Is AI Art Theft?
Вставка
- Опубліковано 1 чер 2024
- In this light-hearted and completely uncontroversial video, I express my opinions about the way #aiart companies like #midjourney and #stablediffusion collect data from artists to make their art generators. I also bash NFTs and Beeple because it is fun.
Tldr; I think the way it is done is wrong and they should ask permission to use artist's work. Feel free to express your opinion about this, but remember that my opinion is the only objectively true one, and fully authoritative. All who disagree with me shall meet my furious judgment. Orangutan Good.
Links to my stuff:
Patreon: / emergentgarden
Discord invite: / discord
Twitter: / max_romana
Sources:
The Orangutan Artist: www.fiverr.com/zlemuuz
Steve Zapatas vid: • The End of Art: An Arg...
Laion non-profit: laion.ai/about/
Stability: stability.ai/
PlaygroundAi: playgroundai.com/
Laion funding: techcrunch.com/2022/08/12/a-s...
Open AI "capped-profit": openai.com/blog/openai-lp/
Greg Rutkowski on Ai Art: www.technologyreview.com/2022...
Midjourney making copyrighted works: / 1588915427018559490
Some useful info about copyright/fair use: ogc.harvard.edu/pages/copyrig...
Timestamps:
(0:00) Intro
(1:52) The Ethical Problem
(4:16) Automation and Tools
(6:46) Data Collection
(10:14) Art Theft
(12:02) Ethical Data and Opt-in vs Opt-out
(14:12) Inspiration and Imitation
(15:43) Fair Use
(18:08) Where to draw the line
(21:08) What can you do?
(24:34) Some reasons to hope
Music:
• 10 hours of silence in... - Наука та технологія
To clarify a few points:
1. I am NOT claiming these companies have done anything illegal (7:53). I AM claiming what they do is unethical, and the two are not the same thing.
2. I love AI and want it to improve. I also want it to be done ethically, and ethics does sometimes require you back up and do things the hard way. Reckless progress at all costs is guaranteed to be catastrophic.
3. I DO love orangutans. I will NOT waver on this point.
Do you think that making an entirely ethical model can ever be ethical at least by the standards that we want to uphold? I have trouble seeing large models able to replace artistic images (which I think is what most people want to make/protect) without the help of pre-existing labeled images. Sure, maybe we will someday have an incredibly advanced model capable of generating digital and traditional art while only being trained on photos of nature. I think everyone can agree that a model like that would be completely unproblematic, the issue is that unless these are somehow more enticing in terms of its capabilities (better results) and accessibility (low compute cost and training data required) then it's gonna be really hard to convince people to not use it, especially locally. But yes, I see that you are talking about the consequences of commercialization and not so much the personal use aspect here. I still agree with your overall sentiment to "vote with your wallet" since that is how I fight against many other exploitative companies as well.
@@Frostyflytrap "I have trouble seeing large models able to replace artistic images (which I think is what most people want to make/protect)"
I reckon the main issue folks have is that the art is taken without permission. If I recall the stephan zapata video (references in this video) correctly, he says he himself would probably opt-in to datasets, just that artists should have the right to choose.
It's also brought up in the other video that stability's other AI for music (dance diffusion) is trained only on copyright-free and voluntarily provided music. If stable diffusion had done the same, there probably would have been significantly less fuss. Problem is, there's a lot less of those, and they tend to be of limited quality.
What I reckon would have been a good call would have been to train the AI using the LAION dataset, as they did, then try reproduce the methodology with a dataset from volunteers or contracted artists, using the existing tech as promotion material to draw them in. Alas, in fairness, all much easier said than done I suppose.
Your entire video is based on the premise that "scraping" is wrong, but scraping is literally just the AI looking at the image the same way that a human does. The only difference is that the process of learning has been programattically rather than biologically defined, and now it's causing an uproar because the machines/programs can do it better and more efficiently than we can. It's an advance and yes, it will result in job losses, etc., but your claim that it's unethical is wrong from the outset. Nothing was "taken" from these artists unless you also agree that you are stealing from artist by looking at their art in the first place. And even if that's the case, they shouldn't have put it somewhere where it can be seen, if they are uncomfortable with other entities (human or robotic) being able to see and learn from them.
@@DavidHagar-mw8vn Exactly. You hit the nail on the head. Although it's not really exactly the same way like we humans look at images. Scraping has been a legal issue for a long time, even before AI Art was a thing. But you are absolutely correct to say that "saving" a piece of art in order to create more unique, creative images isn't "taking" anything from these artists. Nobody steals their revenue, nobody claims said art as theirs, etc The only *valid* complaint one could make is that the AI uses all the hard work and sweat put in by artists to create new, novel and unique art without any much effort from the person who just typed in what he wants to generate. From *that* specific perspective it might be correct to say that it is questionable to shamelessly build on the backs of all those who took a lot of time and effort into creating their art while the average AI art user just generates an image by typing into the input a few words of what he wants to generate.
@@Kaje_ Regarding artists having their hard work lessened by people who can more or less punch in a few keypresses, I'd say it's helpful to imagine another scenario. What if we allowed cloned humans with the same *physical* capabilities as normal humans, but with AI brains, to fight in martial arts tournaments? Would it be "cheating" to allow those AI to watch videos of kung fu masters and learn programmatically from observation? I think we could pretty comfortably say that it would be ridiculous to make such a claim, and furthermore, that it really wouldn't matter in a sense whether AI ended up winning the martial arts tournament by observing human masters or by simply figuring out its own way to win. The part that would really matter, and which would probably cause an outcry, would be that it won.
I think this scenario is going to play out across multiple industries. Music or video are likely to be next, followed by programming.
For me the most compelling argument was: if a human is inspired by/imitates an artist, it is likely to be good for the artist (their work gets more attention and recognition) but if an AI model does it is bad for the artist (jobs will be lost and individual artists are unlikely to gain exposure or recognition)
For me it is always how perfect is computer memory compared to human's. Yes, both will look at the same picture and will memorize it, but we just can't recreate 1:1 copy of what we saw, can't remember everything to the smallest detail. So when we use something from others it is definitely just an inspiration that fuels our creativity and imagination.
Counter argument, the jobs which remain for artists will require more skills and have better working conditions (less grunt work, more artistic choice/ managing the AI) which in turn results in higher pay and better quality of life.
@@ligametis AI does not generate 1:1 copies of what it was trained on. Where does this notion come from? It’s not even technically possible for it to remember things “perfectly”. The size of the latest Stable Diffusion model is 5 gigabytes, whereas the training dataset is hundreds of terabytes.
@@boshi9 there were test when database is more limited, or AI is asked for something very well known, "afghan girl with green eyes".
Either way trainings AI means taking someone's work and using it for your own product without asking permission. For training a true 1:1 capy has been made. That is intellectual property theft. AI couldn't work without true artists.
@@ligametis Sure, if you create an artificial test on with a tiny dataset, you're going to get whatever result you're looking for. Just like a human can have pretty good memory if he/she only has to remember a small number of things.
As for the second part: this is not in any way different from human artists. All art is derivative.
I abolutely agree with the "is for freee?" sentiment. It's the monetisation that's got me scratching my head the most. In an ideal world AI art would be something for artists, by artists. There's something to say for a system where contributing your art gives you access to the entire software, which would be a pretty good deal. But by the magic of the free market the artists get to pay for their own work (???). I think the logical development by programmers, most of whom are not artist, definitely plays a role as well.
What's related in this discussion that I find really interesting the comparison to code. Exactly the same thing as with art is happening with programming. Copyrighted code/algorithms are being scraped all across the board and reused in AI bots. But the sentiment is much less negative. Why is this?
I think people are hung up on the fact that "a computer" did it, disregarding that a human made the program (and the computer). Even if you excluded "your own work" from the data set, you'd still be getting access to images extrapolated from an enormous amount of publicly viewable images. A human could go out, save all the images, maybe print an inspiration board, and then make unique images from them. A human could also hand carry a letter from NYC to LA, but a tool makes it a lot faster.
Like you hinted, all industries imitate or copy from inside and are increasingly automated. We can 3D print concrete houses with an AI created floorplan. Robots can cook, assemble, and deliver a hamburger. People should stop fighting the technological progress and figure out how they are going to adapt the new technology to serve their purpose, or get left behind.
@@alkaliwreck2474 the problems, most people will be left behind. Normal people can't win against machines that are constantly running and improving 24/7
@@kiattim2100 So don't beat the machine; use it. Lumberjacks with hand saws didn't go "whelp, guess I'm out of a job," when the chain saw came out. It made them more efficient, more prolific, at what they do.
@@alkaliwreck2474 Does those millions of handsaws work by themselves, constantly improving themselves 24/7/365 and is managed by a few people, trained by millions of people's data and feedback in a short time span?
Obviously no, what a terrible comparison. literally comparing apple to orange.
@@kiattim2100 Does an image bot cut down the rain forest at an alarming rate? The "AI" doesn't prompt itself either. It's not even a real AI, it's a machine learning algorithm; it does nothing if no one uses it.
Metaphorically comparing a tangerine to a clementine.
the weird thing is that we didnt automate the bullshit jobs.. we automate creative work.. AI should replace bureaucracy jobs.. not creative jobs..
Oh don’t worry, it is
Whelp, i guess in your world industrial revolution didn't happened
Wikimedia Commons has a lot of pictures... Like a lot a lot. All pictures are at least licensed under CC, some PD. It would be very non-controversial to use them to train an AI model. Many of them even come with annotations in the form of RDF.
Sadly there is a huge fear mongering going on and the point about copyright and the „need“ for artists permissions is a non issue and in large based on false understandings of the subject matter. It is the same discussion as: „Is using references cheating?“ or „Is digital Art inferior to traditional art?“ or „can somebody steal an art style?“, „is photoshop cheating?“ and to be honest - after years and years of pretentious elitism in our art community and more than stupid discussions about brainless trash, it is rather annoying to have this new outrage in our community.
For me, AI art plainly means hope. Due to my multiple sclerosis, I lost my ability to work as an art director in the advertisement field and two partly paralyzed hands make it hard to make art. Not only on a physical level, but I only have minutes a day, that I really can concentrate and focus, making it rather time intensive, to create even simple works. In the last few month, AI-Systems helped me so much and while it is true, that those systems can create whole pictures, they are just phantastic in helping with parts of images. Things I could draw or paint myself, but that would take stupidly long times to finish. It takes me several days to finish a piece but that is much better than taking weeks or month and I can tell the stories I want to tell. AI systems and tools are hope for everybody who is challanged in producing the art they wish - no matter the challange. It uses inspirations from all art on the public internet, just like every artist does, but can bring it to a point, the artist in front of the system needs it to be. Just like when creating mood boards and reference sheets and then use those to paint. As our community established over the past years: that is not theft and it never will be. You want to create art that nobody references? Keep it to yourself and it is save. This worked for millenia on this planet. Everytime people publish art, there will be copycats, there will be works that looks kinda similar and there will be unique spin-offs. AI-Art is scary and new and we all watched Terminator and know what AI is planning… but seriously people: Get a grip!
it's really funny the idea that AI art isn't art because "there is no human expression and bla bla", like wtf i'm an artist and an ai hater but they would recognize how stupid they're sounding if they read about phenomenology of art and of aesthetics, that's serious art philosophy saying they're spitting nonsense
Reminds me of the evolution experiment where they setup a bunch of computers in a museum, and "breed" procedural artworks that people look at the most. Would that not count as art?
i mean. per definition, art must be done by a living thing, otherwise the term does not apply.
That oragutan changed my life, i think about it every minute of my day now.
I appreciate the honest and nuanced discussion, I'm not exactly sure where I personally lie on how much I agree or disagree with the points but I mostly see everybody's thoughts for what they are and I don't personally have a huge stake either way.
I think a part of the discussion that you missed (which I don't blame you for, the debate grows exponential in scope every time I learn a new argument about it) and that is of the consequences of the free to use tools. I 100% agree with the commercial angle, I think it is common sense to not claim to be another artist and I also think it is ridiculous to pay companies to generate these for you for many reasons, but mostly due to the fact that anyone with a powerful GPU locally on their computer or on Google Colab.
Artists that depend on commissions will definitely be hit by this, and not even because of the "art theft" issue, but mostly to do with the potential market and demand. I don't think this is that loud of an argument anymore for some reason since, to me, it feels like we're looking for moral justifications because we've realized it's already too late to stop it? (I'm making a big assumption here, I don't fully know what is going on or if I'm totally off the mark.)
You touched upon it a bit in the video, but I think this whole thing is kinda like digital piracy and a lot of the arguments are very similar to that other old and still controversial topic. There's stuff like video game emulation, streaming service password sharing, and listening to music (specifically video game OST) freely uploaded online, which a lot of people will have varying reasons and justifications in doing, not that I'm claiming that these are all equally in morality.The main reason digital piracy is fought against is that it disincentivizes paying for the product itself, but a counterargument to that is that many of these people who engage in digital piracy couldn't afford to pay in the first place and thus were never going to be a customer in the first place. Many people who live in developing countries get their digital media through piracy because of the unreasonable cost it would take in their conditions. AI will act the same way since it is simply faster and less risky to use image generators than it is to pay freelancers you can't be confident or trust yet if they will deliver.
As someone who experiments with image generation a lot for my personal enjoyment and inspirations for designs and to help with visualization as an artist, I think there are good and bad ways to go about this. The bad ways are to be dishonest about what you put out there and how you claim to produce them, especially when you then accuse an honest artist for plagiarizing you after you did img2img on their work. I'm still a bit iffy on whether being honest automatically makes it ok to earn money with, like is being a prompt crafter a skill worth getting paid for? How about computational time and power costs? What if you input your own drawings and visually enhance them using img2img, and then retouch to the AI output? How much do you have to retouch an output image until it's considered ok to monetize? This is a nice shortcut for many amateur artists who only need to use photoshop to adjust the output to fit their creative vision, but then is the fact that an unethical tool was used in the process now suddenly makes the final piece exploitative no matter how many adjustments you add to it?
I don't have any answers to these questions that work for everyone, and I doubt anyone will for quite some time. There are a lot of honest artists who are accused of using AI because their work looks too good, and I think that's pretty tragic. I've been keeping a close eye on what these generators mostly output that I can do a pretty good job in recognizing whether something's generated with DALL-E2 Stable Diffusion, Midjourney, NovelAI, and several other models since they all have their signature "style" indicative of it being generated. (Except Nijijourney though, that is legitimately the only one I could not tell from a glance is AI since it is just too varied in its results for me to train my brain to recognize any patterns.)
In the end, I think it's irrelevant to reality whether any of this is ethical or not since it's going to happen anyway, the best we can all do is to prepare and be aware of the tech to not get scammed and for us to collectively change our standards for art. A lot of image generators look pretty impressive at a glance, but they all lack attention to detail and cohesion which make the raw outputs unfit to be used professionally. I understand that the reality is that most people don't have an artistic eye to even notice or care when the rest of the image is instant eye candy. I don't know how far artists can go without at some point using the tech to their advantage due to personal misgivings with the tech, I think a lot of well-known skilled professionals will be mostly unharmed since there will always be a demand for their work, though the artists who have had difficulty honing their creative skills and depend on having the commissions will be forced to adapt to the changing landscape since you can't control whether your customers should pay you instead of running a personal image generator. You can't stop people from pirating movies or game roms, but you can depend on the honest and financially able to support your work.
I've been following AI image generation ever since its early days with Disco Diffusion and that has made a lot of these advances less of a surprise to me since I've been thinking a lot about this for a long time now. But the sudden skyrocketing progress all taking place within 2022 alone has been astonishing. I can't help but wonder if we are still at the beginning or if progress is starting to plateau while society figures out how to adapt. We only have 2 weeks left of 2022 but I wouldn't be surprised if a few more surprises like ChatGPT come right at the end. What a time to be alive… I've seen a rudimentary audio version of Stable Diffusion, and there has also been steady progress in voice changers/synthesizers and deep fakes, and I don't think the controversy will only start again once those become easily usable.
One thing's for sure, we will not be able to predict what the modern world will look like in 10 years, let alone 5. There will be a turning point when most content online will be completely synthesized and true things online will become harder and harder to distinguish from fiction, the debate with AI art will look so small in comparison. This is what I think we should be most worried about in the near future and I don't see enough people preparing for. Forgive the philosophical speculation here at the end, but I really feel like we are still in the middle a paradigm shift unfolding before our eyes. What a time to be alive...
"You touched upon it a bit in the video, but I think this whole thing is kinda like digital piracy and a lot of the arguments are very similar to that other old and still controversial topic".
It's not though. The legality of data scraping has been upheld multiple times by the courts. If you put it online, it can be used. Copyright is violated if you try to monetize it, but copyright is way more limited than most people think. The courts have been very clear. f you don't want it seen and used, don't put it on a public site.
We can have the ethics conversation til the end of time, but I wonder how much money most of these artists are actually making from their work in the first place. How many of these people are actually making a living with their work, and how many are screaming because they have a dream of being a professional artist and now they don't?
Look, I'm all about seeing artists getting paid when their work is legitimately used, but as someone who works in both the design industry and in AI research, I'd be shocked if the number of professional artists that AI was actually putting out of business was in 6 digits.
@@housellama Ah, I had a feeling that's how it was, I don't really know much about the legality considering how many countries deal with it differently.
Either way, I still think it's similar to digital piracy in how people will continue to perceive it rather than how it really is legally. Albeit a nebulous conceptual way than what is even actionable since most generated images are transformative enough to be legal.
I forgot to mention it, but even if you use an artist’s names in the prompt, the results never quite match the original artist's style, unless people specifically train a model on their works in particular. The legality of making using Dreambooth on specific art styles is getting closer to the line many think shouldn't be crossed (I agree if they only mean commercially and not for personal uses) but even still, the output still looks like the most well-decorated cakes but are still fundamentally misshapen in composition.
Even if certain countries make AI art illegal, it's certainly not going to stop the rest of the world to view things the same way, a lot of people obviously find joy in using the tool for themselves harmlessly so I don't see it going anywhere, nor do I think it's "taking away potential customers" because yeah, probably couldn't and wouldn't pay you anyway and even if they could, you can't work faster than a machine.
I agree that this isn't really going to affect people who are already established and a huge amount of the anti-AI art sentiment comes from those who dream of potential success, they see a faster runner and claim that it's making them reach the finish line slower. The success of others doesn't automatically mean the failure of one's self, or even a sabotage of those who are trying. A lot of egos are getting hit because the established trope of AI often depicted as not being capable of creativity is being challenged.
I think we should definitely keep talking about it though and have the tools become a part of the greater discussion. Yes, we should frown upon fraudulent uses of the tool because not only is it wrong but also because it paints everyone who uses the tool innocently in a bad light if we turn a bad eye to the bad actors. I also think that people are giving the capabilities of the tool way too much credit since a lot of the outputs are incredibly generic and easy to spot due to people using it with the default settings.
I'm incredibly critical of visual quality and I think people should raise their standards if they look at these results and say that they're worth paying for. Img2img is still the best functionality of the tool that gives the most control to artists and, but even still it's kind of a hack to get to work efficiently in established workflows since you're constantly trying to alter things that would normally be corrected in step 1 for a human artist, but instead you're trying to alter the recipe of an already cooked meal. Fundamental design skills are still necessary in making unique and flawless results, and if your livelihood as an artist is supported mainly by customers who have low standards to begin with that they turn a blind eye to minor dips in quality, then good luck because you were never financially stable to begin with.
@@Frostyflytrap I've spent a LOT of time with this.
My partner is a design professional and as he said in a phone call the other day, if he, as a business professional, hires an artist to do the work, it's not the work he's paying for, it's the fact that some human did it. The quality is wholly secondary. It's the fact that some craftsman somewhere took the time to hand-paint that icon. He doesn't care how good the icon is. He cares that he can put on the website "This icon was lovingly crafted by Guseppe from Milan." It's not about the icon, it's about fashion.
But business is business and nobody has the money to hire Guseppe to hand paint the icons. They barely have money to buy them in the thousands off some stock art site, much less hire some artist to design them. If they can get some unpaid intern to type a few lines into a computer and spit out 200 icons for free, that's exactly what they are going to do. That's $12 they saved from paying some stock art site.
That's the dirty secret, see? The professional artists aren't the ones drawing the gorgeous paintings that you see plastered everywhere that, let's be frank, no one actually needs. They are the ones grinding out icons and buttons and boring shit that make up everything we interact with. And those artists (who are actually called designers and are the people I work with most of the time) should know enough to see that AI isn't just a good thing for them, it's an absolute godsend that they should grab hold of and never let go.
Now, the argument is that "art is what makes us human". We need art to be more than just machines clicking away at soul destroying enterprise software. And I agree. But the myth of the professional artist is like the myth of the professional musician, but worse. Bands actually make money, and there are a lot of them. How many modern professional artists have you actually heard of?
@@housellama "How many modern professional artists have you actually heard of?"
I mean, a lot actually? I think there's often a higher bar for it, but professional artists, outside of just logo work & graphic designers, do things like concept art for movies, shows, games, animation (2D concept art is used both in 2D and 3D animation). Then you've got things tangential, like toy product design or fashion design which still use illustration and I wouldn't call boring by any means, and things that are more popular in some south-east/east asian countries like commissioned mascot art. Going back to games/film/animation, you have stuff like splash art, and advertising art. You have people who make picture books, or make pictures for books (ie TTRPG books; also digital TTRPG resources) and people who do art for card games, trading cards, board games. And of course there are people who just do a lot of commission work or have a patreon model (this is usually less financially stable for smaller artists but every job has its entry requirements ig; being said, I do know a guy at my uni who, in their first year, made 10k in a few months off of art commissions, which is not nothing). All this is just specifically 2D illustrators - not mentioning animators, VFX artists and 3D artists, etc, but I digress.
I'm sure I'm missing some but, basically just looking at artstation, it's not hard to find many employed professional artists. Biggest hitter is concept art imo, but point is, it's a much larger field than just 'icons and buttons', though tbf I'd understand where that perception might come from.
@@HiHi-iu8gf I'm not a visual designer. Quite the opposite, my brain does not do pictures. I think in concepts and other things instead, which makes me a great algorithm designer but a HORRIBLE visual designer. My partner at the AI lab IS a visual designer, and has worked in the design industry for years. Most of the stuff I know about art in the industry comes from him. He says that the business doesn't use that stuff and I believe him. He's had many, many years of experience to back it up.
What I have DIRECT experience with and can talk knowledgeably on is how these AIs work, and what I can tell you is that anyone who makes sequential art (not just concept art, but stuff that requires continuity like animation, tv, comics/graphic novels, etc, isn't going to be affected by this for a little while longer. Current AI models don't have statefulness, which means that anything that requires consistency is out of their reach. Furthermore, making them consistent is going to require some SERIOUS redesign. It's not going to be as simple as going from v1 to v3. They are going to have to fundamentally change how the models work to make that possible.
Which means it is SPECIFICALLY one-shot, 2d illustrators that are affected (although 3d generative art is already a thing and is reaching the point where it is commercially viable FAST). This means that it really is primarily those commissioned artists that are affected. Which is why I have been heavily advocating for them to learn to use AI in their process rather than fighting it. Because it is literally THAT GROUP alone that is affected by this currently. It's either get on board or get run over.
I think the main thing is AI can't do "specifics" well unless trained on doing that "specific" thing. Unless you are an artist one would not be able to describe a scene well enough and the details fine enough to actually create the exact image one wants. Sure AI can get close and often for Applied Art that is all you really need. AI makes stock photos really really well but if you wanted a picture of yourself eating an ice-cream cone on a beach you'll be hard pressed to get any AI to do that (that wasn't trained on your image). but a skilled artist? that'd be easy (well not easy it'll still take hours but you get my point)
This is similar to a camera... it allowed anyone to push a button and create art and do it in mass! one can even take a photo of an art peace that someone spent weeks creating (an architect's building for example) and produce it for anyone to see. cameras did not destroy the art world, even if people at the time thought it might, in fact it created a whole new set of artists... photographers. I see AI the same way. sure some will loose their job but people loose their jobs all the time from automation (I personally feel that the need for everyone to have a job in a world where automation has taken them all is a whole different problem and can of worms I won't get to here). At the same time however artists will always exist and their will always be a need for them.
AI won't take your jobs. People who learn to use AI will take your jobs.
(note I will not argue Ethics as those are different on a per person basis as well as on a group society aspect. nothing is good or evil, the world is a blend of everything)
I think, with all communitys that AI will impact, there will be a certain movement we already saw with open source software. Some gonna be open, some gonna be closed; either is fine. An like with open source, there are gonna be options to be payed for your service. There are enough well established open source companies.
As an artist, I only make my art unique to me.
Which means your dataset might be too low for AI to compete against(like with Blackthornprod)
@@revimfadli4666 I believe having a unique art style sets me apart from other artists. While algorithms and AI can generate an impressive range of artwork, they cannot replicate the diversity of ideas, and individual styles humans can create. As an artist, I constantly challenge myself to develop new and innovative ways to express my creativity. Every artwork I make is something completely new, never seen before. Furthermore, I strive to explore different techniques and materials when creating my art, which further distinguishes it from anything else. With every creation, I attempt to push the boundaries of what is possible and bring something fresh into the world.
I'd love to see a part 2 about code and things like Copilot
Copilot is awesome and engineers who use it overwhelmingly love it. I'm a software engineer, I've contributed a lot to open source in the past 10 years, and I also love it.
@@AnnasVirtual The majority of engineers who know about the lawsuit don't really care about it and don't think it matters, win or lose. Most are skeptical there's even an actual case. As an open source maintainer myself, I don't really care. If Microsoft trained on my code without my consent to build a tool I love to use, great! To me, that is the true spirit of open source. While a tiny group may not see it that way, so many more people do.
Joel, i would agree with everything you are saying were it not for the fact that the tool costs money. Taking people’s skill and selling it back to them? That hardly seems like the spirit of open source.
@@PokeNebula I don't see it as taking skill nor selling it back, though. I'm not sure if you've used it or not, but after a couple months of using it, I can confidently say it isn't that. It is really akin to having virtual pair programmer. It's tool than can amplify the skills I already have. What I do feel like I am paying for is the operating costs of the service and further training of the model. At $10/month I think that is reasonable (even on a junior salary).
I don't know what the "spirit of open source" is TBH. The software industry today mainly consumes the fruit of open source software, builds products with it, sells those products, and, rarely, gives back. What Microsoft is doing with Copilot is no different (to me).
@@joelholdbrooks6729 what you “feel like” you are paying for? That’s silly.
At the end of the day, it’s still just taking free work and selling it back to us. And just because the software industry does it all the time doesnt mean i like it.
I just don’t want to opt my art into the AI “training” from it
Just don't show it ..
Agreed on Art theft as taking credit for a piece. I wouldn't call images on AI generators like Bluewillow stealing as most of them are produced from prompts. It's also still just a collection of ideas not an actual concept or art piece being stolen.
Art did not create the core technology. Art is a valuable and entertaining subject of the core technology. Just like someone who poses for a photo did not create the camera, but they did give value and a reason for the camera to exist. But still, a debate on where we're allowed to POINT the technology is still needed just as was the case when the camera was invented.
I think your camera analogy is very off the mark: a camera was not made using thousands of people posing.
An AI art generator is not made the same way a camera is, and you do not “point the AI at a subject”, you get the AI to directly emulate the works that it was trained on. An AI making an image is not like taking a photo.
@@notveryobservant1056 You've incorrectly changed the analogy and are no longer comparing like parts. There's the core technology, the subject, and the product. You've changed how you relate these components not only from my analogy but even within your own post. The core technology can be taken to another planet and still operate as intended. Every bit of training data can be new and different because that is the subject in comparison to the core technology. You've confused this with the product which occurs after pointing the core technology at any valid subject. In the case of a digital neural network the variable subject IS the training data. In the case of a camera it's the patterns of light. In both cases the subject did not create the core technology, but the product in tandem with the core technology. Hence you can move either to another planet and entirely change every bit of the subject. That's why you can't say the subject created the technology. It's used to create a product, in tandem with the technology, but did not retroactively create the core technology itself. People posing didn't create the camera and a particular set of training data did not create the core technology that could be pointed at completely different training data.
@@notveryobservant1056 don't cameras capture subjects almost verbatim, rather than synthesise new data?
My take on the inspiration thing is slightly different. Because before AI existed, we artists had to explain to newer artists all the time over and over again. what is or isn't plagiarism.
It's seems simple to us, but even artists who's been drawing for a couple years struggle with it. so explaining it to people who haven't even thought or studied any art is going to be difficult.
"It's inspiration because you look at stuff and made art too" always forgets that that's also exactly what you do when you plagiarize something. How do you plagiarize stuff you've never seen?
It doesn't look like human plagiarism because it wasn't done by a human. and the rules of thumbs we have are for when a human does it.
Long story short: I'm in the camp that believes one day AI won't be plagiarism. Right now it is, but one day it won't be.
As an artist we don't need AI art generators to make concept art for us to reference when creating our own media. We can pay other artists to do that for us, we don't have a use for AI art as a tool other than a cheap replacement of the middle man then eventually men.
its all fun and games until the next generation of server hardware packs 10 server racks worth of compute power onto one motherboard. We are experiencing an explosion from the inside, and it is seriously not just the artists that should be worried.
Taking jobs away shouldn't be a reason to worry. Work will either exist, and if not enough work exists governments will be forced to take an actual more socially aware approach.
@@Natsukashii-Records it shouldn't be a reason to worry but it is and that's the problem
@@Nxnn132 It really isn't, we are way too close to singularity where social restructuring will be a must. Until then people can get some other type of work. Besides, 90% of artists always whine about not getting enough commissions before AI was a thing, now will just need to get an actual job until the government catches up and bring about a more socialist approach to the problem as that's literally the only option.
Who says non-profits can create copies of copyrighted images?? Non-profits are not protected from copyright claims
Humans when they learn art by seeing a lot of artworks and repeating them: it is fine Ai it does the same thing: reeeee is theft
A lot of these generators are vague on the details, I find it interesting that they study 100's of millions of pieces of art but the artwork it produces seems to mirror the absolute cream of the crop, so exclusively artists who are masters in color composition and style. It's almost as if they told it to study the best in the field and I imagine they were more than aware that this would make these people mad but did it regardless.
Especially Midjourney. This had a very artistic way if "expression".
I beg to differ I'm a graphic devoloper and that happens to be linked to the prompt. something like color and art style are not owned by the artist niether. they are more of an understanding one which can be punched into the ai. the ai is trained on objects and style. and typically trademarked ideas are kept out of it. as a graphic artist i had to learn these styles. Do i feel the ai is stealing or other wise putting my effort down the drain? Not particularly, for one to use the ai in the same way they would have to know many of the same aspects of art lighting coloring and style as myself to do so. What of originality? The ai can not really steal an idea so if i have an idea i can now create concepts of that idea in 10 folds if not more with all the varies artistic knowledges applied to aim the concepts to be created in particular lighting style coloring ECT. This happens to not exactly make my job any easier but rather more creative as the ai can help me think up the same idea 100 times over. Though taking this illistrations of ai generated contents to then become part of a story serve a purpose and possibly be animated and also modeled. The ai can not do this for you atleast not very well and so this part of the job still remains. it's likely not to take jobs rather create more opportunities for product lines and more need for the artist who can help finish the products or help with.
7:19 "How is this legal?"
LAION datasets are just a list of link caption pairs, if that would be illegal then search engine would also be, moreover LAION exists way before they were funded by StabilityAI and others as a project to help the reimplementation of models like CLIP (see openCLIP)
Yaaaaaa!
Meanwhile OpenAI has put forward legal claims that other Text Bots trained on their data sets can't compete with ChatGPT.
The theoretical concepts of how AI is poetically similar to human learning have no baring on whether or not we should treat AI training as theft.
This topic makes me wonder that is it possible to create an AI artist that can produce artwork without any external data, and what the AI artwork would look like. In my opinion, the closest thing to prior art (Art that is not imitate anything) is mathematics, theoretical physic, cellular automata, music, Plato's idea on the world of from and rationalism.
Look up procedural generation. There is some external data (a programmer has to write the generation program and configure it) but it's much less than the gigabytes of data used for training image generation AIs. Procedural generation is used to generate levels in many video games, including Minecraft. It can also be used for generating textures, crowds of people, forests of trees, and many other things.
@@jimmyhirr5773 procedural general are no where near diffusion models
Its like comparing a water rocket to a jet engine
Do you have a Discord we could discuss all those topics ?
As a retired newspaper cartoonist in Utah I appreciate your insight. I was laid off in 2014 and have enough investment money. I have used DALL-E3 to help my Dinosaur Park donated cartoons.
Correction. OpenAI did not release an open source version of their model (DALL-E), much less Stable Diffusion model as you said.
I think the problem of ai art can be solved at a more root source. Ultimately, to train an ai model, it must have copies of artistic works. I think artist should be able to control their art getting copied.
Currently, artists' copyright means you can't use their art in commercial aspects without paying them. But if they show it online, it does allow people to access make copies for personal use. For example, you can download the image and use it as a reference for creating your own art. While this seems harmless, if I switch the context it becomes obvious that the right to access should not imply the right to copy.
For example, if you own a book yiu are free to make photocopies of it. But if you check out a book from the library, it wouldn't be ok to make a photocopy of every page and then return the book. Even though you have the right to access the book, even to access it for free, that doesn't mean you can freely copy.
With this additional protection artists could still reference others work for their own art, but only by accessing it through means the original artist decides. Maybe this will mean the original artist gets add revenue off of the web page you kept up all day for reference to make a parody of their work, maybe the original artist decides you can only view it for a certain amount of time before the webpage stops showing it and you have to work within that restriction. Whatever they decide, it should be up to them to decide.
In addition to restricting what ai can harvest to train to what people have freely given, UT would extend beyond to give artists extra rights over their creations that at least I think they should already have.
How about a "transparent AI" that is able to provide info on what artists it pulled from to generate each image, and they get paid royalties based on how much they were used. So if you want it to copy an artist's style directly, they primarily get paid.
They will go bankrupt because of the already existing ones not giving a cent to those they already copied. if they add an extra fee to use, people will use the free ones, if they dont, they will pay the artist so little and with so many caveats it will make youtube ad revenue to creators look stable and plentiful
I’m a Photographer but I have been using AI to create images for a couple of months now. So I put in a specific thing on my website that says that I am opting in to data sets. I figure I’m giving back what I’ve taken away creating a sort of symbiosis between artist and the art being created. Since AI art is so important to me then it’s only fair that I give my work as part of training.
it has already been hard enough to earn a living making art - or music for that matter. i really fear for the non-digital creative future not least because the AI is that good and progress cannot be stopped. The core problem is the same as it has been: EVERYONE loves art and music, but almost no one wants to pay for it. Im curious if future generations will even have the need to learn how to draw or learn how to play an instrument. its scary.
AI is a poweful tool. Thats about it, so far. However the "free" datasets fo this AI are not really free, so there will be a lot to do in terms of legal usage.
I remember when the deviantart AI launched, you could throw an username along the prompt and have the art be in that artist style, you can imagine what happened to commisions until they "removed it" ie made only the username not work, if they had a real name or it was part of a series, it still worked
I agree with a lot of what you say in this video, but not all. I personally hate data use via commercial company/corporations. But it's a hard topic as AI is just learning, it isn't conscious or have any feeling like us, and it trains 100000x faster than us and produce output 10000x faster than human.
But a lot of your concerns came from the society we live where private property and merit are pillars, and I don't agree such much with that. I personally understand why humans want to protect hard work that is one of the most valuable incentive for them. But in reality, I don't think artists really own as much their work as a mathematician or scientist their equation. Yes they found it, yes it is really valuable data, that either allow humans to comprehend the world and describe/simulate it more accurately, or make human level interaction feelings (tha'ts what art does) but how much do we really own that work ?
Chess is a good example, the ability to found path in a game as been surpassed with AI, we still play it as we like to value what humans can do, and we wouldn't give any credit to one person using stockfish or openai to play chess. Like in Olympic sport, we may have reached a limit of what human is capable of, and the excitement from discover (what makes human feel joy, like reading a book/film/game for the first time is better than the second time, going from child to adult) may just decrease. Wanting to ban AI to learn chess using human match is futile in my opinion, it would eventually discover it anyway, just using more energy as it won't be prémâch and energy will be required to re-discover this.
I understand your pure utilitarisic way of vision those ethical/modal questions, and I share them with you, I just draw the line differently than you. But I agree with you a lot !
What society utopia would we be most happy ? Is forbidding AI medium is similar to trying to avoid automatic checkouts, or lamplighter ? What about hypothetical AGI that would really act similar to our imagination/inspiration, and would be able to replicate someone's voice or art as fluent as we could simulate an ant colony with a bunch of pythons lines of code ? Should ants be mad at us that we can simulate them easily, will knowing that make their life just worthless ?
As the only thing that we do on our planet and found as a goal is: living and evolving (borrow energy from our environment to reduce locally entropy by reducing the kolmogorov complexity) If this is our only thing to do it may also be really futile and a transcendental AGI could just found existence futile/pain and just decide existing is not enough on its own and nonexistence is better than existence and this could be the final Hanson Great Filter.
But I may go too far in theories, it's just there don't seem to be easy answer for us, little dumb chimp colony ^^
If I take an art class and learn drawing techniques (plenty of them online for free too). then I go and draw a "original" work that uses those techniques is that unethical? Anime, for example, tends to be fairly similar across all artists. Some may argue that artist(s) have their own flare which is why you can sometimes tell who drew it by just looking at it but really it just a culmination of styles they learned and combined to create a brand. Sounds a lot like AI art to me. I get why people don't like it but it's not going to stop based on whining.
i agree with whoever said that the real problem with AI art is the fact we need to pay for it, that's absurd.
Artists will indeed hurt with the rise of image generators such as Bluewillow. But they will have to look at it in a different perspective, one point that you've mentioned is to use them as tools.
Artists already have a tool that does the existing job of AI art and it's other artists, your just advocating to replace them as well.
I have been thinking about this topic a lot and here are some broad strokes from an indy film producer/director for what it's worth, which is probably nothing:
First, Ai cannot be stopped and it is going to destroy us; 'reshape' is a euphemism.
Second, what you have going on here EG (and all the others formulating neural networks) is part of the tinkering, experimental stage of that destruction: seemingly harmless.
Third, besides the theft from individual artists, the moral/ethical conundrum is the elimination of the human being from the process which is another kind of theft, and part and parcel of our destruction. I get notes of Monsanto for some reason.
For what it's worth, and it's probably not worth a GD thing, my production company's mission statement is a complete boycott of Ai from all stages of our creative process: inception of the idea, writing, pre-production, principal photography (which we do on actual celluloid film), post-production, promotion and screening. Generally speaking I resist digital aspects of the process but the opportunity for a completely analog workflow is still daunting even now 15 years after my first attempt at it. I work face-to-face with artists and ensure there is no Ai input at any stage of their work. I have anti-Ai clauses in my contracts. I bias to pen-on-paper work as part of every element. Humans are wonderful. I love working with people who create from their own minds. Down with the machines and shame on the people who have brought this zombie apocalypse curse to our doors.
Man with shovel? There is still a man. Generative Ai? There is no man.
Are you aware that you are opposing evolution? Evolution always wins.
You've motivated me to make my own video on this. As a graphic devoloper that uses ai art as a tool. And plan to pay others in future to create art if ai is used in that creation i have nothing against it. I don't feel it replaces any artist, as an artist. I feel there's plenty of applications ai can not fill, and there are plenty of applications for an artist to use ai as a tool to enhance there own ability to help in conceptual pieces and to increase the rate at which illistrations can be created to use as the foundation for charcters objects ect.
In regards to this there needs to be originality which i would scrutinize of those and myself who use the ai as a tool. This means that and is also listed as part of the legal use of ai art for profit, trademarks and copyrighted material. Can not be used as a template or part of the prompt. Only common usage terminology say a artistic style are example of what would be ok, while as a shows name charcters name ECT that already exist as an individual copyrighted and possibly trademarked product, character, onject, ECT would not be ok. I am an expert in trademark and xopyrights / intelectual property legalities.
I would simply not except art made by an AI that jumped over the line of using type to detailed ideas that are the exact character, trademarked product, or object. But if it is something of a style piece and the work is derivative then i may be of oki with considering it's use.
This would be like nikes hightops derivative used as the shoes to a fully original character. The character itself could not be a deravative of Naruto. I wouldn't except it and it would be easily noticed that it was done.
The AI voice actor hypothetical is definitely a good one.
I think copying a persons characteristics, looks/voice etc is definitely not okay. Deepfakes are terrifying and voice cloning while an exciting field for gamedevelopment is equally scary.
Copying a persons artstyle isn't the same as copying a persons physical characteristics.
Identity theft is not okay, and I also think we should be specific about the workprocess when using AI image generators. I don't use the words 'look what I made' instead opting for 'look what I had genned up' when it comes to showing AI image generated images. (though if I take such an image further. like... make a 3d model out of a flower graphic generated by AI, at some point I will start using 'I made')
I think that both of these are okay as long as the person consents to it, for instance an actor gets injured and consents to a studio using their face and voice for a film while they recover. Besides that I fully agree.
Thank you very much for this wonderful video
It is really hard to work as an artist these days, especially since we don't earn much from our work all the time, and we also have to practice drawing daily, so as to maintain our skill in drawing.
The artist community has been supportive of non-artists all these years and have worked hard for everyone
But today, seeing some people stab us in the back, call us useless, and take away all our work is really painful.
I like AI, but not in this way
There are amazing things that AI can do better than us
You can use artificial intelligence when you read a novel and create a visual from the story
We can develop characters within the game for you to speak
Imagine Skyrim ,GTA we can make characters think it's from that era
Imagine sitting with a character from the game and talking to him
Also, very old videos can be revived and given new life.
We can also develop artificial intelligence to simulate the Earth, and you can travel anywhere.
In short, there are thousands of ways in which these companies can earn
The difference now is between artists and AI artists
Real artists own their creativity
But the creativity of AI artists are owned by a company. In the future, companies can take your creativity from you at any time, and they can ban you personally.
The joy of art is not in the results, but in the process of painting
This is what keeps us drawing
As a non-artist my take is slightly different than yours. I enjoy oil painting as an amateur, but would never imagine earning my income with it. Though I obviously don't think there shouldn't be any professional artists, I think it makes sense from a societal perspective to expect people to choose a career that adds enough value to society. When farming gets automated, hand farming will be a niche that might have a market, but there's no real systemic way to hold back automation in leveraging value created in a certain amount of time.
Also, making art so accessible probably means a lot more companies can start affording beautiful design, so hopefully we will see a much more beautiful society in the future.
re: 15:17, I think the best analogy is not one artist being inspired by another, it is more like a factory owner deciding to get into art by hiring dozens and dozens of low-skill, high-volume artists who imitate the work of better known, lower volume, and higher skilled artists.
Which does happen, right? Like, corporate graphic design teams often struggle to match what a single person with a lot of skill and some free time can do, and post on social media.
@@Homerow1 the irony is that the reason something like that doesn't work IRL is art is usually valued by exclusivity rather than skill. Digital assets it's really harder to judge, but I would say the last decade has proven that corporations are more than able to outcompete independent creators online - just look what's happened to youtube!
@@Homerow1 It's still going to shake stuff up a lot. Humans have social limits and might not get good at learning to copy a style as it would be hard to hide that, whereas AI can train itself without any transparency.
Speaking of factories, never mind art, this happens all the time. A new product is designed, and an assembly line of humans will be the first production stage until the engineers can catch up with robots. The end result is more reliable products at a lower price. But where is the pity party for the displaced assembly workers? They move on. Do other things to pay the bills.
@AndrewBrownK what is the point of life as a human? If not even the creation of art is something worth doing for its own sake (even though printers have been able to perfectly replicate it for years), then what is? Why not have a planet of machines that produce and consume goods arbitrarily in perfect efficiency?
Let's compare data collection of a human learning how to draw. Same process. Should we take permission for every picture that being used to learn how to draw?
It's incredible how these art AI's are marketed on art sites as replacement to human artists, while the databases uses those artists works without permission. The legal issues have to be cleared up to protect artists. But after that I think it will change how we humans do art in interesting ways. And I think artworks that are physical will have more value in the future, because AI made art is always digital and can't replace the feel of physical paintings and drawings.
This is going to be a really dangerous area, improving artist copyright. Music is one of the best comparisons, and copyright in that area has measurably damaged the ability to share experiences. Another comparison is movies and IP in general. Disney's lobbying has resulted in copyright lasting far, far too long. This measurably damages the ability to take stories and twist them in new ways, even once they're decades old and unused by the original creators.
@@Homerow1 We should avoid both extremes, not easy though.
As an artist I'm happy I never uploaded much of my stuff online, other than for advertising purposes for games I work on.
(and mostly those are in videos.)
Making what people can scrape of me VERY limited.
Going to keep it that way, too bad my own social media wont grow much but at least my company still get a decent growth every time I upload trailers and gameplay.
Since I want to keep my stuff unique to me it got to be that way for now, since the art style is a big selling can't risk stuff. Kind of sucks but lots of people don't have morals...
Before I would not really worry as people would have to be skilled enough to copy me.
And I thought it was cute how artist would send me their attempt of making my art style, seeing them trying and then evolve into their own styles.
Maybe in the future we get better safety nets in the future and we don't have to guard our stuff like a dragon on a treasure pile.
Could be an interesting project to create the AI that recognizes the original sources of features of the final AI generated content. It could be possible if the "reverse" AI was trained on similar data sets as the "generative" one.
this is not possible
The main point I don't agree with is that trainning models on someone elses work is unethical, human artists have done this themselves since the begining of time. Artists don't ask to learn other peoples art, they see it replicate it and we all agree this is our natural way of learning. Why should it be different for learning machines?
I do agree however with the sentiment of artists, affraid of being on a quikcly shifting industry and the perspective of loosing their jobs. This I understand and that's why believe it's ok to limit the implementation of this tools to let people adjust.
need another video on latent space!! that ai evoloution vid was great
I think what all this means, is technology has now proven this level of AI is possible and never going away, and we need to focus on changing our economy and society, for this new reality. Nothing we do to try and stop AI, will benefit us anywhere near as much as if we spend our efforts adapting things we CAN improve to benefit all.
I agree, these things should be freely available.. Tho I also question many of the accusations being thrown around the community.. Even if they train them on entirely consenting data, they'll still gain this level of skill within a matter of a few extra months, and still be able to recreate any style. So many look-a-like artists out there for every style, the AI can infer how to achieve the style they're based on, without ever having any of that source-artists data.
personally, from what I've seen, AI art companies bad, AI art itself perfectly netrual
Artists themselves get inspired by other artist art, are they paying royalties for their inspiration sources? I don't think so. How is AI training any different ?
It's there one that is trained only on photos?
But but but... that virtual aquarium! Is that (I'm not a gamer) a quiet little nook in Subnautica?
How about those art that has ALREADY BEEN GENERATED? those art are already use to train and retrain the models. Even if you remove the name of the artist, a "likeness" and inspired from images that has then been generated... Is always there for the community to use and learn from.
MidJourney has literally generated BILLIONS of images... And derived variations which are fair use right?
I think it's better to call them images rather than art. I'm using Bluewillow and I woud not them art but rather, an image of which I can use for whatever purpose it will serve me. From the definition, art is a form of expression.
I don't think jobs will be taken, I am not a techy person but I am an artist for a career, and I have an overactive imagination, but I also have artistic talent, and I have been able to reinvent myself many times over with AI, and I have saved money coming up with new concepts to try my physical hand at. And it has revealed the gaps in my knowledge of the craft as well. That being said AI is good, but humans use AI, and we abuse the shit out of everything we touch. What is humanity known for? it's irresponsibility. Mismanaging absolutely everything. Name one thing that we have invented that was used for the purpose it was made for?
Just one.
I'll wait.
Thank you for the Beautiful image that you showed in your video. It'll work well in my next dataset nice and authentic.
"Imagine some company has been collecting data from you for years, and will now replace you for cheap". We don't have to imagine. This is happening in literally every field. It happened with chess playing algorithms, it happened more recently with go. Now it's happening in creative writing and my own vocation; programming. It's interesting that artists seem to be the only ones calling foul. My job is at risk with the recent development of chatGPT. I have a ton of code online, was my code part of the training data for this model? Probably. Yet, the thought of pushing back on this, or trying to hobble it in some way to keep my job has never entered my mind. Why? Because the pushback has nothing to do with copyrights, or using data without permission. Case and point, your copyrighted artwork and images have been used for over a decade in image classification algorithms. In that entire time, there's never been an issue with using the images, and this is because an algorithm that can tell the difference between a cat and a dog isn't a threat to your livelihood. This pushback didn't start with the use of copyrighted images in machine learning algorithms, it started when the algorithms got to the point where they were a legitimate threat to the livelihood of artists. Another important aspect of this, is that the need for human generated art may be a temporary requirement. Think of the difference between AlphaGo and Alphazero; AlphaGo required expert go games to become proficient in Go, and Go players could have made all the same arguments, protesting that their games should not have been used to train the algorithm without their permission. But this protest would have been short lived, as Alphazero allowed for a learning paradigm that didn't require any expert human games. What does an Alphazero like algorithm look like for A.I. art? It's a trickier problem for sure, and might involve approximating the human limbic system to create a loss function for aesthetics, but it's not an impossibility. If that happens, and artists still protest, we'll see if it really was about the copyrights and stealing of art. Knowing how these algorithms work, and how the art gets stored in the model as abstract parameters in transformer and convolutional layers, I don't see it as any different from an artist scouring the internet for reference material. Do you need to inform an artist when you look at their copyrighted work? Do you need to pay them? Maybe I'm full of shit, and feel free to roast me, but just some things to think about.
I very heavily agree on that last part there. It's why I find the argument of AI stealing art kind of a dumb point, as human artists do the same thing already, and have done so for many, *many* years and will only continue to do so, yet that's never been a problem, but is when AI does so apparently :/
Thank you. Yes, this is very much correct. Another example: I started out in the professional world as a transcriptionist. With all the AI voice models, Dragon Dictation, etc. that is out there now, it doesn't exactly take a rocket scientist to figure out where that profession is going. Yet I don't hear any outcry about THOSE people losing their jobs. Somehow artists are sacrosanct and we should all consider art some kind of holy thing that machines should not be able to reproduce, and God forbid monetize.
Absolutely agree. Nothing (minuscule amount to be irrelevant) is created in a vacuum and without learning by watching, hearing or learning from someone before us. AI is able to learn uninterrupted and the ability to retain all that it’s learned. Imagine if it was decided that Google should start to pay for all of the data it uses to be unarguably the best knowledge base the world has ever seen? I find it (ha ha) amazing that instead of searching for inspiration using text and tags on Google, or any number of platforms, AI can create it and even expand on what you thought you wanted in the first place.
Imagine if musicians were required to live their whole life actively avoiding listening to any copyrighted music so that there’s no chance they can imitate any aspect of it without paying royalties (and who can even determine what % of inspiration came from which source?). The reason we even have concepts such as musical genres is people imitating each other’s style.
if i were to look at hundreds if not thousands of pictures and then use those pictures to draw my art I'm technically also using information from other artists without asking them
and the result is as different as one generated by ai
this probably a very weird claim but i just wanna say i don't think it's really stealing since the result is different
and yes technology always killed jobs and made life more efficient
for example: machines
1 machine replaced 10 human workers
i do however think artists will still exist (obviously)
ai is probably gonna be used as a tool to make art
This is a false analogy and people need to stop pretending these things are sentient, they are not. What they do is cobble together- literally taking pieces of artwork, millions of it, and do dice rolls based on prompts. There is no "learning" in the way you describe. It is not "studying" to figure it out, it is taking, putting into a blending, rolling dice and spitting it back out.
It's a failure of the word itself that we gave it. "intelligence" it is not. Did you know that if you know the prompt and the seed number, you can reproduce the generated artwork, every single time? Thats not "learning"
I recommend checking out these videos on the subject:
pt 1. ua-cam.com/video/1CIpzeNxIhU/v-deo.html
pt 2. ua-cam.com/video/-lz30by8-sU/v-deo.html
@@natehorsfall8379 of course they aren't sentient
Because they are basically machines
AI is trained on big databases and they make connections between different images and text similar to how our brain does
Those databases are often made for educational purposes in order to avoid conflicts with copyrighted images
So technically it's not against any law
I also think the same
I agree I create vastly different and new things nobody has seen before eith Midjourney. Heck, even if I upload an image as inspiration for the AI, it splits back a totally different image in every way. Often better.
That's why we have laws against plagiarism. I saw a lot of AI users emulate a drawing in the style of some popular artist, and the AI is using a data bank with said artist work. It is plagiarism assisted by computer.
I just don't see any way to do something about this as long as people can easily finetune their own models.
Short answer: kinda, they just found loopholes around the law, if it weren't for those loopholes they'd be sued into the ground by now
If you want to keep to "free" stuff, you should ditch DALL-E. They claim ownership over all the art the program generates. With Stable Diffusion, your works are yours.
The scumminess of the commercial AI continues past the gathering - They technically don't sell the art the AI generates: the sell the processor costs for the server that runs the AI. That's how they get around the 'can't sell the product from the research'.
Edit: Stable Diffusion is also pretty easy to set up and get running on your local machine. Then you get as many images as you want using whatever models you want whenever you want.
I use Stable Diffusion-UI, Automatic1111 is the main one, but I find UI more user friendly. Great installation instructions as well.
It's giving "you wouldn't download a car"
Loved the dolphin-ball.
An informed and nuanced examination of the subject.
Data ownership and intellectual property have always been controversial, since the only thing that causes scarcity in the world of information is obstruction, either deliberate or by ignorance (i.e. nobody knows the info yet). The typical rules governing physical property, when applied to information, cause a million ridiculous outcomes and divide by 0 errors. And those rules are already messy.
AI is just exposing complications that were already present, really. I don't know how, but we need a renovation of our understanding of data ownership/intellectual property from the ground up, or any system we attempt to use as a band-aid just won't be able to keep up with the changing world.
Edit: That's a very good orangutan.
24:05 but chess is closer to a sport than art
Within art, creating something new and unique is a great driving force, not only raw technique
giving artists, or any workers, the ability to stifle AI development for their own protection is just backwards. We should all want to arrive at a future where automation has made human labor optional. Artists being put out of work is just a symptom of the larger economic problems we will face as we move towards this future. Our solutions should be forward thinking, not obstructionist.
Everyday in which we can delay the intelligence explosion is a good day.
I would happily slow AI development if it meant doing it right, which is to say ethically. I love the tech, I want it to improve, but I want it to be beneficial to as many people as possible. Careless progress could be catastrophic.
Normal people can't use images and words without attributing the source material. It is illegal to do so as a person. Why is this different if you get a program to do the same activity. Some of the AI are spitting out almost perfect duplicates of art they are being fed with out proper licenses or reference credit.
@@HallieEva We do that to protect artist's ability to profit. The point of automation is to move towards a future where artists don't have to profit. We're not there yet certainly, but as we get closer I think we need to be ready to accept that intellectual property shouldn't need to exist.
@@EmergentGarden As someone who runs their own AI R&D lab, AI Ethics is a huge and thriving field. Any college that has an AI department has an AI Ethics person, and so do many companies. It's just a complicated issue made even more complicated by the fact that money is driving most AI development these days, and money doesn't have ethics.
Everything in AI ethics has the same problem. It basically just exist in a vacuum and will not have any effect on the industry. Individual artists don't have power, while big tech companies do. Power determines what happens in the real world. And the general public doesn't care that a few artists are getting exploited, since the new tech is so cool, so there is probably no chance of artists getting any compensation.
the real problem is scammers and Copyrighted works
The concept you're exploring is related to the immense volume of art that has already been generated, a good deal of which is used in training and refining AI models. Even when specific details such as the artist's name are removed, the generated art often retains a "likeness" or draws inspiration from the original images used in the training set. It is this enduring echo or "imprint" of the original that remains accessible for community learning and creative inspiration.
A prime example of this is MidJourney, a prolific generator of digital images. Over time, MidJourney has produced billions of unique images. Beyond that, it has also created a vast array of derivative works. These variations, though inspired by original pieces, are fundamentally unique creations. As such, they often fall under the category of fair use. This means they can be utilized without infringing on the copyrights of the original pieces, allowing a wider audience to learn from and interact with these pieces.
This system of generation and reiteration is not only a testament to the transformative potential of AI in the art world, but also a powerful tool for fostering community engagement and learning. It allows for the democratization of art, opening up new avenues for creative exploration, and presenting countless opportunities for individuals to engage with and learn from the art world.
I wonder if these models can be set up in such a way that they know which specific images are weighted more heavily in a given prompt. That way, all AI art can be "credited" in a way such as "This image is credited 18% to artist X, 12% to artist Y, 4% to artist Z..." etc
There are existing AI's (like using CLIP) that can retrieve artist names from an image. However, it is not very accurate because it will prefer popular big names and will exclude many small artists and newer artists.
Stable Diffusion 2.0+ has now excluded artist names and celebrities to avoid legal and ethical issues, making it much more difficult to reproduce individual works and styles.
No - because the images no longer exist in the AI. Essentially, Stable Diffusion looked at 5 Billion images and learned a bunch of patterns. The patterns are the only things that are stored in the model file. It can know a "style" if you tell it which to use (this can cause some ethical, if not legal issues - and of course could lead to fraud if you claim a piece of art was created by that other artist whose style you used instead of by you with your AI tool).
This would be akin to listing all of a persons ancestors.
Isn't there a text generator that cites sources? Though it's for research literature IIRC
@@revimfadli4666 An Israeli firm just released AI called Wordtune Spices they say cites sources. Haven't tried it.
Great Video, many relevant statement as you say you can go for days , me as designer hobbieest I have to admit that midjourney has make me become lazy , do the idea set the color palette and just do the last retouches and combined other prompts , voila ! I love it honestly, I think tthat to reach a level where you are well know or world wide recognize you need study and time which I can’t afford , so this tool comes realy handy, in the end we can’t stop technology , like the internet, blockchain, Gaudí said, you think you are creating something? Look around , nature did it already,
The comparison you make are very valid we all imitate the work of our favorite artist anyway, you give ,e a lot to think about, but this debate would go for many months I think
I love people like ANTI A.I laugh on A.I ART in 2017. NOW ALL OF THEM SCARED TO A.I
i'm not scared of ai, i love ai. i'm just angry at the unethical practices of these companies.
Oh man. The comments over there have gotten really mad at me
Normal people can't use images and words without attributing the source material. Why is this different. Some of the AI are spitting out almost perfect duplicates of art they are being fed with out proper licenses or reference credit.
Perfect duplicates violate copyright regardless of how they are made. That's plagiarism and no one is questioning that. The issue arises around transformative works: works that don't resemble any existing image. Is that still theft?
It's also not true that "people can't use images and words without attributing the source material". This guy did it _in this very video._ He showed video thumbnails created by fellow UA-camrs and didn't credit them in the video or description. There are entire art movements centered around literally copy and pasting existing images and recombining them. It's well understood that if you draw from enough works, the resulting work is transformative because how the pieces are arranged becomes the main form of expression. Hip-hop is one of the prime examples of this. While mainstream artists have gotten better at attributing where their samples come from, that's not how hip-hop started and this entire genre would have never existed if they had to live up to your standards from the beginning.
When the AI are spitting out near perfect copies, they're being used differently. A normal "text->image" query is what is being discussed here. You give it words (tokens), and it uses those along with a seed to generate an image based on all the patterns it has learned.
You can also do an "image->image" query. In this case, you give it a starting picture along with the prompts. It then uses the picture you gave it as the base, changing it a little or a lot depending on the prompts. I've seen a few people posting those "This is what the AI gave me, look how close it is to this artwork". Well, yeah. It's that close because they gave the AI that art piece as the starting point.
This is how Lensa works: you give it some selfies (hopefully of yourself or someone who gave you permission), and the AI alters them to turn you into a Space Fairy or a Mongoose or something.
Because AI learns and uses that learned knowledge to make new images
It's very different from copy pasting
Well and what with the usage of Ai-generated images as reference pictures for your own man-made art? I did something like this a few times, and I know that a lot of artists do this themselves (even a few recommended it here on YT).
I hate the "stealing art to train the datasets" argument. If it were true then the same can be said for every artist that has seen another artists work, which is practically all of them unless they've live in an absolute bubble.
I disagree with it too because all my life I got inspiration from other media and artists, but it took me 6 years of hard work before I could start selling smth online. So there was no flooding of the market with useless works. My insecurities and perfectionism made me work had until I was sure I had some value to offer.
@FreymanArt Terrible argument. Since the dawn of art, artists have published their works *with the undertanding* their work would be seen by, and used as reference/inspiration for, other artists. Every artist has implicitly signed this clause whenever they release an artwork. The same cannot be said of being used to train AI models. Indeed, the vast majority of artists would be flattered if other artists studied their work, and if asked the vast majority would grant permission.
Moreover, artists learn in a different way from AI. Artists learn concepts like anatomy and perspective so they may dynamically apply them. If one understands anatomy, he can depict a human in any pose from any camera angle. AI as it currently exists does not learn in this way; it merely observes abstract patterns in the visual data (which is a method humans also employ as well). This is evidenced by the fact that AI will often make obvious anatomical errors, and that these errors are inversely proportionate to the prevalence of a given pose in its training set. So Ai will make fewer errors on generic poses like standing, but mess up horribly on less common poses like doing a handstand. As another example, signature-like residue will appear in the corner of AI artworks, demonstrating that AI-as-it-currently-exists copies patterns, not knowing what to omit. This phenomenon is known as overfitting, which Stability.AI has admitted "Diffusion models are prone to memorization and overfitting" www.reddit.com/r/StableDiffusion/comments/xtvaup/how_do_you_feel_about_stabilityai_being/
bullshit, you can't compare the artist effort in taking inspiration and elaborate to a super computer which learning capabilities are infinitely faster than humans
@@davideghirelli5856 It's what the brain does in part, they are comparable. Just because it's better doesn't mean there aren't overlaps. Would you say the same thing about a savant? Being better at something isn't an appropriate reason for disqualification, seems silly
I had never thought about it until artists that I follow starting talking about it even though I have been following AI Art progress. I never thought that a company would try to make money from scraping websites with user-content without consent from the users. This is a very thoughtful and educational video. Definitely the first time I heard the term "data laundering"
Designers and artists are probably the least threatened of all the professions that AI will replace. Their role is to make the client different from all the others. If everyone has an AI design, how will those who want to be different differentiate themselves? In the beginning, until the senses are satiated, it will also take their work. But after that, the need for them will be greater than before. It's like when you look at works of art in the Louvre for the first time and come across huge, realistic paintings with lots of detail. The first picture shocks you, the second delights you, but you just pass by the 76th picture of the same style, because you want to see a different style. So artistic professions will be saved, but there is no hope for other professions, which will suffer according to the same immoral principle.
nope. AI is exponentially getting better. you haven't see AI talk and learn from another AI yet. you see AI is base from human intelligence. just like humans, AI is dumb if it has nobody to learn from. humans learn from communicating with human, we teach, we learn , we argue. AI learned by copying human dataset. we haven't seen them learn from each other yet. the next level would be AI learning from each other. human being thrown out from the equation.
the only thing limiting AI is "computing power" we are reaching the limit on microchip design. if we can't breakout from those limitation then AI tech would be having roadblock. at the very least it won't kill the art industry yet.
Could you tell me the jobs that might be replaced and why? Would love your insight as I have a Debate on this.
Hi there, an actual artist working in the industry here. Like you I've been curious about the latron data set. Ironically those companies legally have obtained those data sets,funny I know but there is something people never read, and that's the terms of service for sites like deviantart ex.. you agree to allow that site to use your data and sell it..funny how people click accept on anything and never read the terms of service, and complain afterwards
Wooooow they fucked us from the start. Only solution is to bomb the servers then.
nice there will be blood reference
I don’t mien to be nasty, but I’m going to point something out. The Star Wars clip is not a good example, in fact it demonstrates the exact opposite situation. The fact that some one had taken that clip and used it to make a joke is no big deal is because every knows that it is from the popular movie star wars and they know exactly who made it and know that he made a killing on it. Now if you what a better example using Star Wars here’s one. Imagen that your George Lucas and this is the seventies you decide to go to the movie theater after a long day of filming your move star wars now there is this new movie that just came out it starts with a girl and two robots on a spaceship that under attack, then you get this sick feeling in the pit of your stomach that turns into rage as you storm out after the robots met this “Duke Sky-dancer!” then you fide out that the night security Gard moonlights at the same big-name-studio that stole your storyline and you can’t prove anything cause nothing was physically taken, so now your blockbuster turns in to a flop and everyone keeps saying gorge whocas. This is a more accurate description of what the A.I. companies are doing to artisans. Not only are they stealing their thunder but like the young George Lucas they are steeling there livelihood as well.
7:35
Stability doesn't commercialize their models. It's not a loophole. They use the data for research and release the model for research open source as it's meant to be. They do offer their own creation tool, which you do not have to use, and making profit on the GPU and server use - you can do the exact same by installing Stable Diffusion on a google colab notebook, for free or speeding it up by paying google. Since they released it open source, they can't monopolize it.
The selling the server cost is the loophole that lets DALL-E et al. get around the restriction of commercial use of the research data.
@@waltlock8805 Commercializing research is what happens all the time. That is the goal of getting companies to finance research. As long as the results stay open source and aren't sold themselves the Law exactly did what it was supposed to.
Great video. Well reasoned. I wanted to ask about a few ethical loopholes:
1. If I hire artists to make imitations of existing art styles and train a new model on that imitation art, is that more ethical?
2. What about training a model on AI art instead of original art? Over many training generations where only AI art is used and is further removed from the stolen original training data, does it then become more ethical?
The more generations you do over AI art, the more bizarre results you are gonna get, if one AI had for example, a tendency to fuse teeth, over a few generations, people will only have a white denture in the shape of all our teeth. The only way to train an AI with another AI is have a perfect one be the source, and at that point you dont need a second one
@@tiagocf1208 sorry, I intended to imply that only good samples would be used in the training. There's probably not much value in using ones with obvious mistakes.
Why not build an application that allows artist to train a modle with their own style and profit if people use it? would that be technically possible?
it could be done very easily in all of the big txt2img models already available. simply by giving a royalty every time their name is entered in a prompt (assuming they're actually in training data first). the artist could even set their price to be included in a training set (more popular artists will likely be able to get a better deal). this will make free txt2img models less likely in the future as it will be inheritably more costly to provide customers, and cracked versions running on local hardware will still exist, but yes this is a possible compromise
@@dfhdf4214 what kind of programming skills are required to play around with these things? Do you know?
@@permanentprogress there's plenty to do with no programming knowledge. dalle 2 and midjourney are completely ready to go just put in your prompt and it spits out the image. and then there are programs that make getting into the cogs a little more or customizing the training data that just require good instruction reading and learning the program. and then ultimately a degree in machine learning would the pinnacle
I believe there's a japanese AI service that does something like this - allowing you to throw in some of your images and create art in your style - but I forget the name. I reckon the concern here is that it's still backed by a bunch of other data which was questionably sourced, so the ethical issues still arise. If you were to make a model completely just from your own data, that might circumvent it, but you'd need a lot images to train it, enough that I don't think it would be realistic for one person to generate, or even a small group of artists.
@@HiHi-iu8gf it definitely needs more than just the artist's complete works to be used for prompts. For instance if the artist never drew a plane you're not gonna be able to see a plane in their style. In fact chances are the only text caption value would be the artist's name and it would have literally no connection to any other word and no way to understand any prompt besides the artist's name. But these services probably have a base dataset of real photos and captions. I'm just curious if only including one artist actually hinders it's ability to understand what the artist's style really means without other types of art styles included in the dataset
Im the future you will be paid for what you can do that ai cannot. There will be those that will purchase only non ai products. But the tiny things that get copyrighted today wont be. It will be the same when it comes to creating anything. Ai will start making cars, tools, etc. You wont pay for the idea but for the cost and materials it took to make it and there will be only a few big companies that will create anything. You will just type in your order and ai and a company will create it and send it be it car, software, art, and probably eventually music and entertainment like movies.
Artists trained on the works of other artists, are they thief as well? By this logic - yeah, pretty much
Steal all you want from me, it will take you years before you can produce something competitive and of value. I have "stolen" the way you sat all my life until I became a better artist.
So the "theft" argument doesn't stand strong, but the bigger problem is scale. While we sacrifice our lives and health working countless hours to improve enough to sell 5 good pieces, now this new tech makes it possible to flood the market in seconds.
@@UnknownDino I'm an artist myself and I know what you're talking about, but still the effort is irrelevant when it comes to judging the result. Just like an industrial revolution happened and many protested it's now time of a digital revolution, unavoidable.
@ETERNAL DREAMER the closest comparison i can think about is the games that make actual achievements placed behind a pay wall. Countless more people are now able to achieve the same or even better rewards but actual skill regarding that subset is disregarded since it's much easier to just buy it then to actually learn how to do it
personally I dont feel making it open source really fixes the problem. Its better but not much because it still ruins the market value of the original artist. This data should work as a form of unemployment insurance or pension fund for artists that would benefit everyone both user and creator.
First) most artists are influenced by the work of other artists. Generative AI that produces Art (like) products is influenced by the art contained within the training dataset. How is that different from one artist being influenced by the work of others? Second) let's assume that you disagree with my previous statement and artists should be compensated when their work is used in the training dataset. And to illustrate the problem, let's assume the training DS contains 100k different images. Are you suggesting that an artist that produced 1 piece of art in the training DS is entitled to compensation? If so, how much? 1/100,000 of what?
'How many fingers, Winston?' 'Four. I suppose there are four. I would see five if I could.
cant wait for a green AI label, or fair fed, or whatever
I don't agree with quite a but with what you said but upvoted because its good to have a healthy debate around the topic and you discuss a lot of valid points with your personal take.
The main thing I agree with is artists pre 2023 did not know about AI and as such did not know their art in the public domain could be used in such a way. There should be a way for them to opt out those art works. Stability Diffusion has in their ToC that you must use the latest diffusion model but that doesn't mean people will.
The main areas I disagree with is how models are trained. The core AI training data is to denoise images based on inputs (prompts or images) until it gets closer to the original. It does not need any specific artist to replicate any style.
Humans and AI learn differently. AI doesn't learn a style, it pattern recognises by denoising images. Through trial and error with human guidance, you will be able to make AAA pixel art, modern art, hyper-realistic art etc. The end message is, AI will be able to copy any style eventually even without the original artist simply based on the fact it can process billions times more data than a human. It obviously needs suitable learning material, but it does not need Picasso art to produce a Picasso, provided with sufficeint Cubsim art and it will eventually be able to mimic Picasso.
I mention I agree pre 2023 that artists should be able to opt out their work, but going forward, I do not believe any artist who leaves their work in the public domain, should be "protected". Think about it, you are "protecting " them from others being able to learn. Learning is something that should be free for all. Such artists should simply have a small public portfolio and keep most of their work private and only show it to clients if they don't want AI to learn - after all, AI needs a lot of data.
You give a good example of copying someone's voice, but at the end of the day, people won't go and watch a movie for a voice, they go for the human behind it. Why do you think Pixar movies use celebrities when they could hire any good voice actor for a fraction of the price. The same with art, the best works have human characteristics behind them like composition, lighting, subject matter etc.
Using someone's voice/ art in a derogative way should be protected though imo as should creating fake news/ harmful content - some is protected by law but legislation should be updated.
The main issue is some people will lose jobs because AI can reduce the number of workers a company will need but that is what technology has being doing forever. Artists need to use these tools to get ahead of the curve or pivot to other career choices - and art won't be the only field affected. The ones to suffer the most will be those who scraped by with hobby art/ semi-profession art/ cheap art i.e. fiver etc.
I agree this tech will inevitably reduce jobs but would end on the fact that those which remain will need a higher skill set which in turn means better pay and quality of life.
I think the key word here is 'inspired'. These models are using art, copyrighted or not, as inspiration. They are not reproducing the art as it is but taking elements from it to create new art. As an artist and photographer that's what I do. I browse other work, find art that I like and create art like it.
I'm also a fan of A. I. art and recently started playing around with the models. It's actually inspired me to take up the pencil and camera again after many years of being dormant. I can draw a rough idea, upload it and prompt the A. I. to make something from it.
I learnt early on that there is no protection online. Once you upload something that's it. Some art sites do try to protect your art or photography but it's hard to do and easy to get around. I took a long time to reclassify my under for Creative Commons or Copyright notices back in the Flickr days. I don't copyright most of my work, it's simply not good enough, but I have a small batch that I've put more work into copyrighted. I don't hold up hope the sites will protect my work but that's the sites fault and not necessarily the models fault.
I don't think the current web is designed well for this. Perhaps web 3.0 might but I don't know and that's a long way off anyway. It's certainly something that should be addressed but that's complicated.
I do wonder if we've not been reading the T&Cs of the sites we upload our art to. It wouldn't be the first time we've missed this. I would definitely look into that. I wouldn't be surprised if we've signed our rights away, at least to a certain degree, and not known we have.
Is it unethical to scrap the art, that is the big question. Well, I believe DALL-E, at least, doesn't scrap but please let me know if this is incorrect? For the others, it's hard to say and I fall on either side from one day to the next Would I be angry if my work was found in an A. I. art? Well I can't find any work of mine from searching so it's not happened so I can only theorise that I would be angry if I saw art identical to mine but I wouldn't if I saw art similar or using my style.
I think this debate needs to go on further and see how it actually affects artists. Technology has always affected someone negatively but benefited lots of others positively. A. I. art might negatively impact artists but positively impact people who have the imagination to write down their visual ideas but never have the skill to produce it visually.
I wonder if ever my art or style was used that it would actually promote my own work. I am not a known artist, not at all, but if I saw an AI art in a style of artist x that I really like I would, and have, go find more of artist x. So they _could_ be a benefit to artists?
Great video, thank you!
I love MidJourney. I Don't have much money, but I'm a person who wants to hire artists, but every artist I'm able to afford is horrible. Now, i can get the quality I want basically for free.
yes, next question.
Human learn from all previous creation, all artist are influenced inspired from previous artists and their work. I don’t agree that AI company steal the work of artists because their AI learn from work that was published. Age of AI is here we will be going thought an evolution on how human create.
I want to become an artist and i'm not going to leave the idea behind. With that said, i personally can't really find any moral problems with these models, it's absolutly awful to lose your job, i'd say it's even worse for artists because of the nature of their pursues, but it's progress and the analogy of imitation works perfectly here, we just feel like it's unfair when ai's do It simple because they're much better at it. When photography appeared, it was bad for artist (or craftsman as i like to call people who just have the technical part, and not the artistic one), when cars appeared, It was bad for those who breeded and sold horses. The part of companies making profit off these models comes i'd say, more from the economical bit than from the laws.
Sorry for my english
Oh, and grat video!!
You may get hate because of your videos, and i Hope you the best of lucks
There is still art in taking photos, it's not because they are using a machine to paint their art that there isn't any more to it. This makes more look like that art for you is just aesthetic. Just a comment, bye
@@yeah8598 no! That was not my point, sorry if It looked like It, i'm saying It wipes out the Craft of painting realistic portraits and that suff
@@habadababa31415 oh right, sorry, that is true tho
If you want to become an artist, you should use AI in your work as an artist. Artists shouldn't hate AI, they should see it as a tool, the same as they do a paintbrush or easel. If I was an artist, I'd be celebrating this, not hating it. My life just got a thousand times easier. I can generate a bunch of images as starting places and then add my personal touch to them, or use them as concept art for physical pieces. This is a godsend, not a demon.
Is too late, regular ppl are creating a lots of models in this moment
Bit of a rant here
I mean, to me personally, I don't think any real "theft" goes on in terms of AI art. Granted, yea, some algorithms use private works from artists without their permission, but is that really any different from a person doing that instead, still without their permission?
Like... if someone really doesn't want their work copied or imitated at all, why even put it out there? If you're putting it out there for the world to see, someone is going to imitate it or be inspired by it and use it as a starting point. It's really no different if an AI does it versus a person. I've had several people make the point to me that its' different because an AI does it algorithmically, in turn being the reason to why it's worse. But that makes even less sense, because it's just simply referencing images for ideas and to "understand" concepts, with next to nothing further than that. Whereas, a person can have direct intent, not limited by coded capability.
Simply said, I personally really don't think it's wrong for data of private art to to be used in AI algorithms, as actual human artists reference and will continue to reference private works with the intent of doing so, regardless of what they do and don't know, whereas an AI does it so it knows how to do something at all.
Something else as well that some people have said is the issue of people then selling off AI-generated art that's at a higher level of quality than some commission artists and say that there needs to be systems in place to fight this issue, and in turn, for some reason, blame the AI art generators. This argument is completely nonsensical, as first; many of the major AI art companies do has TOS rules against selling off AI-generated art like it's one's own, and second; why hate on AI art generation for the actions a person made? The wrong battle is being fought.
Agreed. I think the main reason people are reacting like this is not because someone can copy their style / artistic ideas without obtaining their permission (this has really been the case for as long as art existed), but because suddenly many more people are able to do it, which upsets the economics. Ultimately, that’s what technological progress does, it’s just that it worked itself up from automating away manual labor to now starting to affect the creative fields, which were previously thought of as the last bastion of human employment.
Just because someone posts their art online does not mean they should expect their work to be mechanically copied. Even an artist inspired by another will develop quirks and unique distinctions, because it's creative expression.
Is there a way the AI art companies enforce not selling off AI-generated art? Has anyone had consequences for that so far? If so I've never heard of it.