If you respect the art of game development in any way, please, do not use AI Art.

[Desk]

The main problem with this topic, I feel, is that 1) it's a moral/ethical quagmire and 2) we're trying to judge/measure something "new" (AI art and all the wonderful tech that is slowly coming out on the market) by using something "old" (law, ethics, philosophy and morality as they've been up until now).

Many of our preconceptions are just that: preconceptions. They're born from what is (and more importantly what has been) possible right now (or up until now). Similar to how the cellphone was heralded as the death of society and social interaction when it first came out, or even worse how news of a brand new invention, the automobile, was met with apocalyptic rage and fervor at the same time. I believe we have to recognize that we are kind of ill-equipped to "deal" with something so radically new and different, because not only has there never been something like it before, but those who came before us, the same people that built up morals, ethics, society and law as we have and understand it today, couldn't begin to imagine something like this ever existing.

Both the OP and the people who've answered make valid points. We've seen them echoed all throughout society (like art competitions and schools and of course forums). AI art is surely a big, and very important, step forward in innovation, but it does run the risk of running rampant if not properly understood and channeled.

I don't think I'm knowledgeable enough to offer a constructive/authoritative opinion on the matter, but I will say that as of now whatever opinion we or anyone else offers up will be, by the very nature of this moment in time, incomplete. It is however very important that we continue having such discussions, that we challenge each others' viewpoints if we are to arrive to a tentative consensus in the not-so-distant future.

 

[Iron_Brew]

If AI is a threat to anyone I think it's to publishers and human resource committees. Sub-human gatekeepers running some kind of talent show where everyone is in fierce competition with each other; desperate to make themselves irreplaceable.

This is such an unbelievably terrible take and is so disrespectful to support staff working within the industry. I don't like it when people call people working hard to ensure that video games come out on time and at a high quality "Sub-Human".

I don't think at this point I can really continue to engage, so I'm going to have to block your content.

 

[123edc]

It's doing the same thing I'm doing, just faster.

but exactly at that point, we need to seperate ...
between professional work, done by big studios ... and hobbyist work ... done by indivduals

yes, it does the same thing you are doing ... becouse you already learned your skill

but for me, without that knowledge ... it would mean, to either spend several years of hard training (during which i can not improve my skills in other areas) ... buying dlc's and using standard assets (thus looking, like every other game out there and being kinda limited, to what i can do) ... or paying 60$ to an artist for each and every single picture i'd need [which simply is not ... and probably will never be] inside my pockets ...

so for me, it's not just doing it "faster" ...
it's making the possibility, to vizualize my story affordable

 

[gstv87]

(law, ethics, philosophy and morality as they've been up until now).

the problem can, however, be analyzed objectively: does the AI or does the AI *not* work by copying existing pieces?
and by copying I mean explicitly taking the chunks of pixels from one digital piece and using it to create another.
if the piece just happens to look similar but no actual pixels were lifted, well, we might just be witnessing something that can reproduce something else with such a degree of similarity that there is no way of telling whether the bulk of pixels painted in a certain pattern belong to this or that piece!

and if that is the case, then the matter becomes not the ownership of the piece, but the ownership of the pattern.
and from there, the argument goes all over the place.

 

[Trihan]

I basically developed my art skills just supplementing Rpg Maker's resource library.
I don't think AI is going to replace artists. It's doing the same thing I'm doing, just faster.
But at the end of the day, any work has to be approved by...someone.

The speed is, I feel, part of the potential problem. Right now, the rates artists charge by expectation and necessity include the time taken to create the artwork, because one's time is a valuable commodity and spending it on a particular endeavour for someone else should be compensated accordingly. You will of course still get people who want their art to be made by an actual person who is able to properly assess minute detail adjustments, but you're also going to get a fair few people who take "almost exactly what I want" at a fraction of the time and cost. AI art is always going to be cheaper than that made by human artists simply because there is absolutely no time expenditure component to consider. You enter your prompt, and you get something out almost instantly. The neural network never tires and no human needs to spend a single second on the final product. The flipside of that is that human artists cannot possibly hope to compete with AI on cost, because doing the same work for the same charges wouldn't be remotely enough to live on. The impact may well be exaggerated in some cases, but it's a possibility that can't be completely dismissed, either.

 

[Austrian]

Just throwing my two cents. Like with all new technologies, they eventually replace something. Cars, replaced the horse, machines replaced a lot of human labor, email replaced mail, etc. This will eventually replace human art. There may be a small niche for the human artist, but let's be real; this will replace the artist.

You have to remember this is just the first public (beta) version of A.I. art, and it's only going to improve from here. Soon you won't know the difference between A.I. art and human art.

 

[Desk]

Just throwing my two cents.

I love seeing the coin avatar next to this sentence

the problem can, however, be analyzed objectively: does the AI or does the AI *not* work by copying existing pieces?
and by copying I mean explicitly taking the chunks of pixels from one digital piece and using it to create another.

I understand the reasoning behind this, but let me play devil's advocate for a moment and ask: is there a difference between an AI lifting pixels and a human artist reproducing part of a work (be it an arm, a tree or what have you) to the degree that for the average person there is functionally no difference? Obviously the AI will do a better job of it in a fraction of the time, but is that enough of a difference?

I don't have the answer to this question (or any other I'm afraid), but I am reminded of many of history's greatest artists. Michelangelo started his career by defrauding the Vatican as a forger. Shakespeare lifted 80% of Romeo and Juliet from another author (can't remeber the name) who in turn had copied it almost verbatim from an italian novella. Pablo Picasso himself said that "good artists borrow, great artists steal".

So maybe the problem lies in the use of said art (or in this case said pixels). Let's say you draw a battle scene straight out of Lord of the Rings. A duel between Boromir and the orcs that end up kill him. If an AI takes your work, as is, and simply swaps out the face, then that's something to condemn. Same thing if it only changes the clothes or whatever.
But if it can make something different, if it can, let's say, reproduce the way you made the light hit Boromir's cape and use it for something completely different, like its own version of Washington crossing the Delaware, then wouldn't that be fine?

I mean, try googling Dejeuner sur l'herbe. There isn't one instance of that piece that is wholly original. Something, be it the subjects or the poses or the arrangements or the background or what have you, is always 'copied' from something else.

 

[Trihan]

I think another thing some people are missing is that the licencing issues aren't even so much because of the art that's being produced, but the images being used to train the neural networks in the first place. In a lot of cases, that wasn't a usage of the asset that was authorised by the licence that came with it, and that's a completely separate issue that is in no way affected by what kind of art is created and how.

 

[Desk]

Fair. Very fair.

This is part of what I was talking about in my previous post, when I said that the infrastructure we're using is "old" (i.e. hasn't caught up with the times).

 

[Trihan]

I believe this is the foundation of GettyImage's lawsuit against Stable Diffusion.

 

[coyotecraft]

but exactly at that point, we need to seperate ...
between professional work, done by big studios ... and hobbyist work ... done by indivduals

yes, it does the same thing you are doing ... becouse you already learned your skill

but for me, without that knowledge ... it would mean, to either spend several years of hard training (during which i can not improve my skills in other areas) ... buying dlc's and using standard assets (thus looking, like every other game out there and being kinda limited, to what i can do) ... or paying 60$ to an artist for each and every single picture i'd need [which simply is not ... and probably will never be] inside my pockets ...

so for me, it's not just doing it "faster" ...
it's making the possibility, to vizualize my story affordable

I believe the painter/studio group that did the backgrounds for MZ has different options for professional and hobbyists. On their homepage they have prices - in USD it's like $300 a background, but they also sell packs independently for maybe $7.

I don't know what Gotcha Gotcha Games paid them, but there's no way it was $300 an image. Especially the ones that are just a brick pattern mapped to a plane. "Dirt Wall" doesn't require much design work. It's like digital brush #4 on Brown.
These days people don't even need to learn technique because the software does it for you. I think AI is removing the need to learn composition as well.

I mean check this out. This is terrifying.

DALL·E: Introducing outpainting

Extend creativity and tell a bigger story with DALL·E images of any size.

openai.com

The AI generates a background in different chunks. I have no idea what the rules are that it's operating by. But it's more than just "teacups go on shelves" .

Soon you won't know the difference between A.I. art and human art.

I've been playing a lot ps5 games lately and I can't get over how crazy good the environments are. I'm willing to challenge the design logic behind some of spaces as inhuman. Like, locker-rooms have no business being as photogenic as they are. Maintenance tunnels aren't just concrete hallways with a conduit running through them. There'll be some kind of interior design-sense applied to the opposite wall that's abnormal. They'll fill an empty space with a garbage bin and it's visually appealing. But at the same time it's uncanny because real-world logic kicks in and you think what a nightmare it would be to live here. Some kind of thoughtful-thoughtlessness that a child with no experience would think of. I've seen a "parking garage" that's made for the player to run through, but it gives off this liminal-horror vibe because it's proportionally wrong to anything in real life. And I think "omg, if I had to drive through here".

 

[Htlaets]

The speed is, I feel, part of the potential problem. Right now, the rates artists charge by expectation and necessity include the time taken to create the artwork, because one's time is a valuable commodity and spending it on a particular endeavour for someone else should be compensated accordingly. You will of course still get people who want their art to be made by an actual person who is able to properly assess minute detail adjustments, but you're also going to get a fair few people who take "almost exactly what I want" at a fraction of the time and cost. AI art is always going to be cheaper than that made by human artists simply because there is absolutely no time expenditure component to consider. You enter your prompt, and you get something out almost instantly. The neural network never tires and no human needs to spend a single second on the final product. The flipside of that is that human artists cannot possibly hope to compete with AI on cost, because doing the same work for the same charges wouldn't be remotely enough to live on. The impact may well be exaggerated in some cases, but it's a possibility that can't be completely dismissed, either.

You make a good point, but I'd say AI isn't at the point where I'd say it's instant. Playing around with it (emphasis on playing, because there's a lot of things it can't do well so it's more toy/supplement/reference tool atm beyond nature backgrounds) trying to get something close to what you want most of the time is going to take a lot of generations. Even if you have the prompt right, getting something that actually is in the ballpark of what you want to see will take playing with sliders and a time investment especially if you want to do it locally. Even with a fairly strong GPU generating a 720p image is going to cost you a minute and it likely won't be right, but, alternatively, you won't even be able to generate a 720p image online for free without using a private collab, because that gpu heft and server farm costs money in watts, too.

That's not even to mention no service or collab is going to let you generate images that are more than a little above 1080p because, well, exponential increase in time/wattage, while an artist commission generally is high res. Of course, AI upscalers are actually pretty legit, too, in this mix.

Obviously all that's nowhere close to an artist's time investment, but from the commissioner's perspective there's plenty of time saved (so long as it doesn't take them hours to write the commissioning email).

But then you also get into the fact that none of these drawing AI's can draw transparency (since another channel is an exponential increase in calculations).

As a tool to help your workflow? There's inpainting from stable diffusion that can help with edits. I've seen artists (like actual artists even without the AI) use the AI to quickly color line-art or add macro details to be filled in.

Basically, as things stand now, an artist is still going to be the best choice for professional work for those that value time and quality.

Should emphasize, though, that's just where things are at the moment and it's still very early for this type of AI.

I believe this is the foundation of GettyImage's lawsuit against Stable Diffusion.

More or less. But it's a rabbit hole. Gettyimages has licensed all their content to other AI images services in past. There's also the bit that Stable Diffusion did not gather their images themselves before feeding it to the AI, they used another non-profit org, LAION's, database with permission from LAION.

Which also gets more complicated, as I don't actually think Gettyimages asked the people who upload photos to their service for consent to send that all to the other AI companies. The main image model companies use databases gathered by other companies for the most part, so where does the fault lie? Is it stable diffusion for using LAION's database? Is it LAION for having the images in its database and then giving it to Stable Diffusion to train on?

Of course, another possibility here that I would hope was double-checked is that the model used to generate the gettyimages watermark is actually made by Stable Diffusion the company and not some random person.

Also, can whether or not an AI looks at something even be truly licensed? If it can, can an image hosting service sell that license without their user's consent?

I think this is very much new legal territory all over the world.

 

[Trihan]

You make a good point, but I'd say AI isn't at the point where I'd say it's instant. Playing around with it (emphasis on playing, because there's a lot of things it can't do well so it's more toy/supplement/reference tool atm beyond nature backgrounds) trying to get something close to what you want most of the time is going to take a lot of generations. Even if you have the prompt right, getting something that actually is in the ballpark of what you want to see will take playing with sliders and a time investment especially if you want to do it locally. Even with a fairly strong GPU generating a 720p image is going to cost you a minute and it likely won't be right, but, alternatively, you won't even be able to generate a 720p image online for free without using a private collab, because that gpu heft and server farm costs money in watts, too.

That's not even to mention no service or collab is going to let you generate images that are more than a little above 1080p because, well, exponential increase in time/wattage, while an artist commission generally is high res. Of course, AI upscalers are actually pretty legit, too, in this mix.

Obviously all that's nowhere close to an artist's time investment, but from the commissioner's perspective there's plenty of time saved (so long as it doesn't take them hours to write the commissioning email).

But then you also get into the fact that none of these drawing AI's can draw transparency (since another channel is an exponential increase in calculations).

As a tool to help your workflow? There's inpainting from stable diffusion that can help with edits. I've seen artists (like actual artists even without the AI) use the AI to quickly color line-art or add macro details to be filled in.

Basically, as things stand now, an artist is still going to be the best choice for professional work for those that value time and quality.

Should emphasize, though, that's just where things are at the moment and it's still very early for this type of AI.


More or less. But it's a rabbit hole. Gettyimages has licensed all their content to other AI images services in past. There's also the bit that Stable Diffusion did not gather their images themselves before feeding it to the AI, they used another non-profit org, LAION's, database with permission from LAION.

Which also gets more complicated, as I don't actually think Gettyimages asked the people who upload photos to their service for consent to send that all to the other AI companies. The main image model companies use databases gathered by other companies for the most part, so where does the fault lie? Is it stable diffusion for using LAION's database? Is it LAION for having the images in its database and then giving it to Stable Diffusion?

Of course, another possibility here that I would hope was double-checked is that the model used to generate the gettyimages watermark is actually made by Stable Diffusion the company and not some random person.

Also, can whether or not an AI looks at something even be truly licensed? If it can, can an image hosting service sell that license without their user's consent?

I think this is very much new legal territory all over the world.

You present some very good points. Well reasoned.

 

[Iron_Brew]

https://time.com/6247678/openai-chatgpt-kenya-workers/

The ethical cesspool only gets deeper

 

[Htlaets]

You know chatgpt isn't for generating images, right?
Also, big company exploits foreign workers is, unfortunately, not really shocking news.

 

[123edc]

2$/h â 8h/day = 16$/day â 5day/weak = 80$/weak â 4weak/month = 320$/month â 11 month (let's give em 1 month holiday) = 3520$/year

now, that's just a number without context ... so, to give it context:

(article found via google from buisnessdailyafrika from dec.2022)

The average monthly income for Kenyans rose at the fastest pace in six years to Sh20,123

that are 162,7 USD ...

yes, it is a problem, how capitalistic economics work ...
but you can't realy blame a single company for a global inequality problem and they're paying twice the average income from that country (according to your image)

 

[Iron_Brew]

You know chatgpt isn't for generating images, right?
Also, big company exploits foreign workers is, unfortunately, not really shocking news.

Still generative AI. Nobody seems to care about the human cost of all this. People just want work done for them for free because they don't respect the people doing the work.

Story never changes.

 

[Htlaets]

Still generative AI. Nobody seems to care about the human cost of all this. People just want work done for them for free because they don't respect the people doing the work.

Story never changes.

I mean, do you know how Disney animated their classic movies? Art and worker exploitation has a famous history long before AI. Not that any of that is good, but it's not really a distinction.

 

[Aoi Ninami]

Shakespeare lifted 80% of Romeo and Juliet from another author (can't remeber the name) who in turn had copied it almost verbatim from an italian novella.

I know this is a sideline from the main topic, but I couldn't let nonsense like this pass unchallenged. The plot of Romeo and Juliet is taken from a narrative poem by Arthur Brooke, which in turn borrows from earlier Italian sources; that much is true. But the writing of the play, the dialogue and characterisation, in other words everything that makes it a classic work while these earlier works are forgotten, is original.

You can read Brooke's version here: https://shakespeare-navigators.com/romeo/BrookeIndex.html

 

[123edc]

People just want work done for them for free because they don't respect the people doing the work.

of course, i want to go from A to B without having to pay for a horse drawn carriage with driver (only affordable to a hand full select nobles) ... that's why i'm using a car

isn't it the same?