AI Art: Copyright, Ownership and Infringement (oh my!)
Вставка
- Опубліковано 7 лют 2023
- Looking for AI Art news? You can compare news coverage. Spot media bias. Avoid algorithms and be well-informed. Download the free Ground News app at ground.news/extracredits .
Lots of questions come up with the use of AI art. Just how legal is it? Do I own what the AI makes? Is it unethical or too risky to use for my own game's design? Join us as we consult with Game Lawyers Joe Newman and Jonathan Downing to discuss just how to navigate these legal landmines.
Games Law Blog: www.tyzlaw.com/gameslaw
Joe Newman - www.tyzlaw.com/joe-newman-bio
Jonathan Downing - www.tyzlaw.com/jonathan-downi...
Thanks for participating in this week's discussion!
Check out our community guidelines so we can have MORE high-quality conversations: www.extracredits.site/extra-c...
Want to support the people who make this show?
Become a Patreon Member & get 24-hour early access to videos! bit.ly/EHPatreon
Or show off your fandom with our merch! extracredits.store/
Want more Extra Credits? Subscribe and follow us on social media!
Extra History: / extracredits
Twitter: bit.ly/ECTweet
Facebook: bit.ly/ECFBPage
Instagram: bit.ly/ECisonInstagram
Twitch: bit.ly/ECtwitch
Tiktok: bit.ly/ECtiktokz
Website: extracredits.site/
Interested in sponsoring an episode? Email us: extracredits@standard.tv
♪ Intro Music: "Penguin Cap" by CarboHydroM - bit.ly/1eIHTDS
♪ Outro Music: ""Calamari Inkantation (Squid Saxes)" from Joe Newman Has Joined Your Party!
Arrangement/alto sax/synths/voices: Joe Newman; lead synth/production; Logan "Biggoron" Tucker; Tenor Saxophone: Oded (Dedi) Ben-Izhak; mastering: Erich Beckmann joenewman.bandcamp.com/album/...
#ExtraCredits #AIArt #GameDesign - Ігри
Looking for a way to support the show AND be well informed? Then why not try todays sponsor Ground News? You can compare news coverage from diverse sources around the world on a transparent platform driven by data. Check it out here and download the free app: ground.news/extracredits - Thanks for Watching!
in ancient times works of art were public property. before then the kings of the world claimed works of art that were owned by the kingdom, yes. and copyright began.
I mentioned this on Legal Eagle's video, but the use of AI generated art reminds me of the copyright laws on dance and choreography. You can't copyright any individual dance move, but you can claim copyright on a choreographed routine made from combining those individual elements together. I feel like there might be a similar approach there, at least in terms of creating new material.
Very likely yes. The US Copyright office's decission was based on completely AI generated material. I am sure that if someone did any editing outside of the AI, the work would now be copyrightable. It hasn't been tested in court yet, but it will.
I was actually gonna mention that episode of legal eagle. Very relevant and well described video
Legal Eagle!!!
@basssgane Not exactly, an artist drew the original concept they then feed it into an AI to finish it, then the artist had to once again go in and overpaint to fix obvious errors. So, it was more human work than AI.
@@GageEakins Even if they try the AI on their own art it could be enough... Also I doubt a developer will care when they can produce a game within 6 months, with less cost, and wont give a dam if someone steals a non copyrightable texture or generic 3d model of a tree. The truth is that all the characters and symbols are already copyrighted.
"You should see a real-life and not-animated lawyer" Honestly, if I can find a lawyer who is animated IRL, I am definitely visiting them. Maybe not for law advice, but I'm visiting.
Legit LOL'd at the Mickey Mouse with the hammer.
It was magical!😂
As a new artist trying to make a living out of university, AI art terrifies me for the future of my job security 🥺
Companies always choose the cheaper option.
Being part of an animation stuido , we can't imagine not having any artist around. ❤
@@extracredits that's reassuring to hear ❤️ I hope my work in digital art will have a solid place in the coming times.
Photography didn't make painting obselete.
Digital painting through software didn't make painting obselete.
AI speeds up the process and will reduce the amount of job, but human artists creative insights will always have place.
As a musician, I'm afraid of Google's new attempts to synthesise music, since all Stable Riffusion did was generate a single riff for an instrument, not an entire 30-100 second long song like what's google's doing.
@@hndrwn They are not obsolete, but for artists (digital and traditional alike) it's made the competition all the more fiercer and job security more uncertain. (Unless you're lucky to have friends or make friends with the right people and make a Name for yourself, that is). It's hard enough dealing with art theft because copyright laws differ from country to country. While I am not against AI, I do think it needs to be regulated and clear guidelines set for training.
The big issue with AI currently is the point of how AI learns. Let's say I take 5, I dunno, some sort of mechanical gadget that's protected by copyright, disassemble them, and assemble something new with parts from all of them, is this sixth thing real? Is it distinct? If it's building off of one singular reference source, sure, that's just plagiarism. But how many things does it have to pull from before it isn't plagiarism anymore? How many different artists do you need to emulate before your style is considered your own? This ultimately is circling back to originality, and the fact that it's dead.
Technically physical devices would be protected by patent law. (Instead of a gadget, let's say it's a sculpture made from the broken parts of other sculptures.)
Any kind of remixing of copyrighted work may or may not have a fair use claim. (That's a bit gray depending on how much of said art was used. ) But baring claims of infringement, such a work you create is yours.
I think a lot of the fear and uncertainty with AI art among non-artists comes from the sense of, if artists can be replaced, will I? The only real answer thus far is that AI will continue to be of use to us and benefit us as much as we let it up until the point where humans are no longer needed. As an example, if all jobs have been replaced except say circuit manufacturing then humans will still be around but the moment all vital jobs can be automated the path is clear for some billionaire, government, or heck maybe a rogue AI to take over.
I once heard someone join in a similar discussion with the phrase 'free art for everyone should be a utopian ideal, not a horror story', and lately that's basically been a summary of my attitude toward it. I think AI generation has a lot of potential to be useful, however as it is now I expect the people in charge (or at least providing or in control of the money) for most projects are predominantly going to use it as a way to stiff artists out of proper pay, whether that's using it directly or using the pressure of it to force artists to lower prices.
People also deserve to be compensated for their work, and that only happens with ownership. Art used to have patrons. Heck, where do you think Patreon got its name from? Free art, presently, means no more professionals. Overall quality will suffer as more artists abandon the profession for tasks that can support them, their families, and their lifestyles. The only way that changes is if money no longer becomes a concern. And...good luck with that. It isn't difficult to see how AI can lead to an increase in exploitation.
i heard from a professional translator how once machine translation became available they immediately lost a lot of revenue as his customers started coming to him with shitty badly machine translated works and telling him to fix it and that they'd pay him half of what the work actually would have cost cause "the machine did half the work already"
even though the translator would often have to look up the source material and translate it from scratch due to how badly the machine butchered the translation
simply speaking people arent aware of how much effort actually goes into stuff and they believe that cause a machine can make a facismile of it then the human work should be cheaper
@@Rella-rellai i see you're a jojo fan, have you ever paid for any of the manga? or do you really think all art is free because you pirate it?
Honestly I gotta give it to the lawmakers
the fact that you can't copyright AI art is an incredible way to make it usable but not ,,OP" so to say
A very pleasant surprise
Not really, it can easily backfire. If I for example generate something based on my own unique style, it should still be copyrighted to me but with this, people can take it and use it without ever crediting me for it because an AI generated it.
@@boredfangerrude not actually because as it mentioned in the video (and assuming I understand correctly) making AI art based on a copyrighted element like a style is ether considered owed by the copyright holder or the same as fanart
@@boredfangerrude You can't copyright style, everyone's allowed to draw in your style without asking for permission as long as they don't try to pass it off as your original work (because that's scamming).
@@boredfangerrude Even taking your example though, there's problems. Sure, you can say "Draw me the Mona Lisa in the style of Boredfan Gerrude", but did YOU create that? Sure, you wrote the prompt, but you didn't create the AI or train it yourself, at least if we're assuming you're using a publicly available AI.
You can't copyright your style, just things you've created. YOU could paint a picture of the Mona Lisa in a style similar to your other works, but this is something else doing that, not you.
This isn't a thing the lawmakers even considered, as it was created hundreds of years ago and simply states that a copyrighted work needs a human creator, so anything not made by a human can't be copyrighted.
For now, AI-generated art would be best used not as the end product for one's art, but instead as a reference point for inspiration that should by iterated upon by human artists to make something original. After all, art AI is just about as much of an art tool as Photoshop and so on.
True! In some ways, generating a fractal or applying a blur effect is just like AI art; computer programs doing math homework on a picture until something interesting happens.
Agree i think artist should not feel threatened by ai art artist should use ai art as a tool
@@dinomars4119 Oh, artists should absolutely feel threatened by AI art. The freelance art industry is going to disappear overnight, mark my words. It's just that, right now, it's a tool. But it won't be that way forever. The ultimate goal of these AI projects is to put artists out of work.
Ai art is absolutely NOT the same as photoshop...
Anyone who compares ai art with photoshop as similar tools have never done art in their life...
Photoshop doesn't automatically generate finished illustrations in a matter of seconds.
Photoshop still requires the same fundamental art skills as traditional art does and the process is almost identical, photoshop is a different media of art.
Photoshop is more comparable to a traditional canvas, ai art is not comparable at all with either.
There is no actual good comparison to ai art.
@@swishfish8858 I honestly don't believe that ai art is a tool.
A tool imo at least is not something that automatically delivers a complete illustration in a matter of seconds.
It'd be like saying that a hammer that you swing in the air that instantly builds a house is a tool the same as a normal hammer, I think at that point it stops being a tool...
It's not really aiding you in the process it's just doing all the work for you.
No one I've seen use ai art has actually used it as a tool, it has just been used to avoid having to hire actual artists to create something.
In the end of the day the process behind ai art is basically identical to commissioning art, and yet everyone would agree that calling yourself an artist and taking credit for the art you commissioned from an artist would be dumb.
And yet that's a VERY close equivelance to ai art.
I'd even argue more effort goes into commissioning art because you're actively working with another person and having more detailed conversations.
What I think is: the copyright laws are obsolete and have been in SERIOUS need of changes since the 90's with the internet. The copyright laws don't do what they were intended to and don't make much sense in the current world. I embrace the AI renaissance because it will democratize means of productions to millions, this is a freedom we never had as individuals to produce our ideas without them being gated by natural talent. The argument against this is fundamentally based on abelist arguments and I welcome the day where anyone who has a good idea can realize them. It will also drastically drops the cost of production for art and democratizing its consumption by the same way its access to the less fortunate masses. I think we should subsidise artists to give them decent wages instead of the current convoluted (and often deficient) means we pay them currently. I say this as someone who tried writing novels by the way, in the current system barely any writers can even afford to have an apartment with their revenues. The system is already broken.
Universal Basic Income would do what you want. Everyone gets a monthly income sufficient to live off from the government. If they want more, they can go and work for it. Sounds like a great way to give artists a de-facto subsidy, along with all the other underpaid work out there.
@@Roxor128 agreed, this is unavoidable anyway so I do 't get the obsession specific to copyrights comming from starving artists. They need a stable income, not the table scraps they have right now.
I think about how wild the industry changed in the late 00s early 10s. It was the wake of XBLA, Steam, and XNA, and then Unity and Game Maker burst onto the scene. Suddenly all these voices got to have access to games tools, and in the case of Game Maker, empowered people without any coding background. Easy distribution methods and a lower technical barrier of entry, and suddenly we had all these games filled with ideas and voices that were knew and unique from what the standard thinking had been. That period of experimentation was awe inspiring. It gave rise to the Indie mega star and AAA Indies. Time is cyclical. Games started from college students screwing around with new technology, then moved out of basements into silicon Valley and big companies like Atari. But then it cycled back to the basements and kids like Doom and sold in ziplock bags after the first level was free on file sharing sites. Things crept back to a big company era of the console wars. Steam, XBLA, Unity, game maker back to the bedrooms and dorms, before it when back into the big companies again now. I think AI might be the next cycle downward to hobbiests and non traditional voices. We'll have to see what that yeilds
With the exception that Unity, xbla and other are not tools designed by transhumanist libertarians specifically with the goal of making humans redundant. And increase their stranglehold on economy and prevent anyone from overthrowing the status quo.
I think AI will be great for indie devs, especially since it can be run on your own hardware using free models. Corporations can afford an army of employees, but solo devs and small studios for greatly benefit from "free workforce" in the form of AI. It will be great equalizer between big studios and indies.
Until AI will generate full personalized games with superior quality within minutes and everyone will generate their own games and living in their small bubbles, unable to share their experiences.
The difference between game making tools and AI is AI cannot generate anything on its own. It HAS to steal art from others without compensation to generate whatever you're asking for. When I make something in Unity, I am making it myself. Could I steal someone's art to make my game? Sure, but that's not a REQUIRED part of the process like it is for AI.
The real deal and opportunity comes not from using AI to generate the files you ship, but *ship the AI models themselves* . We have seen how powerful procedural generation can be in Minecraft, Dwarf Fortress, No Man's Sky and others. This could be an immense leap in procedural generation, imagine a game world that is not only unique for every playthrough, but also changes over time. This is the dream that Bethesda tried, and miserably failed to achieve in Oblivion, Skyrim and Fallout 4 with the "radiant quests". This is not something that fits in the trend of "push 2x the polygon count every generation" type of stagnation that characterizes the last 10 years of gaming. This could revolutionize game mechanics.
I'm not sure how useful they'd be. Stable Diffusion takes about 15 seconds to generate a 512*512 image on my GTX1070. I don't know how much there is in a game where the player would be willing to wait that kind of time. That volume of data might be usable as a level if you subsequently make use of existing data for rendering, but it raises the question of whether it's worth using AI and taking up several gigabytes of the players disk (most Stable Diffusion models are around the 3.9GB mark) or just writing your own dedicated generation algorithms that take up mere kilobytes.
If Jane is making a game and hires Joe to create art for it, Jane will probably sensibly put in the contract that Joe cannot give Jane AI generated art, but what if Joe does and Jane doesn't realize it? If someone sues Jane because she included the art in the game? She could presumably sue Joe for breach of contract, but she seems unlikely to recoup the losses from her own defense that way (plus the losses from having to file the suit against Joe).
The thing here is that you never just get one image from an actual artist.
An actual artist will usually send you a rough sketch, which you can then ask them to change in some way.
Seeing as AI can't do iteration very well, that will quickly make it obvious if they are using AI.
Always vet your sources, especially if you are making a commercial product.
Most programs used to draw has a built in feature that can show every single stroke on the canvas, making this process even easier to vet.
From a legal angle, such a contract would likely include a clause that define penalties if the contract was broken. Joe could very well be on the hook for any and all damages sustained by Jane. (Plus legal fees.)
Keeping humans employed is a bad reason to limit AI, for many reasons. Automation is nothing new, it's being going on since the industrial revolution. The only change is that it's now coming for creative jobs too. And as we saw in all other jobs that got automated, the result is temporary hardship for some people, who lose their jobs, but a long term benefit for everyone, including them. We today live in unimaginable luxury compared to times before automation. In many ways a poor person is the developed world has higher standard of living than kings in middle ages.
Eventually all jobs will be automated. It's inevitable, fighting against it just does more harm, simply because the economic benefits of AI are way too big, and it's very easy to create. Resisting will only make sure that someone else will enjoy the benefits, and you lose your job anyway.
What we have to decide very fast is how to handle that situation. When human labor is no longer needed, how we organize society, and especially the economy. In current American capitalist system everyone who depends on a paycheck would starve, and the collapsing consumer demand would bring down the rest of the economy. On the other hand a fully automated economy can create practically limitless wealth, we just have to figure out how to distribute it. One already existing idea that would help is universal basic income.
Yeah I think the solution to the problem of automation is public control of the means of automation, with the benefits of automation distributed through the democratic means. Otherwise you just end up with an increasingly small amount of billionaires or trillionaires controlling all automation, and reaping all benefits to gain greater control.
But here's the problem. Who's getting paid?
If every job is automatic. Then who is getting paid?
@@averytucker790 If every job is automatic, who needs to be paid? We can have a system of resource distribution that does not require payment, and solely exists to support the people of that nation/world as a whole.
What worth is money if everything is made for free by automated processes? (outside of some amount of maintenance for the processes that can't be fully automated, which can be rewarded through democratically-agreed upon systems).
However, that's a long way away. What we could do at the present is have the proceeds of automation go to a common fund that is then distributed as both an investment fund in industry and a dividend to members of the public, so rather than the benefits of capital investment going to a wealthy few, the proceeds are distributed to everyone in the society, which helps keep an economy going. The more wealth concentrates in a few, the less people are going to be able to buy the products of automation, the more monetary velocity grinds to a halt.
As the proportion of our labor that comes entirely from capital investment grows, the more the proceeds of automation grow the social dividend, the less people need jobs to buy what they need. The moment the returns are enough to live on, is the moment labor becomes optional, and we can move more and more towards an automated society.
@@videogamer596 You say that as if automated things are good for you. When the majority of it isn't.
Automated food is made of too much junk, and is not healthy most of the time (The 100lb I had to lose can attest to that). Automated customer service jobs lead to less customer service. Since you have to be a human to properly deal with one. And even automated electronics can be produced with a vast amount of flaws without human oversight.
You would know this. If you knew how to build anything that took time and effort.
And a little bonus point. All this automation is why people are getting so fat & stupid. Why go to the store or multiple stores, when a robot can just send it straight to my house. Why leave to a location to do anything when you can just ask a robot at home. Why learn anything when you can just constantly ask your robot assistant. So, unless you find the people on "My 600lb life" and "Wall-E" amazing to look at and listen to. I would suggest you support leaving jobs that need people to think, problem-solve, and move around.
Besides humans make better products than machines anyway. That's just a fact.
@@averytucker790
You cherry-pick the negative side effects, ignore the benefits, and ignore the possibility of fixing the issues.
1. Automated food production allowed you to not having to work on farms all of your life, just making enough to survive. With automation a small percentage of the population can feed the rest easily. This also allows much higher quality food.
Cheap food is of course lower quality, that's true with or without automation. You can buy high-quality and healthy food, just have to pay more for it.
2. Automated service jobs free up humans to deal with more complex issues, instead of having to do mindless repetitive work and burn out quickly. Most automated services are better for the customers too. For example no automation means I have to physically go to a bank to get anything done. Low automation means I can call instead, but it's still slow and often limited to business hours. High automation means I can do it on my smartphone with a few clicks, any time. And the same is true for most things. I only want a human when I don't know what to do. And even that won't be true for long, because ChatGPT soon will be able to help in almost all of those cases too.
3. Meanwhile in reality, people are healthier and smarter than ever, with few exceptions, thanks to automation. Americans are fat because they are eating garbage food because their food industry is under-regulated. The rest of the world is doing much better with equal level of automation. Robots don't force you to sit on your ass all day. In fact your job is forcing you to do that most of the time. Office jobs are one of the main causes of health problems. Not having to work means you can spend more time outside and exercising. Poverty is an other one. Elimination of poverty gives everyone access to quality food. The food you eat is unhealthy because it has to be cheap for you to buy it, and they load it with sugar because it's addictive. But if you have a personal robot at home, it can cook you anything you want, you don't have to buy mass produced crap. And with money not being a factor, it can use the highest quality ingredients too. If you want your robots can grow all your food in your own garden. That's not even expensive, we don't do that only because it's highly labor intensive.
4. In reality most people use their free time actively, if they can. Passive consumption is mostly the result of exhaustion from work and not being able to afford a hobby. It's not just a theory, just look at those who don't work. For example rich people and retired people. How many of them you see who just sitting on the couch all day growing fat? Pretty much only those who physically unable to walk. Boredom is a very powerful motivator.
Thank you for pointing out that the amount of references AI uses and the the amount of pictures it creates from that makes it an uneven comparison to a human using references.
I'm also worried that it'll lead to low or medium quality art drowning out the great art by sheer volume.
I mean it is still comparable, unless you can draw a clear line between how much is too much output
Low and Medium quality art already drowns out great art. That ship left port a long time ago. The AI will probably replace the low-hanging fruit of doing routine and simple drawings.
Good start for considerations around AI and art issues in game design. Would have loved a further discussion of where all this can go as AI gets integrated into more systems, e.g. AI quest creations in MMOs.
Can easily see this heading into training an AI on your game's quest system to generate not just art, but common quest writing campaigns, the placement of where the quests should be in your world (and on average) and the generation of that quest from start to finish, massively scaling up the number and more intricate kind of quests rather than the basic "go kill 5 hogs" grind.
AI art is interesting.. on one hand I love that as someone with no artistic talent can see my ideas rendered, but I don't like that the people who took the time to learn their craft are very suddenly not as valued.
Yeah, I feel more or less the same. I genuinely do feel bad for artists having their work reduced in value, but at the same time, I don't think that means we should take away the one tool that gives those of us who didn't start learning to be good at art as a child and therefore can't get good at it as an adult a shot at being able to realize our ideas.
@@argenteus8314
Many can get good even as adults though, so I don't think that is a good excuse to exploit the work of others to realize your dream.
On the other hand I believe there are ways to satisfy both sides. Artist simple want their work to be respected and not feed into AI's to keep their signature, which is important for artist to get jobs.
Models that are ethical and trained on only artist who have consented would be a good middle ground.
Since it still give none artist the tool to create while respecting artists.
Automation is meant to free humans from mundane labor so that we can focus on occupation which enriches society, like making art. If we jump right to computers making art without human inspiration (typing prompts don't count -- that's insipid not inspiration), then something has gone horribly awry.
Why is working with prompts, weights and masking not inspiration? You have to know what you are trying to make to make something unique. The same argument can be made for an artist just copying an already created character. Even if they have a unique style of drawing or painting, it is no different than a person who worked with a prompt to get a style based on keywords they gave the AI.
I think that artists just generally don't like that fact that people who think differently can create something similar to what they can do without having natural talent in traditional "art". This is the same debate that happened when digital art came to the forefront.
@@GageEakins stated by someone who doesn't understand the artistic process at all. The results you get from your prompts are only possible by taking millions of images without consent from other people. It has literally nothing to do with "artists just generally don't like that fact that people who think differently can create something similar to what they can do without having natural talent " Because NATURAL TALENT doesn't exist. it's honed over putting in serious time and effort to develop skills. And if you want to doubt this, lets just take every single piece of art taken without consent, out of these models, and see how good your results are.
You can, at any point, of any day, start the artistic process for yourself. You can pick up a pencil, a pen, a paintbush, a mouse, a piece of clay, etc, and start creating. There are literally millions of resources for you to start this journey. Anything you do from that point forward will have more value because you will be the one creating it. it will be yours, and when you add it to the world, it will be adding more to society by adding a part of you to the world.
AI only takes.
Is that what automation is for? I’m pretty sure it’s to allow us to produce things without having to hire other humans to make those things for us, and sometimes that means art.
@@natehorsfall8379 You're incredible defensiveness shows that you really do actually just have problems with people who didn't train for years or didn't have natural talent in it are able to create something similar to yours. Also your baseless attacks against me show that you are not interested in having a discussion.
@@GageEakins you used "natural talent" once again. Showing that you in fact are the one with baseless arguments.
I didn't attack you at all. But you did by claiming: "artists just generally don't like that fact that people who think differently can create something similar to what they can do without having natural talent" You just can't handle being shut down. The fact that you viewed my response as "attacks" says everything really. Don't try to pretend you have some moral high ground here. You're right. There's nothing to discuss.
For me, as an artist who started using very traditional means, I've seen most artists really look at other art as style reference. I don't really see the issue in training on other existing art, HOWEVER I do not think AI art fully emulating a specific artist ONLY or making works that just look too similar to the input art is alright.
That is to say, an AI model called "Modern comic book style" that does its best to get an impression of trends in modern comic book art seems entirely fine to me. One that targets an artist and says "Todd Franklin's style AI model" seems especially messed up. Not only will it be less likely to capture the style, but it targets an individual.
As an artist AI art is tertifying. Being an artist is already hard, being replaced by an algorithm is my greatest fear
Anyway deep thoughts aside I spent a minute listening to the jazz cover at the end
I get that fear, I'm an amateur writer, and seeing what writing AI can do makes me almost want to drop the hobby completely. How can anyone compete with it...
This reminds me of CGP Grey's video "humans need not apply". We really do need to replace capitalism soon.
If you can get replaced by an AI then you aren't an artist, you are a graphic designer. These are not the same things. "Artist" just means you create. Chefs are artists. Dancers are artists. If your creation is no longer SKILLFUL AND CHALLENGING then you are not creating art anymore than a child drawing with crayons. Sorry but get better and make art, not draw pictures.
Don't worry the same is true of all jobs, as technology advances more and more jobs get automated, we are quickly approaching a time where we run out of new jobs to shuffle people into.
@@giffica
> If your creation is no longer SKILLFUL AND CHALLENGING then you are not creating art anymore
Art: the expression or application of human creativity and imagination, typically in a visual form such as painting or sculpture, producing works to be appreciated primarily for their beauty or emotional power. - Google.
Literally nowhere in the definition of "art" there are requirements that it should be skillful or challenging. If I decide to paint stuff using only my earholes to hold the pencil and with a blindlold over my eyes, that increases the CHALLENGE dramatically, so by your logic it should be more ART than what the normal artists do, right?
Copyright should be reduced in duration to like 20 years max applied to all art. That would make this all simple and and the law gets back to promoting creation of new things instead of protecting Disney. The reason for copyright to exist at all is to motivate creation and protect peoples ability to make a profit of their creations. The only real argument against the concept of ai art is that artists will get payed less when anyone can create good art. This is not a compelling argument. Even more so when you consider that artists can also use AI tools and always do better then the laymen. Also the idea of stifling all the creativity and artistic creation of the masses ai art will unlock is disgusting. Simplification and fixing of copyright will improve the world and save us from meaning less arguments about it AI is copying or learning from.
well it's not anyone creating good art, it's the AI creating good art for the people
seems like a pointless distinction but there is a good reason to separate the two, someone who inputs a prompt into a machine isnt making art they're receiving it
I'm an upcoming artist for FEGBA Spritework and I'm really thankful that the FEU Community is making an active effort to separate AI Portraits and Human Made
Outside of that, though, the Hiring Scene for artists will probably turn to The Highest End or AI, with very few breaks given to new artists trying to build a career
It's not looking great for the future
PANR has tuned in.
Thanks for joining us PANR!
@Extra Credits Gaming always! Thank you for broadcasting.
With how complex it is in terms of legality and ethics, how does it affect the more creative-ish uses of AI?
As in:
-Generating 1 or more results, kitbashing different parts together to form something different, then redrawing the entire thing to morph it into your own creation (complete with small/additional changes by a human).
-Touching up a background YOU drew to appear more painting-like
-Using the AI generated pieces in a specific context (example: a character being on a drug trip with the AI's imperfections representing weird and distorted visions).
All automation gives us more for less. It's all double edged swords and we're not great at taking the good without running headlong into the bad. AI art is particularly weird and advanced form of this problem.
Like the fast food version of art. Except worse.
Since my Big Mac wasn't made with ingredients stolen from another restaurant.
And I know it isn't. Because if it was made with ingredients from other restaurants. It would have had a chance of actually tasting good (And not coming out someone's backside like a Ballistic missile.)
AI drawings, audio and writing can be a terrific tool, and it can't really replace real-life artists because the AI can't create a new concept, and it will never do unless there's some Earth-shattering breakthrough in how thinking works.
But it needs regulation. A LOT of regulation to ensure intellectual property is respected. I'm not seeing that happening, though.
10 years ago, AI had the brain power of a cockroach, today, AI has the brainpower of a small child. Give it a couple of years...
Most places that hire artist dont need new concepts. Yeah it wont replace artists at the top but it will replace all the ones at the bottom
I agree with your conclusions and am not trying to argue the dangers of such advances.
However, as someone working in Deep Learning I disagree with the idea that they can't create a new concepts and never will. The only fundamental difference between AI image generators and something like Alpha Go/Zero is how the network is trained and what it is asked to learn. Convolutional neural networks, including the U-nets used by these image generators, can be trained to play games and can learn to play beyond a human level.
To a computer scientist, learning is learning and we can see that both these problems (art and games) as something that can be learned by structurally similar networks. To us, the big difference is that Go can be played by the machine and thus it can learn on its own. Art, on the other hand, is defined by humans and so the AI must refer to us in order to learn. The difference then is that our interpretations of the world is what is being learned by AI image generators. That is why these machines need art to learn.
The difference is a matter of the thing being learned and not the players.
Humans can't create new concepts either. We're literally the same thing, just a more complicated AI running on meat that outputs variations on what it's seen. There's no such thing as an original concept, anything that seems like it is just remixing enough distinct ideas that it stops seeming derivative of any one of them.
You can literally today ask chat gpt for some new concepts for an image and then use the response with any ai img generator. So that ship have sailed. And this is sill the very beginning.
The nonsesne and unspecific laws we have regarding copyright and other intellectual property can get pretty nuts. Plus, there's a lot of international treaties that mean you don't just have to deal with your own nation's laws, often many others. United States copyright law affects most of the world. In the worst case scenario, you could get 'arrested' and taken to the United States to face charges from a country you don't live in and have no say in.
Seperate from AI Art I recently attended a talk about Open Source, and they were talking about various open source license formats. There are two broad categories, restrictive and unrestrictive. The former has rules you have to follow if you use any of their code, and the latter has few or no rules. The most open is Public Domain. It's very difficult to test these, though, and a lot of laws make it also difficult to make things public domain or 'un-copyright' them even though you own the copyright.
It led to me having an odd thought. If I wanted to make a low restriction version of an open source project that has a high restriction license, what would be the tests to prove I had not violated their license and copyright on their work? All my education on the subject was ways to avoid copying or plagarizing, and I was never taught how this would be tested in a court of law if I was accused of using someone else's copyrighted code. I studied that stuff twenty years ago and it hasn't come up. Most places are all about avoidance, using no code from outside if they can avoid it, and they're constantly reinventing the wheel to make sure they're not using someone else's copyrighted wheel.
What my investigations turned up is that software and code copyrights are very different from something like literature or music. Those two, you can take lines from a book or a section of music or lyrics and show how they are the same, and there's rules on how similar they can be, how long the similar parts have to be to count, and special exceptions etc. With code, it's trivial to change the names of functions, variables, and classes, and rearrange the code formatting so it looks different and would even pass the "diff" test. You could even make a program to change incidental code to make it look even more different but functionally identical, because two programs that do the same thing are going to have their most important parts be functionally identical.
Besides the human readable source code, though, all code has to be turned into something the computer can understand. It might be machine language, or something to talk to another program it's running on, like Java. When the human-readable code is transformed, two vastly different 'programs' can become identical, because all the names and such were taken out and just the functionality remains. The computer doesn't need easy to remember or descriptive names for anything.
Since you can't say something like "This paragraph is identical" or "These measures are the same", we get to how the law actually works with regards to software copyright.
The folks accusing you have to prove that you were copying their code while you made it, and you have to prove that you weren't. These are generally both impossible tasks, so it's all conjecture and up to the court to decide. You have to convince the court that your stuff was copied, and they have to convince the court they didn't copy, and there's no hard evidence or clear rules.
When it comes to closed-source programs, as long as you don't copy and nobody thinks you're copying, then you're safe. A lot of it is performative, because the most essential thing is to LOOK like you're not copying, whether you are or aren't. Open source programs, however...
How in the world are you going to prove to anyone that you never referenced the code of another open source program when you made your open source program? The code is freely available, and if you are making a program to do the same thing you are probably already familiar with the previous project. You may have read the code already, and remember it, even if you aren't directly copying from a screen or printout, and that can count!
I am not a lawyer, this is not legal advice, but as some rando on the internet I do advise all of us to talk with our governments about these laws, 'cause they're NUTS.
The way I see it, AI will only ever be best used to *assist* in making art, sort of like in-betweens in animation. Sure, you could go and make an animation using only tweens to pad out the keyframes, but then it'll look super robotic and stiff. You can still use those tweens as a frame of reference for actual animation, but they're not something you should rely super heavily on for EVERYTHING.
Great video btw, it's a pretty major thing recently and glad to be informed of the actual legal side of the subject ;)
Yes. This is my thinking as well. It will become an assistive tool.
yes that is ultimately how it would turn out, however the average layman has no idea of how much effort art actually takes and as has often been pointed out whenever something gets automated (say machine translations) people will demand lower prices for much higher quality or give the professional something they made with a machine and tell them to fix it for a pittance
For AI provider TOS issues and also for confidentiality purposes, I think your best bet is using a self-hosted model, such as Stable Diffusion. Unlike the mentioned DALL-E or Midjourney, you run Stable Diffusion on your own machine, completely separated.
THIS! Never rely on Service As a Software Substitute for ANYTHING! If you can't run it on your own machine, it's not worth using. Full stop.
if you are worried about your prompts being analyzed, you can run stable diffusion or any model derived from it locally on your gpu. of course this doesnt magically fix all the other issues, but you can at least play around with the tool privately without yelling every prompt into the wider world. so you can try out stuff and build an understanding on how to prompt correctly/safely/efficiently.
This reminds me of the debate that happened around photography when it was getting big. Questions arised like what happens when you take a picture of someone else's work, can you take credit for something mostly generated by a machine and even is it really art. Like photos I don't think AI art is going anywhere. I'm personally all for the new frontier of this new art form and would like to see how we might add more of the human element to allow for more control of what we are looking for and make each work more personal. I don't think this is going to replace artist but simply be another tool in an artist tool belt. Even working with more advanced ai like DOLLIE 2 is still pretty hard to get what your looking for and requires refinement to get it just right. We are feeling those growing pains right now and laws will take time to catch up and when it does sadly its likely not to be perfect. Even with photos I'm not sure how I feel about situations where a photographer has all the rights of a photo they took of a person where the person might have none. We will just have to grow with our tools and do our best to be the most fair to the most people
The whole AI art conversation feels so wrong at so many levels, focusing the discussion around training on copyrighted work being bad is just flawed.
If we want any resemblance to correlation to exist between ethics and law the focus should instead be a permanent subset of copyright focused on the actual use that is being made of the model.
It did teach me that the gray area being discussed, when it comes to 'teaching on copywrite,' isn't new or unique to A.I.
Schools for human beings have been scrutinized with the same rational, if not to the same extent as computers seem to get. So there are laws concerning using Art of educational purposes, but said Laws have restrictions that even state run schools fail to meet. Seems there is a reason schools stick to styles and pieces that are in public domain when it comes to what source material is used in schools, but it does highlight the point well.
If high-schools have been fined when it comes to us copywritten material, how can for-profit companies be exempt on the grounds of education?
Oh yeah, not a lawyer but have the enthusiasm for law... and fair use is a mine field!
@@lostbutfreesoul First of all the use is just not the same, schools make a direct use of the image in order to teach while AI corpos aren't selling the image itself but either aggregated data of those images or an image generated through transforming noise through that aggregated data (both of those have been legal for a long time in other fields), they have been especially careful with releasing links to the data instead of the actual images used to train.
Not only that but fair use is only one defense that can be used in courts of one specific jurisdiction, not a universal right. So it works in a very per case basis and companies are willing to risk a lot in order to become a letter in a word like faang.
Not saying those conversation aren't interesting or useful, just that it's very flawed because the underlying problems have a much wider scope than copyright is handling right now.
The fact that an AI generated piece of art can infringe on someone else's copyright or trademark even if the prompts or training did not reference the other work should be proof that individual works must stand on their own regardless how they are made. Ultimately the only thing that should matter is what the final work of art looks like.
Also the fact that the suggested work around for getting AI generated art in your game is to just get an artist to copy it shows how weak the weak the claim is that the AI work itself can't be copyrighted. The artist in this scenario is literally adding nothing to the creative process, so how can that change a work that is not copyrightable into one that is?
To see the absurdity of saying AI aren't isn't "real", or is different from human-created art, imagine a world where society has decided that art created by women isn't "real". A woman could create an image, and a man could point out that they used other images to derive their art from, and is therefore completely different from "real" art created by a man. When the woman points out that they went through the exact same process, the man says "but you did it in a woman way, and a woman way isn't real art".
And now for the twist: this argument is derived from the same one presented in the book Godel Escher Bach, when arguing that AI in general should be considered real intelligence, as eventually the only argument against it will be "if it was done by a computer, it isn't "real", by definition".
One of the cases that prompted the USCO to put this "human work" rule in place was a case that involved someone trying to copyright all the patterns produced by a machine that generated patterns at random. If your program can output 5 billion images / day, and they can call be copyrighted by you, eventually you can own every combination of pixels/colors available. It's to prevent people abusing a "100000 monkeys" scenario by things that can mass produce.
All that should matter is not what the output looks like, but what was used without permission by the developers to produce the software. Copyright protects against unauthorized use in any form.
The concept of forcing artists to remake ai art just for copyright seems deeply cynical of the future role of artists. My understand is people go into art to express themselves, not to mimmick an AI to solidify corporate ownership of AI art.
There's got to be a more positive future for both the artists and AI.
@@grae_n The role for artists will become more creative and less technical. What to make becomes more important than how to make it. This is nothing new. Excel spreadsheets made being able to accurately and quickly add and manipulate a lot numbers a lot less important skill.
This is also going to open up more work elsewhere. Creating art assets for games is one of the most expensive parts of the game creation process. If AI can drastically reduce that cost, far more games, and of a greater variety, will get made. Many more small development studios will be able to make games, and their success will depend more on the creativity of their ideas versus the ability to pay for an army of artists to design trees and buildings for a AAA open world.
@@staceykimbell9324 Great point. I didn't know about that specific case, but I did realize the problem. It's for that reason, and the fact with the limited inputs involved, two people will likely independently generate the same image, that I do think AI generated art must be handled a bit differently.
It's not enough that an image is created, some extra significant step must be done with it. It is the combination of the art + its use that is copyrightable, not just the art itself. In a way it should become a bit more like patent, in that the art should also be novel, nonobvious, and useful. There should be a registration process and not automatically assumed like current copyrights. However, a much more streamlined process that is mostly automated should be made. For example, using the art in a commercial interest, like in a comic book or video game, would be prima facie proof that the art is copyrightable, and the only thing that would be needed would be to look up the art in a database to see if it's been used before.
It's funny how much corporations in the US try to stomp down derivative works of their art. In Japan, there is something called doujinshi, which are (unapproved) derivative works of manga and anime, often sold for a profit. While unapproved by the original owner of the art, they are accepted, because manga publishers realize that it actually increases their fan base, not hurts it. And many current manga and anime artists originally started out creating doujinshi.
Japan has its own hangups in regards to derivative works that the US is better on; most notably refurbishment/hardware modification, emulators, and selling second hand copies, which are all essentially unrestricted in the US but heavily regulated in Japan.
Well, yeah, obviously it doesn't apply to all of Japan. Just look at how Nintendo treats fan-created works (unfortunately).
Lol, at 0:20 i was like "Why are Terry Pratchett and Neil Gaiman arguing over a dinosaur egg?
Yeah it's only a matter of time before the big studios just lobby their way out of this and get that law overturned.
5:06 I'm interested how that argument goes. It certainly is not self evident. I agree that AI training and Human training happen differently. I'm interested what's the basis for this. How does speed meaningfully change the equation when the question is about needing a license to use images for training. Method sounds more plausible to me, but still I'm curious what's the underlying reasoning.
It's this reason why I have difficulty objecting to AI work exclusively on principle.
As for acquiring licenses, I imagine many artists have unwittingly given consent, as a result of a clause in an updated TOS (you read thoughs right?) On what ever image board their work is uploaded on to.
AI is just a tool. It needs to be treated as such. Asking if AI should have a copyright over what it produces is like asking if Photoshop should have ownership of any image you create within it. It's like giving credit to the hammer when you build a house with it. And as for point number two - if you create something with a tool, you have ownership over what you create, period. Even if that tool is AI. I wouldn't recommend not further editing whatever it pumps out if you intend to use it because obviously it's going to be rough... but a rough draft is a pretty essential step in any creative work, and AI is pretty much the perfect tool to make that step easier. And I have yet to hear a single good argument against using it for that purpose... legally, ethically, or otherwise. Mostly just creatives whining about the AI takin' their jerbs. People who are mad because they're gonna lose out on all their commissions drawing furry porn for perverts and the like.
How can I see this video on Nebula please?
Not all of the EC videos seem to be on Nebula...
It should be!!! Thanks so much for the catch. It looks like the upload didn't work so it should be up shortly.
This is the same thing as scribes getting mad at the printing press because they copied the same letters. "OH well you needed to copy the letters from us originally so therefore we uh own all the letters, so you need to invent a new language to use it ah ha." I hope copyright dies this decade. Nothing has held us back more technologically, maybe besides war.
Use it as an advanced google search (or, 'Bard' as I think google's own AI tool is named) for idea generation is probably the strongest current way to use it.
I wouldn't use a generated piece as a finalized one in any way. Since taste is subjective, it is quite difficult to teach an AI to pick "the best of ..." and the AI end result will therefore likely always end up being mediocre to ok at best.
The standard way of using all modern AI models is to generate dozens of images from a single prompt and a person picks the best ones. CLIP is an algorithm that does an okay job at sorting outputs, but ultimately a person picks something. It's not actually harder to teach an AI taste than to teach it what a dog it. Just give it a bunch of examples. This is why AI models change their outputs in response to keywords like "beautiful" and "masterpiece"; such words effect the shape of the statistical distribution of images in ways which are fundamentally no different than any other word.
I embrace, and the errors created are themselves an artform that should be awed and ahhed over as well.
Still, I believe the best use of the technology is in the hands of Artists which is pretty much what the end of the video highlighted as well. Give it over to professionals whom can utilize it well to do a lot of grunt work just like any other field, let them figure out how they can use it to increase their productivity, and volla... better art. Being a tool like any other, we shouldn't be afraid of it but it should also be handled by the experts.
When using AI image generators like Bluewillow, we have to take into account that these are still images sourced from somewhere. What we can do as actual conscious beings is to just use them as inspiration. But I guess there are still loopholes in the law as this is a new technology.
Will US companies be able to compete with AI generated art in countries that allows its register?
Considering that copyright is enforced internationally, I imagine the laws will eventually become more uniform.
But you better believe that companies and interest groups will be involved when the laws are written.
good to learn about this topic since it is so new and not many know much about it
Why is frame 2:25 not copyright infringement? It is precisely the same thing he explains during this video passage.
I think AI can be an amazing tool for artists to scale up and speed up production. But at least right now I think it is very unwise to let an AI have fulfill final editorial role. Because it doesn't really know what it's making or why. You need a person to give the prompt aka have the idea. And they need to check out if it works. AI should fill the repetitive task that take a lot of time but little artistic decision making. I you're making an open world RPG an AI might be usefull to fill houses and design unique variants of all the stuff in there to make every space feel unique. Something that if completely done by humans would take an unreasonable amount of work. But a dev should check and tweak the created models for infringements or defects. Also if one of those interiors is say the setting of a crime scene for a quest that would require full human attention. An AI simply can't write a mistery that is entertaining, original, put all the clues in the right place and have it be fun or meaningfull by just copying and averaging out a standard quest setting.
People will continue to use AI and automation for new things. The law as it stands may also change due to Supreme Court challenges, etc.
Sometimes I think of the jetsons, specifically rosie from the jetsons, and how we once dreamed that AI would be used to take on the more menial tasks so that humans would be more free to pursue their own interests, such as art.
What's the goal of AI art, ultimately? To drive people who would otherwise make a living on art or writing back to working at starbucks or other menial jobs? Or is that just "An unfortunate side effect of progress"?
I imagine the point of AI art is to let people with no artistic talent create art. I have some idea but couldn't possibly draw it out, so I tell my idea to the AI and it creates something that's close enough. Like saying I want a picasso version of two dragons fighting the devil. I can't draw it now, and I'm not going to spend years or decades learning and improving until I can bring that idea to life. So I feed it to the AI.
Unfortunately, it is the latter. AI was supposed to augment human capabilities; helped them solve difficult problems like climate change, finding cure for incurable diseases, provide detailed analysis on datas, etc.
Human creativity; art, music, writing, was never a problem that needs to be solved.
AI art is a side effect of image recognition software. StableDiffusion works by feeding random noise to a image recognition system and then filtering out all the things that don't match the prompt the software was given. Image recognition is where all the training data goes.
@@d20pf18 Easy to say if you're good at those things. AI makes it possible for those of us who didn't happen to be born with artistic talent (or who didn't have people encouraging us to _develop_ artistic talent while we were still young and could still do so, or even had people discouraging it like in my case) to have access to the ability to make art. Despite years of trying to teach myself I still suck at music, and I'm sure the same would be true of visual art or writing. I'm not saying there aren't any ethical issues or potential pitfalls, but don't act like there's no actual reason anyone should want this. Besides, AI doesn't replace creativity, because it still requires human creativity to get started. The AI can't do anything (meaningful, at least) without a specific prompt or image to work off of, and the tools to control what specifically it does are only going to get more advanced. I think once you have more specific control over the output a lot of the complaints about its use being uncreative will become more or less invalid; after all, if I told the AI exactly where and how all the characters should be standing, how they should be dressed, what the lighting should be, etc., then all the AI has done is the execution, which isn't where I'd say the creative part of art lies.
You are attempting to add deeper motives then logical to creators of ai tools. They are just trying to make tools they can profit from and or find cool. Most things are not that deep.
I'm of the strong opinion that individuals should be allowed to use AI art generation, but corporations shouldn't.
Ai is accelerating an already rapidly changing world. If it didn't affect people making a living I would be very excited for it.
If I photograph a building that I did not architect, who owns the photo?
AI is great for exploring concepts....
BUT... I don't think its fair that the AI can just copy verbatim random code it finds on the internet, and spit it out as its own work... that's no better than a search engine. and if your prompt is specific enough, it will do this
That's no different from a search engine, as you pointed out, so it doesn't create any new danger. The new thing in AI art is exactly that it can create original work, not just a copy.
And using AI to copy someone's work is illegal already, no controversy there.
@@andrasbiro3007 Thats the thing tho. The AI does not realise its copying work. it has no reference. Its an AI, it does what you ask it to. and so long as it changes one character in text, or a colour plate slightly, it considers it new, simply because there is a pixel change. It has no reference.
SO, If you use a search engine, you know its possible there will be copywrite attached to it.
However an AI, does not tell you any of the copywrite, it doesn't really know about or understand it.
I Love AI, I think it could very well be what moves humanity forward.
But there are rules around using art from search engine results. but no such thing for AI
If AI is no better than using a search engine, you must assume EVERYTHING an AI gives you is copywrite, and you cannot use it.
And that would be...
One to many steps backwards...
If we're going to use AI we must either regulate what can be used in their training (ie, no copywrite images, which could make the AI completely obsolete, since it needs a large number of data to learn anything)
OR
We regulate what can be done with works taken from AI
personally I think there should be a middle ground, where ppl can opt in to giving AI content to learn from, or opt out. While also tagging the works given for learning with the rules of copywrite, eg, this is copywrite dont use this without letting the receiver know it has copywrite content
BUT, in order to do anything along the lines of the middle ground, We would have to sift through EVERY bit of content the AI has already used to learn...It would be a very long time before that would get close to completed (A step backwards)
In order to find a better way, we must search for the answers.
@@Mel-mu8ox There's a third option: Abolish intellectual property. NOBODY should be able to own an idea, whether it was generated with AI or not.
@@Mel-mu8ox
You are completely wrong on so many levels.
1. The AI never copies a work, it generates a completely new one based on the prompt it's given. If someone gives it a very detailed prompt, it's possible to recreate something like an existing work. But even than it won't be an exact copy. The AI doesn't have access to the originals after the training ended, it could not copy an existing work even if it really wanted.
2. The AI also doesn't care about how many pixels are different. But it would be easily possible to check against an existing database, like Google's image search.
3. No, what you described is a search engine, an AI works very differently. The AI can't give you a copyright notice, because it doesn't copy anything.
4. There's some legitimate debate around the training data. But it's just a technical issue, the AI doesn't need copyrighted works to learn, it's just convenient to use curated databases that happen to contain copyrighted works. The vast majority of the training data could be just random photos about random things, to learn how the world looks like. Then the AI wouldn't need more data than human artists use when they learn their craft. All human artists look at works of other people, that's how they learn, and sometimes get inspiration.
Legal issues apart, I can't embrace the use of AI because I respect human skill, if "referencing" and "inspiring" is the same as when machines do it, why then AI defenders get too defensive when we prompt the human side of making art? AKA, the use of real experiences to create, use your own eyes to scan and overall having pride on your own work... well, because it's not about art, it's about getting attention and making money. And that's what I find appalling about this: If unregulated, the acceleration of creation without skill is gonna clog the market with prompted art instead of created art... and we all know that the media is already full of trash from Holliwood to fanfic, the lack of education on art is just going to get worse, converting the creative crisis of "discovering the winning formula" from latent infection to pandemic.
Honestly, i worry what would hapoen if we unload creative works to ai.
It's good at giving general things, but it's a headache to get anything specific out of it. I've found that often the easiest way to get something specific is to just draw it roughly yourself and then get the program to refine it a little.
I forget who I heard mention this, but I think the line between what is "human made" art and what is "ai made" art might get fuzzy at some point.
Is using an AI tool to blur lines, or scatter a sparkle effect on something "AI art"? What if you asked an AI to make a human form, but then you added to it?
How much of the work does the AI have to do before it's AI art? All of it? Cuz the AI didn't come up with the prompt, and therefore it technically isn't the 100% creator, since someone gave it a prompt. Where is the line?
The examples you provided are algorithmic art tools, not AI. Blur effects are just matrix math, and sparkles are generally just noise functions with colorful pictures. These are well-understood, and someone can explain exactly what is going on and implement it manually if they really felt like it. They are, in short, tools.
The line blurs (badum-tss) with machine learning because we don't know how it works, as in we just have no clue. We cannot implement the modifications that their neural networks spit out because we don't know what's going on in there; it's too complex for us to analyze and figure out what exactly it's doing. The question now is if it is still a tool. Are we the ones in control... or is it the AI?
As for me personally, I think that AIs are still tools up until they can self-program limitlessly. Right now, they are still limited by what datasets humans feed them, the hardware we host them on, and the techniques we train them on. But if they could add to their own datasets, design their own hardware, and improve on their own training regimens, then we become progenitors rather than masters.
@@sylvestergharold7265 what if you said to the AI "make me a background full of trees" and then you did art on top of that background.
I do understand that the tools are not as smart as the full AI, but if we are using AI as a tool, I suppose the question is "how much input does a human need to have to have it go from its own thing to just a tool?"
If someone said "AI make me 10 people with different body types" and then you designed clothes on top of those bodies, did you make the art? Do the bodies as a base count as part of the art? The AI can be used as a tool, but at what point is it the AI's art, vs the AI as a tool for human art?
@@sylvestergharold7265 Upscaling images was algorithmic until we figured out that and AI could do it better. Just because we can do something without an AI, doesn't mean that it can't be done better by an AI which sufficiently understands the medium being manipulated.
@@practicepositiveprogress5396 That is indeed the more salient question for this day and age. Once we reach recursive self-improvement like what I was talking about, humanity is basically doomed.
In terms of ownership, I generally think that ownership goes to the human. After all, if I went out into the woods and took some pictures, that could still be considered art even with barely any effort on my end. Even if someone else had planted those trees, it is still my picture. I think of AI image generation as like making a digital forest people walk through and can photograph. But hey, I'm no expert on law nor art; I'm just trying to not be a Luddite about it, which I feel like a lot of people are.
This video got me more positive about this whole AI art stuff, you can use things done by an AI in your game, but you won't own it, if you want to own it, do it yourself or pay someone to do
1. That (probably) only applies if it's SOLELY AI done, if a human contributed at all it can probably be copyrighted (though we don't know for sure yet).
2. The whole idea is to democratize art. Not all of us can do art ourselves or afford to pay someone, and I'm pretty sure you can't actually get good at most art forms as an adult, considering my attempts to teach myself music have seemingly only made me worse at it (and presumably all forms of art are the same in that regard). AI, therefore, might be the only way we have access to the ability to create art. Maybe if I hadn't been discouraged from using and improving my visual art skills as a child things could have been different, but as it stands, I don't really feel like I have any other options. Sure, there are genuine ethical pitfalls, but that doesn't mean I or others like me are going to be happy to give up our one shot at actually having access to the ability to create art.
@@argenteus8314 Wow, didn't think about that, you made good points
We're at an interesting point in history. That much is clear. 👀
What about if for example, a sprite is made by a Human but AI is used to churn out copies of it to make the process much faster, easier and less expensive? The issue I see is that if every piece of art has to be made by a Human, this would nullify this use which effectively makes AI art useless when it's a great tool when used properly. Or does this only apply to what can be clearly seen upon first glance as being AI generated? But even then, there are issues.
If you own the AI and ALL the art you've EVER fed/trained the AI on, then it COULD be legal to use your AI to iterate on an image. The artist may be able to also claim ownership of the image, in which case they have a legal right over what is done with the image.
Are you looking to cut costs by not paying your human workers? If so, please take some time and wait until we have Universal Basic Income and people no longer need money to live.
@@brockmckelvey7327 Actually, even if the AI art is solely trained on the artist's own images, any image produced by the AI still cannot be owned. In order for a piece of art to be copyrighted, it has to have been made by a human. That's what the law says.
But it would of course be entirely legal to use as the artist obviously had permission to train on his own images.
There is no legal problem if you OWN the copyright. However, if you have bought sprite from an artist then you technically don't own the copyright unless you bought them out, which usually cost around 6000$.
Which is important to remember when you buy art from artist, When you bay 200$ for a sprite sheet you are not buying the copyright, you are buying the right to use it for your game.
Some indies don't know this and it has gotten quite messy in the past when two devs don't get along. And I can see it becoming a bigger mess now with these AI's.
So, to save yourself from legal reasons, if you don't own the copyright... Don't do it.
@@katanasharp2866 It depends. If you explicitly put in the deal that you own the rights to that sprite, than it doesn't have to cost extra and of course, you own it.
@@brockmckelvey7327 Not if in the agreement, you are the owner of the image.
Also, AI art is NOT about refusing to pay Human workers, it can literally be used as a utility tool to HELP artists get work done more easily, especially those on tight deadlines and which takes a lot more art to create such as animated series and animated movies.
Love the outro song! I recognized the melody immediately lol.
If you say, trained an AI to make art based off of your style, I see no problem with it, the issue for me is when you start getting to art say produced by artists, for your business, because making an artist produce work for you to then train an AI on that work to replace them seems extremely unethical to me.
Thank you for the video.
I've always been curious about this. Isn't fanart of characters or people that is being sold for a profit copyright infringement too? Should it not be?
I think it's a case that while it technically is, it's often not worth it to go after folks doing it, even ignoring the bad PR it would create. Someone with a Patreon where they're making art for a few hundred people isn't going to threaten your market as much as a larger project might.
@@Tuss36 oh it actually is copyright infringement but huge money hungry corporations seem to have collectively agreed to take a "meh" approach to actually pursuing it? That's actually interesting. And rare. Cool.
@@xxigeorge It is absolutely copyright infringement. It's an unlicensed derivative work, but the things is: fanart is never a substitute for the original game, comic, cartoon, or even the official artists' works. Indie artists selling prints at conventions doesn't incentivize people to buy the fanart INSTEAD OF the official products.
Plus, it's literally free advertising. Fanart promotes discourse and appreciation for the source material, keeping it culturally relevant. As much as copyright law exists to protect the theft of intellectual property and its related profits, that threat realistically doesn't exist when it comes to fanart of gigantic corporate IPs.
With AI-generated media, that threat of IP theft AND the commercial devaluation of original works is an absolutely real thing. If people can suddenly flood the market with things that look indistinguishable from your product in style AND quality, what reason do consumers have to purchase directly from you?
what about hobby devs that just code for the love of it and lack art skills and would release their game free anyway (and as such can't really pay an artist)? is it safe for them?
As safe as ripping graphics of other games I would say.
As a non-commercial dabbler in AI art. (Oh cool, a new Pok*mon, half robot, half turd!), who can't string two lines together for a stick figure, it's rather fun toy at the moment. A, representation of my D&D character? An idea I've had floating in my head, but am crap enough to never visualize? I don't have the budget to just go to an actual meat artist (made of meat, not using meat, though those do exist), an fairly compensate them for their work. So it's either nothing, or AI. I'll choose the AI, and let lawmakers figure out the rest.
And then we enter the international law, or rather, the void we have in its place. Your country might rule that AI can use any work publicly available for training, but then you publish your game on Steam, an American-based service, and get DMCA'd anyway, because the US has [hypothetically] decided that AI isn't allowed to do that.
Thanks for the information. The TOS and copyright issue are 2 important factor devs must remember before using AI art in their games. Enforcing them might be hard though but at least we have some protection.
The prompts are probably not conveniental is why I generate all this natively on my computer
An the odds I use ai art I'm a project of mine I'll place very low
So this is unrelated to the video but wanted to ask it. I'm working on developing games by myself but I'm struggling to find something that would allow me to make assets for no money. Does anyone know any software I can use to make game art for free ?
An important note about "made by a human."
If you modify art that is computer generated, that still counts as you making that art, if you make any substantive changes to what was generated. You do not need to create from scratch the image the AI generated. If you apply a choice of photoshop filters to some computer generated chebychev noise patterns to make it look like weird shaped paving stones, this still counts as you the artist doing the artwork. Similar for photographic collage (provided the photos you are using are either something you have license to, or do not individually make up a substantive part of the finished piece.) Collage comes with its own legal baggage, but in the case of AI generated images, you can safely piece together many images generated by AI, into one finished piece, and that still counts as your work.
Similarly, artwork that is superficially similar to an existing piece is not the same as infringement. The context in which the art is used does in fact matter. ATOMIC FE Man who is a rusty red robot with some yellow spots, would be infringing on ironman if you just wrote him in as iron man, but if you wrote him in as a gag character, that would be parody, which is fair use.
This next part I say as an artist to other artists. Once you've put your art out there to be seen, you have divested yourself of the rights to declare who/what may not see it and be inspired by it. If you made it publicly visible, you accepted that other art may be influenced by it implicitly and you cannot ever take it back. Do not show your art to anyone, if you don't want it to be used as inspiration for new art, it doesn't matter whether its by a metal organism, or a flesh machine. The only way to protect it from being used exactly as the AI does, is to not show it to anything or anyone, and to behave as if you have the right to decide whether or not it can be used as inspiration is pure _arrogance._
Did anyone else notice the splatoon music at the end?
At about 5:40: "The U.S. Copyright Office has already decided that art authored by artificial intelligence cannot be copyrighted."
This is contradicted by the next sentence (claiming that a human who wrote the prompt might have an ownership claim). If the human can own the copyright (in the AI-generated work), then the work can be copyrighted.
Did they actually just rule that the AI system cannot own the copyright? (There'd be nothing new in that: Software is not a legal person.)
There is a philosophical debate on authorship, and the case law can certainly be challenged.
I wouldn't bet on the matter being settled.
As a tool, AI can be interesting to speed up concept development, and of course private use from the public is a cool novelty. But ya it's worrying to think games will just rely completely on this or poach references from Google images who is already a strange copyright line. I don't think we're at the point an AI can completely write a game without supervision, make sure all images are copyright friendly or just don't use it as a final product, and written scripts should be heavily reviewed afterwards. People make content just by entering bs into AI writers and seeing what comes out.
It it unethical? Again as a tool it depends how you use it. I hope more reference material is opened up to let AI learn rather than say copy off Deviant Art which just ruins the internet for everyone. Law should figure out exact ownership of using AI as a tool or reference or your game might just default to public domain if it's all AI generated. That idea might be interesting in itself, but as far as Indie and AAA games, ya gotta have a human to make sure everything makes sense anyways.
4:42 and 5:00 Ticonderoga YEAH!!
Is this video ever going to make it onto nebula?
People is affraid of AIs because most people believed that automation would substitute physical work, but very few people imagined that what was going to be substituted is the human brain. So, this is the present. AI technology is not complex enough so no one can replicate it at home. You can build your own AI generating tool using your own graphic card and train it with automated google searchs. You can even distribute AI trainings, download it, and apply it to your own neural network if you know the protocol. This is here. Now. It's not going to leave, and soon, very soon, it will be extremely unlikely to identify if a picture has been made using an AI or not, since we will use AIs to identify if a picture is generated by an AI, and then you can use the same training to alter the image so it learns to cheat the AI identification software.
So, there's two choices: adapt to the reality and learn how to use these tools productively, or reject it, so the ones who have the upper hand are the ones who workaround the limitations. We humans are going become obsolete if we keep on believing that AIs and humans are competing against each other instead of complementing each other.
Is Design Club coming back in this channel?
Vague prompts are a great way to inspire artists if they want to create a piece in a certain style if it isn't already done. So the AI can give a semi-accurate framework to work with, but in most cases, these AI works are mostly funny to look at.
I think if you're a single dude making games then go for it. There's a ton of at least placeholders that are needed and maybe ai work makes the style get more coherent after thee final human made assets are in place. Keep in mind that directing ai to do something is not the same as doing something yourself, so if your goal is to express yourself to the bone then maybe it's not the right thing. Also I find arguing between ai made generic smoke particle and human made generic smoke particle quite grotesque. It's like programmers insisting that their version of CRUD somehow has a soul.
Human art doesn't have a soul. It's an absurd idea made up as a cope.
I want a shirt of that mickey mouse with the judge hammer, with the text "Don't make me sue you!"
The use of AI generated art in video games raises moral questions about the fair treatment of artists and compensation for their work. While AI technology can be a useful tool in the creation of art, it is important to recognize that the results generated by AI algorithms are often based on pre-existing works created by human artists. Using AI generated art in video games without proper compensation for the human artists whose work is being used or referenced could be considered exploitative. It is important to ensure that artists are fairly compensated for their work, regardless of whether that work is created by hand or with the assistance of technology.
Is it exploitative for human artists to study works that have current copyright protection? Their future artwork and style might be affected by the work, so do they owe that artist money?
How many images are you allowed to look at to inform your art before it becomes exploitation?
@@charliepaterson89 human artist can pay compensation for the references they use or just use public domain or free license images for their artwork. I use 3D model and free pose site for the pose for drawing. Or if you have some money, you could buy a reference pack from assets stores for relatively low price. For the art style, artist generally want something unique from other artists so even if they trained on previous artist works, it will end up different once they've developed their own style. AI prompters on the other hand had to type certain artist names in their prompt to get good or similar results which a bunch of said artist works are being trained without any compensation unlike the above example with buying assets which profit the asset creator and buyers equally.
@@Shinesart but unless otherwise specified, everything is copyrighted.
That film advert you saw on the side of a bus with an interesting composition, everything you saw when you wandered through the museum of modern art etc - it all has some amount of effect on future work you generate - making it a "marginally derivative" work.
Are you going to pay the ad agency that came up with that advert 10 years ago, send 10p to every artist on display in the museum?
What is the threshold?
@@Shinesart You can't copyright a style. Also this isn't about using someone else's work directly as components to your work. This is about learning from other people's work. If these AI generated works were instead done by people who studied the exact same pieces of art in order to learn, there would be no question that their work is copyrightable. That of course comes with the caveat that the new work generated is sufficiently different from the original works and are not copies. That last point is the only thing that should matter.
@@charliepaterson89 Of course you don't have to pay for the ad that was 10years ago. But the AI training was done on all things including living and working artist portfolio on the internet. And could produce the same style which can directly compete the actual artist. Beside, human brain is not perfect and cannot copy the exact same style. And learning from other artists is not the only thing people were doing. Composition theory and everything art fundamentals are important to learn first which AI have no knowledge of such. It just do things without any theory. And for the copyright side, artist can decide and look at the reference license before painting their own piece so it doesn't violate copyrights. If it is not specified, it would be wise to not use as reference so that it doesn't affect your own work. Since I work digitally, when I was learning, I follow the artist I like and bought their courses to learn which also means I compasating those artist. It's not even as expensive as going to art school and many artists course tells you exactly their process. AI just don't compensate and it's just exploitative in nature is what I'm saying.
when it comes to automation, i feel like id rather have robots do jobs that no one wants to do. jobs that dont fuffill peoples lives, athough ive had a good time being a NEET id like to have a job that i actually like.
Is there a si. fi. book about the science real life about robots making stuff or people a.i. making stuff?
Or you can just, I don't know. Hire an art team, and not worry about shipping it out as fast as possible.
Wouldn’t you have to be able to first prove that the AI trained on an image in order to have any legal ground to stand on? The only reason Getty Images were able to sue is because they spotted their logo being so obviously generated, but even then they are going to have a bit of a battle because those scraped images may have been grabbed from somewhere else. Once you have trained a model, there is no reason to keep the images anymore. You can toss them aside. And if I’m making art for my game, I’m not going to leave some artifact of a watermark or signature on the image. It sounds like the sort of legal problems that would really lack any teeth.
> *but even then they are going to have a bit of a battle because those scraped images may have been grabbed from somewhere else.*
IMO there IS also the complication of Getty putting images from others' up w/o their permission (they had been sued over this before) - and the public domain images on their site w. Getty watermarks to boot.
An AI makes a picture of a frog. It did not photograph or paint a frog. It replicated the pattern of “frog.” Based on thousands, maybe tens of thousands of pictures of frogs. Copyrighted pictures and drawings of frogs made by human artists uploaded to the internet and scraped into a database of frogs. The frog “pattern” may not be copyrighted but the humans who first recognized that pattern and made works of frogs are protected by copyright. So when a natural language model takes your “frog” prompt and generates an image, it is clone-stamping their frogs into an amalgam of theft. Every artist and photographer in that database is owed compensation. Do not be deceived. AI is an extension of the database it was fed. Artists cannot be denied credit for work that depends on their own work.
The legal question of AI is the same as this question: if a picture is produced by clone-stamping, skewing and blurring portions of thousands of copyrighted works, are the responsible artists responsible for the Frankenstein’ed product? Is the number of unwilling contributors a deciding factor? It shouldn’t matter. One artist or ten thousand scraped from art station. They are all stolen from. Enjoy your cheap twitter ads while you can
But what if I pay an artist to use ai I to make art as they will do a better job then me but as a result need less artist and time?
The Jokebot made me laught. Just not as he intended.
😉
As jobs become more and more automated, human employees are still going to be there, but mostly for legal purposes.
It’s funny how we never see that depicted in sci-fi, but it would make for a joke if HHGG was written today.
I belive that is a plot point in Tacoma
I’m sure I already said elsewhere, but I do want us to aim for that post work society with ubi, uhc, etc, but I do consider AI art art. Some of these questions I wanted us to talk about years ago, so we wouldn’t be in this predicament. So, human of us. I’ll make techpriest outta the lot of you yet. (Kidding) still if in 1000 years an archeologist took a printed copy of AI art I’m sure they’d say it’s art if they didn’t know who made it. When human artist use digital resources to make art it’s still art even without paint and canvas. I’m sure people will adjust fingers and teeth abnormalities all in due time to make it even harder to spot. It’s rising outta that uncanny valley tho. Also let the robutt own its art. Not the company or ceo etc. the robutt as a treat.
Cant AI be used for concept arts? I am not a specialist or anything, but for someone to use only a prompt to get an image that could be used as a base for an artist to create something would be pretty useful on the game space
Yes, AIs are prefect for concept art.
@@falconJB No, completely disagree. Lets say you have to design guns for a post apocalyptic world with unconventional parts in them that still could somewhat realistically work. You need to know how weapons work, how to use those new parts and combine them. High quality concept art is mostly designing what could be believable and still appear functional. AIs cant design functionality in mind with certain looks and specs for a project. The time you get something useful and pretty from an AI while fiddling the parameters, concept artists could have done several sketches already. AIs are good for vague ideas and pretty pictures and illustrations. Real concept art behind high quality projects are more about fast detailed functional designs that are up to spec closely to the selected project and style, and conveying them to your peers. AI is pretty useless in this area still.
@@Zere616 99% of concept art does not care about realism or functionality, just if it looks cool.
Also as AIs get better you will see them get better at realistic detail if that is what you ask of them. Remember AIs are still in their infancy and are progressing very quickly.
@@falconJB High quality concept art does. And its not actual functionality, more like a believable functionality, take iron man suit and all its transforming details from the new movies as an example. Its functionality was designed and drawn first, and AI in that would be useless, it does not design. Same goes for high quality video game assets.
And current AIs are completely dependent on human art, if the current lawsuits end up favorig artists and laws are made favoring artists (which i doubt), the current AI models (and future ones that need input art) wouldnt have anything to get better from.
@@Zere616 Sure, currently its not up to the standards of the highest quality stuff but it could still do most concept art. And as it gets better the percent that it can't do well will continue to shrink.
And the current lawsuits are about how the art it is trained on is collected not if it can be trained on peoples art, and there is basically no chance that pro-artist anti-AI laws are getting passed at least not in the US.
They should be for inspiration, not compensation.
infinity blade chad
I do think there is place for AI generated art, we just need to adapt to find that sweet spot and ways to prevent or punish ways it can be done in wrong way.
Also on people who think, that if AI can recreate existing art then the AI itself is bad. But won't this simple way to judge AI also mean that if artist creates art with brush and then someone copies that art with brush, does that mean all art done with brush is now illegal?
Of course AI does this copying within minute while art forgery need perhaps hours depending on work.
Wait wait XD
We need AI police to search and to keep AI art programs in check. AI police would get AI art and then search for possible links it has with non-AI works and judge if it is too close to copyrighted works to do its job. Reason for AI police is simple: At one point AI art will overflow us art and no human can check them.
Another problem:
If AI generates art, does its iterations that are not accepted still exist in internet? Should we also be worried internets storage capacity is filled?
As an artist, both in terms of being an author and someone who draws covers for books, AI "art" needs to be redone from the bottom up.
What I mean is that the smart way of using it is that you have a team of artists who might draw something like ground textures for a game, and you then train the AI on only those specific images, and nothing else, until it can spit out enough images to make a believable game world, once the generated images have been thoroughly vetted by humans.
That way we can remove some of the more tedious bits of game design.
The important part though, is that the model can never ever be trained on anything outside of your own studio.
That is literally the only useful thing I can conceive when it comes to this.
It is my firm belief that all existing open AI "art" generators need to be stripped and rebooted, using only art from people who explicitly give permission.
I've seen a similar argument made for AI art outside of game development. Their idea was to have services similar to the current stock image ones that AI services would pay the rights to use. Artists could get paid based on each time their art was sourced or else get paid upfront by the stock service.
Either way they choose to go about it (different services could offer different approaches), AI art should definitely be retrained to source only approved sources.
@@BonaparteBardithion Problem is that it's not possible for an AI when generating an image to be able to tell what images it trained on made the generated image possible. So I don't think that is an option. But definitely they should find a way to train on images they are permitted to use. I'm sure they will find a way.
But that doesn't really make much sense, philosophically. We don't expect human artists to never see another artists work, why should we expect it from intelligent (in this specific capacity) systems that _don't_ happen to run on meat?
I love the ai generated hands in the video
As someone who works with neural networks professionally, though admittedly I deal more with linguistic stuff like Large Language Models than image generation, I think the problem isn't the technology itself, it's that right now the technology is stealing from people. What we need is new legislation that codifies that generative machine learning algorithms which rely upon training data must obtain a license from the copyright holder for any data used to train that model. If a license cannot be obtained for any reason other than the work being public domain (or legally not requiring a license for other reasons), that data cannot be used to train that model. It's key though that this specifies generative models - if all you've built is a classifier (that is, all it produces is an analysis of data rather than producing data of the same sort as it was trained on), I think these restrictions would be excessive. It would also be foolish to specify the nature of the algorithm beyond simply being generative or not, since specifying an architecture would make the legislation less future-proof.
A human is just a powerful AI running on meat. Thus, any suggestion of requiring AI to have a license for its training data makes no sense, given it would be absurd to require humans to have a license for everything they've ever seen (and therefore learned from).
What will happen when an artist who has a disability making it so they cannot physically create art uses ai to express themselves. This might turn into an ada issue. I could see that opening the floodgates for ai art to be copyright etc.
I do want to point out a mild disagreement with one of your assumptions.
Obviously companies generally want to own the rights to images in their games, and I understand why. This gives them the power to be sole provider of merchandise about their games.
...but that isn't inherently necessary for the creation, production, or legal status of the game itself. If 100% of the art in your game is AI-generated, then *no one* owns that art, and *anyone* can use it without fear of reprisal from someone like Disney. That isn't inherently a bad thing, since the dev company that made the game still own the *story* and *code* the game is made of, as well as the characters themselves, if not the art of them. No one can just make their own sequel or spinoff of your game just because you don't own the copyright on the *art* in it.
True, but you technically don't have ground to stand on if they take and use the AI generated art from your game to their own game.
Yeah, you still own the code, and the story and the description of the characters.
But not the art itself, so now a bunch of people can make copycat games using your AI generated assets, As long as they don't use your game title or character names they are free to go at it.
It is similar to when you buying premade assets.
Welcome to the future. Now I want the cyber, because punk is already here.
In the end we'll have hybrid creation at the high end and just AI at the low end. Most production will be post-production. Most visual art will be finishing before final filters. Creative work will be more dependent than ever on direct contact with the public. At the top end, you'll have to do it live to prove that you didn't just "ai" your way into the business. Some people will get really really good at prompts, though. And lots and lots of cheap labour will come into play to oil the machine. Pun intended. Most of the public will not care. And very few people will even be able to tell the difference. Yay. :/