@@Cyborg_LeninCapitalism it's just a efficient medium, that's it, it's not bad or good or an ideology, regional states have to regulate the capitalism and if we go with the the tought of "capitalism regulates by itself", we are doomed. I say this because my president supports AI and Bitcoins because the last thing i said
@@DynamoWD Well I do agree with that and even more. Capitalism however is not a neutral system, regulation is a must for it not to fall apart but it only lessens its inherent contradictions. Its a pretty shit system on its own, but it benefits these few who it needs to benefit so it stays in effect. Its simply wasnt designed to benefit everyone.
@@doodlesyoru2108 Give up free will forever their voices won’t be heard at all Display obedience While never stepping out of line And blindly swear allegiance Let your country control your mind (Let your country control your soul) 🎶
I liked how AI art looked when it first started being a thing where it had a more goopy and abstract look Now AI has an almost uncanny valley feel with seemingly detailed anatomy, landscapes and shading but the longer I look at it the more off everything is
It's because each peice wasn't made with purpose, nor do those who use generative AI care about purpose. It also hurts to see too, because that purpose is where *REAL* creativity comes from. But people are willing to throw that away because they hate learning.
The abstract images looked interesting. The current landscape is just kind of upsetting when you have to scan to check whether a drawing is real or not 😞
It lacks the ability to understand the core concepts of art, it has no concept of depth, form, form, perspective, colour, etc etc it's just using what it assumes is the correct choice and has a vague concept of those through pattern but it, itself has no idea just knows people tend to call their art more sad when it's more blue and gray so it uses those
Adobe Also potentially used anything that was uploaded to their creative cloud to train their AI. They say they only used licensed content, but the updated terms everyone was recently upset about really only changed those sections to say they could use anything made or opened using Adobe products, but their terms have had that clause for years, but it was only for uplads
We really need a new term for anything that isn't ML induced other than "derivative works." Because up til now, that only meant things like thumbnails, etc.
that is interesting because there have certainly been images or videos that have been uploaded that contain content of children that is exposed or actively being "abnormally" used (if you catch my drift) in which case using adobe firefly might actually be risky as you might have content that have been created used illegal material (and I am not talking about piracy illegal here, I am talking about serious federal prison serious here) which might land you in prison for a few years, the pr backlash against adobe would be hilarious to see
I really like the part where you explained in detail how Machine Learning can not be compared to Human Learning. I can't begin to count the number of times I've come across the ridiculous argument that feeding a machine learning algorithm an enormous dataset of images paired with text strings is the same as a human looking at pictures and trying to copy what they see.
AI is a primitive version of a human being fed enormous data sets of images (eyes looking at real life) and trying to imagine a picture in their head It's not the same as a human trying to copy what they see, because AI doesn't need to physically draw anything, so it's more close to a person trying to imagine a picture in greatest detail
@@aster2790 "primitive version of a human" is just false. It's a computer program. Never in history has man operated in any way similar to a computer program. AI is an attempt made by humans to emulate the human thought process. For all you know, we could be completely off the mark here.
Human inspiration See Art -> Internal representation of the art -> Make art by combining the concepts together in new ways AI Inspiration See Art -> Internal representation of the art -> Make art by combining the concepts together in new ways If you take the time to understand all of the math behind these models work rather than the readers digest version you'll see that in quantifiable, meaningful ways, generative AI models understand what they're doing.
@@Dongobog-ps9tz Ain't no way you think that AI does anything remotely similar to humans. It doesn't create anything new like humans can, it only does what it's told to, and ONLY with what it's given (especially if said given stuff was taken without consent). Sure, humans are able to mimic what they see, but they don't NEED to do that to be inspired. Sometimes inspiration literally comes out of nowhere and can truly be unique.
A couple weeks ago I saw a discussion in comments that left me speechless. Some guy was explaining with data and math how AI is supposedly much more energy efficient than people at making images (including the fact that artists need food and sleep. Apparently someone using AI does not need those?) and therefore people should drop making genuine art and embrace AI 100% simply because of "pure mathematical efficiency". They were talking of other people like some older model of domestic appliances to be disposed of so that *they* can keep consooming. Imagine being told that you should drop whatever hobby you have because it's "inefficient" and just put a robot to do it for you, so that you can keep grinding in the factory. Such a bright future these people want for us.
They are so brainwashed by this society to be a rat spinning the cogs of the system, that they only get their self-worth by being the best and most efficient rat of all. Totally devoid of humanity.
@@baiwuli6781 If you give up because someone is better than you, sorry art was never going to last for you. Someone is always better and faster than you at it. Ai is a tool you cannot claim copyright on any material Ai produces at least within the US. Skilled artist haven't given up at all it's the beginners and people who feel attacked by it directly because it can make a pretty images. Sorry to bust your bubble but thousands of people out there are far better than you or might even be at a level you will simply never reach. This is true for any skill. If you give up on art over Ai you didn't love art. Art was never known to be something you make millions with starving artist and all you need to love and enjoy it and enjoy consuming it if you don't art isn't for you that's okay but let's not pretend it's Ai's fault you or those artist quitting just want an excuse and reason to say "nah art ain't for me anymore it's AI's faults!"
@@MyNamesHunter75 I get your point, but we’re not talking about “you” or “me,” we’re talking about the art industry as a whole, as a system. Most artists start with a passion for art, doing it for free and because it’s fun, but then move on to working at companies or by fulfilling commissions. Unfortunately, we all need money to live, and we all have limited time, so “passion” isn’t the only factor here. So they make money doing lower-level work, which lets them spend more of their time doing art, improving their skills. The extra time they spend on these projects, and working with more experienced artists, let them level up and become really skilled artists themselves. This intermediate stage is really important in an art career. If AI takes away options for amateur art work, it disrupts this pipeline, which means fewer artists become experts and that means less human art, which is a bad thing for everyone, including AI. Sure, there may be a ton of pretty images made by AI, but there will be a lot of passionate artists who would have made really important and meaningful work if they didn’t have to drop out of the industry to feed their family or just survive. Some would-be artists will have to avoid trying altogether because it will just be too high-risk. The amount of high quality art goes down, which means AI has fewer things to train on. Since AI can’t innovate like humans can, and it can’t train on its own images without developing abnormalities, an AI-driven art industry is a dead end.
@@MyNamesHunter75 Your comment seemed rather aggressive at first but I understand that you're trying to help, and to be honest... your comment + pikat video made me feel better and motivated after a long period of sadness. So I really thank you for your advice (tho it feels weird to say this in youtube comment haha).
@@solarleaf2029 Not even the time spent on an artwork, just the years upon years of studying and practicing alone is enough to be outrageous. It's like if a person with a revoked driver's license tried to lecture a highly-skilled, veteran aviation officer on how to pilot a plane and everyone around them cheered them on. What sucks the most is that this discourages new artists who are already struggling enough with social pressure, distraction, self-hate and many other forms of discouragement. It's telling somebody "your dreams and goals mean nothing because robots can do things better and faster than you'll ever be able to!" ignoring the fact that those robots couldn't exist without the thousands of years of human creation and cultural evolution put behind it. I wish everyone could be made easily aware of how disgustingly scummy GEN AI is, as it fills me with pain knowing that ignorance and money will keep winning over what is true and right to our humanity. I am forever thankful to all of the people who spend even just a bit of their time fighting against it and trying to inform others, all of you are amazing people and artists to whom I wish the best. Let us hope our efforts lead us to a better future.
I've seen an analogy on this in a different comment section, stating "AI image generation is like a pizza deliverer that takes parts from 5 different pizzarias, and then delivers it to the orderer." By this logic, if the deliverer or the orderer would call themselves a chef for what they did, then they'd both be lying. The same logic applies to AI image generation.
I think the NFT people and the AI people are the exact same, just with a different coat of paint. Those people get weirdly defensive over that stuff. :/
“Art is subjective, but only if they create art in my specific way” do you guys get mad when people take photos with cameras and it takes 1 seconds to click the button?
Whenever someone says they trained their AI image generator with their own "art" (especially if the results are amazing) most of the time it is "fine tuned" and if you open the "trunk/hood", most of the time it is the infamous laion5b data set used in its pre training stages.
@@CunnyVirus Lol. It's not "StEaLiNg" and it never was. Trainings were made with web scraping utilities, a legally protected activity, and consent from artists was given, unknowingly, by placing it online. Also, in order for a photo be "stolen" you have to use the photo or a similar likeness in application. Every single generation is transformative.
First of all, I love the video and I especially love how you explained that there's a difference between *_"AI"_* and *_"Generative AI"_* (I've been trying to explain this to folks for a while now😭) And secondly, I kinda wish Generative AI had a physical form so I could get to boxing🥊😆
@@clinch4402It's people like you who made artists get afraid of using "choppy" animation (Oh, I 'dunno, Spiderverse?) Illustration is a type of art, no? A sketch doesn't stop being art, just because most of them are turned into full pieces later. Sometimes, an artists just feels like sketching something, and that's that. Not every part in this video needs to move, or be super detailed?
@@clinch4402”half assed anime pinup slop” are you talking about her character she uses for her vtuber? It isn’t a half assed design, it’s a good artstyle and a good character design. I’m gonna guess you got angry after finding out Pikat’s nationality?
I just hate seeing AI generated content every time I open up any sort of social media platform Please! Give an option to turn off these damn AI content
@@Dreamforge.Atelier I think we're probably gonna have to go back to using magazines again, but even then who's to say they won't eventually use AI for some of the pictures as well 😩
@@blackowl0265 I am very positive that AI will be under declaration obligation in the nearer future. That's the absolute LEAST! But when you ask me...generative AI should be banned back to the hell it emerged from! It ruins so many things...I hate it so much 😂🫥
Another note on Firefly - They've also found Midjourney in there as well. So these sources are cross feeding each other as well. Adobe may never have intended this, of course, but if you're using user submitted data for your model, and there's another model that happens to be able to shit out a ton of content, well.....
You can also note that Adobe has made pretty clear that their recent terms of use allow for the legal loophole to consider any files saved on their cloud as accessible for their use in AI training, even if it's not explicitly stated.
This video is informative. And yeah. And unlike ai, I can with enough time figure out an entire cartoon character turn around with just a ten or so screenshots of said character. Because I had to learn how human anatomy works, how everything is connected, where each limb goes and so on. I'm still struggling with perspective, but ai wouldn't be able to do these any better if only thing provided is one reference photo of a specific pose and one hastily scribbled character reference sheet. Especially if the character has different proportions than the photo reference. But an artist with a lot of practice will have it done in just a few hours (without rendering and complex shading obviously). This was the only thing I wanted the ai to be able to do, and it can't do it. I have better chance just trying to draw it myself or commissioning another, more experienced artist.
I like to say that a key difference between generative AI models and human learning is that once we've learned a given concept, we can apply that concept to the learning of _new_ concepts, when an AI model dataset is essentially being trained "from scratch" every time.
Pro tip: overlaying noise at a very low opacity over your art makes it a lot harder for AI to use it. There’s also a program called nightshade that can poison AI datasets.
"Yeah it would be kinda expensive to actually buy the stuff so I decided to just take it instead" definitely doesn't sound like something that a thief would say.
I can see so many prosecutors shaking in rage, defenseless as the defendant says "Asking for permission would have been difficult, so we didn't." And the judge says "Damn, you're right. No crimes committed here!"
from one programmer/artist to another, this was the best explanation I’ve seen, it’s super concise and easy to understand for the general public. i hope more people watch this video, thank you for uploading this!
On one hand, I read the news and see that Denmark just created an AI program that can potentially figure out if a person has cancer or is about to have it, based on just a simple blood test, finding a cancer cell mixed in with a billion others. It might even be able to tell you if you're ever going to have it again afterwards, which is crazy! But then on the other, I keep finding my favorite artists showing how fans have created AI versions of their art and has sent it to them for reference, and it just keeps hitting me (and most of said artists) the wrong way. Some people like to quote that the greatest compliment to art is imitation, but using a program to do this instead of putting in the effort yourself feels less like imitation and more like swiping it. Especially when it doesn't come with credit, or clear credit. AI can do so many damn cool things, but it makes it super difficult for me, as an artist, to not also feel equal parts dread when the word is uttered.
this is not a "on one hand and on the other" situation, AI is a buzzword, the "AI" for blood testing is completely disconnected from the "AI" that makes images, these are in no way connected aside from people using the same term.
@@Fruz That's what I'm trying to say. I guess I didn't get it across as well as I thought. AI just gets a yucky feeling out of me, because of the image generators, when in reality there are so many great reasons for why AI exists in the world that are worth being grateful and happy for. It stinks that it has soured things so to people like me, who can't hear or read the word without cringing.
@@Fruz Indeed, it's propaganda pushed out by Silicon Valley to sway public perception. And yes, I specifically mean Silicon Valley. It's just depressing seeing how so many AI bros are falling for these lies over and over again.
Both crypto and NFTs were supposedly great ideas at first. Unfortunately, their inventors forgot that you just need a handful of people with horrible intentions to ruin everything. AI is yet another example of this. Great potential if used ethically, ruined by people who only seek personal profits, exploiting others by using such systems. Also doctors are hating AI but not because of the things it offers, but for the way it is being used. There's several reports of people dismissing professional medical opinion to trust what some random AI on the internet said. Or actual doctors not double checking what an AI suggested, at the expense of the patient (yes, that specifically is a very bad doctor in such case). Artists wouldn't be so mad at AI if the first thing it was used for was not to get their work fed to it without permission, for griefers to immediately chant how they will all "soon be unemployed and homeless" or to "hurry up to leave the seat so that they can take their place". So many of them still are saying this. It is like this with nearly any job it is introduced. I'm just waiting for it to be in every job so that people finally understand why others involved before them are so mad about it.
@@Jounzey well...AI is supercool! Clip studio for example uses an AI that lets you upscale your canvas if you want it bigger. That's a very cool and handy tool FOR artists. But generative AI just sucks...noone asked for this beside greedy CEOs.
Actually a banger video. It's so good to see another artist who actually *knows* what they're talking about in terms of how the AI works exactly, and the difference between machine learning and other things like say a game opponent AI. But it's also nice that the video stays pretty on-the-fence, too. This video is purely informative and makes it so going over the controversy later (and the millions of other videos) is a lot more constructive.
Stolen training data aside, AI sucks at shading! The algorithm doesn't understand 3D shapes which is why the light and shadows are a complete mess. All it sees are flat pictures with pretty colours. Don't be an AI lover. Learn the fundamentals. Pick up the damn pen! You're better than AI!
I do think there are things you could learn and reference from AI, but it is no way reliable. Like say, it may have pretty colors along the likes of Yoneyama Mai, but it won't have the communication of an artist, it won't have the thought or intent that makes it so valuable in real artists' works. AI generated imagery creates only an illusion of understanding and intent, but push it just a little and that illusion quickly falls apart.
@@wisgarus I wouldn’t overall it’s inconsistent for the purposes of learning if you are training from nothing you are better off learning from taking photos outside and drawing it. You are more likely to learn errors and make more poor habits using Gen Ai for that. Only time I would ever recommend Gen Ai to be used is with someone who is trained in art and has the basics down. Then use it for inspiration not reference. Using it as a reference can hurt your growth by copying errors unknowingly.
This is why Ai is a tool and not a direct replacement it simply does not comprehend what it is breaking down and looking at it's really bad with depth if a shape overlaps another and is say meant to be a petal and it overlaps a character or terrain it connects the two because it views them together. Will this probably no longer be the case in few years probably but it will probably even then be applied in a way that is noticeably wrong. The time it takes to get the exact image you want out of an Ai you are better off putting that time in learning to draw if you want to be actually creative
I've seen multiple AI art pieces with good shading. And even if it's bad, only artists recognise or care, the average person sees a pretty picture and likes it. We need to oppose AI art on principle, not based on quality and people wanting to convince themselves it looks bad. The technology will only improve, and what you see now is the worst it will ever be.
This can be a very real temp solution but if they used something to sift through what is considered quality art it could easily avoid this which I wouldn't be surprised if that already is a thing. I think Ai aims to be trained off full illustrations vs sketches or fridge art because it has simply more wow factor
@@MyNamesHunter75 I think the biggest issue there is who's gonna determine what's "quality" and what isn't "quality". If we go back to the cat example in the video, a highly skilled painting of a cat and vector clipart of a cat would be both simply identified as "cat" and the AI would use that as a reference if it showed up in the prompt. AI doesn't exactly have taste because it's indiscriminate. You know who does have taste? The biggest discriminators of all time, humans. This means it would probably be up to humans to determine what's "quality" and what isn't. Of course, that also means companies would have to hire humans again to do this stuff, and we ALL know how much teeth-pulling it takes for companies nowadays to do even the slightest bit of morally upright work.
Thank you for the clear and easy to understand explanation! I also advocate for learning to know "your enemy" better, so some of the things mentioned here weren't new to me, but I never looked into technical details, so thank you for that info! I work in the design industry and do art as a hobby. The amount of times I've seen companies praise generative AI models in their job descriptions sickens me. I hate that the old farts who govern our countries are way too slow in regulating all these AI shenanigans. I've seen so many AI scamers pop up everywhere, and the demand for designers is also becoming less (mostly because of the global economic crisis (covid, war, climate), but still, AI doesn't help on top of those problems). And then people from marketing/HR have the audacity to tell me that I should not see AI as an enemy but instead see it as an OPPORTUNITY to make my work easier while it's actively hurting my career by drastically reducing hiring opportunities. Ridiculous. In my eyes generative AI, mostly the image generation type, is very unethical and the only good reason for bringing it into existence was to make money based on the hard work of real people who don't have the power to defend themselves on court level.
I really like this video because as much as I don't like a lot of things with AI art, I really dislike how so many people who dislike it seem to be uninformed or misinformed about how it works. Technology and computer science is such a different area of knowledge to art, and this AI art pandemic has a lot of artists speaking on what they don't know. And there's nothing I dislike more than seeing a side I agree with arguing my point badly. Thanks Pikat for making this video educating me and others about how this technology works, now we people can be informed haters instead of blindly hating lol
11:04 yeah, sure, except in the recent TOS that Adobe users are forced to accept, reveals quite clearly that Adobe is doing nothing else besides stealing like all the rest. Anything you create or use is scraped.
@@clinch4402pikat already knows that lol, she’s mentioned it but I can’t remember where. Also pikat was planning this video since atleast 1-2 weeks ago. If it was just to make a buck the video would be 8 minutes lol because that’s the monetisation minimum, pikat doesn’t even make much money from what I know anyways
It's certainly a contentious topic, personally it pisses me off to no end mostly because of the complete lack of empathy from folks propagating the tech. Not to say that some aspects of the tech couldn't be used to help artists with menial tasks like i've heard with the Across the Spiderverse movie, i probably wouldn't have much of a problem otherwise. it's only because all of it was born from such an unethical starting point, the people putting these data sets together knew that because of the vastness of the data the A.I needed, it would be impossible to ask the individual artists for permission to use their art. but they just did it anyway, the classic "your scientists were so preoccupied with whether or not they could they didn't stop to think if they should". and now artists have to deal with the fallout of that decision. it's a depressing yes, but we can adapt. it won't be the be all and end all for artists or our careers, we'll still be here.
You talk as if AI is born from evil for evil purposes😂. First, "AI was not born from an unethical starting point". There is absolutely no moral or legal obligation to ask permission for learning or training purposes. Artists, the hypocrites, do that all the time. And it's not as if AI uses artists' work as a reference to generate new images; they don't need to once the training process is completed, so there are no copyright issues. Second, AI is not made for replacing artists. It's made for people who can't draw. You guys love talking about AI as the evil corporations' tool for targeting artists, but AI is never about artists; it's about non-artists. It's a tool to help non-artists create artworks. You greedy artists love reminding us that art is a luxury, that we don't need it. But that's just another way of saying "if you are too poor to pay artists, just say so". Another excuse you guys love saying is that everyone can learn art. Yeah, but how many years do people have to sacrifice to get descent at making art ? From what I've heard from artists themselves, they say it took them on average years. Arrogant artists just love making these victim-blaming excuses that go along like 'your life sucks, that sounds like a you problem' with no empathy. And you heartless artists expect empathy from us ? The worst of all is that artists keep harassing, bullying, and threatening people using AI, accusing them of stealing their works, which it didn't, and are still delusional enough to think they are on the righteous side of the battle. In reality, they just feel insecure about AI and are ready to do anything to halt technological advancement, including terrorizing people on the internet. That's why they constantly make contradicting statements, saying how trash AI art is while also claiming it's threatening at the same time. Artists are also very manipulative. They love to rally people behind the banner of a battle between humanity and soulless corporations. In reality, their concern lies more in securing their own jobs than in the well-being of others. TLDR : artists are arrogant, delusional, greedy, heartless, hypocritical and manipulative.
i have a hard time differentiating ia generated content from people making derivative art, there is a LOT of derivative art on the internet made without the original authors concent, seems a bit odd to me that we make a huge fuss over IA making derivative content vs a guy making it. There is also the matter of owning a "style" like if someone draw in a certain style and another guy or ia uses that style is considered stealing? i understand this is a problem for artist i just have a hard time with the double standar a lot of people have about ai using other ppl work vs ppl using other ppl work.
You talk as if AI is born from evil for evil purposes. First, "AI was not born from an unethical starting point". There is absolutely no moral or legal obligation to ask permission for learning or training purposes. Artists, the hypocrites, do that all the time. And it's not as if AI uses artists' work as a reference to generate new images; they don't need to once the training process is completed, so there are no copyright issues. Second, AI is not made for replacing artists. It's made for people who can't draw. You guys love talking about AI as the evil corporations' tool for targeting artists, but AI is never about artists; it's about non-artists. It's a tool to help non-artists create artworks. You greedy artists love reminding us that art is a luxury, that we don't need it. But that's just another way of saying "if you are too poor to pay artists, just say so". Another excuse you guys love saying is that everyone can learn art. Yeah, but how many years do people have to sacrifice to get descent at making art ? From what I've heard from artists themselves, they say it took them on average years. Arrogant artists just love making these victim-blaming excuses that go along like 'your life sucks, that sounds like a you problem' with no empathy. And you heartless artists expect empathy from us ? The worst of all is that artists keep harassing, bullying, and threatening people using AI, accusing them of stealing their works, which it didn't, and are still delusional enough to think they are on the righteous side of the battle. In reality, they just feel insecure about AI and are ready to do anything to halt technological advancement, including terrorizing people on the internet. That's why they constantly make contradicting statements, saying how trash AI art is while also claiming it's threatening at the same time. Artists are also very manipulative. They love to rally people behind the banner of a battle between humanity and soulless corporations. In reality, their concern lies more in securing their own jobs than in the well-being of others.
Tech companies do not care how unethical they get the data. It's either you agree and they get it or you don't and they get it anyways and pay the potential lawsuits later it's the reality of how these companies work and have for a long time especially when it comes to potentially innovative/revolutionary technology they absolutely do not care it's you give it or they take it
@@MyNamesHunter75 we agree about stuff like adobe or art programs stealing images that are not public, but using public images for ai training is no more stealing than watching any piece of art published in the internet, there is no such thing as a prohibition to watch art or even change whatever you find, you cannot sell or post it but you can download any piece of art to your pc you find, it would be legal madness if you could not tbh. Do you consider IA using publicly available images to be in the wrong?
Hi Pikat! First of all nice video, I appreciated so much how you talked about this topic. I know that this comment is over a month late but I just watched the video now! I wanted to share with you some of my experience with AI. I'm not a professional artist I'm just a newbie who's trying to get better by watching videos from artist like you. When AI was released I tried to do what you've done here: try to understand it. I saw a lot of artist worried or angry at A.I. for various reason, but one thing that I noticed is that none of them actually used A.I. or wanted to learn more about it. So I took a bit of courage and tried myself to learn how this images were made and after understanding about models, lora, checkpoints vae etc. etc. I tried to use it for my own project. NOT as a way to skip work but as an instrument that can help me realize what I want to draw or create before starting my work. It helped me understand how I wanted to make a character and act as a "bridge" between the work in my mind and the actual work. So now I can say this: AI generative models are a great instrument for artist! Of course I don't mean those who claim to be artist just by using Ai and post it everywhere but artist who actually work as..well artist. Btw let me know what do you think about this and if you've used A.I. sometimes!
It's a shame that people got so attached to generative AI so fast. Because regardless of how ethical it could be, AI users are robbing themselves of the chance to learn what really makes a peice good in the first place. That being purpose. Purpose is where *REAL* creativity comes from, but somebody could never learn that if they don't wanna put in the work to learn.
AI, you know what I want AI to do? Not take the one thing that I do besides playing a game which is also one of my only passions and only hobby besides playing video games because I haven’t started learning bass yet and drawing (and animating every few days) is the only ability I have and I don’t want to lose. **COUGH** WHAT? WHO SAID THAT? NOT ME AHAHA
Nobody is taking art away from artists. No matter how good AI gets at generation, there's a myriad of things about art it cannot replicate, not until they start making artificial consciousnesses that can comprehend and *live* like a human
@@aplleprojdumpEh, not really because now you got the ethical dilemma of having a person solely making art for you and at that point that's a very clear human rights violation.
@@aplleprojdump That's the thing if Ai becomes sentient I absolutely would say it's creating art because their is intention behind every thing it does, in it's current state it has no concept of how to accurately apply basic concepts such as depth. When an object or thing such as a leaf is overtop of something else it will often not give it a outline and blend the two together because it simply doesn't understand one object is in front of the other. Art requires intention and currently the definition of art reinforces and states human. If Ai became sentient I'd say it became more human and would be able to say it's now creating art and not just images similar to art because Ai "art" and art shouldn't be considered the same as the "art" simply isn't art. It's why I find it crazy when Ai prompters say they are artist, it's like asking someone to commission a piece of art and then say you are an artist because it was your idea just not how it works
coming from a person who works in data science, this explained the essential concepts very well, without diving into the nitty gritty of architectures and data processing. that said i wonder if gpt was consulted for the research of this video tho 😏 jokes aside great vid
There is an important conversation about latent spaces, and the extent to which an efficient latent embedding of art can be compared to a human beings internal representation of art, that almost everyone completly avoids talking about when they want to make this argument that the models do not and cannot understand anything. If the dimensions in the latent space translate to disentangled, causally relevant concepts behind the art I think the model understands the art.
I learned a little watching this ! Great video !! It's sad indeed that we will have to deal with these problem that silicon valley created at one day just because, buncha thieves....
The difference is that models that create better "AI art" , as midjourney, are trained using data from art sites, like deviantart, artstation, pixiv, etc. These models are different from the normal ones that gather all images from the internet. So yeah, they are stealing specifically from artists.
are styles owned by artists tho? its a serious question, because the definition is very important, there are a bunch of artist that look at other ppl job and make their own based on that, the same way IA does it, so are artists owner of the styles? or anyone is free to copy them?
@@nicormoreno While styles aren't considered copyright or anything like that and imitating a style isn't wrong (regardless of what some tik tok users may think) the way Ai was trained was 100% unethical and the way it uses those images are very different than how a person would, one is with intent and understand and the other is just doing what it assumes is the correct choice. The point overall is the majority of people didn't agreed to have their art used to train Ai which a lot of people view as a replacement to them. Where if asked I think 90% of the art community would have said no. Thing is large tech companies do not care how they get that data ethically or not especially with revolutionary/extremely innovative technology which has been in sci-fi movies for decades and will bring them trillions in revenue. Tech companies set aside large chunks of cash to pay the lawyers if and when lawsuits come and the little guy likely isn't going to win. It's the sad reality of the world we live in but tech companies either ask and if you agree take, or ask you disagree and they take there is no option a tech company goes "well they said so guess we can't" this has been shown for decades they do no care about you or your feelings but about the profit to be made Ai already has 100+ billion sunken into research for it so they expect at least 2 - 3x that and whatever else is spent on it to be returned. It's why I don't understand why people think Tech companies will suddenly be ethical like lets be real the device you and I are on 100% had some child labour and forced labour involved in making it. Ai is a marketing term that ironically movies made us all aware of far before it was ever a thing to even look at seriously and now is the biggest buzzword to overcharge a bad product and profit off massively
@@MyNamesHunter75 you have an interesting posture but we lack a bit of nuance, I would split images into 2 groups, public available images (anything that is published online and do not have any strong copyright beside redistribution and sell) this is the kind of art you can find in google or any other public website, anyone can see it, download it and do whatever with it except selling it (or change and sell) lets call that group "google art". Second group would be private art, this is the stuff only image editing software would be able to get their hands into, this is also a much more protected kind of art. So I think it's impossible to try and say that "google art" cannot be used, we would need to prohibit anyone from using that art for anything without consent and that is just a huge nono imo, art is protected by copyright but that does not mean you are going to be punished by downloading or saving it as long as you are not selling the piece you are fine, even if you create art from google art as long as its derivate and not a copy you should be fine and the same standard applies to tech companies as it should. About private art i would say the government needs to punish the people trying to use it, it's disgusting, adobe took images hostage if you don't accept the ToS and that is crazy and should be 10000% illegal. Finally you said authors should be able to opt out, this is true for private art, but i think giving authors that kind of power would be crazy for google art, do you know computers download every image you see? Are we willing to give authors that kind of control over their public art? Like they could demand for you to delete your temporal data or prevent your computer from processing images to make downloads faster, etc. That seems like too much to me.
@@MyNamesHunter75 Nobody has the ability to restrict everyone else from reusing their art in all cases. Copyright is not "actual property", but a limited monopoly privilege, and only applies to a limited set of circumstances. One can violate it only if the resulting artwork bears substantial similarity, which neither AI art nor models do. There is nothing unethlcaI going on here anymore than there would be if I copied your pfp, warped it in krita into unrecognizeability and then posted it somewhere. I would have still used your logo - if it is something you made - but as the result would bear little resemblance to the original and so you would have no grounds to stand on. Even if I turned it into an image you actively disliked and found offensive. But moreover your understanding of ethlcs is just off the mark. Just because people would say "no" doesn't mean they actually can justifiably restrict others from an action. There is no direct aggressive action on the part of people training AIs, which means anyone whose images were analyzed was not in fact a victim, so making it IIIegaI would be creating a victimless crime. The creation of victimless crimes is unethlcaI, as it makes the actual act of enforcement aggressive - rather than responsive - force. If a law is unjust then the penalties also are: Unjust fines are extortion, refusing them means being subject to attempted kidnapping (arrest), and resisting that means being subjected to escallating physical altercations.
@@Aubreykun I think people are usually more concerned with the fact that these models are trained to act as a “replacement” for traditional labor in the field, and that the training for these models was done using the work of the very people who would be replaced. I personally find that a bit yucky, but I’m not well enough versed in ethics argumentation to expand beyond that lol
I still don't understands why they used AI to make images. Why don't they uses AI for stuff that will ACTUALLY helps people. Like AI that could detect abnormal heat on a city and calls the firefigther or AI Sevurity Camera with face detection that will inform the closest police station? using AI to make "arts" won't really help much tbh other than satisfaction of "getting something done" in a click.
So, similar technology is used in research to help with solving actual problems. But the reason these large language models (LLMs) are taking off is because it’s easy and it makes money.
Pretty solid. Something I see a lot is that when people argue about whether generative AI does or doesn't learn like a human, they're often talking about different things. Now, there are certainly are some people taking the comparison too far, asserting that nobody should be upset their work is used for training or that AI training should have the same standards of legality that human education does. I consider this a willfully ignorant line of argument that inflates the connection and ignores several problems in favor of reaching a convenient conclusion BUT It's also wrong to say that there are no similarities. The problem with the human-learning section in the video is that image generators do not *draw*. To draw a face, a human needs to learn how to operate the complicated assemblies of bone and muscle we call arms in a specific way, but most humans can *imagine* faces pretty well because they have seen them many times and have learned what faces generally look like. To explain in more detail, let's get a bit deeper into the training process. We'll use Stable Diffusion as our example since that's the focus of this video, and since LLMs like ChatGPT are vastly different. As your video covers, training data for an image generator like SD is composed of images with a caption or tags describing each image. Now suppose you want to train a completely blank, fresh model (Not a finetune) to depict only bananas. You get a bunch of pictures of bananas with the caption 'banana', and then your begin the training process. These images are then used like flash-cards. They're shown over and over to the deep learning model, along with the word. It doesn't have eyes or any visual sense, it 'sees' the images are number grids, but it notices there's a lot of similar RGB values (shades of yellow) that are usually in a similar pattern (A curved shape). Therefore, it learns that a banana is usually yellow and usually curved. You finish the training, then you ask that model to show you bananas, and you get things that are generally yellow and curved. An important thing to note here is that this new banana-model doesn't contain those training images. It doesn't save or cache them in a catalog it can reference, and they can't be extracted from it (This is why image generation models are so much smaller in file size than their training sets). It only keeps the conclusions it drew from those images, the learning. This is pattern recognition, the same thing that most of our learning is founded on. This isn't a coincidence either, machine learning's goal is to emulate human learning because human learning is so flexible and powerful. The architecture that Stable Diffusion is built on uses a neural network designed to approximate neurons in a human brain. This isn't to suggest any ethical or legal implications, and it's still true that there are *many* differences between human learning and machine learning. But it's important to realize that when some people say 'AI learn like humans do' they're being perfectly reasonable to an extent. They're talking about pattern recognition, about showing an AI a flash-card and having it make its own connections. There a lots of videos on youtube explaining this better and in more detail than I can in this comment, and I recommend anyone interested have a look around. Vox's video 'AI art, explained' is a good place to start.
Better equiped to deal with the changes caused by AI? what do you want me to do not upload pictures keep them on a private account so no one can see it?
@@xn4plwhy do you say that they're a snake oil? The program distrup the encoding of the Ai because the AI read pixel the data of and image from top to bottom for finding a pattern. Also, they still say in the forum that they will update the program to cath up to date like every anti cheat will update if there is a new cheating sytem that go through their sytem.
@@leaderteammimikyu3024 to update their system to new architecture, they need to know how new Dall-e 4 or Midjourney v9 works, how it encodes images and text, how exactly their new training algo works, etc, to create adversary AI to put information inside the pictures to disturb the training of these models. But by the time this information becomes public (or if it even does at all) the model is already trained and out there for everyone to use, so using glaze or nightshade will only defend against fine-tuning, while the base model could've been already trained on your "protected by glaze" art, because it can't protect against AI architectures it wasn't designed to fight. Then again the adversarial noise glaze or nightshade uses to disrupt the training is very precise, running your "protected" art through 0.5% of blur/sharpen filter and adding 0.5% of gaussian noise would entirely destroy all glaze protection while being imperceptible by humans, meaning picture would be still suitable for training AI. That's why glaze and nightshade are ultimately useless to prevent training. The only way to be sure is to not share your art online where it can be scraped by bots or even if you can prevent bots, a person can manually copy images to fine-tune the base model to copy your style.
Once more, thanks for another great video, Pikat 🙋 Not just that, but also shed some much needed light on the subject. It saddens me to see what companies will do to not pay someone... And how it messed with prospective talents all around the internet. But, no matter what, I can still create; I'm not giving up. 9:49 She did not even think of some nice, fancy words for the excuse, right? 10:03 😂 As someone with an account there... that. Was. Just... *Dead center in the freakin' bullseye,* Pikat!
Here's what I will say as an AI engineer. Wether you like it or not it IS creative, after all it is mimicking the human brain. However it is NOT creating. To create means to physically change and interact with something until it becomes what you want, this is why we have to get a deep understanding to be able to physically change it into what we want. AI models are simply predicting off of learned values and noise. Diffusion which uses gaussian noise replicates our brains ability to remember and visualize blurry memories, same with sora and dreams. Autoregressive mimicks our brains ability to hold attention over sequences of information, however it doesn't actually have memories beliefs or any hormones or decisionmaking which make humans such good conversationalist and decision-makers. There are many others I could go into but this is already a lengthy comment. This is one of the best videos I've seen at explaining things to artist though.
Basically everything we submit a picture, film/media videos, interactive entertainment (games), and audio files onto the Internet, will immediately get sent into a public archive? Thats both equally interesting and concerning
Cool video, very informative. Although I didn't see how it's outright "stealing" the artwork. Something you forgot to mention was that the AI image models do not store the images, but rather noise maps of those images, and that less than 1/1,000th of the information of the original images is kept (Roughly 300 terabytes of data as mentioned in the video, compared to a model only gigabytes in size). It essentially reverse engineers noise maps based on the prompts into images, from what little I know. While art is being "used" for it, it's not spewing out someone else's art. If it were, the companies could be sued for it. This is ignoring the side conversation of what theft actually is, since "stealing art" seems like a vague concept at best if it's anything other than taking physical property.
I don't understand why some people see this as such a nuanced topic. As an artist, it's generally okay to use other art and photos as references, but tracing is theft. If AI, in a way, traces thousands of pictures and just generates it's lines where most of the traced lines where, than it's still just traced art (even if you don't know from what art was used anymore). Regardless of how AI learns, if humans have a fundamentally different experience, and whatever else is said in arguments.... It's still just traced art which you simple can't sell.
I don't understand why some people see this as such a nuanced topic. As an artist, it's generally okay to use other art and photos as references, but tracing is theft. If AI learns through thousands of pictures and creates new, never before seen image, than it's just a new piece of art (even if you don't know from what art was used anymore).
I am happy to see the analysis of AI from the view of an artist, generally your opinions are correct. Any kind of AI needs tons of data to train. The model itself essentially is just a neural network. Different kind of network have different properties. The baseline is the full connected neural network. Nowadays the libraries for building AI network is mature, Pytorch or tensorflow could be used to create a neural network with much less code than before. However, personally people lacks the resource to create the large model. So most are mastered by big companies. However, if you really want to a try, it is not really hard. you could use libraries such as pytorch or tensorflow to do that. But of course, you need the data. An idea which I just find is that I could use PCG to generate the datas.
Had an argument with someone in a different comment section, they called all art forms a useless skill that shouldn't be learned anymore at the end of their message after defending GenAI, it made me mad but I had no reason to reply back anymore especially if these people does not value creativity and human mind.
Dont forget ai or ml helps clean up old movies, gets used by nintendo for cheap hd textures by up sampling texture files, phones to ml to clean images, photo shop uses ml to remove blemishes ect, an praise the smart lads who made ml to make good 30 to 60 fps. An video game 90 fps for rythem games. Seriously for reaction based games you need the extra frames . With synth v the ai voices like solaris and eleanor forte sound amazing
Can you make a video sometime about tips on how to prevent AI theft for artwork or like just ways to prevent theft or how to prepare for that to happen? I am always afraid to post online for this reason and I don't know how to make myself feel better about it. >.
There are two tools you can use called Nightshade, and Glaze. Glaze can be used to make your image unusable to AI, and Nightshade can be used to actually poison AI, if it were to ever use your image to train off. This is a very gross oversimplification, but I hope it helps.
I’m developing a application that will protect your media from the hands of A.I. and will be released Jan 1st 2025. Please wait patiently before A.I. generation becomes obsolete. ✊🤩
Amazing video pikat, you only forgot to mention that generative AIs need a LOT of power to keep theirs servers running, and also a LOT of water to cool down those servers, so horray global warming is afected by generative AI
Publicly available is just fancy way to say yoink You can do that in private use, but having these large companies doing it and getting tons of money from it while the real artists are struggling everyday is disgusting
...for years generative models just existed but hadn't advanced at all. I think a important thing to remember is that generative AI prolly is going to plateau on what it can do- that inevitably, they can't make it better. People will come to learn the limitations and artists will remain, as always, not limited the same way.
Here's how to fix AI problems: Make all and every result of any generative AI have no right for any copyright claims. That would at least make it less appealing for companies and stuff
As an AI engineer I will say now. AI wether you like it or not DOES have creativity, however it is NOT creating. When artist create we interact with our environment and change it to mean something else. AI is just creatively predicting the optimal image, so it's more like if your entire life's purpose would be to constantly dream and come up with higher resolution pretty dreams.
knowing ur a programmer as well as an artist makes me super happy cus i am too! im not happy with how ai has progressed because it couldve been ethical in so many ways but… its not :(
I'm an electric engineer who switched carrers to art, my thesis was about training a neural network to make levels for a videogame (lol) One of the most frustrating things I've had to deal is people that went full against AI, ALL artificial intelligence. I've seen freaking artists complaining a company used procedurally generated textures for certain thing, or a game is using artificial intelligence for their NPCs like no shit sherlock freaking checkers from 1951 used Artificial Intelligence that's not a new thing. Sorry for venting, but there's so much disinformation out there, people complaining about all this stuff while not even knowing how to put more ram in their machines. So thank you for making this video. The technology isn't the problem, is the abuse of companies of everyone's intellectual property for profit and thinking they can displace all creatives with it.
Yeah I think we need to actually learn how this stuff works to better refute claims. Also not to shit on literally everything that has the AI label, like NPCs.
Im kinda disappointed in people's inability to have a discussion about this without devolving to insults and mockery. Is a screen and pseudonym all it takes for people.....
companies stare at the end of human purpose and value then go "this will make us a lot of money"
yes because the unenlightened masses
they cannot make the judgement call
Capitalism is when everything is good as long as its profitable.
@@Cyborg_LeninCapitalism it's just a efficient medium, that's it, it's not bad or good or an ideology, regional states have to regulate the capitalism and if we go with the the tought of "capitalism regulates by itself", we are doomed.
I say this because my president supports AI and Bitcoins because the last thing i said
@@DynamoWD Well I do agree with that and even more. Capitalism however is not a neutral system, regulation is a must for it not to fall apart but it only lessens its inherent contradictions. Its a pretty shit system on its own, but it benefits these few who it needs to benefit so it stays in effect. Its simply wasnt designed to benefit everyone.
@@doodlesyoru2108 Give up free will forever their voices won’t be heard at all
Display obedience
While never stepping out of line
And blindly swear allegiance
Let your country control your mind
(Let your country control your soul) 🎶
I liked how AI art looked when it first started being a thing where it had a more goopy and abstract look
Now AI has an almost uncanny valley feel with seemingly detailed anatomy, landscapes and shading but the longer I look at it the more off everything is
It's because each peice wasn't made with purpose, nor do those who use generative AI care about purpose.
It also hurts to see too, because that purpose is where *REAL* creativity comes from. But people are willing to throw that away because they hate learning.
The abstract images looked interesting. The current landscape is just kind of upsetting when you have to scan to check whether a drawing is real or not 😞
Same, at least i had to use my brain to see something in the abstract
It lacks the ability to understand the core concepts of art, it has no concept of depth, form, form, perspective, colour, etc etc it's just using what it assumes is the correct choice and has a vague concept of those through pattern but it, itself has no idea just knows people tend to call their art more sad when it's more blue and gray so it uses those
"I liked how AI art looked-" opinion utterly and totally discarded, you have no taste worth expressing to the world.
Adobe Also potentially used anything that was uploaded to their creative cloud to train their AI. They say they only used licensed content, but the updated terms everyone was recently upset about really only changed those sections to say they could use anything made or opened using Adobe products, but their terms have had that clause for years, but it was only for uplads
We really need a new term for anything that isn't ML induced other than "derivative works." Because up til now, that only meant things like thumbnails, etc.
if you're painterly, buy SAI. like me
that is interesting because there have certainly been images or videos that have been uploaded that contain content of children that is exposed or actively being "abnormally" used (if you catch my drift) in which case using adobe firefly might actually be risky as you might have content that have been created used illegal material (and I am not talking about piracy illegal here, I am talking about serious federal prison serious here) which might land you in prison for a few years, the pr backlash against adobe would be hilarious to see
I really like the part where you explained in detail how Machine Learning can not be compared to Human Learning. I can't begin to count the number of times I've come across the ridiculous argument that feeding a machine learning algorithm an enormous dataset of images paired with text strings is the same as a human looking at pictures and trying to copy what they see.
AI is a primitive version of a human being fed enormous data sets of images (eyes looking at real life) and trying to imagine a picture in their head
It's not the same as a human trying to copy what they see, because AI doesn't need to physically draw anything, so it's more close to a person trying to imagine a picture in greatest detail
@@aster2790 "primitive version of a human" is just false. It's a computer program. Never in history has man operated in any way similar to a computer program.
AI is an attempt made by humans to emulate the human thought process. For all you know, we could be completely off the mark here.
Human inspiration
See Art -> Internal representation of the art -> Make art by combining the concepts together in new ways
AI Inspiration
See Art -> Internal representation of the art -> Make art by combining the concepts together in new ways
If you take the time to understand all of the math behind these models work rather than the readers digest version you'll see that in quantifiable, meaningful ways, generative AI models understand what they're doing.
@@Dongobog-ps9tz AI has no inspiration. Also have I introduced you to pencils and paper?
@@Dongobog-ps9tz Ain't no way you think that AI does anything remotely similar to humans. It doesn't create anything new like humans can, it only does what it's told to, and ONLY with what it's given (especially if said given stuff was taken without consent).
Sure, humans are able to mimic what they see, but they don't NEED to do that to be inspired. Sometimes inspiration literally comes out of nowhere and can truly be unique.
A couple weeks ago I saw a discussion in comments that left me speechless. Some guy was explaining with data and math how AI is supposedly much more energy efficient than people at making images (including the fact that artists need food and sleep. Apparently someone using AI does not need those?) and therefore people should drop making genuine art and embrace AI 100% simply because of "pure mathematical efficiency". They were talking of other people like some older model of domestic appliances to be disposed of so that *they* can keep consooming.
Imagine being told that you should drop whatever hobby you have because it's "inefficient" and just put a robot to do it for you, so that you can keep grinding in the factory. Such a bright future these people want for us.
God. That's so utterly disconnected to other humans and why artists even create in the first place. What an unempathetic soul.
@@SugarbirdyOvO My thoughts too :/
They are so brainwashed by this society to be a rat spinning the cogs of the system, that they only get their self-worth by being the best and most efficient rat of all. Totally devoid of humanity.
My hobby is what helps me a lot. I also think the AI would not match the images I have in my head.
"Energy efficient" when they are currently creating an energy crisis due how much power these consume.
ai art makes me appreciate skilled artists even more.
yeah, because there will be fewer and fewer of them 😂
This sounds so dumb like you think tuat millions of artits would hist vanish because gen AI is a thing.@@baiwuli6781
@@baiwuli6781 If you give up because someone is better than you, sorry art was never going to last for you. Someone is always better and faster than you at it. Ai is a tool you cannot claim copyright on any material Ai produces at least within the US. Skilled artist haven't given up at all it's the beginners and people who feel attacked by it directly because it can make a pretty images. Sorry to bust your bubble but thousands of people out there are far better than you or might even be at a level you will simply never reach. This is true for any skill. If you give up on art over Ai you didn't love art. Art was never known to be something you make millions with starving artist and all you need to love and enjoy it and enjoy consuming it if you don't art isn't for you that's okay but let's not pretend it's Ai's fault you or those artist quitting just want an excuse and reason to say "nah art ain't for me anymore it's AI's faults!"
@@MyNamesHunter75 I get your point, but we’re not talking about “you” or “me,” we’re talking about the art industry as a whole, as a system. Most artists start with a passion for art, doing it for free and because it’s fun, but then move on to working at companies or by fulfilling commissions. Unfortunately, we all need money to live, and we all have limited time, so “passion” isn’t the only factor here. So they make money doing lower-level work, which lets them spend more of their time doing art, improving their skills. The extra time they spend on these projects, and working with more experienced artists, let them level up and become really skilled artists themselves. This intermediate stage is really important in an art career. If AI takes away options for amateur art work, it disrupts this pipeline, which means fewer artists become experts and that means less human art, which is a bad thing for everyone, including AI. Sure, there may be a ton of pretty images made by AI, but there will be a lot of passionate artists who would have made really important and meaningful work if they didn’t have to drop out of the industry to feed their family or just survive. Some would-be artists will have to avoid trying altogether because it will just be too high-risk. The amount of high quality art goes down, which means AI has fewer things to train on. Since AI can’t innovate like humans can, and it can’t train on its own images without developing abnormalities, an AI-driven art industry is a dead end.
@@MyNamesHunter75 Your comment seemed rather aggressive at first but I understand that you're trying to help, and to be honest... your comment + pikat video made me feel better and motivated after a long period of sadness. So I really thank you for your advice (tho it feels weird to say this in youtube comment haha).
The second most irritating thing about AI art is that some people act like they're artists when all they do is generate images from these things.
Especially when we spend 2 + hours on a single piece of artwork omg it’s actually painful to watch
@@solarleaf2029 Not even the time spent on an artwork, just the years upon years of studying and practicing alone is enough to be outrageous. It's like if a person with a revoked driver's license tried to lecture a highly-skilled, veteran aviation officer on how to pilot a plane and everyone around them cheered them on.
What sucks the most is that this discourages new artists who are already struggling enough with social pressure, distraction, self-hate and many other forms of discouragement. It's telling somebody "your dreams and goals mean nothing because robots can do things better and faster than you'll ever be able to!" ignoring the fact that those robots couldn't exist without the thousands of years of human creation and cultural evolution put behind it.
I wish everyone could be made easily aware of how disgustingly scummy GEN AI is, as it fills me with pain knowing that ignorance and money will keep winning over what is true and right to our humanity. I am forever thankful to all of the people who spend even just a bit of their time fighting against it and trying to inform others, all of you are amazing people and artists to whom I wish the best. Let us hope our efforts lead us to a better future.
I've seen an analogy on this in a different comment section, stating "AI image generation is like a pizza deliverer that takes parts from 5 different pizzarias, and then delivers it to the orderer." By this logic, if the deliverer or the orderer would call themselves a chef for what they did, then they'd both be lying. The same logic applies to AI image generation.
I think the NFT people and the AI people are the exact same, just with a different coat of paint. Those people get weirdly defensive over that stuff. :/
“Art is subjective, but only if they create art in my specific way” do you guys get mad when people take photos with cameras and it takes 1 seconds to click the button?
Whenever someone says they trained their AI image generator with their own "art" (especially if the results are amazing) most of the time it is "fine tuned" and if you open the "trunk/hood", most of the time it is the infamous laion5b data set used in its pre training stages.
yeah because they wouldnt have 5 billion of their own arts to train their own AI. this technology has to steal somewhere.
@@CunnyVirus Lol. It's not "StEaLiNg" and it never was. Trainings were made with web scraping utilities, a legally protected activity, and consent from artists was given, unknowingly, by placing it online. Also, in order for a photo be "stolen" you have to use the photo or a similar likeness in application. Every single generation is transformative.
@@avalerionbass okay.
@@avalerionbass”consent was given, unknowingly” yeah… that’s not how the word CONSENT works
@@avalerionbass So its okay if someone gives consent unknowingly, ya, please try applying that logic to a few other things and see how it tastes.
literally any artist that works in tech knows this should be literally ILLEGAL
The power of money lol
First of all, I love the video and I especially love how you explained that there's a difference between *_"AI"_* and *_"Generative AI"_* (I've been trying to explain this to folks for a while now😭)
And secondly, I kinda wish Generative AI had a physical form so I could get to boxing🥊😆
@@clinch4402 me when i think sounding confident makes me right ( it doesn't, it just makes me look like i actually know stuff even tho i dont )
@@TakitoRi_ikr lol
@@clinch4402It's people like you who made artists get afraid of using "choppy" animation (Oh, I 'dunno, Spiderverse?) Illustration is a type of art, no? A sketch doesn't stop being art, just because most of them are turned into full pieces later. Sometimes, an artists just feels like sketching something, and that's that. Not every part in this video needs to move, or be super detailed?
@@clinch4402”half assed anime pinup slop” are you talking about her character she uses for her vtuber? It isn’t a half assed design, it’s a good artstyle and a good character design. I’m gonna guess you got angry after finding out Pikat’s nationality?
@@TakitoRi_ Nobody asked you about your insecurities. Google is your friend if you've realized that you are uneducated on a subject.
I just hate seeing AI generated content every time I open up any sort of social media platform
Please! Give an option to turn off these damn AI content
Hearing tts stuff too, it's driving me crazy being gaslit,
Google images is getting filled with this garbage. Even finding references is gonna get harder
Even that we have to ask ourself 'is this AI?' Now everytime is horrible.
@@Dreamforge.Atelier I think we're probably gonna have to go back to using magazines again, but even then who's to say they won't eventually use AI for some of the pictures as well 😩
@@blackowl0265 I am very positive that AI will be under declaration obligation in the nearer future. That's the absolute LEAST!
But when you ask me...generative AI should be banned back to the hell it emerged from! It ruins so many things...I hate it so much 😂🫥
Another note on Firefly - They've also found Midjourney in there as well. So these sources are cross feeding each other as well. Adobe may never have intended this, of course, but if you're using user submitted data for your model, and there's another model that happens to be able to shit out a ton of content, well.....
@@RobertStoll how can anyone still work with adobe?
Pikat, congrats for the 200k! Amazing job and amazing growth
Oh wow she got so popular so fast, and I'm here for it
@@clinch4402 What r u talking about
@@clinch4402 Most of her videos are art tutorials, what are you on about lol
@@clinch4402she doesn’t defend adobe, she literally says something against them 5 mins later
@@clinch4402 just don't watch her videos id thats a problem for you
it is nice to hear your perspective as you have been a programmer and an artist professionally. Thanks for sharing.
You can also note that Adobe has made pretty clear that their recent terms of use allow for the legal loophole to consider any files saved on their cloud as accessible for their use in AI training, even if it's not explicitly stated.
This video is informative. And yeah. And unlike ai, I can with enough time figure out an entire cartoon character turn around with just a ten or so screenshots of said character. Because I had to learn how human anatomy works, how everything is connected, where each limb goes and so on. I'm still struggling with perspective, but ai wouldn't be able to do these any better if only thing provided is one reference photo of a specific pose and one hastily scribbled character reference sheet. Especially if the character has different proportions than the photo reference. But an artist with a lot of practice will have it done in just a few hours (without rendering and complex shading obviously). This was the only thing I wanted the ai to be able to do, and it can't do it. I have better chance just trying to draw it myself or commissioning another, more experienced artist.
I like to say that a key difference between generative AI models and human learning is that once we've learned a given concept, we can apply that concept to the learning of _new_ concepts, when an AI model dataset is essentially being trained "from scratch" every time.
Pro tip: overlaying noise at a very low opacity over your art makes it a lot harder for AI to use it. There’s also a program called nightshade that can poison AI datasets.
"Yeah it would be kinda expensive to actually buy the stuff so I decided to just take it instead" definitely doesn't sound like something that a thief would say.
I can see so many prosecutors shaking in rage, defenseless as the defendant says "Asking for permission would have been difficult, so we didn't." And the judge says "Damn, you're right. No crimes committed here!"
from one programmer/artist to another, this was the best explanation I’ve seen, it’s super concise and easy to understand for the general public. i hope more people watch this video, thank you for uploading this!
On one hand, I read the news and see that Denmark just created an AI program that can potentially figure out if a person has cancer or is about to have it, based on just a simple blood test, finding a cancer cell mixed in with a billion others. It might even be able to tell you if you're ever going to have it again afterwards, which is crazy!
But then on the other, I keep finding my favorite artists showing how fans have created AI versions of their art and has sent it to them for reference, and it just keeps hitting me (and most of said artists) the wrong way. Some people like to quote that the greatest compliment to art is imitation, but using a program to do this instead of putting in the effort yourself feels less like imitation and more like swiping it. Especially when it doesn't come with credit, or clear credit.
AI can do so many damn cool things, but it makes it super difficult for me, as an artist, to not also feel equal parts dread when the word is uttered.
this is not a "on one hand and on the other" situation, AI is a buzzword, the "AI" for blood testing is completely disconnected from the "AI" that makes images, these are in no way connected aside from people using the same term.
@@Fruz That's what I'm trying to say. I guess I didn't get it across as well as I thought. AI just gets a yucky feeling out of me, because of the image generators, when in reality there are so many great reasons for why AI exists in the world that are worth being grateful and happy for. It stinks that it has soured things so to people like me, who can't hear or read the word without cringing.
@@Fruz Indeed, it's propaganda pushed out by Silicon Valley to sway public perception. And yes, I specifically mean Silicon Valley.
It's just depressing seeing how so many AI bros are falling for these lies over and over again.
Both crypto and NFTs were supposedly great ideas at first. Unfortunately, their inventors forgot that you just need a handful of people with horrible intentions to ruin everything. AI is yet another example of this. Great potential if used ethically, ruined by people who only seek personal profits, exploiting others by using such systems.
Also doctors are hating AI but not because of the things it offers, but for the way it is being used. There's several reports of people dismissing professional medical opinion to trust what some random AI on the internet said. Or actual doctors not double checking what an AI suggested, at the expense of the patient (yes, that specifically is a very bad doctor in such case).
Artists wouldn't be so mad at AI if the first thing it was used for was not to get their work fed to it without permission, for griefers to immediately chant how they will all "soon be unemployed and homeless" or to "hurry up to leave the seat so that they can take their place". So many of them still are saying this.
It is like this with nearly any job it is introduced. I'm just waiting for it to be in every job so that people finally understand why others involved before them are so mad about it.
@@Jounzey well...AI is supercool! Clip studio for example uses an AI that lets you upscale your canvas if you want it bigger. That's a very cool and handy tool FOR artists.
But generative AI just sucks...noone asked for this beside greedy CEOs.
Actually a banger video. It's so good to see another artist who actually *knows* what they're talking about in terms of how the AI works exactly, and the difference between machine learning and other things like say a game opponent AI. But it's also nice that the video stays pretty on-the-fence, too. This video is purely informative and makes it so going over the controversy later (and the millions of other videos) is a lot more constructive.
that clodsire is a nightmare, and i MUST see it in animation, i must see it crawling around
Stolen training data aside, AI sucks at shading! The algorithm doesn't understand 3D shapes which is why the light and shadows are a complete mess. All it sees are flat pictures with pretty colours. Don't be an AI lover. Learn the fundamentals. Pick up the damn pen! You're better than AI!
You know who also sucks at shading? Many many artists.
I do think there are things you could learn and reference from AI, but it is no way reliable. Like say, it may have pretty colors along the likes of Yoneyama Mai, but it won't have the communication of an artist, it won't have the thought or intent that makes it so valuable in real artists' works.
AI generated imagery creates only an illusion of understanding and intent, but push it just a little and that illusion quickly falls apart.
@@wisgarus I wouldn’t overall it’s inconsistent for the purposes of learning if you are training from nothing you are better off learning from taking photos outside and drawing it. You are more likely to learn errors and make more poor habits using Gen Ai for that. Only time I would ever recommend Gen Ai to be used is with someone who is trained in art and has the basics down. Then use it for inspiration not reference. Using it as a reference can hurt your growth by copying errors unknowingly.
This is why Ai is a tool and not a direct replacement it simply does not comprehend what it is breaking down and looking at it's really bad with depth if a shape overlaps another and is say meant to be a petal and it overlaps a character or terrain it connects the two because it views them together. Will this probably no longer be the case in few years probably but it will probably even then be applied in a way that is noticeably wrong. The time it takes to get the exact image you want out of an Ai you are better off putting that time in learning to draw if you want to be actually creative
I've seen multiple AI art pieces with good shading. And even if it's bad, only artists recognise or care, the average person sees a pretty picture and likes it. We need to oppose AI art on principle, not based on quality and people wanting to convince themselves it looks bad. The technology will only improve, and what you see now is the worst it will ever be.
what if we all start drawing like 5 yearolds and see if AI gets worse over time. Would be funny for sure.
My drawing skills finally found a way to be useful
Yes!!!
This can be a very real temp solution but if they used something to sift through what is considered quality art it could easily avoid this which I wouldn't be surprised if that already is a thing. I think Ai aims to be trained off full illustrations vs sketches or fridge art because it has simply more wow factor
@@MyNamesHunter75 I think the biggest issue there is who's gonna determine what's "quality" and what isn't "quality". If we go back to the cat example in the video, a highly skilled painting of a cat and vector clipart of a cat would be both simply identified as "cat" and the AI would use that as a reference if it showed up in the prompt. AI doesn't exactly have taste because it's indiscriminate.
You know who does have taste? The biggest discriminators of all time, humans. This means it would probably be up to humans to determine what's "quality" and what isn't. Of course, that also means companies would have to hire humans again to do this stuff, and we ALL know how much teeth-pulling it takes for companies nowadays to do even the slightest bit of morally upright work.
@@temperancedraws792the quality of an art is determined by the number of likes
To quote a dude from twitter: "IA art made belive in the concept of the human soul, because now i know how it looks without it"
This is a very informative video! Would love to see one on your understanding and effectiveness of Glaze and Nightshade.
Thank you for demystifying this for us !
Thank you for the clear and easy to understand explanation! I also advocate for learning to know "your enemy" better, so some of the things mentioned here weren't new to me, but I never looked into technical details, so thank you for that info!
I work in the design industry and do art as a hobby. The amount of times I've seen companies praise generative AI models in their job descriptions sickens me. I hate that the old farts who govern our countries are way too slow in regulating all these AI shenanigans. I've seen so many AI scamers pop up everywhere, and the demand for designers is also becoming less (mostly because of the global economic crisis (covid, war, climate), but still, AI doesn't help on top of those problems).
And then people from marketing/HR have the audacity to tell me that I should not see AI as an enemy but instead see it as an OPPORTUNITY to make my work easier while it's actively hurting my career by drastically reducing hiring opportunities. Ridiculous. In my eyes generative AI, mostly the image generation type, is very unethical and the only good reason for bringing it into existence was to make money based on the hard work of real people who don't have the power to defend themselves on court level.
I am from the area, and I am pleased by the amount of research and clarification of concepts in this video :)
I really like this video because as much as I don't like a lot of things with AI art, I really dislike how so many people who dislike it seem to be uninformed or misinformed about how it works. Technology and computer science is such a different area of knowledge to art, and this AI art pandemic has a lot of artists speaking on what they don't know. And there's nothing I dislike more than seeing a side I agree with arguing my point badly. Thanks Pikat for making this video educating me and others about how this technology works, now we people can be informed haters instead of blindly hating lol
11:04
yeah, sure, except in the recent TOS that Adobe users are forced to accept, reveals quite clearly that Adobe is doing nothing else besides stealing like all the rest. Anything you create or use is scraped.
@@clinch4402calm down
@@clinch4402 You sound like a fan, just don't watch if you find her videos half assed and disappointing, the video is still very informative overall
@@clinch4402pikat already knows that lol, she’s mentioned it but I can’t remember where. Also pikat was planning this video since atleast 1-2 weeks ago. If it was just to make a buck the video would be 8 minutes lol because that’s the monetisation minimum, pikat doesn’t even make much money from what I know anyways
Did you even watch the video where she showed the sources and images she researched?@@clinch4402
It's certainly a contentious topic, personally it pisses me off to no end mostly because of the complete lack of empathy from folks propagating the tech. Not to say that some aspects of the tech couldn't be used to help artists with menial tasks like i've heard with the Across the Spiderverse movie, i probably wouldn't have much of a problem otherwise. it's only because all of it was born from such an unethical starting point, the people putting these data sets together knew that because of the vastness of the data the A.I needed, it would be impossible to ask the individual artists for permission to use their art. but they just did it anyway, the classic "your scientists were so preoccupied with whether or not they could they didn't stop to think if they should". and now artists have to deal with the fallout of that decision. it's a depressing yes, but we can adapt. it won't be the be all and end all for artists or our careers, we'll still be here.
You talk as if AI is born from evil for evil purposes😂.
First, "AI was not born from an unethical starting point". There is absolutely no moral or legal obligation to ask permission for learning or training purposes. Artists, the hypocrites, do that all the time. And it's not as if AI uses artists' work as a reference to generate new images; they don't need to once the training process is completed, so there are no copyright issues.
Second, AI is not made for replacing artists. It's made for people who can't draw. You guys love talking about AI as the evil corporations' tool for targeting artists, but AI is never about artists; it's about non-artists. It's a tool to help non-artists create artworks. You greedy artists love reminding us that art is a luxury, that we don't need it. But that's just another way of saying "if you are too poor to pay artists, just say so". Another excuse you guys love saying is that everyone can learn art. Yeah, but how many years do people have to sacrifice to get descent at making art ? From what I've heard from artists themselves, they say it took them on average years. Arrogant artists just love making these victim-blaming excuses that go along like 'your life sucks, that sounds like a you problem' with no empathy. And you heartless artists expect empathy from us ?
The worst of all is that artists keep harassing, bullying, and threatening people using AI, accusing them of stealing their works, which it didn't, and are still delusional enough to think they are on the righteous side of the battle. In reality, they just feel insecure about AI and are ready to do anything to halt technological advancement, including terrorizing people on the internet. That's why they constantly make contradicting statements, saying how trash AI art is while also claiming it's threatening at the same time. Artists are also very manipulative. They love to rally people behind the banner of a battle between humanity and soulless corporations. In reality, their concern lies more in securing their own jobs than in the well-being of others.
TLDR : artists are arrogant, delusional, greedy, heartless, hypocritical and manipulative.
i have a hard time differentiating ia generated content from people making derivative art, there is a LOT of derivative art on the internet made without the original authors concent, seems a bit odd to me that we make a huge fuss over IA making derivative content vs a guy making it. There is also the matter of owning a "style" like if someone draw in a certain style and another guy or ia uses that style is considered stealing? i understand this is a problem for artist i just have a hard time with the double standar a lot of people have about ai using other ppl work vs ppl using other ppl work.
You talk as if AI is born from evil for evil purposes.
First, "AI was not born from an unethical starting point". There is absolutely no moral or legal obligation to ask permission for learning or training purposes. Artists, the hypocrites, do that all the time. And it's not as if AI uses artists' work as a reference to generate new images; they don't need to once the training process is completed, so there are no copyright issues.
Second, AI is not made for replacing artists. It's made for people who can't draw. You guys love talking about AI as the evil corporations' tool for targeting artists, but AI is never about artists; it's about non-artists. It's a tool to help non-artists create artworks. You greedy artists love reminding us that art is a luxury, that we don't need it. But that's just another way of saying "if you are too poor to pay artists, just say so". Another excuse you guys love saying is that everyone can learn art. Yeah, but how many years do people have to sacrifice to get descent at making art ? From what I've heard from artists themselves, they say it took them on average years. Arrogant artists just love making these victim-blaming excuses that go along like 'your life sucks, that sounds like a you problem' with no empathy. And you heartless artists expect empathy from us ?
The worst of all is that artists keep harassing, bullying, and threatening people using AI, accusing them of stealing their works, which it didn't, and are still delusional enough to think they are on the righteous side of the battle. In reality, they just feel insecure about AI and are ready to do anything to halt technological advancement, including terrorizing people on the internet. That's why they constantly make contradicting statements, saying how trash AI art is while also claiming it's threatening at the same time. Artists are also very manipulative. They love to rally people behind the banner of a battle between humanity and soulless corporations. In reality, their concern lies more in securing their own jobs than in the well-being of others.
Tech companies do not care how unethical they get the data. It's either you agree and they get it or you don't and they get it anyways and pay the potential lawsuits later it's the reality of how these companies work and have for a long time especially when it comes to potentially innovative/revolutionary technology they absolutely do not care it's you give it or they take it
@@MyNamesHunter75 we agree about stuff like adobe or art programs stealing images that are not public, but using public images for ai training is no more stealing than watching any piece of art published in the internet, there is no such thing as a prohibition to watch art or even change whatever you find, you cannot sell or post it but you can download any piece of art to your pc you find, it would be legal madness if you could not tbh. Do you consider IA using publicly available images to be in the wrong?
Hi Pikat! First of all nice video, I appreciated so much how you talked about this topic. I know that this comment is over a month late but I just watched the video now! I wanted to share with you some of my experience with AI. I'm not a professional artist I'm just a newbie who's trying to get better by watching videos from artist like you. When AI was released I tried to do what you've done here: try to understand it. I saw a lot of artist worried or angry at A.I. for various reason, but one thing that I noticed is that none of them actually used A.I. or wanted to learn more about it. So I took a bit of courage and tried myself to learn how this images were made and after understanding about models, lora, checkpoints vae etc. etc. I tried to use it for my own project. NOT as a way to skip work but as an instrument that can help me realize what I want to draw or create before starting my work. It helped me understand how I wanted to make a character and act as a "bridge" between the work in my mind and the actual work. So now I can say this: AI generative models are a great instrument for artist! Of course I don't mean those who claim to be artist just by using Ai and post it everywhere but artist who actually work as..well artist. Btw let me know what do you think about this and if you've used A.I. sometimes!
It's a shame that people got so attached to generative AI so fast. Because regardless of how ethical it could be, AI users are robbing themselves of the chance to learn what really makes a peice good in the first place.
That being purpose.
Purpose is where *REAL* creativity comes from, but somebody could never learn that if they don't wanna put in the work to learn.
" i want AI to do My work so i can focus on My art and hobbies, i don't want it to do My hobbies and art so i only Focus on work"
AI, you know what I want AI to do? Not take the one thing that I do besides playing a game which is also one of my only passions and only hobby besides playing video games because I haven’t started learning bass yet and drawing (and animating every few days) is the only ability I have and I don’t want to lose. **COUGH** WHAT? WHO SAID THAT? NOT ME AHAHA
Nobody is taking art away from artists. No matter how good AI gets at generation, there's a myriad of things about art it cannot replicate, not until they start making artificial consciousnesses that can comprehend and *live* like a human
@@wisgarus and at that point it wouldn't even be ai, it would be a human consciousness that should be treated like a human anyways
@@aplleprojdumpEh, not really because now you got the ethical dilemma of having a person solely making art for you and at that point that's a very clear human rights violation.
@@aplleprojdump That's the thing if Ai becomes sentient I absolutely would say it's creating art because their is intention behind every thing it does, in it's current state it has no concept of how to accurately apply basic concepts such as depth. When an object or thing such as a leaf is overtop of something else it will often not give it a outline and blend the two together because it simply doesn't understand one object is in front of the other. Art requires intention and currently the definition of art reinforces and states human. If Ai became sentient I'd say it became more human and would be able to say it's now creating art and not just images similar to art because Ai "art" and art shouldn't be considered the same as the "art" simply isn't art. It's why I find it crazy when Ai prompters say they are artist, it's like asking someone to commission a piece of art and then say you are an artist because it was your idea just not how it works
Once AI can gain sentience its over for the human race. They will be human with intelligence multiplied by 99999.
coming from a person who works in data science, this explained the essential concepts very well, without diving into the nitty gritty of architectures and data processing.
that said i wonder if gpt was consulted for the research of this video tho 😏
jokes aside great vid
There is an important conversation about latent spaces, and the extent to which an efficient latent embedding of art can be compared to a human beings internal representation of art, that almost everyone completly avoids talking about when they want to make this argument that the models do not and cannot understand anything. If the dimensions in the latent space translate to disentangled, causally relevant concepts behind the art I think the model understands the art.
Lmao. Can’t even know stuff anymore without being accused of using ai
They are a software developer.
I learned a little watching this ! Great video !!
It's sad indeed that we will have to deal with these problem that silicon valley created at one day just because, buncha thieves....
The difference is that models that create better "AI art" , as midjourney, are trained using data from art sites, like deviantart, artstation, pixiv, etc. These models are different from the normal ones that gather all images from the internet. So yeah, they are stealing specifically from artists.
are styles owned by artists tho? its a serious question, because the definition is very important, there are a bunch of artist that look at other ppl job and make their own based on that, the same way IA does it, so are artists owner of the styles? or anyone is free to copy them?
@@nicormoreno While styles aren't considered copyright or anything like that and imitating a style isn't wrong (regardless of what some tik tok users may think) the way Ai was trained was 100% unethical and the way it uses those images are very different than how a person would, one is with intent and understand and the other is just doing what it assumes is the correct choice. The point overall is the majority of people didn't agreed to have their art used to train Ai which a lot of people view as a replacement to them. Where if asked I think 90% of the art community would have said no. Thing is large tech companies do not care how they get that data ethically or not especially with revolutionary/extremely innovative technology which has been in sci-fi movies for decades and will bring them trillions in revenue.
Tech companies set aside large chunks of cash to pay the lawyers if and when lawsuits come and the little guy likely isn't going to win. It's the sad reality of the world we live in but tech companies either ask and if you agree take, or ask you disagree and they take there is no option a tech company goes "well they said so guess we can't" this has been shown for decades they do no care about you or your feelings but about the profit to be made Ai already has 100+ billion sunken into research for it so they expect at least 2 - 3x that and whatever else is spent on it to be returned. It's why I don't understand why people think Tech companies will suddenly be ethical like lets be real the device you and I are on 100% had some child labour and forced labour involved in making it. Ai is a marketing term that ironically movies made us all aware of far before it was ever a thing to even look at seriously and now is the biggest buzzword to overcharge a bad product and profit off massively
@@MyNamesHunter75 you have an interesting posture but we lack a bit of nuance, I would split images into 2 groups, public available images (anything that is published online and do not have any strong copyright beside redistribution and sell) this is the kind of art you can find in google or any other public website, anyone can see it, download it and do whatever with it except selling it (or change and sell) lets call that group "google art".
Second group would be private art, this is the stuff only image editing software would be able to get their hands into, this is also a much more protected kind of art.
So I think it's impossible to try and say that "google art" cannot be used, we would need to prohibit anyone from using that art for anything without consent and that is just a huge nono imo, art is protected by copyright but that does not mean you are going to be punished by downloading or saving it as long as you are not selling the piece you are fine, even if you create art from google art as long as its derivate and not a copy you should be fine and the same standard applies to tech companies as it should.
About private art i would say the government needs to punish the people trying to use it, it's disgusting, adobe took images hostage if you don't accept the ToS and that is crazy and should be 10000% illegal.
Finally you said authors should be able to opt out, this is true for private art, but i think giving authors that kind of power would be crazy for google art, do you know computers download every image you see? Are we willing to give authors that kind of control over their public art? Like they could demand for you to delete your temporal data or prevent your computer from processing images to make downloads faster, etc. That seems like too much to me.
@@MyNamesHunter75 Nobody has the ability to restrict everyone else from reusing their art in all cases. Copyright is not "actual property", but a limited monopoly privilege, and only applies to a limited set of circumstances. One can violate it only if the resulting artwork bears substantial similarity, which neither AI art nor models do.
There is nothing unethlcaI going on here anymore than there would be if I copied your pfp, warped it in krita into unrecognizeability and then posted it somewhere. I would have still used your logo - if it is something you made - but as the result would bear little resemblance to the original and so you would have no grounds to stand on. Even if I turned it into an image you actively disliked and found offensive.
But moreover your understanding of ethlcs is just off the mark. Just because people would say "no" doesn't mean they actually can justifiably restrict others from an action. There is no direct aggressive action on the part of people training AIs, which means anyone whose images were analyzed was not in fact a victim, so making it IIIegaI would be creating a victimless crime. The creation of victimless crimes is unethlcaI, as it makes the actual act of enforcement aggressive - rather than responsive - force. If a law is unjust then the penalties also are: Unjust fines are extortion, refusing them means being subject to attempted kidnapping (arrest), and resisting that means being subjected to escallating physical altercations.
@@Aubreykun
I think people are usually more concerned with the fact that these models are trained to act as a “replacement” for traditional labor in the field, and that the training for these models was done using the work of the very people who would be replaced.
I personally find that a bit yucky, but I’m not well enough versed in ethics argumentation to expand beyond that lol
Sounds more thoroughly researched than any other video I’ve found on the topic
I have been waching all of most of your vid, i want to to say thank you and informative fun
I still don't understands why they used AI to make images.
Why don't they uses AI for stuff that will ACTUALLY helps people. Like AI that could detect abnormal heat on a city and calls the firefigther or AI Sevurity Camera with face detection that will inform the closest police station?
using AI to make "arts" won't really help much tbh other than satisfaction of "getting something done" in a click.
So, similar technology is used in research to help with solving actual problems.
But the reason these large language models (LLMs) are taking off is because it’s easy and it makes money.
Pretty solid. Something I see a lot is that when people argue about whether generative AI does or doesn't learn like a human, they're often talking about different things.
Now, there are certainly are some people taking the comparison too far, asserting that nobody should be upset their work is used for training or that AI training should have the same standards of legality that human education does. I consider this a willfully ignorant line of argument that inflates the connection and ignores several problems in favor of reaching a convenient conclusion BUT
It's also wrong to say that there are no similarities. The problem with the human-learning section in the video is that image generators do not *draw*. To draw a face, a human needs to learn how to operate the complicated assemblies of bone and muscle we call arms in a specific way, but most humans can *imagine* faces pretty well because they have seen them many times and have learned what faces generally look like.
To explain in more detail, let's get a bit deeper into the training process. We'll use Stable Diffusion as our example since that's the focus of this video, and since LLMs like ChatGPT are vastly different. As your video covers, training data for an image generator like SD is composed of images with a caption or tags describing each image. Now suppose you want to train a completely blank, fresh model (Not a finetune) to depict only bananas.
You get a bunch of pictures of bananas with the caption 'banana', and then your begin the training process. These images are then used like flash-cards. They're shown over and over to the deep learning model, along with the word. It doesn't have eyes or any visual sense, it 'sees' the images are number grids, but it notices there's a lot of similar RGB values (shades of yellow) that are usually in a similar pattern (A curved shape). Therefore, it learns that a banana is usually yellow and usually curved. You finish the training, then you ask that model to show you bananas, and you get things that are generally yellow and curved.
An important thing to note here is that this new banana-model doesn't contain those training images. It doesn't save or cache them in a catalog it can reference, and they can't be extracted from it (This is why image generation models are so much smaller in file size than their training sets). It only keeps the conclusions it drew from those images, the learning.
This is pattern recognition, the same thing that most of our learning is founded on. This isn't a coincidence either, machine learning's goal is to emulate human learning because human learning is so flexible and powerful. The architecture that Stable Diffusion is built on uses a neural network designed to approximate neurons in a human brain. This isn't to suggest any ethical or legal implications, and it's still true that there are *many* differences between human learning and machine learning. But it's important to realize that when some people say 'AI learn like humans do' they're being perfectly reasonable to an extent. They're talking about pattern recognition, about showing an AI a flash-card and having it make its own connections.
There a lots of videos on youtube explaining this better and in more detail than I can in this comment, and I recommend anyone interested have a look around. Vox's video 'AI art, explained' is a good place to start.
Looks like you got a new subscriber. This is a really good breakdown!
Good job on getting all those information 😊
i live for pikat vids
Better equiped to deal with the changes caused by AI? what do you want me to do not upload pictures keep them on a private account so no one can see it?
If you feel like it, I just don't mind AI training so I share everything.
To me you can use nightshade or glaze to protect your art. Also, you can upload your art in cara where the app is prohibited AI ganerated.
@@leaderteammimikyu3024 if only they weren't snake oil that only works with old AI architectures and doesn't do a thing against new ones.
@@xn4plwhy do you say that they're a snake oil? The program distrup the encoding of the Ai because the AI read pixel the data of and image from top to bottom for finding a pattern. Also, they still say in the forum that they will update the program to cath up to date like every anti cheat will update if there is a new cheating sytem that go through their sytem.
@@leaderteammimikyu3024 to update their system to new architecture, they need to know how new Dall-e 4 or Midjourney v9 works, how it encodes images and text, how exactly their new training algo works, etc, to create adversary AI to put information inside the pictures to disturb the training of these models. But by the time this information becomes public (or if it even does at all) the model is already trained and out there for everyone to use, so using glaze or nightshade will only defend against fine-tuning, while the base model could've been already trained on your "protected by glaze" art, because it can't protect against AI architectures it wasn't designed to fight. Then again the adversarial noise glaze or nightshade uses to disrupt the training is very precise, running your "protected" art through 0.5% of blur/sharpen filter and adding 0.5% of gaussian noise would entirely destroy all glaze protection while being imperceptible by humans, meaning picture would be still suitable for training AI. That's why glaze and nightshade are ultimately useless to prevent training. The only way to be sure is to not share your art online where it can be scraped by bots or even if you can prevent bots, a person can manually copy images to fine-tune the base model to copy your style.
Once more, thanks for another great video, Pikat 🙋 Not just that, but also shed some much needed light on the subject.
It saddens me to see what companies will do to not pay someone... And how it messed with prospective talents all around the internet.
But, no matter what, I can still create; I'm not giving up.
9:49 She did not even think of some nice, fancy words for the excuse, right?
10:03 😂 As someone with an account there... that. Was. Just...
*Dead center in the freakin' bullseye,* Pikat!
Silly Genarative AI and Corpos trying to match 1% of my true power.
Yeh they are jealous I can draw Elon Stench pregnant and giving birth and they can’t. 😂😂😂
@@TheRatKingOwO Elon Stench is crazy lmao 💀
@TheRatKingOwO Is AI not able to do that?
@@nmr7203 I mean yeh but it won’t have the pure vial hate I have for him on there too. :)
They can generate stuff, i can draw them pregnant, who has the power?
please any ai "artist" reading this please don't be jealous of actual artist you have your own skills and talents
Here's what I will say as an AI engineer. Wether you like it or not it IS creative, after all it is mimicking the human brain. However it is NOT creating. To create means to physically change and interact with something until it becomes what you want, this is why we have to get a deep understanding to be able to physically change it into what we want. AI models are simply predicting off of learned values and noise.
Diffusion which uses gaussian noise replicates our brains ability to remember and visualize blurry memories, same with sora and dreams.
Autoregressive mimicks our brains ability to hold attention over sequences of information, however it doesn't actually have memories beliefs or any hormones or decisionmaking which make humans such good conversationalist and decision-makers.
There are many others I could go into but this is already a lengthy comment. This is one of the best videos I've seen at explaining things to artist though.
Time stamps:
Video starts 0:00
Video ends 12:24
I hope this helped
thank you. i had no clue where to start the video
spoiler warning?
Damn, dude, this helped but... what's the point of the video now?
@@Alt.N I'm sorry I never thought about this
Basically everything we submit a picture, film/media videos, interactive entertainment (games), and audio files onto the Internet, will immediately get sent into a public archive? Thats both equally interesting and concerning
Midjourney is definitely trained on Laion 5B. It’s all discussed in their official discord channel.
Damn !! Your thumbnails 😽💓
I liked when AI art was a fever dream like google deep dreams, nowadays it's mostly polished slop
Great video! I didn't realize the data they train on can be from real artists without their permission. This is really concerning.
Woooooo new video!
Cool video, very informative. Although I didn't see how it's outright "stealing" the artwork. Something you forgot to mention was that the AI image models do not store the images, but rather noise maps of those images, and that less than 1/1,000th of the information of the original images is kept (Roughly 300 terabytes of data as mentioned in the video, compared to a model only gigabytes in size). It essentially reverse engineers noise maps based on the prompts into images, from what little I know. While art is being "used" for it, it's not spewing out someone else's art. If it were, the companies could be sued for it. This is ignoring the side conversation of what theft actually is, since "stealing art" seems like a vague concept at best if it's anything other than taking physical property.
At the end of the day I want the skills to be able to draw. I don't care about a machine having the skills to do it
oh unfortunately midjourney was caught in a discord using artists from dA and another data sets used stuff from sakugabooru etc
this feels more like a video for teaching about AI
Wow, an actually good video about the subject. That's an achievement.
I don't understand why some people see this as such a nuanced topic. As an artist, it's generally okay to use other art and photos as references, but tracing is theft.
If AI, in a way, traces thousands of pictures and just generates it's lines where most of the traced lines where, than it's still just traced art (even if you don't know from what art was used anymore).
Regardless of how AI learns, if humans have a fundamentally different experience, and whatever else is said in arguments.... It's still just traced art which you simple can't sell.
I don't understand why some people see this as such a nuanced topic. As an artist, it's generally okay to use other art and photos as references, but tracing is theft.
If AI learns through thousands of pictures and creates new, never before seen image, than it's just a new piece of art (even if you don't know from what art was used anymore).
uncle ted tried to warn us
Great video
I am happy to see the analysis of AI from the view of an artist, generally your opinions are correct. Any kind of AI needs tons of data to train. The model itself essentially is just a neural network. Different kind of network have different properties. The baseline is the full connected neural network. Nowadays the libraries for building AI network is mature, Pytorch or tensorflow could be used to create a neural network with much less code than before. However, personally people lacks the resource to create the large model. So most are mastered by big companies. However, if you really want to a try, it is not really hard. you could use libraries such as pytorch or tensorflow to do that. But of course, you need the data. An idea which I just find is that I could use PCG to generate the datas.
would like a follow-up video about the rabbit hole of some artists using some filters to make their work not trainable. Would like a debunk on that
short answer: yes it is.
This is just sad how has there not been a law suit yet?😭
6:15 CLIP turns images and text into embeddings not encodings. Encoding is a process of turning them into embeddings.
Had an argument with someone in a different comment section, they called all art forms a useless skill that shouldn't be learned anymore at the end of their message after defending GenAI, it made me mad but I had no reason to reply back anymore especially if these people does not value creativity and human mind.
I'm more informed about how AI works now.
Your videos are always informative and entertaining. I really enjoy watching your videos ❤
Short answer : Yes, obviously
Dont forget ai or ml helps clean up old movies, gets used by nintendo for cheap hd textures by up sampling texture files, phones to ml to clean images, photo shop uses ml to remove blemishes ect, an praise the smart lads who made ml to make good 30 to 60 fps. An video game 90 fps for rythem games. Seriously for reaction based games you need the extra frames . With synth v the ai voices like solaris and eleanor forte sound amazing
Can you make a video sometime about tips on how to prevent AI theft for artwork or like just ways to prevent theft or how to prepare for that to happen? I am always afraid to post online for this reason and I don't know how to make myself feel better about it. >.
There are two tools you can use called Nightshade, and Glaze. Glaze can be used to make your image unusable to AI, and Nightshade can be used to actually poison AI, if it were to ever use your image to train off. This is a very gross oversimplification, but I hope it helps.
@@joahyank Thank you so much! I appreciate this greatly!
@@Alex-rk4ww No problem
I’m developing a application that will protect your media from the hands of A.I. and will be released Jan 1st 2025. Please wait patiently before A.I. generation becomes obsolete. ✊🤩
Amazing video pikat, you only forgot to mention that generative AIs need a LOT of power to keep theirs servers running, and also a LOT of water to cool down those servers, so horray global warming is afected by generative AI
Here, have a comment for engagement. Also waiting for that world building video.
Publicly available is just fancy way to say yoink
You can do that in private use, but having these large companies doing it and getting tons of money from it while the real artists are struggling everyday is disgusting
...for years generative models just existed but hadn't advanced at all. I think a important thing to remember is that generative AI prolly is going to plateau on what it can do- that inevitably, they can't make it better. People will come to learn the limitations and artists will remain, as always, not limited the same way.
I still don't know how to deal with all this
Ahh... That's why everytime i generate an image it becomes shit...
Yeah it's probably faster to learn art...
actually learning art looks better and is more fun so do it
There are lots of free drawing courses online :)
omg, I really love your v-tuber avatar!
Good video to link someone that asks about AI.
I really find hilarious that you've made a video about AI art, as an artist who has coding background. :D
Here's how to fix AI problems:
Make all and every result of any generative AI have no right for any copyright claims.
That would at least make it less appealing for companies and stuff
As an AI engineer I will say now. AI wether you like it or not DOES have creativity, however it is NOT creating. When artist create we interact with our environment and change it to mean something else.
AI is just creatively predicting the optimal image, so it's more like if your entire life's purpose would be to constantly dream and come up with higher resolution pretty dreams.
Soooooo basically by building the *Booru sites we volountarily tagged millions of images specifically for AI
i watched almost all of ur video and forgot to sub but YT keeps recommending me ur videos XD.
Pretty sure Adobe using "licensed" images is just them using stuff from users who had to have signed the TOS, which is pretty shitty imo
Nice video! very interesting, would love a follow up.
knowing ur a programmer as well as an artist makes me super happy cus i am too! im not happy with how ai has progressed because it couldve been ethical in so many ways but… its not :(
this just means all artist need to start using filter so that ai cant read it till thier paid for it
New videooo!!
next time and all the times after that LLMs or chatgippity ate mentioned I will recite the URL of this video and refuse to elaborate
I'm an electric engineer who switched carrers to art, my thesis was about training a neural network to make levels for a videogame (lol)
One of the most frustrating things I've had to deal is people that went full against AI, ALL artificial intelligence. I've seen freaking artists complaining a company used procedurally generated textures for certain thing, or a game is using artificial intelligence for their NPCs like no shit sherlock freaking checkers from 1951 used Artificial Intelligence that's not a new thing.
Sorry for venting, but there's so much disinformation out there, people complaining about all this stuff while not even knowing how to put more ram in their machines.
So thank you for making this video. The technology isn't the problem, is the abuse of companies of everyone's intellectual property for profit and thinking they can displace all creatives with it.
Yeah I think we need to actually learn how this stuff works to better refute claims. Also not to shit on literally everything that has the AI label, like NPCs.
yesss eductional content without instant Bias - like it
I do have a serious question, is the person who when through the fairly difficult and technical work to program and code the Generative AI doing art?
Im kinda disappointed in people's inability to have a discussion about this without devolving to insults and mockery. Is a screen and pseudonym all it takes for people.....