I actually think watermarks is a pretty good idea. It doesn't fringe on censorship or denial of services. of course, people can simply not use them or remove them, the same way people do now. but saying it will help nothing is going too far I think.
Gun laws, copyrights, banning of drugs, murder being illegal, etc definitely work. Of course not to 100%, but to a reasonable point that doing nothing would be far worse.
lotet
that's fair, but laws are an ever changing construct, they get adjusted, they get adapted.
most of the times too slow.
but law is nothing that is written in stone and can never be touched, often enough they get changed in the favor of the people with the big money, so yeah there is always a risk that this goes sideways.
there is no law, does not mean there can not be one? there is no law that says ai art has to get marked as such, until there is.
and yes lawsuits between artists and artists or artists and companies exist, plagiarism is a thing.
I think the one thing that is missing from the people thinking only in "tech terms" is the fact that just because something is now possible to do, doesn't mean that it is "too late to stop it" and that it is the new normal. For instance in France there is a law from 2017 that stipulates that a photograph of a person used in advertising needs to have a written mention that it has been touched up if it was. The fact that liquify and the clone brush exist does not matter - the law is simply there to enforce the will of the people to not get unknowingly subjected to touched up photos without knowing about it. Similarly, just because research could be made on human cloning, doesn't mean that it can go unregulated. And so on.
Of course everyone can try to bypass the law for whatever reason, and of course no law is perfect and is subject to interpretations. But there is no need for some crazy bLocKcHaIn or wEbThReE tech to authenticate if a picture as not being photoshopped to erase fat rolls - the law just needs to exist, and then if gets violated then it can lead to an investigation.
pior
i am not trying to get rid of this competitor, as i know i can not.
I do what i always do. make nice art and hope it will all be good, that there will be people who we work with who value art done by hand and that there will be people who will not. there are already clients who we do not work with for various reasons, thats fine.
i have no hurt feelings, maybe because the damn database doesn't have my work in it!!!!111 ;)
i do think there is a lot of value in AI stuff, there are people who would have NEVER commisioned to an artist, who have great ideas (not just ideaguys getting nothing done ever besides lots of talking) and now have the possibility to tell their stories. Use the tool by all means, as an artist or studio, train it yourself, but doing it on the backs of thousands who have not even been ask, is unethical at the least.
you already admitted that there are ethical reasons why it's not in a great state, so why refuse to make it more ethical?
The tech is here, nobody will take it away. But there are a few things that need to be fixed.
The first and foremost would be making sure how its sourcing its inputs. Using an open source library that was made to study, in a commercial sense will have to stop or there will have to be systems in place to compensate.
I feel like as an artust asking for AI images to go away is like asking for cars to go away as a horse carriage driver back in the days. The times without AI supported Art is over, it will not come back.
But it doesn't mean it can and should not be regulated. But the topic is so big I don't see how a platform like ie. Artstation alone could tackle it.
I don't see how they could enforce a ban on AI art, the easiest they can do is an optional Tag that AI images should use, and people not interested filtering those out, treating them differently on artstations end.
Atm images made by AI are often simple to recognize, but that might not be the case forever. A forced on watermark that the suppliers of those platforms have to use, can solve that issue.
The other thing they can do, is give an opt-in function, marking all existing pieces opt-out. If the image ends up in a database used for profit, there are then ways to go against that.
That wouldnt solve automatic scraping as a whole, and certainly not manual scraping. But it's a step in the right direction
samnwck
Hey,
Check out my new work:) This time i did a likeness study based on Jay Samuelz
ArtStation - Jay Samuelz - Realtime Likeness Study, Marlon Wolf
https://www.artstation.com/artwork/QnDoRx
Custom Sig Sauer MPX game-ready model with aftermarket parts rendered in Marmoset Toolbag 4. Thank you to the Polycount community for allowing me to flood their forums with questions and for providing help!
ㅤ
Rendered in 4K
73K tris
Maya to ZBrush Live Boolean workflow
PBR Spec, Gloss
Hands from Eugene Petrov
ㅤ
Optic - Sig Sauer ROMEO5 1X20 MM TREAD
Foregrip - M-LOK® MVG® Vertical Grip
Suppressor - Dead Air Wolfman
Stock - Sig Sauer Telescoping / Folding Stock
Laser Device - Steiner Optics DBAL-A3
Iron Sights - Troy Industries Tritium Battlesights
https://www.artstation.com/artwork/QnDoRx
Please like if you enjoy :)
coolguyslims
The "cat is out of the bag" position is IMHO a rather short-sighted and somewhat unimaginative admission of defeat in regards to the way current models are using data from non-public domain images, without obtaining any license from the artists who got scraped.
One could very well imagine a situation where providers of AI-trained generators (for images, music, text, 3d models ...) would be legally obliged to state whether a model has been trained on public domain images and/or images that they explicitly acquired the rights to use. This is one of the many practical things that could very well be done today, and would solve a major part of the current problem. Similarly, it could be made illegal to publish an AI-powered generator without explicitly giving access to the entirety of the training data. There is no gray area there, and no reliance on tech to enforce it.
Of course there is no way to know in advance by how much the output would differ (although it's obvious that the generator wouldn't have a clue how to spit out a "Cyberpunk anime city skyline, Trending on Artstation") ... but it doesn't matter really.
Obviously there is no way to know what the next controversy would be - as even under such constraints, AI generators trained on public domain data could still be perceived as a threat. But that's another story really.
pior