Generative AI Art

[doofus-01]

Wesnoth does not accept AI-generated art, this post is an explanation.

I've wrestled with the issues of AI art and what the official Wesnoth policy should be. My basic thinking hasn't changed much from viewtopic.php?t=56294. The AI performance has improved since then (sort of [*]), but the general landscape hasn't changed much.[**]

I've tried to see what other indie and open-source games projects had to say about this publicly, and the main takeaway I got is that we would be industry leaders if we had a well-defined stance with clear guidelines. A notable example of a project in a similar situation is https://github.com/endless-sky/endless- ... sions/9325

Creative Commons seems generally supportive of the technology, but wary of how it will be executed [***]. OSI seems to be similarly ambivalent.[****] I'm trying to ask other people involved in open-source game art, including former art leads for Wesnoth, what they think. That will take a while but so far, wary ambivalence seems to be the feeling.

In the absence of clear guidance, I have to go to with some level of gut feeling and introspection: What are we doing here? Wesnoth UMC is easy to create (making it high quality isn't easy, but it's still easy to understand). Part of the joy we have to offer is accessibility. If this space were flooded by Generative AI art, the art novice branch would be lost. As an open-source project, we must have some level of idealism. For this reason, regardless of future legal rulings, there are two (related) issues we should have with AI:

1. AI collages, it doesn't create, but it competes with creators - It's great that you get a painting of a wizard or whatever, but it's a Faustian bargain[+] We're not a game studio trying to make payroll, we shouldn't sell our souls. If there is nowhere for amateurs to get exposure, we are eating the seed corn, there is no future.[++]
2. AI companies aren't benign - The people pushing AI abusively consume human effort[+++] and incredible amounts of energy[++++]. The general technology is a tool, like fire or fission, it's not inherently good or evil. But we shouldn't embrace the offerings and frameworks of the current industry leaders.

tl;dr - I'm against allowing AI art here; maybe that will change, but I see no reason why we should be pressured into accepting it. We would gain some assets, but we would lose some of our community function and we would be embracing the framework of immoral people. I make no attempt to address human art that used an AI tool somewhere, or what the threshold would be.

-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------
[*}Stable -Diffusion couldn't draw people anymore once the porn was removed from the training data(,https://arstechnica.com/information-tec ... dy-horror/) which was predictable if you don't buy into "AI training is just like an artist visiting the museum".
[**]https://www.polygon.com/ai-artificial-intelligence ; https://www.copyright.gov/ai/ai_policy_guidance.pdf (Note the dates)
[***]https://creativecommons.org/tag/artific ... elligence/
[****]https://opensource.org/press-mentions/h ... y-not-very (Like most of you, I don't have a subscription to the Financial Times, but there are other links in that roll, and an implied endorsement.)
[+] https://www.theatlantic.com/technology/ ... se/677595/
[++]https://crooked.com/podcast/social-medi ... f-slop-ai/ (there's some fluff in the beginning, but they do get to the point eventually)
[+++] https://www.businessinsider.com/openai- ... tstxt?op=1 & https://brighterworld.mcmaster.ca/artic ... velopment/
[++++] https://arstechnica.com/gadgets/2024/07 ... ive-years/
https://www.nature.com/articles/d41586-024-00478-x

[Pentarctagon]

Thanks for researching and writing this up

[doofus-01]

Thanks. It's no expert whitepaper, but it is what I will link to when future AI quibbles come up.

If project members object to something written here, please let me know.

The gray area of human creator art that used an AI tool somewhere is probably the most important case, as far as what would actually crash hopes into a brick wall, and is not addressed here because it is a thorny issue - probably best dealt with case-by-case, as long as everyone enters into the situation with their eyes open. But AI art presented as-is or minimally touched-up by a human is the greater flood risk, and I think this post makes the stance on that sort of thing clear enough.

[Saizo-Luz]

Interesting! I hope I understood correctly, as I'm not a native speaker!

Are there still artists here creating art for the game? From what I've seen, everyone always takes artwork from each other - the only downside to this is that sometimes a character makes history (epic character and epic story)... and then there's the same image in another campaign. Now the part about it becoming a flood, that's true! Probably because it will be much easier, and you won't need to marathon campaigns and eras looking for artworks for characters.
Also, there aren't images for different schedules (rainy, stormy, cloudy, and maybe others). I used an AI image generator tool a while ago for a presentation I needed to do - and I was amazed at how easy it is to create images so quickly and with good quality.

I will try to find the tool again and attempt to create some images of different schedules using as a basis the ones the main line has - and I will send them here (if allowed) for comparison to try to help you reach more concrete conclusions. It's a big issue, and I know how frustrating it is to have to abandon our roots and change something that's already part of our environment.

[Saizo-Luz]

So... I took the liberty and brought here some schedule images created by an AI image generator. For some reasong I have to put the images at the end...

It's way different! I don't think schedules images would be a problem - it's not like everyone is gonna create an infinite number of schedules.
Now... portraits...
Even if someone were to use AI-generated portraits, it doesn't even make sense to use them in eras - it would have to be campaigns and probably all the portraits would have to be the same style - otherwise it would look really weird. Now... story images... I don't think that would be a good idea either!
When people here want to use images for stories, sometimes they use images from Google Images ! Furthermore, it makes much more sense to use a map as an image (like mainline campaigns). Landscapes and images for story are just to take up space and weigh down the add-on (in my opinion).

What do you guys think about that?

[DuncanDill]

So... I took the liberty and brought here some schedule images created by an AI image generator. For some reasong I have to put the images at the end...

It's way different! I don't think schedules images would be a problem - it's not like everyone is gonna create an infinite number of schedules.
Now... portraits...
Even if someone were to use AI-generated portraits, it doesn't even make sense to use them in eras - it would have to be campaigns and probably all the portraits would have to be the same style - otherwise it would look really weird. Now... story images... I don't think that would be a good idea either!
When people here want to use images for stories, sometimes they use images from Google Images ! Furthermore, it makes much more sense to use a map as an image (like mainline campaigns). Landscapes and images for story are just to take up space and weigh down the add-on (in my opinion).

What do you guys think about that?

Its not so muc the ability of the ai, than the ethics.

[Saizo-Luz]

Its not so muc the ability of the ai, than the ethics.

I know! I'm just highlighting more negative points besides that! And I actually think that new schedule images would be good, because we have very few.
As I said:

I don't think schedules images would be a problem - it's not like everyone is gonna create an infinite number of schedules.

That wouldn't be an issue in my opinion - now, portraits and story images...

[DuncanDill]

Its not so muc the ability of the ai, than the ethics.

I know! I'm just highlighting more negative points besides that! And I actually think that new schedule images would be good, because we have very few.
As I said:

I don't think schedules images would be a problem - it's not like everyone is gonna create an infinite number of schedules.

That wouldn't be an issue in my opinion - now, portraits and story images...

Ah I see, it was a bit hard to understand what you were saying. As for schedule images, I really think no ai art is good as it is, wesnoth should have a sort of skill that was put into making it. Anyways, we have all the schedules we need for now.

[Saizo-Luz]

Ah I see, it was a bit hard to understand what you were saying.

Sorry!
I'll talk to doofus in private about schedule images.
I want to know something... how do you put the name of the speaker in the quote? Like... Saiz-Luz wrote:

[white_haired_uncle]

Ah I see, it was a bit hard to understand what you were saying.

Sorry! ^_^
I'll talk to doofus in private about schedule images.
I want to know something... how do you put the name of the speaker in the quote? Like... Saiz-Luz wrote: :?:

I hit that button to create this post.

[Saizo-Luz]

I hit that button to create this post.

Thanks!

[name]

If project members object to something written here, please let me know.

This is less of an objection and more of a suggestion - why not just stick to the argument that the legality of this technology is currently unknown? Perhaps most of the community will understand the risk to a zero budget project of getting sued for copyright piracy and not want to risk one of their favorite games on such a needless gamble.

Probably years from now,the legality will be cleared up, but likely many other things will have changed by then as well. We may have become inactive developers with younger successors needing to make the call, the technology may have plateaued or advanced far past our present expectations and the public's opinion of generated creations may have matured (perhaps either becoming bored of its limitations and better valuing "hand crafted" work or otherwise seeing it the same way we presently see digital photography and human made digital art versus hand painted art on physical canvas).

So by the time we know if we even can legally use this technology at all, we should also have a more informed opinion of what is or isn't desirable. Which is especially important if we are talking about enforcing a subjective policy not just on mainline but on the UMC community too.

The gray area of human creator art that used an AI tool somewhere is probably the most important case, as far as what would actually crash hopes into a brick wall, and is not addressed here because it is a thorny issue - probably best dealt with case-by-case, as long as everyone enters into the situation with their eyes open.

Another grey area is things like voice acting which are outside the community's motivation or capability to do well or at all. High quality generated voice acting could have enormous potential to improve our narrative experience, accessibility for those with vision impairments, and competitiveness with games which can afford a professional cast and audio engineer.

[doofus-01]

So... I took the liberty and brought here some schedule images created by an AI image generator. For some reasong I have to put the images at the end...

It's way different! I don't think schedules images would be a problem - it's not like everyone is gonna create an infinite number of schedules.
Now... portraits... :|
Even if someone were to use AI-generated portraits, it doesn't even make sense to use them in eras - it would have to be campaigns and probably all the portraits would have to be the same style - otherwise it would look really weird. Now... story images... I don't think that would be a good idea either!
When people here want to use images for stories, sometimes they use images from Google Images :lol: ! Furthermore, it makes much more sense to use a map as an image (like mainline campaigns). Landscapes and images for story are just to take up space and weigh down the add-on (in my opinion).

What do you guys think about that? :hmm:

Its not so muc the ability of the ai, than the ethics.

That's right. Ethics, legality, and project goals are more important than whether an AI output is a nice image.

Probably years from now,the legality will be cleared up, but likely many other things will have changed by then as well. We may have become inactive developers with younger successors needing to make the call, the technology may have plateaued or advanced far past our present expectations and the public's opinion of generated creations may have matured (perhaps either becoming bored of its limitations and better valuing "hand crafted" work or otherwise seeing it the same way we presently see digital photography and human made digital art versus hand painted art on physical canvas).

So by the time we know if we even can legally use this technology at all, we should also have a more informed opinion of what is or isn't desirable. Which is especially important if we are talking about enforcing a subjective policy not just on mainline but on the UMC community too.

Thanks for bringing this back on track. If the technology or ethical situation changes, policy can change, especially if the reasoning is available. Hiding behind legality alone seems pretty shaky and contributors should know what they are contributing to.

[Edenray]

Hi all, I would like to ask a question about time and accessibility.
I am building my first campaign that is story heavy, it has a lot of imagery for story parts. For now, I am using AI in photoshop to collage together various elements for my backdrops. If I were to draw them all to a decent enough quality, it would take me a year (I jest, but I am a slow drawer). Until I can find someone to help me create all the imagery I need I don't personally see the harm in producing AI place holders.
If the project is clearly labelled as a WIP, and has been uploaded in order to get feedback, what does the OP think?
Perhaps even a separate server for WIP's would solve the issue?
Am I to currently ask in the forums for play testers, and send them the project through email or WeTransfer?
Am I to remove all the guide imagery before I can upload?
To argue against myself, maybe having guide imagery would be to restrict the imaginations of those who would help..?
Thanks for any help and opinions
PS: I am quite willing to remove the offending imagery if needs be, but if it is labelled as a WIP, does that make a difference?

EDIT: Sorry I think I have asked this question in the wrong thread!

[Ravana]

It is not allowed in any wesnoth hosted storage. Whatever you do in email and other file host sites is your business.