today, but as the Internet becomes infested with AI generated content today's methods will become ineffective.
Can't spell Nazi without AI. **taps temple**
Good point
This is why I hate New Zealand
Disco elysium dialogue lmao
I mean, their "intelligence" WAS artificial. 🤪🤷♂️
Big uh-oh for AIr New Zealand
It amazes me how ai keep proving over and over it's unreliable black box of hallucinating shizo algorithms yet everyone keeps putting it everywhere.
Chat gpt suffered a mental brakedown recently and all chatbots relying on it went to shit for a day. And nobody even reflected it might not be a good idea to 100% relay on it.
At least we seem to have moved past trying to put everything on the block chain. It too had fundamental problems that made it impractical to use in basically any setting.
in this case google literally just had a piece of code to slap "black" "female" "diverse variety" etc in front of the prompts, which, is the wrong way to do diversity lol
@@kalex22029 It looks like google were being a bit more subtle than that - they were probably doing some fancy math stuff after the prompts, involving manipulating embeddings. Same outcome, but with less in the way of side-effects.
Recently Patrick Boyle made a video on AI. He’s a finance guy and reports on actual studies and university research. Very interesting insight on the pros, cons and how companies are approaching the use of AI.
"Quick, this guy is having a stroke, use the ER AI bot to check him out and have it deliver the diagnosis, stat!"
"The black diversity of the nazi regime indicates the hyperoboloid coefficient at the heart of Arkansas is grappling rapidly with the bluian Faust whomst I once fondly diabetes"
We need more representation in the ranks of the third reich😂
Very Star Wars Empire.
We already got enough Arabs in the Wehrmacht’s ranks for that BlackRock esg score
Honestly there were people with Jewish roots in Wehrmacht because (surprise, surprise) Germans mixed with Jewish people over the centuries. Before scientific racism became a thing Jewish people could get accepted by Christians by simply converting to Christianity. Quite a lot of them get assimilated this way. From what I remember you could have one Jewish grandparent and still be accepted into the army. So far I only saw black and asian representation but given the historic reality we should strive for more diverse representation by adding a Jewish soldier
@@angelikaskoroszyn8495 hmmm im Bosnian :D live in Austria and we learned in Scool that you could have Jewish Ancestors "no Problem" but you would not get Adis "mein Kampf" with his Signature if you married someone with Jewish Ancestors, you could not be considered "Arisch" enoth.
@@arcturionblade1077To be fair, the Empire was hyper racist against aliens instead of different skinned humans. Think the Impirium but much more subtle.
Here’s an AI generated image of strong Roman white men being gay.
Fascists: Mmm gay... no, WAIT
what?
@@Cobbido In fact, Roman empire last a little more than 1500 years. There were period where being gay were fine and other that it was forbidden.
Modern Fascist: ah, the return of the why boner.
@@pierregravel-primeau702there is no evidence it was never not okay
Adolf Hitler: Google AI is slandering the German army.
Himmler: Well...
I dont get it
Me neither
Possibly a reference to the fact that there were Arab Muslim divisions of the SS so brown people in Nazi uniforms are historical fact.
@@zephyr8072They were called "Legion Freies Arabien".
People always have been and always will be dumb beyond all reason it seems
@@zephyr8072Instantly jumping to "muslim = brown" is kinda sus ngl weird way to describe that division
Honestly I embrace the silliness. Throw these diverse AI generated slop back at conservative websites.
anime pfp
@@sleeve9097 thanks
thanks on google for acknowledging Blitler, Aisler, and womanler
This is only a PR problem because Google provides Software-as-a-Service instead of products you can own. Nobody takes a pencil manufacturer to task if their pencils are used for offensive drawings, but the moment these drawings are created Gemini still exists on Google's servers so the "art" is made by them in a very direct sense. As a SaaS hater, I really enjoy when companies suffer in unfixable situations caused by their own shitty business model...
There’s a VERY old computing concept which covers this situation, and Tay, and many others:
Garbage In, Garbage Out
What do you mean?
@@vipcypr8368If the input is biased the output is as well. Of course this is slightly more complex as this is actually Google overcorrecting for the inherent bias of art towards white people.
The way that these chatbots are usually made to produce results like this is that the prompt is automatically modified before being given to the AI. So if you put in "WW2 german soldier" the actual prompt given to the AI will be "WW2 german soldier with diverse ethnicity".
Ironically when I asked it to draw ornithologists banding birds a few weeks ago (because that requires them to draw both hands and birds in hands, both of which they mangled) they were all white unless otherwise specified.
how is that ironic?
@@B-fq7ff
It got in trouble for drawing diverse Nazis (something that didn't exist) but it could only think to make white ornithologists, even though those fields being inaccessible to minorities WOULD make more drawings of black ornithologists good.
Did they not realize the Singularity was a melting pot?
but there were black and "olive-skined" kings in medival era... just not in England and France
Yes. But what medieval people may have referred to as black might just be a slightly more tan person than the average nobility at the time. Skin color in history sources has to be taken in the context of the time and region it was written.
"Medieval" is a part of European history, not just a range of dates. It's like how it doesn't make sense to refer to modern America as Reiwa-era America because the Emperor of Japan doesn't affect us
@@TheRenegade... While that is true, it does however still include the Ottomans who weren't exactly white europeans.
@@TheRenegade...no the term is also used in India
The dates are a lil different though
@@TheRenegade... I wanted to be smart as about olive-skined people like Italians, Spanish, Croatians etc.
I'm sorry but I couldn't generate a response to this prompt, please contact support if this issue persists.
I think it illustrates just how important race is as a concept in this society. There is nothing correct about those "medieval kings" images. Nothing. One of those things is race/ethnicity, but it is the only thing that people get upset about
I mean if anything, a middleastern/near east depiction is relatively reasonable given the ottomans were around during the middle ages with territory well within what is considered as Europe. If anything the bigger issue is that the clothing is often much more anachronistic, with full plate armour only really existing late into the medieval period, and not looking the way it does in the image.
Because the AI is programmed to change what it does based on race. Thats not the case for other features. They come out wrong "naturally" Also yes, race is a pretty big deal. A humans whole skin being depicted wrong is much bigger of a mistake compared to some detail on their clothes. And again, its made wrong intentionally.
Vaush's reaction in the Thumbnail pretty much sums up how I feel about this too. 🤦🏿♂️
Well this stems from liberal ideology
Imagine publicly admitting you were using an AI image generator to produce "1943 German soldier," like it's a perfectly normal thing someone would do.
Well it's better than asking it to generate pictures of "Totenkopf SS" members.
I mean yes? After it started making everyone black, people were pushing it limits
You might want them for a meme
Character in a visual novel
Character in a game
for a historical YT video....ect ect
@@fortheloveofnoise9298you fucking suck at history if youre using ai images, any "history" channel who uses them should not be taken serious.
-sincerely, someone who wrote a paper about a local historic event and had to spend hours in an archive to find even 1 picture
@@fortheloveofnoise9298i second this. I like to make political memes sometimes to share amongst my friends, and I could see myself generating and photoshopping a maga supporter onto a Nazi soldier or something. I don’t have enough artistic talent to draw political cartoons lol
Bing ads on UA-cam for me are advertising their AI image generation with "anime girl in sweater" pictures lmao
Targeted advertising at its finest
Was she cute?
@@scottfromspace I skip at the 5 second mark, I only have time to hear an AI voice say "anime girl in sweater" and in the screen it's like the prompt, never watched the full ad. I downloaded it though, the Bing chatbot seems interesting, it's better than chatpt 3.5 and it's free.
Sucks people try to use it for racism and pornography but all AI is way over correcting. Midjourny won't let me do anything. If any women is in the picture it accuses me of breaking community guidelines.
When you're so feminist you think women existing is pornographic.
Fr AI is so fucking stupid. All the AI boom online has been the product of marketing guys trying to sell the public an innovation that didn't exist yet. These people need to stop having so much control over science and research.
Maybe “no women” is in the community guidelines
@@rockmycd1319 Considering the types of people who run tech companies that sounds about right.
Have you considered learning to draw? You can draw any woman you want, midjourney won't stop you.
@@Vaishino Yes. I enjoy all art mediums. I decided to mess around with ai for a minute and found it bizarre how overly locked down it was. Nothing sexual in any prompts and still get warnings all the time. Re-rolling an image the AI made by itself gets another warning. Inability to make what you want because it forces out of place diversity doesn't help anyone. I don't think we should dive into unrestricted AI too fast but it's stupidity locked right now. I just wanted a cool 4k cyberpunk style wallpaper for my PC and almost got banned immediately with no refund.
This made it a game, who can make AI break its own rules? Even if you're not into hate, who doesn't want to break the AI?
I don't know how to fix this, however, I think it will fix itself by stopping one day.
The greatest evidence that Virtual Intelligence isn't Intelligent is that *it can't do context.*
It can't do context.
Like corporations. :v
It's funny because without this, this kind of AI is a stereotype machine. Obviously it's designed to replicate the most likely human response, so it falls to a stereotype most of the time.
Not a lot of people know this, but L. Ron Hubbard was a black man!
Even though the only real photos of him are grayscale, I can still see that he was white.
5:47 "It's like Battlefield 5" 😭
Would be funnier if it returned every human being request as the mascot from community
It's like tge Battlefield V multi-player where you could fully customize your character with no limitations so you could play as a black nazi if you wanted to
It's called an echo chamber, you can brainwash a human just as easy as exploitjng a AI with garbage input, so it's still the fault if Google either way.
Is it just me, or is this awfully close to when ChatGPT also started shitting the bed?
I don't think this is actually "woke" except in the far right sense that minorities existing is woke. More like colorblind because it acts as of race was just a meaningless variable that you can always roll randomly regardless of context. Sure, the roll is weighted, but that only means it's also corpo tokenism too.
Google is just doing what they do with their regular search, they add terms or replace your own.
what benefit does a corpo gain by tokenizing nazis as black and asian ?
@@Senumunu
What do you gain by bitching about a imagine of a black and Asian nazi?
Who cares
@@LoopyLemon775
The existence of this video should clue you in.
It was all fine and dandy until ~diversity~ went too far.
"Does preventing Nazis from posting really racist shit fix anything, since all it seems to do is give them something to cry about." - Vaush
Honestly, based on your training of logic, since Nazis are the only people who really care about AI art, the notion that you would take away their ability to use it the way they want is pretty fucking funny. Also, Nazis cry about everything, so there doesn't really seem to be a difference either way.
Its actually hilarious bc its so silly.
I imagine some racist guy desperatly trying to generate some white medieval king and only getting some indian looking guy with a turban or native americans with warbonnet as results. Lol
we should just let the nazis use it exclusively. so that generative ai art will be known as a nazi thing. like gold standard and removal of age of consent laws.
Yeah see... people complain that if AI just reflects back the world as it is, that this is a problem as it contains the biases and prejudices that exist in the world, but if you try to correct for those elements you just get this kind of absurdity. But really... this whole predicament is just an extension of society in general, people complain that theres insufficient representation/access/whatever in various areas, what do you get... various attempts at "affirmative action"-esque corrections, and people making the little mermaid black, which then get opposed or mocked.
These "failings" of genAI, are really just our own failings but produced at the massive scales and speeds that computers can attain and we cant.
Because no one wants to address that it is society that is broken
it gets harder to "correct" for some of these elements when humans with opposing beliefs and interests can't agree on what the "truth" is too.
@@Kwatches1190 Ofc they do, but all too often this is how they go about trying to address it with some kind of artificial solution to a set of problems that were not artificially produced.
Ironically enough, by making certain whites colored, you also erase their oppression they were in by their fellow whites... like the British vs Irish.
Yea it's more of an issue with affirmative action than AI
Google AI, generate a well endowed stallion.
critical support for this because its incredibly funny and equally funny to watch nazis out themselves as immediately going to see if a generator meets their white person quota whenever a new ai thing comes out which makes them look weird
This community fetish to call everything outside thei bubble "nazi"/"facist" is so funny.
The majority criticism for this chatbot was that he was still racist but towards white people and it forced wokeness on people whitch is amazingly unconvienient if you seriously consider using it.
Are you seriously so brainwashed that you think people being reasonable are racist just because small group amongs them is retarded? You are doing the same generalization that compares whole left to tankies.
Not as wierd an entire race getting a whole ass month celebrating them.
Not as weird as queers getting a whole ass month celebrating them...
AI is the death of everything
Please show us on this doll where the AI hurt you.
@@zephyr8072They can't, your doll was ai generated and thus lacks human anatomy or anything close to it, cause your ai's a piece of shit ;)
Welcome to the Desert of the Real (thunder crack)
me when i have a boring take with nothing to add to it
Ngl it's very funny that the Google AI thought Hamilton was a documentary
The amount of young people in his chat who seem to have no interest in, or care for, historical accuracy, is quite concerning.
It's part of how we handle text encodings, there is no "entirely white" or "entirely black." It happens with text to text too but it's more noticeable in images because you can see how its gaming it's performance measure by trying to find something that'll score points for getting both positive likelihoods (0.7, 0.3) rather than the intended (1, 0). More data structured to challenge it in this specific category should push those likelihoods closer to 1 & 0 respectively, but with how it's set up it can only get really really close like (.999, .001).
This is an over-simplification & the functions used in these are a lot wilder but trying to get at the concept rather than specifically how Google may handle their text encodings on Gemini (though it may be possible to find out with Gemma, their open source version)
Funny enough I've been talking to friends for a few months about an art project idea I was going to call "does background infer foreground?" based around this specific quirk highlighting a variety of outputs to the prompt "show me a solid white or black background"
looks like Google's image generation AI will replace that one gimmick account.
Great video
Chat, Are we reaching the Singularity?
I feel like people are going to look back on this period of time and laugh at us for our collective dumbass-ness about AI and crypto and NFT's and shit.
I played around with Bard/Gemini's pic generator before the retooling and it was indeed too "woke" and this is coming from someone who is "woke". It seemed to be forcing me to only ask for images of POC, LGBTQ, and people with disabilities. I saw something like this coming a mile away and I'm glad they're making attempts to fix it. You can't do anything about stopping racists misusing your platform, but at least give it some effort instead of overcorrecting from the start.
The editing to keep that roasting of the chatter in at the end Kkomrade
It doesn't need to be fixed... It needs to be purged and start from scratch...
*"Google Ai"* - brought to you by Netflix 😂
(Me:Google)Builds search engine as if some reliable source of documented authority...
(Me:Google)Builds AI inherently unreliable into the search engine not expecting everything to go haywire costing Billions in reputational self damage.
(Me:Google)Looks at investment profit from promoting AI, looks bigger than loses I deem this a success...
Rip everyone who got negatively affected as a result.
The "search for white family" on google is such a strawman. The results clearly have white backgrounds or white clothes. Lots of white keywords. Try caucasian family next time =)
Hopefully AI goes the way of NFTs in a hurry
I really hope so
It won't, because AI is actually useful.
The difference is in who is using a new tech. Random cryptobros and scammers or every single big corporation in existence.
@@boydvanoosterhout7966 so was Blockchain according to the NFT shills
@@silaskuemmerle2505 The difference is that the blockchain boys couldn't actually utter a single use case that wasn't an obvious pyramid scheme. AI on the other hand is already being used in plenty of situations. I'm obviously not defending all uses of AI, but there are plenty of situations where it has real world value.
A mouse of science
bro contradicted himself 😂😂
one day, a history ai will bookeep everything
Bro that medieval king is literally black Aragorn from wizards of the coast they got mad AF over that
I think there might be a fundamental flaw with the whole concept of AI, at least for some purposes.
In Science, often a useful aproach to a problem is to build a model of some physical process in such a way that we can know HOW and WHY the model gives a particular forecast or output. We also want the model's forecast or output to be reproducible, meaning that the same input gives you the same output. Not only does it seem impossible to know WHY AI comes up with a particular result, but the result can not be reproduced because every time you give it a particular quarry, it comes up with a different result. It is then impossible to know WHY the new result is different. Surely this must reduce the usefulness of AI for many applications.
When AI is used to make a decision that impacts peoples lives, there seems to be no way to question the decision and accountability is impossible. This seems uniquely bad for the functioning of democratic and civil society.
So we can't do science on human neural networks (brains)? We understand far less about them than we do these vastly smaller and simpler artificial neural networks. You can actually break open the program and read out the exact neurons, exactly what they're connected to, and what their weighting is, ect, with any of these AI. You can't do that with a small insect brain, let alone ours.
Wait.... So we can't do science on animals either? Literally no branch of biology is legitimate because we can't have reliably repeatable outcomes?
I think someone may have lied to you about how science works.
Edit: and if you think AI is uncountable and seems poorly functioning, you've clearly never interacted with any *humans* in the government. Try and hold a DMV worker accountable for blatantly lying to you about requirements, for example.
@@ASDeckardGood point about neural networks. I think my statement about "Science" was too all encompassing, and different types of models ie determanistic, statistical, and yes neural networks can be used for a given problem. However, I do suspect that too many people are using AI for problems when they are unaware of the innate drawbacks with that approach.
As for humans in any government, I would rather argue with them any day of the week than be told "sorry, that is what the algorithm says...end of story".
it is absolutely possible to figure out how AI reaches decisions. Just not trivial. It's an entire field of research. But I can guarantee you that is exactly what Google is doing rn to correct this overcorrection.
@cHeroxInteresting. Is it possible for them to do something more systematic then trial and error ? Why did they not know that trying to correct for racial bias etc would create such useless and nonsensical results ? They just seem to be trying to correct as they go along. The whole process seems to be stirring up more distrust and suspicion.....hardly what we need in 2024.
14:36 “Inept anti-racism being employed to provide a corporate cover.” Couldn’t have said it better myself.
The black SS guy's helmet looked like Dark Helmet from Spaceballs.
It was internal and unintentional. It was very difficult to even get one European character. And very funny. Even if you specifically asked it would refuse. The grifters insisted it was intentional, for it did play right into their hands.(dalle and midjourney used to do the exact same thing). And then they shut it down.
I´m cracking up :D
The Asian soldiers weren't so inaccurate
And medeivial era wasn't just in Europe
5:30 - Well no. The controversy of Battlefield 5 was people being mad that there were Black people and women (With a literal mod coming out that mad every man a woman and every black person white), which was historically accurate. It was just good ol fashioned racism and sexism disguised as "woke"
Jesus Christ I didn't realize this community had so many Luddites. Obviously the tech needs a lot of work, but it's to be expected considering it's still very much in the early stages (that being said, considering these are the first iterations I find the technology to be pretty impressive with what it can do already). As for the notion of artists being replaced, all I'll say is that these reactions are very similar to how many reacted during the industrial revolution or (perhaps more pertinent in regards to artists) the development of photography. Of course, with new technology things change, but ultimately we adapt.
Sounds like you don’t care about real art
@@Chiefqueef91It sounds like you're being ironic, but from what I've seen that's something a lot of people here would unironically say as well.
Here's a list of things that at one point in time (at their inception) were labeled as not being "real art:"
1) Music using electric instruments
2) Electronic Music
3) Photography
4) Film
5) Street Art
6) Most art movements, in fact
Honestly I can go on and on. To me, art is far more than just the techniques employed. In the case of ai art, I'm seeing many similar arguments that were used against the examples I provided: that it's soulless, that it takes no skill, that it's limiting; I say we let people cook and see what they can come up with. The one constant is that art purists have pretty much always been wrong lol.
Except we have seen company replace human workers with chat gpt already. Many have gone back to hiring humans now but it is only time before they try again.
@@JoelFeilaOh 100%, in fact what these companies will do is fairly predictable, workers don't matter to them and they're all about the bottom line. That being said, there's a distinction to be made between the technology and the companies employing it. Just because a lot of CEOs have this wet dream of replacing all personnel with robots doesn't mean we should all start smashing machines like the original Luddites.
But yea, we can't be afraid of moving forward with technology and science, I think we can all agree life has steadily improved over time thanks to our advancements in these areas. We may lose some jobs in the short term, which sucks for the people who spent time developing these specific affected skills (I also believe we can do a lot more for people whenever these situations come up, with retraining, education costs, etc. but that's another topic), but time and time again we've historically seen that things improve, not just in overall quality of life but also in regards to more jobs often being created due to the new tech as well. The nature of our labor just changes is all, there'll still be plenty of work for people to do.
@@Scientiavore I agree with most of your points, however I would argue that this technological revolution is different in the sense that intelligence itself is being automated. We are closing in on AGI and once it happens pretty much every job can be automated including the new jobs that are being created. That means we'll need massive systemic changes to actually spread the whealth that is being created.
I suspect part of what happened with Google's is they over corrected trying to compensate for the fact that white isbthe default person in a lot of datasets. They just nudged it to far in the default person = diverse direction.
I found last year if you ask it to generate art of petty theft, it was always depicting black people as stealing, if you tried to prompt white or asian, it would usjally just change the colour/style of the clothes.
5:23 The one on the left remember me of the Magi.
damned if you do, damned if you don't XD
"Its the same as drawing." Vaush on AI art
Minors
Maybe it can AI generate your kiddy corn now Vaush
Garbage in, garbage out. The universal law of programming.
It still doesn’t know why you have to have 5 fingers, so it puts them anywhere and it’s amazing to me.
Finally! Hale Hortler-chan!!
This shit is beyond comedy
The bbc article reads like the onion
soembady teach him how AI is trained
This was the truth of the world before yakub
Basically, generating images with "ai" algorithms is really stupid v:
Make an AI image of your gravestone, I'll take a leak on it after I saw that interview 😂
smh I can't believe the nazis have gone woke, now who am I gonna make little dark age edits of?
This is mostly just hilarious.
Love from a Xanderhal and TwistyTheTiny fan!
Better than back when i would have to put in prompts like black and dark skin, multiple times, to get someone black. There's still some ai like that
I get what they were trying to do, but I have no idea how it started over correcting THIS badly.
This reeks of anon.
Meanwhile, DuckDuckGo mischievously rubbing their hands together, like a housefly on rotten fruit.
Remember when a nazi statue with rammstein playing became a meme? How did that get so big
There actually were diverse hires for the nazi military. Not directly but were extra disposed mercenaries that were outsourced rather than minority residents/migrants such as from Turkey, China and Sudan.
The topic is overblown. Those compankes tried to counter a racial bias in the cheapest way posisbl3 and those are the results.
When you try to algorithm social justice, basically
@LimeyLassen not even that. They basically just put in that it randomly also puts in people of colour regardless of the context.
The most primitive methode. Works in most situations but when it fails the results are very strange.
Because that bias doesn't exist
But the whining about it did
That was what caused the overcorrection
@@hoominbeeing sorry but this is just wrong. There is a lot of bias in ais since those are trained by human data which is extremly biased.
The overcorrection is caused by them using the easiest way to fix the problem instead of a far more complex way.
@@emilsinclair4190 Which bias are you referring to? I wasn't referring to bias in general
They are "correcting" for it because of whining about non-existent bias, or at least bias that does not matter or is irrelevant.
2:53 “Now un4chanatly …”
Ai did something dumb/ problematic? Imagine my shock!
It's not AI but Google. It's different, but i guess Vaush community doesn't care
@vipcypr8368 yes, Google is generaly dumb.
But can you name an ai that never did something bad or stupid?
@@davitdavid7165Can you name a human? Weird question lol. Of course any widely used AI will fuck up at some point.
@XMysticHerox that is true to an extent. But some twch have higher failure rates than others. And some fail worse than others. AI constantly messes up and does so so bad, that would get an employee fired or other tech investigated. With all tech you have to ask yourself "is this worth it" and with ai, it looks like it only has niche applications and shouldn't be trusted with things like customer service or art.
@@davitdavid7165 It's literally in AI nature to do dumb things and then learn on it, so no
People can just create a bot that is racist and get it to interact with a chat bot , and after enough interactions can sabotage the chat bot
Well, there actually were plenty of AFRICAN Nazis.
So, the AI isn't off there.
Click clack pow racist google down
Vaush's AI rants are one of the few things that bring me pure elation.
I bet eventually, he'll support AI, once he realizes it's the best justification for ending capitalism we've ever had.
Ha ha, Vaush just said at a certain point AI is no different then DRAWING.... Aka: a form of art.
The right calls it woke, the left calls it fucked up.
Common AI L that i enjoy
The right calls it woke, the left calls it fucked yo and woke, both mald over the dumbest shit imaginable
He was right, this is very very funny.
10:35 try searching “family” though, most results will be all-white
Aaaaah, can't wait for you to deep dive into the Lex--Tucker Carlson convo! Many will benefit from your take, and hopefully begin to see Carlson for the manipulative, opportunistic sycophant that he is. I'm hoping for the same level of depth and clarity as your analysis of the Saagar-Crystal convo. Cheers from the West Coast!