AI art in compos
category: general [glöplog]
IANAL, but this Wikimedia uploader claims that "There is no copyright in AI-generated images"
https://commons.wikimedia.org/wiki/Commons:Deletion_requests/File:Bathingbayc_a_mechanical_dove_by_beeple_low_camera_angle_cute_2daa1f69-e280-45d0-956d-8b89208ac53e.png
Isn't it kind of a rule that you have to have a copyright to your compo entries? Again, IANAL.
FWIW, I don't see it as a terribly bad problem, if there are many entries in a compo, and the audience has to wonder how it was made. :)
Obscure technical limitations are always good, for example that it has to run on a C64.
https://commons.wikimedia.org/wiki/Commons:Deletion_requests/File:Bathingbayc_a_mechanical_dove_by_beeple_low_camera_angle_cute_2daa1f69-e280-45d0-956d-8b89208ac53e.png
Isn't it kind of a rule that you have to have a copyright to your compo entries? Again, IANAL.
FWIW, I don't see it as a terribly bad problem, if there are many entries in a compo, and the audience has to wonder how it was made. :)
Obscure technical limitations are always good, for example that it has to run on a C64.
Quote:
@farfar: My take is that the line between drawing and (AI) generated image is extremely blurry, and these tools make it so blurry it practically speaking becomes meaningless. How do you even phrase compo-rules that disallow AI techniques with clear lines? Does this mean we can't use content aware fill, or the repair-filters in Photoshop? Will you provide a list of approved software and plug-ins/filters?
I don't think you can do this in a meaningful way...
while I acknowledge the content-aware fill and whatever, when was the last time you saw that feature used to any relevant effect in a graphics compo? It's a valid enough example of course, but in practice it's kind of a non-issue.
Regarding how to enforce a no-AI content rule, I guess we would have to use our judgement in the same way that orgas do in music compos with preselection? E.g. through more focus on "worthy" working stages. - Assuming that it's even something people want to deal with or think about - that's by no means a sure thing.. that's why I asked the original question in the first place :) Could be that we should just let things run wild and let the votes decide.
anyway .. since most gfx compos are constrained somehow (system, palette, size, whatever) it's only really a potential issue for the freestyle compos, I guess?
Well one can say nakedly that AI art generators are verboten, ask folks to be honorable and obey the spirit of the rules, and ask for credible working stages . . .
as long as you can still generate AI-assisted seamless textures and AI-assisted retopology for your 3D :P
I wonder how long it will take before some of the generators come with a way to deconstruct the images in a way to show creation steps...
I think that we simply do not automatically know if this is going to be a change for bad or change for good. Invention of photography eliminated many portrait-drawing jobs, but created many new types of jobs around the photography. You may want to argue that photographs do not have a soul like paintings do, but then again, the art of photography is now sufficiently widely accepted for us not to have this argument here.
The invention of AI-generated visuals is the event of the same order of magnitude. Things won't be the same again. And although I think it is easy to suggest that graphics compo should have steps to participate, I think the task of generating plausible set of steps is not much harder than a task of generating good-enough quality art in the first place, so mid-to-long term you probably need to accept that you won't always be able to tell if the submitted art is human-generated or AI-generated.
Fundamentally though, I think that this also means that an artist would have to shift their focus from the technique to the concept, so this is not all that dissimilar to the transformation caused by the introduction of photography. Artists who will find ways to embrace it and use it to increase their range would become originators of new genres. I personally definitely see myself generating some concept art for my demos using an AI. It solves the problem of sourcing original clipart and is already flexible enough to be used in production.
The invention of AI-generated visuals is the event of the same order of magnitude. Things won't be the same again. And although I think it is easy to suggest that graphics compo should have steps to participate, I think the task of generating plausible set of steps is not much harder than a task of generating good-enough quality art in the first place, so mid-to-long term you probably need to accept that you won't always be able to tell if the submitted art is human-generated or AI-generated.
Fundamentally though, I think that this also means that an artist would have to shift their focus from the technique to the concept, so this is not all that dissimilar to the transformation caused by the introduction of photography. Artists who will find ways to embrace it and use it to increase their range would become originators of new genres. I personally definitely see myself generating some concept art for my demos using an AI. It solves the problem of sourcing original clipart and is already flexible enough to be used in production.
The way we deal with the harms of new technologies is with effective regulations and enforcement.
Sometimes.
https://www.vrt.be/vrtnws/en/2022/08/15/unconscious-man-drives-for-23k-on-the-left-lane-of-the-e314/
Unconscious man drives for 23k on left lane of E314
Sometimes.
https://www.vrt.be/vrtnws/en/2022/08/15/unconscious-man-drives-for-23k-on-the-left-lane-of-the-e314/
Unconscious man drives for 23k on left lane of E314
Quote:
as long as you can still generate AI-assisted seamless textures and AI-assisted retopology for your 3D :P
well.. if you look at edgeloops for a human face (and body) for example, there is pretty much a formula for how to do it, so it shouldn't be that hard to train an AI to solve this problem if you feed it a bunch of example meshes - at least for humanoids.
there's also stuff like https://www.youtube.com/watch?v=8oIQy6fxfCA (AI generated character animations ... this is vastly different from e.g. Midjourney though since it's evolutionary and not based on training data)
@farfar: I think you're missing my point. I don't think you can define what is on the "good" side of the algorithm vs AI spectrum in a meaningful way even if you want to. And if you can't define the rules in a meaningful way, enforcement will be arbitrary and unfair.
There's very little fundamentally different between something like a content aware fill and a machine learning based image synthesizer apart from quality of the result. And just because you can't see the content aware fills, don't mean they're not there.
The rules needs to make sense IMO. Not liking AI generators isn't a good reason to ban them.
There's very little fundamentally different between something like a content aware fill and a machine learning based image synthesizer apart from quality of the result. And just because you can't see the content aware fills, don't mean they're not there.
The rules needs to make sense IMO. Not liking AI generators isn't a good reason to ban them.
All AI generated crap is more awful than all those bland fractal demos, the only category it deserves is 'Invalid'. Remember that guy who posted GIFs converted into ZX-Spectrum "demos"? What Midjourney does to the art world is much-much worse!
Kusma: It's not that I don't like AI generators - I think they can have a place as a tool for ideation and stuff we don't realize yet.
What it is, is that I believe graphics compos should be a competition where you demonstrate skill in stuff like composition, lighting, visual ideas, rendering, harnessing technical constraints and so on... with an AI you kind of skip all that and arrive at a finished result through a more or less vague description. 99% of the creative decision-making (which darya and metoikos both formulated so well) is gone, along with most of the artistic skill - all you are left with is "result", which is just fundamentally uninteresting to me... you might as well just participate with something you found on Pinterest :)
Anyhoo.. that's my view of what a compo should be, doesn't mean it's the Truth.
Also: I appreciate that I've got a little skin in the game re: the Black Valley graphics compo - probably that's a contributing factor as to why I started thinking about it in the first place - but not the only factor. As I mentioned in the OP, AI art (AI turds as they are lovingly called) are everywhere on social media .. people are showing them off like they used to do kitbashes, except with even less artistic decisionmaking..
What it is, is that I believe graphics compos should be a competition where you demonstrate skill in stuff like composition, lighting, visual ideas, rendering, harnessing technical constraints and so on... with an AI you kind of skip all that and arrive at a finished result through a more or less vague description. 99% of the creative decision-making (which darya and metoikos both formulated so well) is gone, along with most of the artistic skill - all you are left with is "result", which is just fundamentally uninteresting to me... you might as well just participate with something you found on Pinterest :)
Anyhoo.. that's my view of what a compo should be, doesn't mean it's the Truth.
Also: I appreciate that I've got a little skin in the game re: the Black Valley graphics compo - probably that's a contributing factor as to why I started thinking about it in the first place - but not the only factor. As I mentioned in the OP, AI art (AI turds as they are lovingly called) are everywhere on social media .. people are showing them off like they used to do kitbashes, except with even less artistic decisionmaking..
i think AI generated prods are an interesting topic.
its more than fair to put it in an own compo since its a different thing (and probably requires less creative talent) but i wouldnt blame it as THE DEVIL, its simply a new way of content creatiion.
so, if theres enough interest in it why not make it a compo and see what folks come up with? maybe it works, maybe it looks generic and bad... but thats a lesson learned.
if you are afraid that an algorithm will make everything handmade wothless you might want to think about the creativity involved in the first place ;P
its more than fair to put it in an own compo since its a different thing (and probably requires less creative talent) but i wouldnt blame it as THE DEVIL, its simply a new way of content creatiion.
so, if theres enough interest in it why not make it a compo and see what folks come up with? maybe it works, maybe it looks generic and bad... but thats a lesson learned.
if you are afraid that an algorithm will make everything handmade wothless you might want to think about the creativity involved in the first place ;P
In many ways this reminds me of all the "$(NEW_TECHNOLOGY) is killing demos" discussions that we have had over the decades. But mp3 didn't kill modules and make music easy to make. hardware acceleration didn't remove the need for programming skills even if you didn't need to code the semi-trivial triangle routine anymore and Amiga didn't make demos easy to make even if it has a lot more processing power than the C64. Things might be different though, and maybe in the future the graphics competitions will be akin to the tracked music of today - relatively archaic but fanatically loved by a subset of sceners.
That said, the prevalence of putting an AI in everything makes me really uneasy. I wholly second what Gargaj wrote on the first page. I'd rather wish this technology did not exist.
That said, the prevalence of putting an AI in everything makes me really uneasy. I wholly second what Gargaj wrote on the first page. I'd rather wish this technology did not exist.
If an AI generated prod wins a compo, send the prize to the AI and not the person who submitted the prod on their behalf.
I agree with Kusma that the line between "AI" and "no AI" is very fuzzy but I disagree it is is a problem.
For example, demos that are purely video players are considered cheating, and rightfully so. At the same time, 8088 Corruption is purely a video player, and it's fucking amazing. It all depends on the context.
The demoscene is, by and large, made up of well meaning people (after all, even top demos only get a few thousand views on YouTube - there's few niches that have less ulterior motive to want to win than the demoscene). Competition is always friendly. Cheating is always frowned upon, except when the cheating itself is the performance. Then why can't we just say "no AI" in the compo rules, and leave it up to actual cases to decide whether it's fair?
I recognize that people can still cheat because it's not always obvious to see that something is AI generated, but you can cheat in a myriad of ways already in most compos and still it's seldomly done.
For example, demos that are purely video players are considered cheating, and rightfully so. At the same time, 8088 Corruption is purely a video player, and it's fucking amazing. It all depends on the context.
The demoscene is, by and large, made up of well meaning people (after all, even top demos only get a few thousand views on YouTube - there's few niches that have less ulterior motive to want to win than the demoscene). Competition is always friendly. Cheating is always frowned upon, except when the cheating itself is the performance. Then why can't we just say "no AI" in the compo rules, and leave it up to actual cases to decide whether it's fair?
I recognize that people can still cheat because it's not always obvious to see that something is AI generated, but you can cheat in a myriad of ways already in most compos and still it's seldomly done.
Indeed, the same question comes back with every innovation.... If only the result counts, if only the spectator's experience counts, then unleash AIs and whatever professional tools one may think are unfair. Now, if you think that the pure tech skills and/or artistic skills matter in a competition, you can start worrying.
Think about chess: Computers have been better than humans for decades now. Did we allow computer-assisted players in competitions? No. I believe matches would be more spectacular with computers assisted players, faster, better for the show. But we don't, because we want players to win the hard way, like it was intended to be when the game was invented.
I can't wait for the day when AIs will discuss in this very forum is humans should be banned
from competitions because the quality of our prods is not good enough.
Think about chess: Computers have been better than humans for decades now. Did we allow computer-assisted players in competitions? No. I believe matches would be more spectacular with computers assisted players, faster, better for the show. But we don't, because we want players to win the hard way, like it was intended to be when the game was invented.
I can't wait for the day when AIs will discuss in this very forum is humans should be banned
from competitions because the quality of our prods is not good enough.
It's trending right now, so there's no doubt we're gonna see this everywhere the next few years. Allowing it by giving it a separate category seems like a good decision IMHO.
Allow everything, then when a winner gets chosen put him on stage and give him pen and paper and ask him to draw his own compo image live by hand while everyone is watching!
Quote:
Allowing it by giving it a separate category seems like a good decision IMHO.
I agree. In the long term, as we become more culturally familiar with it I don't think "words I typed into a box made something nice' is going to remain an interesting category, but in the short term there are some folk who are really into the novelty of the tech. The genie is out of the bottle, and we can have some fun while still emphasising that one thing is an art and the other is a complex mechanical reworking of prior art.
Or not!
soon AI will start making music too
Quote:
Quote:Allowing it by giving it a separate category seems like a good decision IMHO.
I agree. In the long term, as we become more culturally familiar with it I don't think "words I typed into a box made something nice' is going to remain an interesting category, but in the short term there are some folk who are really into the novelty of the tech. The genie is out of the bottle, and we can have some fun while still emphasising that one thing is an art and the other is a complex mechanical reworking of prior art.
I still think that's normalizing an otherwise explicitly malicious intent and moral rejection is the only viable option.
You know.. that’s not a bad idea. Fuck it, I can spend the time on setting up and oldskool gfx compo instead.
Quote:
I still think that's normalizing an otherwise explicitly malicious intent and moral rejection is the only viable option.
I don't know what you do and/or don't know about the whole situation, but I want to say explicitly, in the interest of fair representation of what is going on, that the guy you just quoted (RJ Palmer) is very obviously clueless and has no idea who he talked to and what was actually involved in the project that he accused of stealing other artist's work. So your wording of "malicious intent" is pretty strong and is really not supported by your own quotation.
Sure, they made an AI that explicitly knocks off existing specific artists' work BY ACCIDENT.