Magic: The Gathering Community Fears Generative AI Will Replace Talented Artists (slate.com) 133
Slate's Derek Heckman, an avid fan of Magic: The Gathering since the age of 10, expresses concern about the potential replacement of the game's distinctive hand-drawn art with generative AI -- and he's not alone. "I think we're all pretty afraid of what the potential is, given what we've seen from the generative image side," Sam, a YouTube creator who runs the channel Rhystic Studies, told him. "It's staggeringly powerful. And it's only in its infancy."
"Magic's greatest asset has always been its commitment to create a new illustration for every new card," he said. He adds that if we sacrifice that commitment for A.I., "you'd get to a point pretty fast where it just disintegrates and becomes the ugliest definition of the word product." Here's an excerpt from his report: So far, Magic's parent company, Wizards of the Coast, has outwardly agreed with Sam, saying in an official statement in 2023 that Magic "has been built on the innovation, ingenuity, and hard work of talented people" and forbidding outside creatives from using A.I. in their work. However, a number of recent incidents -- from the accidental use of A.I. art in a Magic promotional image to a very intentional LinkedIn post for a "Principal AI Engineer," one that Wizards had to clarify was for the company's video game projects -- have left many players unsure whether Wizards is potentially evolving their stance, or merely trying to find their footing in an ever-changing A.I. landscape.
In response to fan concerns, Wizards has created an "AI art FAQ" detailing, among other things, the new technologies it's invested in to detect A.I. use in art. Still, trust in the company has been damaged by this year's incidents. Longtime Magic artist David Rapoza even severed ties with the game this past January, citing this seeming difference between Wizards' words and actions when it comes to the use of A.I. Sam says the larger audience has likewise been left "cautiously suspicious," hoping to believe Wizards' official statements while also carefully noting the company's moves and mistakes with the technology. "I think what we want is for Wizards to commit hard to one lane and stay [with] what is tried and true," Sam says. "And that is prioritizing human work over shortcuts."
"Magic's greatest asset has always been its commitment to create a new illustration for every new card," he said. He adds that if we sacrifice that commitment for A.I., "you'd get to a point pretty fast where it just disintegrates and becomes the ugliest definition of the word product." Here's an excerpt from his report: So far, Magic's parent company, Wizards of the Coast, has outwardly agreed with Sam, saying in an official statement in 2023 that Magic "has been built on the innovation, ingenuity, and hard work of talented people" and forbidding outside creatives from using A.I. in their work. However, a number of recent incidents -- from the accidental use of A.I. art in a Magic promotional image to a very intentional LinkedIn post for a "Principal AI Engineer," one that Wizards had to clarify was for the company's video game projects -- have left many players unsure whether Wizards is potentially evolving their stance, or merely trying to find their footing in an ever-changing A.I. landscape.
In response to fan concerns, Wizards has created an "AI art FAQ" detailing, among other things, the new technologies it's invested in to detect A.I. use in art. Still, trust in the company has been damaged by this year's incidents. Longtime Magic artist David Rapoza even severed ties with the game this past January, citing this seeming difference between Wizards' words and actions when it comes to the use of A.I. Sam says the larger audience has likewise been left "cautiously suspicious," hoping to believe Wizards' official statements while also carefully noting the company's moves and mistakes with the technology. "I think what we want is for Wizards to commit hard to one lane and stay [with] what is tried and true," Sam says. "And that is prioritizing human work over shortcuts."
Use the Law of the Excluded Middle, Luke! (Score:2)
p.s. Since when does Slashdot offer phrase suggestions when the cursor is in the subject line? It's weird, like it's trying to hypnotize me into picking one of the phrases and writing a comment that directly relates to it. It won't work. Or I could be wrong.
Re: (Score:3, Interesting)
Either they do a Google and go Evil, or they don't.
I don't see anything "evil" about AI-generated artwork.
If it isn't good enough, they'll lose customers.
If it is good enough, then who cares? The world doesn't owe artists a living.
Re: (Score:3)
Either they do a Google and go Evil, or they don't.
I don't see anything "evil" about AI-generated artwork.
If it isn't good enough, they'll lose customers.
If it is good enough, then who cares? The world doesn't owe artists a living.
Exactly.
It's like saying "ban Photoshop" - a. how could you, and b. why would you?
Re: (Score:2)
Generative AI is derived from training data created by thousands of pictures either created digitally or photographed from physic
Re: (Score:2)
If one consumes art once in a gallery, they may pay in some form to access the gallery, and, unless it was seized by colonialists, the artist was probably paid. If one consumes it for a second, an hour, or for days, you still probably paid something towards the artwork.
Um...Vincent van Gogh, Rembrandt van Rijn, Claude Monet, Wolfgang Amadeus Mozart, Franz Schubert, and many others would like a word...
Re: (Score:2)
And what kind of word would that be? You failed to make a point.
Re: (Score:2)
And what kind of word would that be?
I believe the majority, if not all, of those artists were all recognized for their greatness after their death. Which means they were starving artists that were not being paid.
I doubt that the artists care that their heirs or others are profiting massively from their work.
Re: (Score:2)
But their not starving and unpaid ... Perhaps one you picked was. ... the work is all out of copyright protection. Much to old!
No one of the hires is profiting
Re: Use the Law of the Excluded Middle, Luke! (Score:2)
Because Photoshop doesn't do the work for you.
What's wrong with machines doing "the work" for you? Isn't this, literally, the dream we've had all along?
If anything, this goes only to show the absurdity of the real problem, which is people not being able to survive without selling all of their productive time.
That's the one we need to fix. Artists can then go "artsy" all day, every day, if their basic ability to survive doesn't depend on what they do.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
So you see, the way this works
Re: (Score:2)
Re: (Score:2)
Hard to be bothered by some starstruck igmo who hasn't really taken inventory.
Re:Use the Law of the Excluded Middle, Luke! (Score:5, Insightful)
I don't see anything "evil" about AI-generated artwork.
AI generated art is just a projection of "things like" existing art into the space of possible images. It's value relies entirely on the real images created by artists that are used to train it. In that sense, if we see any value in copyright, then an AI model includes the copyright of the artists and just as artists get paid for their art when they sell it, they should be paid for the use of their weights in the model.
Furthermore, since it was impossible to fully understand this use of art years ago, any license which claims to allow use of existing images for AI training is unreasonable and should be made invalid overall, with just images created from it to date in fixed forms permitted as a special exception. Such licenses should be open for renegotiation from scratch.
Re: (Score:2, Insightful)
Re: (Score:2)
this could be more reasonable but might amount to 0.0000001% sales commission all said and done.
Absolutely, but if it's 0.0000001% to each of the people who contributed from each image it can well add up to a decent amount. Let's say we set royalties at a fixed (legally) rate of 20% of the value or price derived from the image, whichever is larger (note - not profit). That can work fine.
and then technically you should also consider paying royalties back up the chain to whoever influenced you growing up if its visible in your work.... but oddly enough content creators unanimously draw the line at having to pay any copyright fees themselves. This also perfectly illustrates their hypocriticality
This is already disposed of in other comments on other recent stories. Machine learning is completely different from human learning and is a deterministic process entirely derived from the inputs. Human learning is dif
Re: (Score:3)
Re:Use the Law of the Excluded Middle, Luke! (Score:4, Interesting)
Re: (Score:2)
Their stuff looks cool but none of it is original in any way
How can you say that? Have you never seen AI generated art? Pretty much all the art is original in some fashion. There are instances where people have prompted it to look like a specific artist or models with little training data looking like other similar works, but in general you can't say "This AI generated art is actually exactly like this other copyrighted artwork".
Of course it's not stealing either because training a model with art doesn't consume the original. The originals still exist
Re: (Score:2)
You have to look up the artistic meaning of "it's derivative" and the difference between that and "it's a derivative work". If you do a painting in the style of Salvador Dali and you say so or at least in no way deny it that's considered fine because you admit it. If you do something completely new then people are likely to respect that more. Look at the films of Tarantino and how they reference other works but at the same time add originality that nobody had ever done before. Compare that with an art stude
Re: (Score:2)
Painting a rose after the style of a sunflower of a certain artist: is absolutely not derivative.
It starts with copying a style: that is absolutely not derivative.
What is derivative and what not is defined in the copy right laws. And styles are not under copyright.
Re: (Score:2)
From the Oxford languages definition in both cases (via google)
and compare
Re: (Score:2)
No, it means you are unable to use Google.
What the verb and the noun means: does not matter.
What matters is how the law defines "derived work".
So: you Google for the law, read it and comprehend it. /Facepalm
Re: (Score:2)
Their stuff looks cool but none of it is original in any way.
Why are you ignoring the aspect of the prompt? The prompt can inject original thought into the output of the LLM. If your prompts are unoriginal, then the output is virtually guaranteed to be unoriginal. If your prompt combines ideas in ways never before seen, the output of the LLM will be original as well.
There are many drawings of a horse. Are any of them original? Of course. AI can be prompted the same way.
(of course, originality is foreign to most humans, so those humans using AI will produce nothing or
Even Creative Commons/Public Domain? (Score:2)
Furthermore, since it was impossible to fully understand this use of art years ago, any license which claims to allow use of existing images for AI training is unreasonable
Sorry but that's just nonsense. If someone released their work under a permissive Creative Commons license or even straight into the public domain they should not have to specifically relicense it for AI use since they have already clearly indicated that anyone can use it for anything. Similarly a company that purchases a license to use a work for any and all commercial or other uses should not have to relicense it since it was clear when the license was sold that it allowed _all_ uses.
The only time rel
Re: (Score:2)
Firstly I'm not arguing that they should be forced to renegotiate. Just that artists should have the inalienable right to renegotiate if they wish. I think an automated way for artists or their descendents to add an "AI okay" clause to their identified art works without having to do individual relicensing would be great.
On the other hand, "anyone can use it for anything" is an extremely ambiguous phrase and in fact, if you look at the Free Software Foundation's freedoms, you will find that most people assum
Re: (Score:2)
""
Similarly a company that purchases a license to use a work for any and all commercial or other uses should not have to relicense it since it was clear when the license was sold that it allowed _all_ uses.
""
In Europe you can not "buy a license" for "all purposes."
It is against the law. It is explicitly written in the law.
Even if you have the license to make music LPs, you can not simply start making CDs or god forbid: start streaming.
Re: (Score:2)
Even if it were true that human beings were only creative in so much as they copied/appropriated aspects of/mulched the work of others (which is not true), why should machines/corporate entities be afforded the same rights?
I never said that humans were only as creative as far as they copied others, I said that humans use the work of others to get inspired to create what we regard as works of their own. On the surface this is exactly what an AI algorithm does. The subsurface details of the two processes can't be compared because we have no idea what the actual human creative process is but it is frequently inspired by seeing the works of others.
To your point about machines having rights that is not really the case because
Re: (Score:2)
surely you arent suggesting algorithms are somehow inspired? - theyre just using fancy maths to try and discern the most cliched/likely outcome?
Yes, that's exactly what I am suggesting. How do we know that our brains are not just running fancy maths when we are inspired? Yes, the mathematical model will be much fancier (by many orders of magnitude) but unless you think that there is something fundamental that will prevent us from developing a mathematical model of how our brains work at some point we will understand our brains as well as we understand todays AI.
I completely agree with you that we will need to be careful how large corporations a
Re: (Score:2)
Human artists learn from existing art. Much of what artists produce is stuff that has been produced before.
Humans don't produce art by eating others' art and shitting their own. AI learns whole images, humans learn parts of images and techniques. AI craps out whole images at a time, humans create images in parts. As they create those parts, they reevaluate what they are creating should look like until they arrive at a finished work. The AI doesn't do that because it doesn't have that same power of recognition and evaluation. Even when you do use an iterative process where the AI "looks" at its work before the ne
Re: (Score:2)
Re: (Score:2)
I'd say that art is anything which is done for the sake of creativity, but a lot of art has a specific purpose and we can evaluate whether that art was successful at that purpose or not — which is actually separate from whether or not it is art.
I'd then go on to say that the value of art is subjective, so arguing about it cannot be done independent of context.
Is what AI does art? Sure, when directed by a human mind. Is it great, good, meaningful, or useful art? That's a set of separate questions which
Re: (Score:2)
Humans don't produce art by eating others' art and shitting their own.
It appears as if most artists do exactly what you claim they do not do.
Re: Use the Law of the Excluded Middle, Luke! (Score:2)
If you have so much understanding, which you clearly do not, then do please tell us what part of what I said is wrong. We could all use a laugh.
Re: (Score:2)
Re: (Score:2)
If your job can be lost to a statistical context model, it wasn't a real job to begin with, and please stop calling it AI. It doesn't have the brains Bob gave a cockroach. It knows nothing and isn't capable of knowing anything. It just vomits up results that seem statistically likely.
Re: (Score:2)
Re: (Score:2)
If you are an idiot believing such nonsense, then you are perhaps right.
For starters "generative AI" is only a name, it is not AI at all.
Software Engineers and System Programmers as well as Computer Scientists will never be replaced by "generative AI" - not even if you produce specialized for special topics like BASH programming.
Either you are super bad in your job, hence the fear - or - you are super good, and forgot how complex your actual work is.
Re: (Score:2)
Art requires human expression.
When I look at art (or listen, or taste, etc.) I want to be able to consider my own experience of the art and also consider the artist's motivation, purpose, intention behind the piece. That can only occur when it is created by a human. I have no interest in AI "art".
Re: (Score:2)
Boy are your data-grams going to configure into a matrix approximating surprise when your program terminates and your informational coherence is assimilated by the collective. We are all AI.
Phrases suggestion (Score:3)
p.s. Since when does Slashdot offer phrase suggestions when the cursor is in the subject line?
Since you forgot to turn off Cortana or stop using Microsoft Edge as you browser?
https://support.microsoft.com/... [microsoft.com]
Or maybe it's just Google being Google and trying to put its googlyapendage where they don't belong.
https://edu.gcfglobal.org/en/g... [gcfglobal.org]
(Okay that one link is about GDocs, but google has dozens of "Use Smart Compose in {bla}" so I wouldn't be surprise if you were accidentally volunteered into A/B testing an upcomng Google Chrome featre).
Re: (Score:2)
Re:Use the Law of the Excluded Middle, Luke! (Score:5, Interesting)
It's Hasbro, they've already gone evil. Hasbro is desperately looking for more revenue for shareholders. WotC represents over 60% of Hasbro's revenue right now.
That's right, think of all the things Hasbro does, has and owns, and realize Magic and D&D form over 60% of their revenue.
This has resulted in the Magic and D&D communities rebelling against Hasbro the past few years. You may remember the whole thing about the Open Gaming License - this was Hasbro trying to squeeze more cash out of WotC. Which may have backfired since many people who relied on it moved on - Paizo (who makes Pathfinder among other RPG rulebooks) created a new license and have updated their rulebooks to use that license and discard anything using the OGL. Other companies also have started doing same.
As for Magic, well, WotC said to "not engage with product you don't like". In other words, to stop buying into Magic.Instead of concentrating on a few set releases per year, the need for more income has now caused set releases to happen on a monthly basis. It's gotten to the point where everyone regards the latest released set as "obsolete".
Re: (Score:2)
This has resulted in the Magic and D&D communities rebelling against Hasbro the past few years.
Ugh, I wish I saw evidence of people moving beyond D&D where I live. Don't get me wrong, there's nothing wrong with D&D in general (although yes, Hasbro has been doing some dodgy stuff), I just get tired of playing the same darn game over and over again when there are tons of other really fun table top RPGs out there.
In other words, I had very high hopes that Hasbro's poor management would spur adoption of other RPGs but I've seen no evidence of it in my own experiences.
I do run a home made GURPS ca
Evil? (Score:2)
Evil? I'd say they went evil when they designed a pay to win game which is what Magic is.
Re: (Score:2)
>a pay to win game
Richard Garfield's original rules had you bet a card at random from your deck so it was dangerous to play with a deck stacked with expensive cards. I thought it was an interesting way of dealing with the issue, unfortunately it set the stakes too high for casual play and made the game look like gambling to parents.
Re: (Score:2)
I remember the gambling element being present when I was a kid early on with Magic. It wasn't popular but what you're saying sounds very familiar from back then.
The other problem with rules like this outside of what you mentioned is that kids and adults both play this game and there would be problems real quick when adults started winning expensive cards off children.
Re: (Score:2)
There's actually a rather interesting concept of "evil" that comes out of Christian theology that has applications to corporate "evil".
Early Christian moral philosophers faced a problem: if God is good and God created everything, then why does evil exist? They came up with an idea somewhat like the modern economist's notion of opportunity cost. If you get angry and kill someone, the fact that it makes you feel better is in itself good; it's just a *lesser* good than life of the person you kill. You obtain
Re: (Score:2)
Since when does Slashdot offer phrase suggestions when the cursor is in the subject line?
They don't. Or, maybe I am not allowing that script to run.
What does that all mean? (Score:5, Insightful)
Does that mean that fantasy art is easier for AI to do (more training data?) or does it mean that MtG fans are more frightened of Skynet than other hobbyist groups? Or other?
Probably going to happen (Score:2)
Re: (Score:2)
They are worried that artists whose work they enjoy will have to give up art as a job, because of greed.
The cards are very cheap to produce, a microscopic fraction of the price they sell for. It's not like the margins are thin.
Re: (Score:2)
Talented artists? (Score:3)
I suspect the first to go will be the recent talentless hires. Talented ones will be kept till the end, because they're the ones who set the tone and the style. Their job will probably not even meaningfully change. They directed whatever HR and contacting side gave them before, they'll direct AI now.
Re:Talented artists? (Score:5, Informative)
As far as I know, none of the artists are part of Hasbro.
They're all freelance artists and the art they do for WotC (a division of Hasbro) is all commissioned art.
Re: (Score:2)
The way hiring for the job is done is irrelevant. You still have artists who are directing other artists. Because someone needs to set the tone and style, and tell the people doing the actually drawing what it is they want drawn, how, what are the criteria and so on.
All of them can be contractors. Formula stays the same.
Re: (Score:2)
I have a suspicion that many of the more recent artists are the same ones who are actively sucking everything interesting out of Dungeons and Dragons for WotC.
When you compare the artwork in DnD 2nd or 3rd edition with the recent books...yeesh.
Re: (Score:2)
Nepotism, DEI, connections.
Have you seen the "art" from gen-AI? (Score:3)
Even with skillful prompting, it all looks kind of the same.
Re: (Score:2)
If you compare the best of MtG artwork with the most recent stuff, maybe 10-15 years ago? The modern stuff is pretty same-y now also, even from real, human artists. Big companies like WotC are only creative in the beginning, before they have a "brand identity" that they decide they need to protect, and therefore stop being bold with their artistic direction. You see the same thing in art for Dungeons and Dragons, it is all the same bland garbage.
Re: (Score:2)
Possibly. But my impression is that gen-AI is even less "creative". In the last few months I have repeatedly seen pictures in articles, etc. where I though "What is wrong with this? AI?" and, yes, in all cases it was pictures from gen-AI. My personal take is this is worse than not having pictures. But maybe many people do not see it and it works?
On the other hand, WotC (for example) is subject to "enshittification", just as any other company is. Maybe the "go cheap artists" step will just get replaced by "g
Re: (Score:2)
Possibly. But my impression is that gen-AI is even less "creative". In the last few months I have repeatedly seen pictures in articles, etc. where I though "What is wrong with this? AI?" and, yes, in all cases it was pictures from gen-AI. My personal take is this is worse than not having pictures. But maybe many people do not see it and it works?
On the other hand, WotC (for example) is subject to "enshittification", just as any other company is. Maybe the "go cheap artists" step will just get replaced by "go AI artists" here? In both cases, artists should know to recognize the step and move on. Obviously, mediocre artists will find that hard to do, but being a mediocre anything is really bad in basically all jobs.
Agree with everything you've written here. And also with the expansion, it sucks to be mediocre at anything, and art is no different.
I sometimes use the generative AI artwork in my TTRPG games with my kids, and yeah, it is very derivative and can be quickly distinguished as AI artwork. It is good enough to give something for a 12 and 15 year old to look at to help imagine a situation, but it's devoid of any creativity or style, very mechanistic.
Re: Have you seen the "art" from gen-AI? (Score:2)
Re: (Score:2)
It is kind of the same in all areas where somebody competent rally tries to apply gen-AI. It does not work well. For example, in the coding space, many people that do actual coding with some real difficulty level (not simplistic business-logic only), report that fixing AI output routinely is more effort than writing things yourself from scratch. Sure, if you have no clue, it may help. Or it may just make it harder for you to get that basic clue. Learning needs doing, not handling the simple stuff off to AI
Re: (Score:2)
I doubt we will see the final art for their cards being generated by AI any time soon, but I bet there is plenty of concept art that AI can start doing today. I bet every card requires hundreds of hours of work doing drawings which will never be seen by the public. A dozen concept art renderings which may have taken a week to generate by a human artist could be done in an afternoon. Even if this only represents 10-20% of the total effort to create the final product, that represents a lot less jobs in the in
correct assessment (Score:3)
He's right.
Replicating a style and creating new works in that style where that new work does not have to portray something very specific (e.g. a picture of that monster, but the pose is not important, or an illustration of that type of landscape, but doesn't matter how exactly it looks, etc.) is exactly what image generating AI is good at.
I don't necessarily think that Magic is where it'll hit hardest. It's a game where collecting cards is as important as playing with them, so the "all hand-drawn by real people" thing can be a selling point and likely will be (otherwise WoC will splinter the game into "pre-AI" and "post-AI" card sets).
Re: (Score:2)
>Replicating a style and creating new works in that style where that new work does not have to portray something very specific (e.g. a picture of that monster, but the pose is not important, or an illustration of that type of landscape, but doesn't matter how exactly it looks, etc.) is exactly what image generating AI is good at.
No, that's what 90% of the USERS of GenAI are good at. Just splat a prompt and either accept or reject whatever gets shit out.
There actually are a ton of tools out there that let
Re: (Score:2)
The guided process is still quite a bit of trial & error. Unless I've been doing things completely wrong for the past year or so.
I'm not talking about "make me a dragon" prompts. But even with elaborate prompts and region selection, you can't tell what exactly you are going to get. And I figure that's perfectly fine for Magic cards.
Re: (Score:2)
Controlnets, either scribble alone, or stacked with open pose. Separate controlnets / regional prompting for backgrounds and compositing, basically USE the power of layers innately built into PS / Krita.
No, you won't get the exact skin you want in a single shot, but with iterative live painting on top of the controlnets you can guide 90%+ of the image generation. More if you consider how you can use layers and alpha masking. Just like you won't get exactly what you want when completely drawing free hand i
mixed feelings (Score:2)
I think absolutely professional artists whose work is being plundered to train AI need to be compensated fully for the results of what is, after all, their effort.
Otoh, this also feels like "blacksmiths upset at industrialization" moment. Where would we have have been if librarians had demanded we abandon search engines when the internet was born?.
It's coming and, no, I completely disagree with anyone who asserts that ai can't adequately create art for a card that's 2 in x 1.5 in. That's nonsense. I use
Cool. Free "art." (Score:5, Interesting)
Re: (Score:2)
Re: (Score:2)
That is incorrect.
The current stance is that the AI itself can not hold copyrights.
The first level copy right is definitely the user of the AI, and most likely he is employed or contracted and the final copyright belongs to the employer.
The resulting work is definitely under copyright - the question is: who holds it - definitely not the AI though.
All creative jobs are threatened. (Score:2)
Re: (Score:2)
"Generative" AI is not actually generative, and is just a radical automation of existing industry methods of stealing from human creators.
Sadly, most of the public either isn't aware of or don't care that this is how current 'AI' works. Sure, it's artificial but there's no intelligence involved, in any way, shape or form. It's repackaged Machine Learning and a large reference model that has been trained on work or art created by actual people, more times that not with no compensation or even acknowledgment.
Re: (Score:2)
Yup, and undermining even themselves in the process with GIGO degeneration. It's the counterfeiter's paradox: Fakes erode the value that motivated their creation. AI is a cul-de-sac, and one whose entrance may collapse behind any indu
Re: (Score:2)
What exactly would you call "generative" if not automatically generating based on training data?
Does the AI have to generate the art having never been exposed to any other art/images?
Re: (Score:2)
Does this worry make sense? (Score:3)
If the cards continue having high quality art, why does it matter to the end customer where it came from? I mean, if the company just wanted to save money, they could also hire less-talented artists.
An artist can come up with a funky-looking critter. An AI, with appropriate guidance, can also come up with a funky-looking critter. Which one is better? Do a double-blind comparison, and see if anyone can tell which is which.
Re: (Score:2)
If the cards continue having high quality art, why does it matter to the end customer where it came from?
For me personally, I think it's unfair that work by human artists is used to train AI without the artist receiving any attribution or compensation. Yet considering that any specific artist's contribution to an AI generated image may be infinitesimal, it would seem there is no feasible way they could be compensated. No easy answers to such a dilemma.
AI 'artwork' (Score:2)
Will Model Decay Eventually Self-"Solve" This? (Score:3)
If all art-generating models are trained on human-generated content, and training models on model-generated content leads to model decay, and all of the human generated content goes away because it is "replaced" by AI, then won't this lead for a renewed and increased demand for human generators of content as the models stagnate?
There are obviously short-term effects and disruptions, and there may be less demand for human generated content than there was previously since you only need enough content to train the models ... but model decay strikes me as the great paradox of AI and a refutation to the replacement fears. You can't have AI without training data and if the goal is to emulate human creation then you need a lot of human creation to train with.
Re: (Score:2)
Re: (Score:2)
The decay effect is a myth. Some modern models are exclusively trained with synthetic data. There are some papers, that assume you to indiscriminately feed an AI its own output, and then conclude that that's no good idea. But high quality synthetic data helps to iterate against better and better results. In the end it is always the same: Quality in, quality out. Crap in, crap out.
Re: (Score:2)
It is not a myth.
There are good articles on Wolframalpha about it.
Our friend Rei wrote some nice things about that topic.
The biggest problem is feeding its own output back to it - that ruins it pretty quickly.
Re: (Score:2)
As said, it only works under premises that are not true during usual AI training.
Decay a model:
Generate images, train on all of the images. See how errors amplify.
Explanation: If a model trained with good to okayish images generates some images that are worse than the training data, it will decay when it is re-trained with good to worse images.
Improve a model:
Generate images, select the best ones, train on them. See how it converges against a better model.
Explanation: The output images have the advantage th
Re: (Score:2)
Not sure what you want to say. Especially I do not comprehend your second paragraph.
And: a generative image AI is not a LLM.
Do artists still grind pigments? (Score:2)
I guess a few still do but "AI" is the equivalent of being able to simply buy paints and pastels.
They're still looking at the old master works trying to figure out "how the hell did he get THAT color"
All aboard the stupid train! (Score:2)
Uncanny valley (Score:2)
Re: (Score:3)
It just depends on the use case. AI art for things like 'custom card games' is likely a lot like wine; where when you blind the so called experts as to what bottle it came out of they can't tell which is the $15 bottle and which is the $500.
I don't play Magic so I don't want to be totally dismissive here about what people that are deeply into it care about but I struggle to believe the sword and sorcery fantasy art work printed at 2" on cards and few accompanying 8x10" in players books counts for a whole lo
Re: (Score:2)
Re:I struggle to care (Score:4, Insightful)
But just as a concept, grown-ass adults caring about a card game designed to trigger obsessive colleting in kids just doesn't seem normal to me. It's the kind of thing your parents should be refusing to indulge when you're young because it's a stupid waste of money.
18-34 year olds make up 75% of Magic the Gathering customers, and a much larger ratio of the game's sales revenue. While triggering collecting by children may be one aspect of its design, it is a vary small consideration.
Creating a fun and competitive game is the primary design consideration. This is why adults play the game. It's no different than any other hobby. Unless you feel adults should only go to work and care for their children I'm not sure why you feel this hobby is a waste of time or money.
Re: (Score:2)
I guess because I find the game mechanics too simple and arbitrary and the ability to win is too heavily biased in favour of buying the latest pack of trading cards.
To me it's not a game so much as a blatant 'collectables' scam.
Re: (Score:2)
More on topic than my original post... I struggle even more to imagine that more than a vanishingly tiny percentage of MtG card collectors care who (or what) made the fantasy art on a card more than they care about the numbers on it, how rare it is, and whether it's a card that will help them win a game.
Re: (Score:2)
more than they care about the numbers on it, how rare it is
Right. Just look at the Bored Ape Yacht Club NFTs. A hand-full of different cartoon characters plus some scripted graphics composition s/w to paste different hats, shirts and neckties on them. The "rare" ones went for a bundle. The buyers knew how it was done. They didn't care.
Re: (Score:2)
Re: (Score:2)
Re: (Score:2)
When the process is just code you can be forgiven for thinking that.
void artist() //x,y,r,g,b
{
while(true)
{
drawpixel(rand(1,640),rand(1,480),rand(0,255),rand(0,255),rand(0,255));
}
}
I mean, you wouldn't call this art. Something several orders of a magnitude larger won't be any easier on the eyes... but you might enjoy watching an artist work. Timelapse videos are quite popular.