Valve quietly not publishing games that contain AI generated content if the submitters can’t prove they own the rights to the assets the AI was trained on
It just seems Valve wants to avoid the legal minefield that is AI art, so the stance they take is just not allowing such things until there is legal precedent and with the advancing field I imagine something will occur within the next 5-10 years (if not in the next year or so). We can question the ethics of AI art and the commercialization of it but things do get a bit murky when we try to shove AI art/AI generative tools into a singular box. It would be like I insinuate that a selfie portrait is in any way comparable to a higher forms of photography like the “Saigon Execution”, it would be downright insulting to have a photo that embodied many people’s feelings of the Vietnam war in such a macabre photo to someone doing fucking duck lips at a black mirror for updoots or what the fuck ever people do selfies for. It seems rather unrealistic to say the process of using generative AI poisons the well (even though some argue it should) but where do we draw the line, doing touch up or drawing over it in a photo manipulation software does that make its own original work now? Like said don’t know until there is legal precedent.
Yeah, given the shit that they allow on their platforms that is barely or not at all working asset flips, the only reason they’re doing this is the legal risk.
Some of the AI generated upscaling has been fantastic, especially some of the generative images that I’ve seen for game assets (such as dynamically creating rusty metal or overgrown bushes).
It’s a bit of a minefield right now but that type of improvement definitely has a place in game dev, especially when the demand on indie devs gets higher each year.
yeah video games is something Im really excited to have ai in. Im actually hoping old games can refactor to a newer engine with a small enough team to be worth it.
Not that AI should be treated with the same rights and dignity a person, but is this not a sort of double standard? I mean, do they publish games with art made by humans who learned from works the human artists did not own?
Based on the language from Valve, it sounds more like legal protection for themselves than a judgment from an ethical perspective.
Your question isn’t a bad one, but the battleground over copyright ownership probably isn’t one they’re weighing in on here.
If a human artist learned by copying paintings, they still create original work. An AI simply copies.
Yeah, algorithmically copying one’s style with out permission isn’t the same thing as a human mirroring art. It’s not a skill.
You can create art with AI for sure but it’s nothing but a tool (at least for now). And it’s unethical to use art without permission in this context where it literally algorithmically copies the material.
When somebody uses a ML model to generate content, the skill is not their goal. The end result is.
If an AI simply copies, it should be easy as pie to tell me what artist they copied here.
If someone told me a human drew this, I would believe them. Looks original as anything else people have made.
It doesn’t copy from a single artist. It’s an amalgamation of a bunch of different artists’ work. That’s literally the entire concept of a model.
Isn’t that literally the entire concept of a person?
That is what people do. I like to write stories and my ideas are a mashup of books that I read.
Describe the criteria you use to determine whether something is “creation”.
I think I’m starting to understand… If I go to an art gallery that allows photos, take some photos, and share them with a friend who is learning to be an artist, that seems to be generally ok and does not feel unethical. But if I take those photos to an underground sweatshop and use it to train a thousand people who are mass producing art for corporate use, that seems wrong.
If I think of the AI as a human analog, then I have trouble seeing the problem with it learning from the same resources as humans, but if I see it as a factory then I see the problem.
And that’s why the companies behind these algorithms are so intent on selling the lie that it’s “revolutionary human-like artificial intelligence” and not just a plagiarism algorithm regurgitating a mashup of the work it was fed.
I mean, do they publish games with art made by humans who learned from works the human artists did not own?
You know plagiarism is a word right? Artists/Writers still strive to have a style unique to themselves…
Every day my love for valve grows stronger
Good.
The ignorance here about how AIs work is staggeringly high, almost as high as the confidence with which some users lecture based on their own beliefs.
Good. Until a studio can point to a known-dataset that isn’t just ripping art illegally from sources they don’t have the rights to use then it’s just not worth the risk.
It’s not 100% unrealistic that large studios like Blizzard and Riot (who have very clear styles that “work well” with AI generation weirdness) will eventually have huge in-house datasets that they own since it’s all created under the umbrella of their employees and contractors who already sign away all the rights when they make content for the games they’re working on. But until that happens, it’s so obviously a red flag / great area that Valve’s move is just a no-brainer.
When I learned to play Piano, I did so by playing music I did not have the rights to and that was fine. I could take my learned skills and even use it commercially. If an AI does the same, its suddenly a bad thing.
If you can’t tell the difference between learning as a human being, and selling content that you don’t own the rights to, then I don’t know what to tell you.
But you do know, and you’re just being disingenuous intentionally.
He wasn’t conflating those two. He was conflating the process of learning for humans and modern AI. You’re just being a dick about a really subjective subject.
Please be civil
A human can “learn” to play an instrument in a vacuum with no access to anything other than the tool itself.
An AI is literally only able to “learn” when fed pre-made works by someone else.Acting like there anything close to the same process is absurd.
AI is perfectly capable of mastering something by itself. Whether it’s chess, or playing an instrument.
AI just has no inherent notion of what is “good art”, because that is a human concept that has no set in stone meaning. The reason AI is trained against our tastes is so that it can produce content that appeals to us.Here is how I learned to play Piano: I watched videos people posted online and then paid money for someone to guide me.
Here is how an AI learns: It analyzes videos people post online and then has someone who has been paid money guide it.
The similarities are obvious. I don’t know about other people, but if you threw a tabula rasa me (someone with no idea of what a piano even is) in the wild with a baby grand, I would never have learned to play it never mind play it well. I am willing to bet that goes for just about everyone here.
Its a scary thought seeing us approach the singularity. Like I said 5 days ago, The AI Revolution is going to be on the scale of the Agrarian and Industrial Revolutions in terms of change. We all like to think we are special, unique, and the pinnacle of life. If a computer can not only do what we do, but faster and better, than what does that make us?
The fact is that Humans are moist computers wrapped in a fleshy case and we have managed to design something that will inevitability be superior to us in most ways. It will learn faster than us, it will think faster than us, it will create faster than us. I am seeing it before my very eyes. I remember about 10 years ago when Nvidia published their Canvas AI that would take a basic drawing and make art from that. Now I am watching it upscale my old DVD collection into 4k quality. Ten years, twenty years from now I expect it to be able to be able to create whole shows from scratch. Imagine watching Babylon 5 and saying “Computer: Change it so Sinclair stays the commander of the station” and it rewrites the show, revoices it, and reshoots it for you. We so often complain about how bankrupt Hollywood is. AI will make each of us our own Hollywood. A creative renaissance!
Haven’t procedurally generated maps been a part of gaming for a long time now?
Procedural generation is not the same thing as assets created by “AI” tools. Procedural generation still has to use proprietary assets created or owned by the devs.
I think OP made it pretty clear:
if the submitters can’t prove they own the rights to the assets the AI was trained on
I would also say that hopefully gamedevs are designing/tweaking their own procedural generation too. Though I won’t disagree that lazy procedural content can/has been used for shovelware (and in a wider sense, filler). But I would say that AI can take that to a whole new level, and one that may fool some people on the surface (like having a really high-quality asset pack that can’t easily be pointed out).
Or worse when they can use AI to pump out content with even less effort than before. For an example, the new wave of (likely all related) fake science video spam channels on YT that are a step above older tactics (like a low-quality Text-to-Speech voice reading an existing article).
(on the other side of the coin, you can still use AI as a tool that is no longer turn-key… but I suspect in instances like that the artist would/should be able to prove that with their workflow steps. Then again, that probably doesn’t cut it as Valve likely means no tainted training data can be used even if original art was added in some way)
.kkreiger, a game that generates almost everything procedurally, came out in 2002
I really hope that lawmakers and AI companies can clear this up soon, because I think AI art could be a massive thing for gaming. In particular by generating small variances so that the world doesn’t feel so copy paste.
For example, consider a map with a large office building (like in the game Control). There’s so many assets needed to avoid feeling copy paste. You’ll notice if the game reuses the contents of whiteboards, which isn’t realistic. In real offices, we can expect every single whiteboard will likely have different contents (with the exception of blank ones). They probably will have lots in common, but they wouldn’t be exactly the same. A human creating dozens of hundreds of unique whiteboards isn’t a very good use of time, especially if we’re talking about one of many minor assets that aren’t even meant to be paid close attention to. An AI, on the other hand, could generate the many variations we’d expect to see. We can even have a human design a couple and ask the AI to make similar ones.
This isn’t even all that new. We’ve had procedural generation (which is not AI) of stuff like height maps and trees for ages now. But we’re finally able to generate entire textures (and perhaps eventually entire 3D models) very easily and while fitting into a specific theme.
Finally, for indie games, developing art can be a major challenge. There’s countless programmers who want to make games and are good programmers, but they’re not good artists. AI generated art could help make being a one person dev more viable. And even when the dev is an artist, it could simply save them a lot of time on what’s a very time consuming part of game dev. eg, AI would be good at generating the profile pictures of characters that RPGs often show during dialogue.
won’t last, before long all games will have some elements of ai generated content
Until the first commercial title gets sued and then publishers won’t touch any game with AI generated content
Not likely.
Studios could and probably will train their own AI models to avoid legal trouble and achieve custom results.
Beyond AI generated textures, I think it’s just a matter of time before AI generated maps, NPCs, game mechanics, etc. become commonplace.
Though, I suspect there will be actual companies that train completely on data they have legally purchased that then sell it legally to other companies. They greyzone of not knowing if or how much art was taken without permission is the issue at hand.
@birlocke_ What about AI generated programming code from Github Copilot (or other similar tools)?
Valve can’t review that source code & would have no liability from that.
deleted by creator
Based beyond belief.
I would like to see re-releases of games with textures upscaled using AI upscalers. Nobody is going to go back and scale these up by hand, but with computer assistance, it might be viable.
With upscaled images, you can prove who owned the original images, which is fine.
Why are people so much against AI?
It takes jobs away? As every progress humankind has made in history.
It copies artists styles? As artist already do. Artists always copy other artists, it’s how art work since forever.
It copies other people’s code? As coders already do. People copy blobs of code without understanding it all the time.
It produces less quality products? It depends on the people using it as with every other tool. People can produce shitty art and incredibly good art with the same tool, ex: ma paint.
I just don’t get it. It reminds me so much to the beginning of digital art and people complaining about it, saying physically made art was the only real art.
As for myself I can’t wait for AI to get even better. I have so many ideas about what me, or others could made with it. It’s a tool with so much potential to throw it away out of fear.
It copies artists styles? As artist already do. Artists always copy other artists, it’s how art work since forever.
What AI does is take artists’ hard work and directly uses it to generate something using their style(s), without their permission.
Artist do that too. Haven’t you notice that most popular songs sound very familiar to each other? And I assure you they do not ask for each other permission either. AI may do things faster, automatically and lower the skill requirements for the user but doesn’t do anything artists weren’t already doing.
The AI Revolution is going to be on the scale of the Agrarian and Industrial Revolutions in terms of change. People, on a whole, do not like change since it means uncertainty. Uncertainty about their careers as you mentioned, but also uncertainty about our very nature. We all like to think we are special, unique, and the pinnacle of life. If a computer can not only do what we do, but faster and better, than what does that make us?
That is why people are so adamant in trying to say AI cannot create art. Art is the one thing humans have over all other known forms of life and if we lose that… like I said: what does that make us?
That sounds like a positive thing as a way to verify that the content was designed by humans, but concerning that AI has any input at all, unless it’s for finding issues with the gameplay mechanics and nothing to do with game designing.
Possible, with AI the single player campaigns might closer to playing with real people but AI can never duplicate human behaviour and instinct, only imitate it.
I’m not sure this is to ensure the content is made by humans, that isn’t the goal. Valve just wants to ensure that the game dev owns the rights of the content created for the game. Using AI, you can still own the rights in some scenario’s as long as the AI doesn’t use inputs that it doesn’t have the rights to.
This is a very good development, it ensures that creators and owners of content are safeguarded, while at the same time ensuring that gamers get fresh and new content.
I have to agree here. Generative AI has so much potential for games. Especially RPG style games for believable NPC characters. But the rights environment is very murky.
I expect it to be resolved relatively soon though. a combination of generally trained AI with subject specific training should do the trick. In the same way we would train a helpdesk bot on company specific information.
The remaining question though is what of the original broad dataset the source model was trained on. There things are less clear.
I think that it is quite feasible to do though. Take for example Lord of the Rings. If the game dev has the rights to Lord of the Rings and its books, then it can be completely fine to write prompts for NPC text as: Produce a response to question X as if the NPC is living in Mordor, with his background as a blacksmith etc… AI can then generate that text under the IP of Lord of the Rings just fine. And it will always be the right tone of voice. Same can be done using dynamic events etc.
KoboldAI and similar do that today, but it’ll soak up all the capacity on a computer and then some just for the text generation. Needs to be more efficient than it is today if one’s going to be generating text on the player’s computer.
If you mean the studio using it to generate static text, then sure.
@tal Or it could make an API call to a server, the way ChatGPT does today. Unfortunately, that will mean the player has to be online to use the text generation, but the tech of it isn’t what we’re discussing anyway. We’re talking about the ethics of it, not the means.
It’s like we’re talking about whether robbing a bank is OK or not, and then someone goes and talks about how hard it is to rob a bank. It’s a non sequiter, it’s not what we’re talking about.
And that is where things gets interesting. The ethics of the situation. Even beyond copyright issues. Was your AI trained on data that you have the rights for, or not?
We then have to think of the base model. How was that trained? I have not formed a well reasoned opinion yet as to the ethics of training on social media and forum style data.
For me, personally, I don’t have an issue with my own posts and responses ending up as AI training data. We can also argue that those posts were made on public forums, therefor in public. But does that argument hold true for everyone. Underlying that question, we have to consider the profit motif off the companies. There is a major difference between training for academic purposes and for corporate purposes.
Valve is probably smart in steering clear of the entire mud bog at this time. Not enough is known of how it will play out in both the courts and in public opinion.
Yeah that’s what I mean, if the game devs can show that the AI language model is fully trained on its own IP, then it should be fine.
Using AI, you can still own the rights in some scenario’s as long as the AI doesn’t use inputs that it doesn’t have the rights to.
This is still very much a gray area. I can understand stating that you cannot copyright AI generated content right now, but who even should own the rights for such content?
Generative AI’s are using such a broad variety of training data form various different sources that it’s impossible to prove that one piece of generated content is actually using some specific piece of copyrighted content as an inspiration. Images generated with prompts of super mario would be affected by source content completely unrelated to super mario.
Additionally at what point is the AI intelligent enough, that it can create original content? If a human was to look at a thousand paintings, and then created one inspired by those paintings, the human would own the copyright for their own work. How is training a neural network different from looking at existing work? It’s simply learning ques from existing material and using that as a baseline for creating new content with the user’s guidance.
