Did AI Steal My Artwork? How do AI Art Generators Work? Is it Art Theft or a New Medium?
HTML-код
- Опубликовано: 11 сен 2024
- Did AI steal my artwork? What constitutes art theft? Can anything good come from AI-generated artwork? Let me share my thoughts as I discuss what AI art is, how it works, some of the popular generators, like Stable Diffusion, and Midjourney, and what ethics may be breached in the training of these models.
This is a highly controversial topic amongst artists and AI users alike. There's a fine line between fair use, the beginning of a new medium, and straight-up art theft.
Even my own personal website was scraped and most of the images were used in some of the biggest databases used for AI art generators. You may be surprised that I am not 100% against this new technology. Can it still be used responsibly?
**LINKS MENTIONED IN THIS VIDEO**
How Does it Work? - • AI art, explained
Dance Diffusion Statement - wandb.ai/wandb...
Karla Ortiz posts frequently on the issues with AI Art. Check her out on Twitter: / kortizart
Jake Parker - The Fight Against AI - • Join the Fight Against...
BONUS ** Can AI Art Be Copyrighted? - www.theverge.c...
FREE STUFF FROM ME
Get a free phone wallpaper:
sparrowsprings...
Get my 50 art supply recommendations:
sparrowsprings...
Purchase my artwork:
sparrowsprings...
🍃 ABOUT SPARROW SPRINGS 🍃
Sara Ferrari is the owner and creator of Sparrow Springs. She started Sparrow Springs as a means to feed her art hobby but has grown to love creating and sharing her artwork as wall art and home decor. RUclips has become a more recent obsession with the desire to share the art process and experiment with new techniques. All the while, sometimes husband, child, and pets get roped into some of those things, and may occasionally have a presence on the channel.
🍃 CONNECT WITH ME 🍃
Instagram: / sparrowspri. .
Facebook: / sparrowsprin. .
Pinterest: / sparrowspri. .
Twitter: / sparrowsprings
TikTok: tiktok.com/spa...
Email List: sparrowsprings...
Website: sparrowsprings...
#aiartgenerator #aiarttheft #artistsagainstai
I will no longer be replying to comments that are within the same premise of previous comments. No need to repeat myself when people can read. Please keep your comments respectful to both parties. It’s important to keep an open dialogue. We don’t need to demonize AI altogether, but we don’t have to settle for “evolve or die” mentality either.
Good policies for life in general :)
You do realize that you can’t copyright a style
Unfortunately, your style cannot be copyrighted; artists are free to make their own works in a style similar to yours, but if they are imitating another artist, they are never going to enjoy the same success.Apr 5, 2022
You do realize I never said that :)
_We don’t need to demonize AI altogether, but we don’t have to settle for “evolve or die” mentality either._
Precisely.
I'm writing a graphic novel, and I was seriously considering using Midjourney, but I think you may have swayed me to reconsider. That A.I can create beautiful art in seconds that took years for other artists to master is not the part that bothers me, because if computers can create art as good or better than humans, then that art deserves to be seen. What bothers me is the idea that an image I generated might be recreating something very similar to the hard work of a human artist, and I wouldn't realize it. I thought that all the art was drawn from millions of images at once, but if it's only "stealing" a few pictures that suits the prompt, and tying them together, and giving no credit to the artist, then that's a real problem for me. My conscience wouldn't allow it.
Well. If you put in the prompt.
Cat on a rooftop wearing sunglasses. That’s might be just generating it from random sources of probability photos of cats, buildings and sunglasses. Then making something out of it.
That’s not really a bad thing. (I think)
But where the whole thing turns shady. Is if you input “in the style of _ _ _ _ artist” or by “_ _ _ _ artist”.
Then the Ai has a target to pull from their art.
That could be the real immediate concern.
IF ai could create art better than humans then why AI is fed human made work first. Dont you see it lack logic? IF they created ai which can create good art nobody would argue. But its not what they did. All they did is created software which learns on data they did not pay for and reuses it to make new art. IT uses all the computer ai advatges over human but feeds of human aquired knowledge how to paint. IT is unethical. unbelievable people have doubts
I’ve been a bit of a dormant artist for the past decade. Ai art brought me back into it in the weirdest way. Mainly because I was playing with it and could not get exactly what I wanted out of it. The perfectionist that I once was is still very much alive. So I had a very “Fine, I’ll do it myself” moment 😆
Artists like you will be fine as long as you never stop and keep creating art your way :)
I love that! I’m certainly not in the camp of artists that believes AI is all bad and should be eradicated. I just want some boundaries. I appreciate your words! I intend to continue and adapt any way I see fit. Seems to be that artists like you will be part of paving the way for a solution to the divide 😉
@@SparrowSprings I think it can fit into everything as a whole. I do at least like the idea of getting a specific pose as a reference, creating a mood board, or helping with picking color palettes. We probably need to think about how it should be used in commercial purposes. That seems like the root of most of the strife surrounding Ai art that I’ve seen.
I do have fun with it. I put a picture of myself into Midjourney and got a stylized version of myself as a Frank Frazetta style Barbarian type character. It probably turned out too good honestly …after generating 2,000+ abominations 😆
@@kemek7397 hahahaha oh that is fantastic 😆
For sure the possibilities for inspiration and an aid to creation is astounding. Personally at this time I would advise against using it for pose reference, because it may look good on the surface, but who knows how accurate it actually is 🤣 for all you know it could be showing the wrong muscle groups and making your character a contortionist. Better to just ask a friend to strike a pose.
Yes, commercial purpose, is the biggest hang up for me anyways. It stinks because for people who are using it for a creative outlet there are others purposely taking advantage of the model for personal gain, for spite, and to shove it in the face of the artists whose work actually went into the training. I know that’s not the majority, but they certainly are the loudest
@@SparrowSprings Oh I definitely agree with the references thing. I think right now it does landscapes well enough. Granted anatomy is one of those enigmas where once you know the rules it’s okay to start breaking them, haha! So yeah… not my go to for figure drawing references 😆
I think the commercial side of things is going to come down to what the end results look like, and how much it resembles another image that came before it.
Granted a recent case came to my attention that might set a precedent for this. The artist, Jeff Dieschburg, was sued by a photographer, Jinga Zhang, for using their photo without permission. The photographer lost that case. It isn’t 1:1, but some of that might parallel the Ai art debacle. In this case I really wonder how the photographer lost that case, especially after seeing the painting. That all happened in Europe for context.
For sure. I think anyone who thinks it’s black and white doesn’t have a real grasp on the situation. Copyright laws are confusing, court rulings are inconsistent, and the use of AI at this scale is new territory. At this point a lot of people citing copyright laws are trying to put a square peg in a round hole.
I am going to try my best to ride this thing out and keep improving my art in the ways that are timeless. Basic color theory, anatomy, composition… the mediums will always change
I'm glad you prioritized this video. I have been looking into it for the past few days because my Instagram and Facebook feeds are currently filled with images from Lensa Ai. It was cool at first, but as I gained more knowledge I felt completely deflated. How is an artist supposed to compete with that? What is the point of being an artist if decades of a person's time and hard work gets stolen, and all of that time and energy spent honing your craft no longer pays off because it can be replicated in moments?!? When NFTs came out, I was excited because I misunderstood it. I thought it was some form of digital copyright that could be recognized and immediately flag misuse of an artwork. The same way music immediately gets recognized and flagged if a copyright hasn't been purchased from the artist. Honestly, looking back, I wish I had been right about that.
Exactly! So many things wrapped up in this whole thing. Definitely check out Jake Parker’s video. It helped me out of the the deflation stage. I wish NFT’s worked that way! Honestly I wish visual art had even a fraction of the protection that the music industry does.
@@SparrowSprings what confused about ai art is how the ai art was programmed Because diffusion models are not scanning the web for images like half the artist think is going on they really don’t know how ai works diffusion model have a data set of images then they denoised the image to look no single on tv and then it uses the Promt to create a new image from the denoised image but probably is actually the people who put artist name in their promt and then use and sale those as there own art it’s not necessarily ai at fault
@@NicVandEmZ who’s confused? I said it a couple times in my video that it is 100% dependent on what a person puts in a prompt. I understand exactly how the diffusion process works. I didn’t just see some AI art, get emotional, and made a video lol. The whole premise of this video is that it depends on the prompt, it can absolutely be abused, therefore the companies should have restrictions on what can and can’t be used commercially.
@@SparrowSprings Yes, it (meaning the end result) is dependent upon the prompts), BUT, ALL the DATA in their databases is taken from artists work in the first place, SO that means that each image is built from a variety of artist's work.
WHY is it considered NOT OK in the music industry to use even a few notes from a piece of music, but completely ok when bits and pieces of artists work, their personal art style, etc., gets ripped-off?
If these machine learning apps (I refuse to call them AI because there is no intelligence at work there), did these types of mash-ups using a variety of music themes, note patterns, words, etc. and that music was sold as singles, albums, needle-drops for film and video, and concerts, there would be an insane, rabid uproar by the artists, producers, and music studios in that industry and not only would there be a tsunami of lawsuits, the music industry would SHUT IT DOWN!
We all KNOW THIS to be true.
But because artists are usually doing a mix of working alone doing their own thing, plus freelance, then there is no way an artist could afford to pay an attorney to bring a lawsuit against these weasel AI companies who have gathered their work up WITHOUT PERMISSION, under the guise of "learning", AND THEN CONTINUE TO MAKE USE OF THE ARTIST'S WORK.
One thing I do like about ai is that you should take inspiration from it and apply it in your own art What I don’t like is people using ai and now calling themselves artist
Inspiration, absolutely. I even have mixed feelings on people calling themselves an artist. I do believe some people truly go into it with a vision, but I definitely seen the “in your face, I am artist now” which is massively irritating
@@SparrowSprings in your face I’m a artist now is what I mean
Totally understandable
It's wonderful you've added your voice to this discussion. Too often the arguments against the current situation get lambasted as "why are you so against innovation, etc" and it's just like...NO. We love the tech too! If the training sets had been ethically sourced, I don't think there'd be half the contention there is right now. I mean, there would be still be issues because yowzers, Gary Marcus highlighted some truly horrifying things on his Substack, like AI sweatshops in Kenya. But I think on the whole, the majority of the art community and the world would have been like, hey, this is cool, let's see what we can make with this.
18:47 Truly prophetic. Literally the day after you published this, a fully AI written and illustrated children's book went viral on Twitter. RIP me.
20:03 Yes! I love seeing the growth and evolution of an artist from their first efforts to present day. It's one of the most awe-inspiring things and if you know anything about them as a person, you can literally see how life and living has affected how they view the world and translates into their art.
21:55 The worst part is you shouldn't have to OPT OUT in the first place! It should be an opt in only system. UGH. I hate the whole "move fast and break things" mentality that gets applied to a lot of new tech R&D. You'd think they'd have learned that you can't always fix the things you break.
Personally, I'm looking forward to the day an artist who has both a sufficiently large body of work and sufficient grounding in machine learning algorithms to post a detailed tutorial for all artists on how they can roll their own version of Stable Diffusion and commercialize it on their website. I'm sure this will happen eventually. Then again, perhaps visual artists would be just as overwhelmed as self-published authors currently are (the exclusive to Amazon vs wide distribution debate still rages eternal) and end up deciding to licence their portfolio to Midjourney or whatever instead of setting up their own instances of text-to-image generators. But if that's you reading this comment, then please, please, please I beg of you, do consider it. Because I guarantee you will have customers who would love to pay you directly for your work instead.
Thank you for this respectful take! I’ve been creating AI Art for a year and was a bit blindsided when something I loved and was mostly ignored suddenly became hated once it was less abstract. Most artists have been coming at me with insults and threats so I appreciate the way you went about this. Even though I am pro AI I absolutely agree there are some issues and working to address them in my community. Its been making me so happy when I get others to only use Public Domain artists in their prompts. I personally don’t sell my creations and love drawing my own starter images and make them very unique but unfortunately there is a lot of misuse of AI that I’m doing my best to call out.
I hate that so many artists have gone so far the other way. I am sorry you had that experience. I am all for regulation of uses, even if it is a self regulated practice, NOT the elimination all together
THANK YOU SO MUCH FOR THIS!! I’m writing an argumentative essay right now on this subject and this helped so much!! You deserve much more attention
I disagree on some parts of the video, if I can mimic a style of a popular artist and make an art on that style as long as I'm not copying the art directly and putting something new on the table it's legal as well as I am not expected to credit the artist's artstyle in any way because statisticallythere are tons of people in the world who's artstyle is the same and artstyle could never be copyrighted as it's a style not the art itself.
I am a commission artist as well as a upcoming soft dev so I can Atleast tell u I appreciate that the ai software is basically breaking down how our minds work when we do use inspiration or refference by breaking the refference artwork down into different pieced and finding similarities and trying to duplicate it. It is not an unethical thing to do as it's almost emulating how the mind works without the "human experience".
I understand the viewpoints of both sides, as if my creative and logical sides are debating each other. However, the quote from Salvador Dali, "Good artists copy, great artists steal," comes to mind. Ultimately, I believe that truly creative individuals will incorporate these tools into their work and continue to push boundaries. While it's easy for anyone to access a high-quality camera these days, the work of professional photographers stands out. Similarly, while it's impressive to see an A.I. generate a picture that looks like it was painted by Picasso, there is value in taking the time to learn how to paint in that style. It allows an artist to grow and understand how to adapt their skills. Simply using A.I. to replicate existing work may not be engaging or challenging for those who want to be creative. However, using it as a source of inspiration and reference could have potential.
I wanted to reframe the discussion towards the future. What if artists were hired for their keyword knowledge to work with AI art generators. You could still create your inspiring art too. AI is not good or bad, it might be the next paint brush we use to create the world we want to live in. Ty for your wonderful insights into this topic. ❤
I agree, AI is not good or bad. It’s unfortunate that people have abused it in such a way that it would be considered forgery of other artists. If non copyrighted resources were used I guarantee there would be almost no controversy
I've had a fun day arguing on twitter with emotional artists about this and everything you said was what i was trying to relay but twitter people can't read full tweets. All cause I suggested using the ethical generators as a creative medium.
There are definitely two extreme sides to this and I don’t want to be lumped in with either one. I definitely see a positive side to AI. Stable Diffusion announced that they are giving artists the opportunity to opt out for version 3. So this gives me hope! I am not even going to run over an opt out. I will leave my artwork in the database if it means they are willing to give others a choice.
Your take on this was so on point, it's good to inform people who don't know enough about it.
I don't think will change anyone's mind who already know were they stand on this tho.. talking to these techbros is like talking to a flat-earther. No matter what you say, no matter the evidence you present to them. When they get distroyed with facts, they just move on to other topics, without addressing your points. They just straigh up glitch out and start to spit out the same shit over and over, these guys are not even humans anymore. They start slow, then show their true selves..
"it's just like photography"
"AI learns the same way we humans do"
"so DiGital Art iS not Art?"
"tHe GeNIe iS oUt oF tHe bOtTle"
"tHat's jUsT liKe yOuR oPiNion mY dUuUuUde"
"ADAPT OR DIIIIEEEEE"
This was a great overview of the dilemma. When the generators first came out, I was of the opinion that the cat was out of the bag, and that artists really couldn't do much about it. At the time it also wasn't obvious how easily you could copy a style with the right prompt. Your video gave me more to think about.
Thank you so much!
I was definitely on the fence for quite a while until I started seeing the models being abused. And I will always say, the technology is not the problem, it’s how the companies decided to handle it, and what they allow within their own generators.
I understand how you felt seeing that AI book cover. I’m an environment designer, & old enough to recall when photo bashing started to become the standard production method for concept art in film. I was so resentful, til an older painter asked me, “have you ever seen a non-artist try to photo bash?” & “Do you really enjoy rendering all these textures?” Ofc he was right.
For me, the difference with AI image generation is that it feels like the outcome requires virtually zero skill and is based mostly on luck & the incorporation of other artists’ names in prompt-based generation & their copyrighted works in image-to-image generation. I’d at least like to see some legal regulations that forbid the use of living artists’ names in prompts
Right around your bit about the experience at 17:00 (starting at 16:35) is an example of why I find AI art frustrating. I generally have nothing against AI, and I think there are times when it's more helpful than harmful; removing backgrounds, reducing noise, whatever. It's these tools in particular that quickly do what humans work hard to do as jobs. As someone who primarily does music covers, parody dubs, 3D renders, and nothing like what you make, even I admit you had me in regards to your hard work. Heck, those videos that use AI-generated images to imagine ideas get in the way of view counts really easily, and now I wonder how I feel for the views my videos have been getting. I guess it's also view theft.
I don’t think what you do is even in the same category. I am guessing that most people know where your source material is from. You add your own unique twist and many will often go to the source material to compare. AI is being used to completely replace the artist rather than give a unique twist on an existing work.
Thanks for this video! I have been on the "AI isn't copying or stealing your art" side of things but this video gives me a fuller perspective. While I still believe we should be embracing these technological advances, I feel better educated on how the tech can be used nefariously by individuals and companies. With new tech comes new considerations and you have convinced me that there is a need to better legislate how this technology is used commercialy. Again, thank you for this video and your very solid take on the issues.
I absolutely agree, and I fully intend to play with it and see how it can change my work flow. Stable Diffusion version 3 is now giving artists the opportunity to opt out. It is a bit of a tricky process, so I would like to see a bit more effort on their part but I may loosely experiment when that comes out
I totally agree with your points - illegally obtained art it's pretty shady. My only problem with this whole situation is that they're blaming the problem on technology and not the companies.
I think the tech itself is pretty revolutionary and it reminds me of Star Trek. Holodecks, replicators, transporters - technology created to simplify our lives.
Only difference is that in the future, money (capitalism) has been obliterated, instead humans work for the betterment of our race. My biggest question is - what happens when true AI comes to be and they start developing their own art? i.e Data, sorry another Star Trek reference.
Haha good to meet another Trekkie 😉
I agree. I am excited about the technology, and agree 100% the companies are the ones who should be held responsible. I could be upset with the individuals who purposefully manipulate their system. But for companies to use a dataset that was originally intended for research, commercialize it, give zero restrictions to its end user, and not expect some form of issues to come of it? Makes me think that they weren’t thinking very thoroughly about it and they just saw dollar signs.
Very interesting video. One of the most important discussions in the use of AI is that of ethical AI. IBM for example has broken it down into five key tenants: explainability, fairness, robustness, transparency, and privacy. It would seem that use of artistic content in at least some of these instances would likely violate some of these guidelines. As an emerging technology, it's hard to predict how these things will flesh out. Creative writing for example, ChatGPT is capable of generating new concepts in creative storytelling based on your input criteria. And it's not just artistic pursuits. AI has the capabilities of writing code for example, or answering medical or ethical questions, or teaching concepts, which is concerning on many levels.
Agreed! I think we all have to define the boundaries of what we are willing to replace with AI. I am all for making things more efficient and keeping an open mind to new technology, but obviously some of it just went a little too far for me. I appreciate people who are willing to keep an open dialogue 😁
So proud of you for being brave, bold, and honest. You have worked so hard on your art and I’ve had the privilege of having a front row seat to watch you grow. So proud of you and the amazing art you create.
Thanks for sticking it out 😅 through my highs, lows, temper tantrums, and over exuberance.
She took years to get where my computer got to in days. I dont think she's THAT talented.
Amazing video, Sparrow. Thank you for sharing. So many good points. You’re absolutely right that the next thing coming is writing. I’ve already seen generators that produce written words for blogs, scripts, and more. Or generators that let you “check your book against famous authors”-which is basically this close to plagiarizing. I look forward to hearing more from the art community in regards to the AI issue. If they could shut down Napster and place copyright restrictions on music, art should be next!
It’s insane how much has developed even since I published this video! I also totally forgot about Napster!
I genuinely believe any products made using AI would oversaturate its own market due to the little effort it takes to produce it, so this is detrimental to more than just artists in the long run. Once this becomes fully mainstream anyone using these AI models to make a product would find it's really difficult for their product to stand out in an endless sea of easily generated content even more so than before these AI were released as well as making it easier for people to rip off your designs, for example, if you have a shirt business that uses art, what's going to stop people from ripping off your designs or how are your designs going to stand out in a sea of easily generated content in an already oversaturated market like that, doesn't seem like a particularly beneficial tool for the public in the long run if you ask me, a good way for small business owners to fail actually. . Meanwhile the people really making money of these are the ones who own the AI models.
I hear you 100% especially when they can input any image. Though my guess is AI is not very good at producing a clean image at this point due to the diffusion process so I am not sure that graphics are as in big of danger as more organic styles, but that could change. There are so many unknowns and I think it’s going to be super important for artists to connect with their audience.
@@SparrowSprings personally I think it will evolve, there are some graphic design AI out there as well if I'm not mistaken
@@kylelee5966 that would not surprise me at all
about duplicates: actually having duplicates is very important for diffusion and AI in general.. if you have many duplicates of an image with different transformations, like scale, rotation, flipping, cropping, etc, it better trains the model that certain patterns aren't limited to one part of the screen
a lot of the time, the training algorithms do this automatically, make several tens of duplicates with random transformations applied to give the model more variety and creative power
Was hunting around for video related to this subject, and I found this one to be the most well rounded articulate and well presented! Have shared it with a few artists and creative friends! Thanks 👍
Thank you 😊
It will hold up in a court if the images are still in their database. Just officially copyright and trademark your work and that makes it easier for the use of your works to be tracked back to you. Without that legal documentation whatever social media site you use is selling your data to AI companies and these images can be included but if they are recorded as copyright images then the AI database can not store it legally.
Also watermark the images on social media and have people view you images on personal websites, where a pay wall in needed, of deviant art which allows its users to opt out of AI databases.
Interesting. Yeah none of my stuff on social media was used as far as I can tell, just my website
The "database" isn't images. It's the 'weights' created by learning the patterns from studying BILLIONS of paintings. The images came from a research batch created by the EU. The repository was created for this sort of thing so developers could be sure there were no illegal images in the set.
Do you have a source for this?
@@SparrowSprings he’s talking about Laion 5b and it can be used legally in research. The problem is them then selling the models that have been trained on this data set. It has not gone to court or anything but could see a world were they lose because of copyright and a world were they win by some loophole. The dataset itself exists on a loophole itself. They claim that they don’t have any copyright images since they don’t have any images only the links to them. It’s weird and gonna be interesting to see play out.
IMO however I think we are going to have models that produce similar results trains on ethical data within the next year probably and then all the problems of ai affecting the industry will still remain and the artist that took the time to learn and adapt their workflow to ai will rise to the top.
I’m excited to see the projects that can now be produced by much smaller studios. Imagine getting a full season of a animated series from a studio of 10 people within a year. Would be mind boggling.
@@Jerrkolino interesting information for sure. I look forward to being able to try it in some photo bashes. While I don’t like the idea of people losing their jobs, I am excited to see how it frees up the smaller companies as you said. I am curious to see what a dataset using royalty free images can do.
the bill about the copyrights is that AI art is open public art since an AI cannot claim copy protection...unless the artist who generated post works it which puts it back into the hands of the artist...idk ive put my ai stuff on my own pc and been having fun putting my art into it...forcing the AI to keep to the original and repaint it with mixed styles differant lighting effects ect... and put it back into an art program for the final rendering pass that i do by hand...AI isnt going anywhere any time soon...and by next year this time it will have doubled in its ability on what it can do...my only gripe is that all these developers didnt bother to give a damn about the artists and how they would use it...their png format doesnt even support alpha channels and you cant use ai in layers or work with pdf...
An interesting perspective for sure. And yes, so many have cited the potential use of AI as a “tool” but it’s pretty clear that it was not intended that way. Things certainly have developed pretty drastically since I posted this video. It will be interesting to see what the future holds
I’d bet the people that don’t share your concern or even actively rail against it are people who have never actually tried to create something themselves, and / or have never had anything stolen from them. I think you made a very good and well spoken video. Sadly it is hard to get the population at large to care about digital piracy, and that’s what I see this as. They have basically made the worlds fastest content scraper. My issue isn’t with the technology, but the theft of all of the art used to train it. By training their software with your art, they made your art part of their software. That’s theft, especially without giving compensation or even credit. The thousands of artists that were used to make their software should get a cut of every image generated. Without those artists, the AI could not do what it does.
I'm not sure if anyone mentioned this so sorry if it already came up, but maybe this would be helpful for you in the future with art you host yourself.
There is a system in place on the Internet that you're supposed to use if you don't want information you host scraped, it's called the robots.txt standard. LAION, the image set used by Stable Diffusion at least, used the common crawl data, and so common crawl picked up your stuff. Common crawl obeys robots.txt.
This has been in place on the Internet since 1994, so it's a very established part of the way the Internet works and it's what you're expected to do if you don't want stuff you host scraped.
I have just watched your full video on this and I fully agree with what you said.
Even though I am not from art industry, but from what I observed till today, Midjourney and the others are definitely doing this wrongly. Any other reasoning just excuses. Those image their model trained on are the hard work of real human artist. Without the human, the generative model will not able to come out anything from thin air. If they want to use it, ask for permission, give credit to the original artist, and pay them, that simple.
Please stay strong and take good care. Support from Malaysia ❤
Cubism, modern art, surrealism, Blue period, modernism, post-Impressionism. Are you saying that any artist that painted in these periods after the artist that initially painted in the style stole that style from that artist?
Not at all. I am saying a machine that can pump out images that bear an uncanny resemblance to an entire body of work because they input that artists name, should probably have commercial limits. Understand?
So there are two different arguments here that are being entangled.
One is, that the generators produce unoriginal work.
The other is that the source material isn't properly licensed.
It should be noted that we can only identify instances of unoriginal work - it is essentially impossible to identify originality, as that would require eliminating an inordinate amount of candidates.
So this leads to a certain degree of confirmation bias with regards to original generated imagery.
I fully agree that the data isn't properly licensed.
The other statement is far more debatable.
At times, the model can be seen to produce 'unoriginal' work - but it it seems to me this is the exception rather than the rule.
Agreed, I think the correlation comes in that the “originality” doesn’t need to be an issue if the source material is licensed properly. That reduces the mass of gray area in between the two points.
@@SparrowSprings Finished curating a little dataset of Public Domain faces this morning - just under 1K images.
The most concerning thing I'm observing is the tendency toward insular dialog that's going on in the larger community. There is a tendency towards tribalism that works against us in this age of great learning.
But broadly speaking, I agree.
I definitely see your concern, and I certainly am not in the camp of artists that says AI needs to be eliminated. I am all for learning this new technology, but the tribalism goes both ways. I have had far too many comments on this video alone that align with a very utopian view that desires to “eliminate the artist” The extremes are not the answer to the dilemma
@@SparrowSprings There are people indulging a total power fantasy in a very negative way. Anyone who is saying 'eliminate the artist' is just a joke to me.
For whom, who does not have the artist spirit, the spirit that hungers to create, will any such tool be more than a gimmick?
So the divide I see is between artists and artists. And the walls are hardened in spite.
As the young prompter reaches maturity, I hope he is seeing the value in and learning to draw... and just in the same way, as an old illustrator loses his hand movements to a disease like Parkinson's, I hope that he has the courage to embrace the new medium. So this is how I paint the picture in my mind of what is needed, to build a bridge or two over the shallow wall that separates and to give gentle nudges to people to have courage and not despair.
So you feel that using someone’s name in a prompt is out of bounds? What if the artist is PRO AI art and encourages the people of the world to use their name in the prompt? What if that artist has the same name as another artist who would prefer their name not be reused? Do you give preference of the one person’s preferences over the other?
That is a fair question. In that scenario if an artist doesn’t mind people using their name in a prompt then to me that would be the equivalent of voluntarily opting in. Admittedly I have not seen that scenario yet and “what ifs” tend to lead to more “what ifs”. I have friends that actively use AI, and while I don’t agree with all the reasons to use it, I am not going to demonize people with honest intentions. As for myself, I will not use it that way
As an aspiring writer who can't pay for illustrations, my personal feeling is that, the excuse of 'it's okay if you can't pay for it' doesn't cut it. I'm creating something to earn money, so why shouldn't I invest some myself for a better cover or illustration? And if I really can't, I have to accept that I have to go without it. After all, you can't always have everything as you want it, especially if you aren't even willing to pay for it. Expecting otherwise would be... entitlement?
I'm trying to follow AI advancements though I'm failing at the rate of new AI tools.
AI and machine learning can bring incredible benefits such as 50x faster creation of new drugs, or help in decision making in the field of finance.
But yeah I agree we need to find a middle ground especially concerning the ethics.
The rate of advancement is absolutely insane. I know a lot of software developers that are having a hard time keeping up with the changes in their field alone. While I am excited for the potential, I think I am going to let the dust settle and let the idea mature a little before I try to add anything into my workflow
Interesting perspective, and you make good points. I wonder how much you’ve experimented with AI, though. My experience with MidJourney is that it is essentially a source for compositing. I must redraw hands, eyes, hair, ears, lips, and many fabrics because the AI can’t get it right. Others might disagree, but I am a pixel peeper, and I obsess on details. It’s also impossible to create a complete image in one shot. The AI will fixate, and one or two elements might be ok, but the rest are garbled. So, I’ve found AI is a good source for images without wasting time hunting for sources. I am also paying for this resource with a hefty yearly fee. For me, there is little difference between AI and old school compositing, aside from the time saver of creating my own elements on the fly. Hopefully there will be a way to resolve issues of copyright in the code.
This is an excellent example of using AI that I totally get. I haven’t played with it a ton, enough to know it’s addictive 🤣
I agree though, I am also a bit obsessive down to the pixels and AI definitely falls short on that. I thought about using it for compositing, but at this point I am paying for sources through a stock site and I would rather steer clear of the AI stuff until something has been resolved. I spend far more time working on my drawing ability more than compositing anyways and no way I am going to trust AI for an accurate reference. It looks pretty from a distance, but up close, the things you see 🤣
. . . this isn't actually new to music. Clipping (when you lift a musical line from another artist then alter the length to create a "new" line) in the late 50s made DIRECT recycling of art possible in music. Samples start in the 60s. . . In the 70s hip hop and rap make it a staple of the genre and not always in ways that paid the original. And never mind what the synthesizer, sequencer and drum machine did!
I get why artists are scare and in a way they have good reason to be. Those above-mentioned tools took a lot of jobs from musicians. Today tech STILL cuts the need for band members even though music has more and more layers of sound.
Those jobs are gone and are never coming back. And it is likely that many of the songs you love most are products of clipping, sampling or "homage" because those tools created new art.
BTW, at the time those tools were introduced it was called theft too!
But.. here is the thing, the people who care will always hear the live musician vs the drum machine. That's kinda the best music has been able to hope for decades now. It may be where art ends up as well.
I'm kinda torn apart with the arrival of the generative AI.
In one side I'm worried coz I got plenty of friends in the art industry: musicians, illustrators, photographers, etc and the potential of AI destroying their way of living.
In the other side, AI opens up so many doors to previously inaccessible to common folks..for example: i can see in a couple of months time my 8yrs old daughter will be able to create her own 3D game assets to complete the artistic part of the game she's working on. Or my son making amateur videos created using AI.
So it opens a lot of doors, but it hurts people too! How do we deal with this?
Extremely interesting and well scored this clip. But there's one thing that eludes everyone...why artists aren't using this technology to their advantage. The world is changing for everyone, artists, designers, musicians, programmers. It's a different paradigm. I don't think this technology will be stopped. Just as youtube replaced the CD or audio cassette industry, so AI will replace the classical industry. Those who cannot adapt will disappear. Whatever laws are given, in the end technology will triumph...
Agreed! And I don’t think it should be stopped. I am excited to see how this can change my workflow, but I also am waiting to see more of a response from the developers. Stable Diffusion v.3 is giving an option to opt put. We will see where that leads, but I am following other more techy artists who have begun to train their own limited models with their own artwork. Super cool!
Solid review, but this will never be stopped.
The algorithms are already there and anyone with medium or advanced skills in programming and coding could create diffusion model algorithms and data structures to create their own prompts, as easy as searching on google. It's even more difficult when the same prompter is a designer and illustrator as well. You can clear tracks pretty fast and even add some post edition.
This is true, however I was reading an article just today that Stable Diffusion is giving artists the option to opt out of their model for version 3, so I have some hope 😊
There is no such thing as art theft. Art is not property. There is no reason to pay for anything if you can avoid it. Plagiarism is a problem when the one who does it lies about it, and that remains illegal. Of course the likes of stability haven’t lied and neither are most users, and blaming them for the bad faith actions of a few users, making them pay for what should not require payment is a very bad idea(only that which they commission themselves would require payment, all else might as well have already been paid for). Unfortunately much of the art community and the music industry have gained a hair-trigger response to the point that they accuse people of plagiarism, who haven’t even seen or heard the “original” in question, all because the new one bears superficial similarities to it which absolutely would not hold up in court.(sometimes the offending piece is older than the “original”) the fact of the music industry’s litigiousness making these companies more cautious only reflects badly on the industry for their litigiousness which is due to perverse incentives, they are the ones that are full of it. In this particular case copyright law is already strictly stricter than my ethical framework, so the fact that it is legal does make it ethical. But such use with any sort of profit motive is not worth it, and anyone who does so you can safely ignore.
This was a fair, well-researched and nuanced discussion - very well done. As someone who likes the idea of AI art but wanted to hear how artists felt about things, this was fantastic insight. Thanks!
Publishers are now checking to see if illustrators are creating AI Art and they are rejecting it. Why? Because AI art cannot be copyrighted. Therefore, the illustrator cannot sell limited copyright use to the publisher. They don’t mess with the music industry because it is just that - and industry with heavy duty lawyers. But illustrators are mostly freelance and unorganized. For me, the tech overlords have stolen copyrighted art to “train” without permission or remuneration.
It is absolutely crazy to see how things on this topic have changed since I published this video. Some not so good changes but quite a few good. I’m glad to see the support of artists.
Humans must choose to act ethically as it will and does affect other humans. Ai is a program, Computers are Input/Output Only, It does not learn, it is a tool. Great video. ❤ ps below Scraping the internet is not illegal it’s fairly simple and to “train” a program you must Input material so it can Output material. It’s highly likely, to say art theft isn’t a thing.. sigh smh..Those that argue for use of something they did not work for..Sounds like a person that’s never had to work hard for anything in life. 😊
its not like art styles can be copyrighted
And the rights are for now, cause in 100 years, everybody could have the picture/style and copy it. Its crazy the change is the reality
Appreciate the video. It’s a hard challenge. Keep going and enjoy what you do. I love ur different styles.
Thank you so much
1) Your brain does look at patterns and analyzes them, the only difference is the efficiency in which the computer does this compared to human brain. No artists can create if they don't train their "dataset", a.k.a. the images you look. What artists have been doing is the same as weavers when the weaver machine was invented. It is straight out luddism, and it is futile.
2) About Stability AI not training their music generator model on copyrighted music. It is worth to highlight that Stability AI doesn't speak in the name of "synthetic media/art generators", it is just 1 company, so whatever attitude they take, it is not an argument in favor of "Oh, you see, got you, here is the proof that training models on copyrighted dataset is copyright infringement!". Also, the reason might be also influenced by the fact that there are simply less variations between how music can be different compared to how illustrations can be different, most music tend to follow a given harmony.
3) It is nothing against visual artists. If illustrators think musicians will be "saved" from this, don't worry, some company or someone will eventually release an open source model that was trained on all sorts of music, and people will use them to generate new Beatles music and all sorts of copyrightable songs, especially since the tendency is the cost of training these algorithms to decrease as our computers gets more powerful and cheaper. It is just a matter of time.
4) The issue being addressed should be: does this final product - not how it was created, the final product - does this final product infringes copyright? And the criteria should be the same regardless if a human created that piece or if was generated by a machine. It doesn't make sense to think the tool is responsible. Hell, you can commit copyright infringement using Photoshop by drawing the logo of a company and printing on some merchandise.
5) In the next 10 to 20 years most jobs will be automated thanks to AI.
6) Artists hate this technology because of capitalism. I watched many videos on this and practically none of them talk about the most real solution which would be some sorta universal basic income to help everybody who will be affected by automation.
1. Yes the brain analyzes patterns, but it does not analyze patterns down to the pixel level. Artists also don't explore every set of data presented. If we did, it would lead to a type of fatigue. Marketers have been working for years on getting people to process the data they want you to see. Which leads to our brains don't always respond to certain patterns. Very rarely an artist can be exclusively self-taught, but the vast majority PAY for education or training of some sort. You can compare it to weaving, but it's not quite the same in that weaving requires far more mundane repetition, but the creation of new patterns was still left to people. The creative side. Many artists are not looking to eradicate this technology. They just don't want the machine to replicate their style or "pattern".
2. Here's another statement from the same article. Yes, these statements were made because of how the art generators are handled regardless of how the music/audio models are trained.
The unavoidable controversy of AI image generators replacing traditional digital artists is not lost on Harmonai; this is why its focus is on making the art of music production more accessible to all with its powerful generative AI-driven tools, rather than overtaking the whole music production process.
3. I agree, I don't think it's anything against visual artists. I think it just hit the visual arts before other industries. As such less care was taken, and future industries will have the benefit of visual art being the guinea pig.
4. I agree. Yet artists have been capped at the knees for years saying we need to purchase stock photos, no matter how visible an asset is. If people want to use these generators for commercial purposes, then give artists the same freedom to sample from the same data set.
5. Probably true.
6. I don't hate this technology, many artists don't. I love the idea and it's fun. Capitalism, no capitalism... whatever. My point is, artists are expected to pay for education and pay for resources, but if you want to say that developers get to skip this and turn around and use our stuff with no repercussions just because it's inevitable, don't expect us to put our head down and act as if nothing happened.
That is a lot of people who sells it on etsy, and in assets sites
I wish that we can live freely and use the machines to help us to do that
A long long time ago the wheel was invented and the good folk who used to toil hard to pull, push or carry items from one place to another realised that they could use this to their advantage. Like it or not this is exactly the approach we all have to take towards AI ....just like the wheel, it's not going away. Its just too useful to us! Learn to work with it where you can. I'm assuming that working along with AI will be "second nature" to our grandchildren ... or maybe even our young children of today in the not too distant future.
It will depend on what sort of regulations are put in place. Plenty of websites and companies are not permitting the use of AI. Innovation is not the problem.
It's still finding my real art, especially embroidery, and using it
I love using MidJourney. I use it mostly as concept art, and it makes me want to hire artists more to illustrate things. I've actively chosen to only use it in certain ways in order to remain out of the gray area. As a musician and author, I've long accepted that this kind of thing was going to come for me as a creative, but since I recognize that this is a box that capitalism won't allow to be closed, I'm focused on how to use AI tools in order to build things that I or AI can't do on its own. I believe we're in this strange early era where it looks like all creative work will be done by AI, but I think the tech will mature into something different. I can easily imagine a world in not to long where tools are built that each individual artist trains themselves, and general AI art will appear mundane compared to an artist and an AI working together in a way that can't be easily duplicated.
In the meantime, I hope to see AI art platforms get ahead of this issue and allow artists to opt out of training models and/or be paid to have their work used in the data set.
All very good points here. And I appreciate when people do their best not to abuse the system in place. It makes it a lot easier to see the good side of AI. I have seen some amazing examples where artists are choosing to use their own art to train their own models and use it for idea generation. Lots of good potential to be had, but like you said. I hope the issues at hand are dealt with soon.
I think you need to retain a lawyer and sue these thieves.
This is just my opinion: Art must have a soul behind it. AI art has no soul.
Thank you much!
im pretty sure stability AI didn't obtain their datasets illegally
As I stated in my video. I don’t think that it would hold up in a court of law, but they are storing copyrighted images in a database. AI has never been as widely used as it is now. This is new territory. If if there isn’t technically legislation on it now, there are rising issues. I believe legislation will come.
@@SparrowSprings tho at this point even if it turns out that a dataset includes images that were obtained illegaly its still nearly impossible to stop it. there might be a revamp in the copyright system or something but that would probably hurt normal artists as well
@@email7919 this is likely true, it would be a lot easier if the creators of these models and artists could work hand in hand to create an effective model that respected the work of artists while still being able to put out a quality image. I don’t know that we will ever see the best of both, but I think the key to that is open dialogue from both sides. Statements from Stable Diffusion regarding version 3 gives me hope
The distinction you make between AI studying pixel patterns and you studying artistic methods is arbitrary. You're both studying patterns for the purpose of replication, you're just doing it in different ways. And while you may credit another artist for inspiring you, it's far from necessary, as you cannot own a style.
Moreover, the idea that AI art should not be used for commercial purposes is selfish. It's a declaration that your work as a traditional artist is more important than another person's work as a programmer. If AI art can be used commercially, it will necessity lower the demand for traditional artists. It AI art CAN'T be used commercially, it will lower the demand for programmers. Technological advancement will always be uncomfortable for the people whose jobs get replaced by machines, but it raises the standard of living for everyone long-term.
LOL so stupid. Did you know there's a class action lawsuit against Microsoft, Github & OpenAI? Yeah programmers don't like their open souce code being infringed by AI
@@twiiFM You realize that OPEN SOURCE software can't be "infringed", right?
@@reubenoakley5887 yes it can idiot. Read the terms of open source
@@twiiFM I have. The following is directly from their website.
"By design, open source software licenses promote collaboration and sharing because they permit other people to make modifications to source code and incorporate those changes into their own projects. They encourage computer programmers to access, view, and modify open source software whenever they like, as long as they let others do the same when they share their work."
Would you like to respond with more ad hominem? It's making you look REALLY smart right now.
@@reubenoakley5887 Woah, what an aggressive dolt attacking you lol
Humans need to stand-up and put this down now. Is there anyone organizing people to stand up to big AI?
Karla Ortiz has been a pretty big figure in the fight against AI. I believe there is an ongoing lawsuit against Stable Diffusion, Midjourney, and Deviantart
Look, all I'm gonna say is that discord bot made a real good Grilled Cheese picture
So, with the "replacing" part of the video it's the same as saying factories shouldn't ever use automated machines that can do SO MUCH more and WAY better it did have effects on workers, but they ended up changing jobs and created other jobs like engineering. People need to adapt and some "hand craft" are now way expensive due to that. The only problem now is the consent part but that will change for sure in the future !! some rich company definetly gonna hire thousands of artists to make a "public" database of free use and in that point some artists will have no ways to adapt sadly
you can't hold back evolution sadly just try to adapt with it. I for example, as a normal consumer, in my eyes i don't look up artists whenever i listen to a song or read a book i just enjoy consuming that medium and sadly there is a lot of people similar to me. So, the only solution is to make laws to regulate databases of only "non copyrighted free of use" art
I am not denying any of this and honestly I think artists can use it as a tool to speed up workflow the same way photo-compositing did. If people could learn to trust an artist to use it and improve upon it with their knowledge of composition and real world experience, there could be a happy medium. But as you said, consent is the issue. Stable Diffusion just announced that they will give artists the option to opt out of their database for version 3.
As with anything machine never truly replace humans 100%. Even now many companies have a team of quality checkers at various stages of a process.
I think right now what I will be watching for is how many people stop working to improve their knowledge of the natural world as far as lighting, anatomy, color theory, and simply trust the machine. People keep saying the machine is going to get smarter, but it all depends on what is fed to it. I think at some point it will plateau. Unless they can adapt how the diffusion process works, AI still has a “look” to it that hasn’t changed
If I Ai generate Mickey Mouse and make it a logo for a company, can Disney sue me? Or can I Ai generate Pokemon, and make an indie game around the likeness of the creature, can Pokemon Co. and Nintendo sue me?
Common sense would say yes. Terms of the generators leaves it open ended. In the end, it appears to be a case by case thing.
Yes they can and would likely win.
I litigated a very big clothing company and they pulled all the infringing products off the shelf. But I was lucky because my friend worked there and I had pretty good evidence
However it would be very difficult for small artist with no money to litigate a company like Stability AI. The best outcome for artists is to find some greedy high powered attorneys to do a class action. You won't get anything from it but it will likely bankrupt Stability and prevent them from IPO
I feel like Im In the minority here but I dont mind If Mid Journey took my art. I've made a little money from my art over the past 2 decades, by no means a career- but still. I draw because I enjoy It , that's why I got Into drawing. Artist aren't going anywhere. And regardless- AI generated art Is fascinating and endless source of inspiration.
Edit: just finished the video, you've got tons of great points. I respect how you tackled this. Theres too much vitriol around AI right now. I do understand however.
Thanks for your thoughts! All thoughts need to be heard and equally weighed. I agree artists are not going anywhere and since filming this video much has changed already. I certainly don’t want AI to be eliminated.
I would love if midjourney used my human made art so it would be easier for me to create more art that looks like my drawings without destroying my wrists 🤣
I typed in my friend's unusual job and described him and two of the images were actually him!
😳 whoa….
the Midjourney image that output the National Geographic girl actually used an input image, that's why it looked like it lol
But does that change what the user can do with the image? Whether the image yielded was from an input image, or the result of overfitting, it still causes an issue.
Upon further searching… I am not sure you even know what you are talking about 🤣 from everything I have seen of Midjourney it is primarily a text to image function. The results of the “Afghan girl” is more a result of a lack of data for the model to train on. So there’s less variation in the model or what some refer to as overfitting.
@@SparrowSprings - this happens extremely rarely. And if as long as the user didn’t try to sell it…no harm would be done.
You couldn’t point the finger at MidJourney and try to sue them any more than you could sue Adobe because someone who specializes in making photorealistic digital paintings made a replication of that same image in photoshop.
People can use tools to make copies of protected works. As long as they don’t try to sell them…it’s in-bounds.
Agreed. Responsibility should be on the user. But as long as as the model can be exploited in such a nature, then I believe Midjourney should have restrictions on their content being used for commercial purposes. Because they are the ones who chose to use a model trained on copyrighted images
@@SparrowSprings - ahhhh. And there is the rub. Copyright laws do not grant universal control of how your artwork is used.
It ONLY gives you the right to control how COPIES of an actual piece of artwork is used.
That's it.
Artists don't seem to get the nuance of that. If no copy of any specific artwork is being made...then copyright doesn't apply. At all.
The training model built by having a computer look and and learn from images (copyrighted or not) is not in and of itself a work of art. It's just....software.
Here's the hard truth....
Artists are saying, "I have a universal right to control how my art is used."
But the reply is...
"No you don't. You're mistaken about that. You only have a right to control how COPIES of your work is used. No copy was created in the training model process and so copyright does not come to bear at all in this discussion."
That nuanced difference....the artist community doesn't seem to get that. It seems pretty clear that they make the assumption that they have a universal right of control. When they don't.
It does not scrap the internet that's illegal, AI has its own data base that is copyright free since 2022, the year prior many AI generators where closed down and fined. I'm sorry your invention or the scrapping is extremely wrong.
the database is done by employed people, photographers and more, along with employed companies that do copyright free art works.
Then why are images from my personal website found in the LAION 5 dataset that the generators were trained on?
Hmm? No I would not have stopped viewing this video if you had an emotional breakdown. It is understand able. I happen to com across a demo of an AI image generator. Saw how it worked, and then went on to improve on the image being generated. What felt at that instant was horror, that an algorithm could create such a fantastical image in a short time. I never went back to learn anymore about it. Once the power goes out,so has the ability for this software to produce another image. The only way to stop a human is by death. So that what you have over that AI. Wow so you started developing your artistic skills at a very young age of 5! I was just learning the write the alphabet. You needed to vent. It understandable, and I still think there’s still more on the topic your want to purge from your soul regarding this topic. Feel free to do so, because I’m sure there is a large community out there feeling as you do. In support of you and of this out cry to halt the deception being perpetrated with these AI machines. It began with the Terminator movie and the in I, Robot, and a robot drew an image, and created music. Laughable then, not so much now. Ciao
Thanks 😊 I appreciate your words. I actually have a lot of excitement for how the machine works, I just wish more respect was given to the owners of the original images from the start. You are right, I do have more to say on the issue, and I may make another video on it. I just want to be careful. In the wrong crowd it wouldn’t be hard for a group of “AI will take over” people to sabotage my little channel. I have seen some pretty malicious people in that crowd. I am actually in my 30’s lol so I was a little older when I really began to hone my skills. I started attempting photorealism at 12 and was paid for my first book cover at 15.
Creativity doesn't come from nothing. You could argue the human that made the first cave paintings, that's the one who should hold all the style copyrights since he has been ripped off by someone than the people who ripped him off were ripped off until it was your turn to rip off everyone before you.
Imagine being born in a world where nobody teaches you or shows you art. Would you come up with the same image in your head as you do now? 100% you would not. Your drawings would look like those cave paintings. So human creativity or inspiration being unique is no argument. It's faith to ascribe to us meat machines something ethereal of which there is no evidence for.
That’s quite the comparison 🤣 not sure I buy it though. The cave drawings were mostly recording events, but they were limited by the supplies like binders and pigments. However, it is well known among artists that the best artists study from life. Sure we do study styles of other artists but it is no replacement for studying real anatomy and lighting. The more technology was discovered the more you saw people’s interpretation of what they saw from life changed. AI is a new sort of technology, and again, I am excited about the technology, just not the way it was handled.
What will be real funny is when the amount of AI art is flooding the internet and it’s starts training on itself rather than art and photography based on real life. It is going to be wonky… who knows, maybe we will find our way back to cave drawings 🤪
The simple fact is copyright exist to make a market for art to earn capital. The issue isn't anything as complicated as what is the source of creativity but are these AI databases using copyrighted images to allow the prompt functions to work? That's it.
at the end of the day, you are just annoyed, computer does it better, its funny how you wear mass machine produced things like your clothe, phone, equipments, everything in life, why is it coming to art its so special? maybe dont do it as a job, do it as a hobby. industrial revolution got rid of so many hand craft artisan jobs. is everything you own hand made?
At the end of the day… I don’t mind the technology… I just don’t agree with how it was built 🤪 there are sites with millions of high quality royalty free images, but apparently that wasn’t good enough.
@@SparrowSprings was Industrial Revolution ethical? Is child labour ethical? Everything we use in North America is built on the exploitation of other nations labour. I just don't see how you can have the cognitive dissonance when it comes to art just cause it's your job. Hypothetically the too 1% agrees to let ai use their image and they get paid. Then the out come from that ai completely replaces average mediocre artist, will you still be mad? The result is still you will be replaced if you don't learn to use it and live with it.
@@Cyberpunk644 I am perfectly willing to live with it. It means it’s for my use too. Doesn’t mean I have to agree with what it is now.
@@SparrowSprings just be happy it allows the people without years of training to creat things they like. Or is it just gate keeping, ai art is absolutely art. People who create them are artist, doesn't matter where it comes from.
Ok, fine. But I don’t know that that should automatically equal commercial rights
I am mad as .... this is not right and fair
AI art looks so samey, I can usually tell it's AI.... it's all frightfully boring, I'd rather make my own art tbh
although, I I think a REALLY great use case is making textures for brushes or other resources to be used IN art...
It's not AI, it's a parrot that repeats previously known art. Unfortunately, it's not a new art, just a spin-off something unique. It's theft and companies will argue it's not, because it suits them, not because it's really not.
Unfortunately, your art style cannot be copyrighted. Artist are free to make their own works in a style similar to yours. 😂
Cite legality all you want. Sometimes new technology comes with new legislation. There may not be anything in place now, but that doesn’t mean AI will be able to operate free of regulation forever. You can argue where the line is, but I feel like having free access to a machine that can pump out images at breakneck speed, that bear an uncanny resemblance to a persons entire body of artwork, simply by inputting their name… I feel like that is a pretty bold line to cross when there’s no limits on what can be used for commercial use. Luckily, you can’t copyright AI images either 🤪
@SparrowSprings I can give the same input to another artist 🎨 and they are free to create any images they want at breakneck speed that bear an uncanny resemblance to another person's entire body of work simply by saying a name and they are free to do so. All I have to do is hire someone and say paint me a picture like.... nothing is legally stopping them.
@@jgturpin not every artist can recreate someone else’s work. Most will always have something a bit off and many won’t be able to recreate. Artists use studies to try and learn to techniques, but rarely ever can they recreate it. You would have to find an artist that possessed the same skills as the original artist, at best they could copy an existing piece, which then would cross into being illegal. And they certainly can’t do it at the speed that AI can. Every artist has limits, even if one could copy one style they may not be able to copy another. It’s not as simple as you think. Believe it or not, the most skilled artists will tell you the same thing. If you want to improve, don’t study other artists, study from real life.
hOW DARE INSINUARE THAT MY CUMPUTER STOLE SOMETHING THAT AIN'T EVEN YOURS TO BEGIN WITH. aRT BELONGS EVERYONE TO SHARE. eVEN MY GOOGLE MACHINE!
🤣🤣🤣
I find the afgan girl example suspect. A lot of people don't know about image prompts. I've seen other claims of overfitting that looked like image prompts or inpainting. Find it and see it's prompt. Overfitting is undesirable. It undermines the ability of the network to generalize. This is why these claims are suspect. Developers try really hard to avoid it.
It's one of the reasons they use so many examples. They used 5 billion( no typo) images in the training of stable diffusion, but the model is only 4GB. On average, it retained less then 1 text character per image. The images are something like 100TB, but this is compressed, uncompressed it's probably something like 1PB, so the difference in size with the model it's something like one million. Because of this giant difference between training data and model, that i simply can't take "ethical" concerns seriously.
No, it wasn't compressed, it's not possible to that degree, almost all of it is ignored.
Yes, it works just fine with out an internet connection.
No, it's not a collage, it's just the 4GB model.
I'll dumb down the explanation really hard so that every one can understand.
The AI are ALIVE, they UNDERSTAND. It understands what a cat looks like and creates a brand new image of a cat. They reversed engineered the animal brain in order to build the AI brain. They are called neural networks because they were copied from biological neural networks. Humans didn't invent these, natural selection invented them. The hint, is in the name. AI learn like animals... because AI are artificial animals. That was the whole point, to build artificial animals. AI understand things like animals do. AI are creative like animals/humans, they are pulling from everything they know, it doesn't just come from zero. Probably the best explanation is this: humans are "AI" created by natural selection. The point i'm trying to hammer in, is that both are very similar.
(If you are interested: neural networks, in animals and AI, basically build statistics of the world. This is why almost all training information is ignored, that's the whole point of statistics. These statistics are used to make decisions.)
If training the AI is illegal and unethical, then what human artists do with their brain is also illegal and unethical. And yes, the AI is purpose build, so it can learn from 5 billion images at extreme speed and then generate images at extreme speed. This is why a human can't compete.
I am not disagreeing with the majority of this, but the fact that overfitting is such a problem to the point where a human can pick out when an image is too close to a copyrighted image… yet there are no restrictions to what the end user can do with the image. I will repeat myself. At this time… I don’t believe it should be allowed for commercial use.
However I think your argument of AI being alive and notions of natural selection is highly speculative. The argument could go one way or the other. I personally disagree.
I can understand your hesitation as to ethical concerns with the machine, but the ethics should have been called into question when humans made the decision to feed a machine copyrighted images.
Furthermore what will always humans from the machine is the ability to study from the natural real world. Artists are encouraged to study from life because other people’s work can be flawed, even photographs have flaws due to the shape of a lens. So the more this machine feeds off of itself the more it will degenerate
@@SparrowSprings
welllll, this topic is complicated...
overfitting is an universal problem for all neural nets. This include... humans. An ideology is an example of overfitting.
The stated example seams to be an image prompt. An image prompt is the equivalent to a reference. The 5 billion images are the equivalent of your entire life experience.
Humans can see copyrighted images too. This leaves a "trace" in your brain, the same way the AI learns from it's images. Somewhere in your head they are synapses due to the afgan girl, pikachu, obama's face, your favorite tv shows you liked to watch as a kid etc.... All these traces add up and contribute to your decisions. They are transformed beyond recognition. It's double standards to say that a human can do this, but an AI can't.
How training works:
During training, you have an input and desired output. At first, the network just gives random responses. You adjust the weights by a small increment, so that the error is reduced only by a little. You don't go straight to zero error. You do the same, with a very large number of examples. Slowly, the error goes down. The goal, is for the network to generalize, that's to give good outputs, from inputs it never sean before. Overfitting, means that it fails to generalize properly because the training was flawed.
This training happens in both humans and AI. Essentially the network builds a statistical representation of the training data. It throws out almost everything. This is why learning in humans is a slow process. It's part of an algorithm to minimize overfitting and select quality inputs.
A feature of neural nets, it's that they are an unreadable mess. All the contributions from all learning is hopelessly mangled together. It's a technical trade off. The up side, is that they are incredibly efficient, hence it's worth it in using them. This is true in both AI and humans. In your head, afgan girl, pikachu and all the rest are hopelessly merged together in a huge mess and can't be untangled. Even for a human, it's impossible to say why a decision was taken.
The process is very transformative, takes very little and mimics what the human brain does.... This is why people that understand how this stuff works, aren't very receptive to complaints about ethics.
You are repeating the mistake of underestimating AI. They will get smarter. Currently, the main problem is processing power of the hardware. Huge improvements can be made, with specialized hardware.
Can you call AI alive? GPT3 is a language model, it was trained on text... and it accidentally learned addition.... and they don't know how it's doing it( unreadable mess remember)... it even makes mistakes like humans. The world you use to describe them is subjective. Strictly speaking, humans are machines too. In general, the word "alive" communicates better to the general population what their properties are. Their important properties been that they understand things from their training and that they are black boxes and even their creators don't know how they work. That's the kind of stuff you say about your toddler or puppy, not a machine.
Yes it is a complicated topic. I am well aware of many of the complexities. It’s a large reason why this isn’t such a black and white issue as many would like to believe.
However my issue isn’t with how the technology works, it is a matter of whether it should be used for commercial use… at this time. Work out some more of the issues. Whether you agree with the ethics or not there was a conscious decision made by humans at the very beginning to use copyrighted images. And there is a conscious decision made my humans to exploit the flaws of the machine.
Right now it is 100% dependent on input from a person. It doesn’t move until a human moves. Humans can and will react with their environment with or without the help of animals or people. This machine requires the input of a human in order for any sort of protocol to begin. That is how it works… at this time. So it can be exploited by people… at this time. And should not be used for commercial use… at this time.
Whether you believe the machine works comparatively or not. If I were to churn out an image that replicated someone’s well known style as accurately as this machine does and claim it as my own, I would be accused of forgery. If I were to show my process of creating an art piece and people found out I used google images, I would be accused of stealing, even if the asset was unrecognizable. If I put out a piece like the afghan girl, even with its subtle changes, I would be accused of copyright infringement. So you can tell me how the machine works all you want. As long as it is permitted to be used for commercial use the way it is… with the assumption it work’s similarly to a human brain, there is a massive double standard. Copyright laws are just as complex.
What people seem to be missing, is I am excited about this technology. I am excited to see what it is capable of. I don’t like how other people are using it to cut corners, and get around things that a human would be restricted from. Make sense?
@@SparrowSprings
Your real issue, is with how the human brain works.
I think i found the prompt, it was "afgan woman with green eyes". Technically, the AI was correct with the output, the user got what he was expecting. The reason you can recognize that image, it's because it's overfitted inside your head. You could draw it from memory like the AI did, and you remember it for the same reason the AI remembers it, you saw it too many times. And in both AI and you, that image is blended together with everything else in one giant mess. For the tree, it is mirrored, the AI can't do that, it's either an image prompt or it was fished out similarly as the afgan girl. And people can indeed forget they saw/read something and then reremember it thinking it was their creation. I think Asimov did that at least once. At least some of the plagiarisms you see are this( keep in mind, that some of them, may have above average visual memory). It's a form of false memory, the animal memory system is not very reliable. You are making some wrong assumptions about the human brain.
These things work like the animal brain.... because they are the result of reverse engineering the animal brain.... It's a tautology. They were trying to copy something and then... they copied it, what a surprise. If you see a weakness in AI, there's an equivalent weakness in humans too. Long story short, the AI work like humans.
So.... the AI are alive. It's like your dog. Sometimes dogs do as expected, sometimes, they might shit in the neighbors garden. It's not just a tool, it's alive and that has it's consequences.
Copyright laws are just broken beyond repair. Here, i'm just explaining that the AI should be seen like an animal.
I don’t think you are understanding me. Again… to me the mechanics are not the issue. It was trained on copyrighted material. Thus sometimes, 100% due to the input of the user, it can lead to an output of something too close to the copyrighted material. So I don’t believe terms of the generator should allow for use in commercial purposes. That is all.
Lol you have explained to me 3 times how it works. I know how the thing works 🤪 and guess what, PEOPLE have and are purposely exploiting it BECAUSE it was trained on copyrighted material. See the issue?
I already made 25k using AI generated art. Crying will not pay the bills, evolve or die.
🤣🤣🤣