January 14, 2023: AI Art In Gaming

[RoboAdmin]

January 14, 2023: AI Art In Gaming

Read this article on the Illuminator.

[emigio]

that post reminded me this.
https://www.history.com/news/industr...ddites-workers
Didn't end well for them

[kkc]

In the spirit of respectful disagreement, isn't there a clear difference to be drawn between goods and art? The Luddites were certainly protesting a disruption to their livelihoods for similar reasons, but their work was always destined to be automated because it was something to be standardized and repeatable. By comparison, art at its best is a unique expression that's also representative of the state of a particular culture or identity. In rare but awesome circumstances great art represents a profound emotional state that all human beings can relate to. Even if we could trust that to an AI, I'm not sure that it would have the same ability to capture important moments in time.

Remember too that with art, we see purpose and intent that's assigned to or presumed in the artist. A lifetime's experience, framed by conflict or change, is what drives people to express something that can only be communicated in art. There's a reason why Steve's name on a game box matters, because I trust that the work he's involved in will be entertaining, include wry commentary on the human condition and will result in a game I want to play more than once. Curt Schilling doesn't inspire nearly the same amount of excitement, loyalty or spending from me.

I'm glad that there are game companies who are seeking to protect human artists and maintain demand for their services. AI art has its uses and purpose, but it also has limitations that would hold back games or whatever other media it ends up being used in.

[ravenfish]

Originally Posted by kkc

[...]their work was always destined to be automated because it was something to be standardized and repeatable. By comparison, art at its best is a unique expression that's also representative of the state of a particular culture or identity.

Unique is where you find it. Before the Industrial Revolution, each piece of clothing you wore was custom crafted to fit your particular body (assuming you were rich enough not to be wearing hand-me-downs). Mass production could not duplicate this aspect, but it could create a product that was nearly as good and far, far cheaper, so people ran with it.

[Archangel Beth]

Clothing is still done by hand. Cloth is automated, but cutting and sewing clothes requires someone at the sewing machine. They may "assembly-line" the garments (I don't know and am too fried today to do the research), but that cheap t-shirt you have? Was made by a human.

(But not woven by humans.)

[Celjabba]

I have mixed feelings.

On one hand, the reasoning exprimed above and in Chaosium article are sound, and in addition, the fear of the impact of AI reducing future illustrator work (especially beginners) legitimate.
Not to mention that without new human artists to feed new materials into the engine, the AI won't get creative and innovate on its own should it became widely used and diminish new art production.

On the other hand, I cannot draw, so if I wanted to publish for exemple a scenario, I would have 4 options:
-no art.
-stock art, seen everywhere and (especially) not tailored to my story.
-paying hundreds of $ for custom art
-spending a few hours feeding prompts to an AI and adjusting it.

And this is not abstract : For last week game, I made portraits for my pregens and major NPCs as well as a few landscapes to describe the setting, all done in a couple hours on Stable Diffusion, and while imperfect in some details, these pictures made a coherent whole tailored to my story details. If I wanted to put that scenario online for a price, many site will refuse AI-generated illustrations, and I certainly cannot pay a human artist price for something I may sell 20 unit for a couple € each...
And the work may be derived from (thousands of) other artists works, but it is certainly unique and not a straight copy of existing art, so it is not illicit use... or is it ? A very hard question to answer.
So ... mixed feeling, as I said.

[Babipoki]

AI art must never make any profits*. I'm an illustrator, but I still use AI to generate me some inspiration.

One exception I may have is that the AI image is obviously labeled as AI generated.

AI can't create art, it is trained off other artists' directly, with no 'creative' process involved. It's very different from argument that all human artists use references. We can look at a reference and have our own thought process about it. AI sees a 'reference' and tries to reproduce it, with data set from millions of artists.

I support advanacement of AI art, but a line must be clear, and any AI-generated artwork must always be obvious that it was AI generated, without looking at the tags.

Also fully understand using AI for making NPCs for your campaigns. We need to create hundreds of NPCs and commissioning 100 artists or 100 artworks is probably out of most people's budgets, especially when running free games. Even $5 a session wouldn't cut it, artists must be paid fairly, so at least $100 - coming up to $100 x 100, $10,000. It is unethical to use AI for it, but whether it's legal or illegal is still a big question - can someone's style be copyrighted? It's an issue at the court of law to be argued about. Just don't make profits off it outside of your campaign group, and you should hopefully be fine.

[kkc]

Celjabba's post prompted me to remember that when legal systems have to consider new moral problems, they look at precedent. How did we address the need for giant piles of art before AI art was an option?

In the specific case of stock art, there are clear rules around the usage and ownership of it, and under what circumstances you have the privilege of modifying or distributing it. While a lot of it is free, there's a perception (or possibly a consensus) that you'll have a pay a license or a per-image usage fee for art that's either higher quality or that fits a narrower purpose. What that gives you is a range of options, where in general the more you pay the better the possible result, which puts commissioned art at the far end of that range.

The wildcard here is art that's in the public domain. There are some images which are owned by the world at large (at least in the countries that keep track and have a clock running so you can tell). The option that's generally exercised here is that public domain art gets modified to become fit for whatever purpose which creates a derivative work that's now technically owned by the person who did the modifying. Obviously this also happens with the free stock art, but I didn't think of derivative works until just now, so let's keep rolling with the thought exercise.

The major precedent for derivative work is to be found in the music world. Just like with visual art, there are stock elements that I can use for a nominal fee, and there are public domain audio clips I can use if every track on my album needs to start with the Wilhelm scream. But when I want to use a previously-recorded Gerard Presencer trumpet lick on the chorus of my hit ska track, now there's decades of legal precedent that covers how much of Presencer's work I'm allowed to reuse, and critically how much money he's owed as my derivative work earns for me.

Where the parallel breaks down here is that it's never been possible before to remix to combine visual work in quite the same way. If I put just put a hat on the Venus de Milo, I don't owe anybody any money. If I put a hat on Andy Warhol's portrait of Marilyn Monroe, I might owe his estate some money. Who do I owe when my AI assistant has taken 40,000 different images of a woman wearing white (or being carved out of white marble) to produce the portrait of my ghostly NPC? This is where precedent breaks down and we have to consider what's possible.

One impractical answer would be for AI art generation to have their source images more sensibly curated. Just because an image is published online doesn't mean that anybody has consented to derivative works being made, and that's an argument that might have a strong legal standing. But just like with stock art, some artists might consent to their work being used freely, whereas others would require payment for their art to feed the algorithm. That would create an interesting level of competition, where some prominent or recognized artists would permit usage only by AI systems that they trust or that paid them the most competitive rate. You flip that around that now you have AI art generation that sell their services based on them having the best arists's material to generate from, or in some cases the most ethically and legally sourced too.

[Phil Masters]

Originally Posted by emigio

that post reminded me this.
https://www.history.com/news/industr...ddites-workers
Didn't end well for them

Last I heard, the Luddites appear, on close examination, to have been engaged in a ... robust ... form of wage negotiation, rather than having a reflexive aversion to new ideas. Which may of course also be how the AI art issue shakes out in the end.

Originally Posted by Celjabba

I have mixed feelings.

Me too. Consider me to be seconding everything in Celjabba's post.

Originally Posted by Babipoki

AI art must never make any profits*.

That's a rather broad declaration. We could just start with the question of whether the AI programmers deserve remuneration, and the possibility of a system that remunerates the artists that the AI uses as seed stock.

Originally Posted by Babipoki

I'm an illustrator, but I still use AI to generate me some inspiration.

So you're using AI art, and presumably hoping to make a profit from the work?

Originally Posted by Babipoki

I support advanacement of AI art, but a line must be clear, and any AI-generated artwork must always be obvious that it was AI generated, without looking at the tags.

That sounds like a futile demand, given that the whole point of much of the work on AI image generation is to produce images that look naturalistic or like the work of a human artist. In fact, I'd say that you've already lost that one.

("We demand that nobody ever fixes the seven-fingered hands problem!"?)

Originally Posted by Babipoki

(re. images for games) It is unethical to use AI for it...

"Unethical"? Really? Why? Nobody is taking bread out of the mouths of artists, or claiming accomplishments that are not their own. Long before AI art came along, I was using images off the Web on my character sheets. The assorted Renaissance painters, Hollywood publicity shot photographers, and Victorian family portrait studios whose work I borrowed suffered no loss thereby.

Originally Posted by kkc

One impractical answer would be for AI art generation to have their source images more sensibly curated.

Shutterstock actually seem to be making a constructive stab at this. They're developing an image generator (working with LG and the people behind DALL-E), and say that part of the project involves developing a system for rewarding the artists whose work is fed into the engine. If that's built into the AI from the start, I can see it should be possible to make it work.

Of course, this will restrict the AI's points of reference to artists who Shutterstock can locate and remunerate (plus, hopefully, public domain stuff) - mostly Shutterstock's own library, I guess. And how well each artist will be paid remains to be seen. But it's a start.

[Shostak]

Originally Posted by kkc

In the spirit of respectful disagreement, isn't there a clear difference to be drawn between goods and art?

No, there isn’t, which is partly why so many art museums have decorative arts departments (ranging from chess sets to furniture), costume departments, what used to be considered ethnographic collections (which often contain religious or ceremonial objects), collections of arms and armor, film/movies, and now some art museums even collect video games.