Skip to main content

Indie game storeFree gamesFun gamesHorror games
Game developmentAssetsComics
SalesBundles
Jobs
TagsGame Engines

I'm dismayed that itch.io decided to hide which games/assets are AI and which aren't

A topic by Kasper Hviid created 39 days ago Views: 444 Replies: 16
Viewing posts 1 to 7
(+2)

Itch.io requires us to tell which of our projects are AI, under the project pages AI generation disclosure section. Failure to comply can result in delistening. As told by the guidelines , this is "due to legal ambiguity around rights associated with Generative AI content". They're preemptively covering their hineys, basically.

Since they have gathered this information, it saddens me that they didn't take the next step of actually letting us see what is AI tagged and what isn't. As a guy who sometimes pay for content, I'd very much like to know if it is created by a human or by an algorithm.

(+2)

I'd like to give them the benefit of the doubt and say this is a feature they're working on implementing. It would be helpful to know what is AI and what is not

(1 edit)

Being able to see which games and assets packs are using AI would indeed be nice. So, I personally am heartened by the fact that the mods/admins have added that check field and hope it'll be easy to filter out all AI-tagged results! 😁

I am only interested in supporting human creators and among them only those who do not employ any automated means of plagiarizing the work of other human beings whatsoever. I have a zero tolerance opinion of it. I don't want even one single grain of sand of the games and assets I support and buy to have ever been touched by any generative AI. It is already hard enough to get a viable market even without the prospect of being drowned out by mass-produced/automatically-stolen AI slop. It seems to me to be essentially an elaborate copyright laundering scheme disguised as an "AI" tech advancement, mostly. The rhetoric of it being "higher level intelligence" used by AI advocates has long seemed questionable to me and I suspect it to be a distraction tactic to focus attention away from the real biggest issue: systematic theft that in effect unwillingly enslaves the labor of the rest of humanity to the few companies and people in control of these AI systems.

The main difference between AI-generated images that closely resemble well-known IPs (such as Mario, for example) and images that don't appears to most likely be that the images that aren't clearly of extremely well-known IPs seem to be likely based largely on artists who are simply too little-known for many people to recognize the very close (often near verbatim) nature of the imagery. It could easily be the case that almost all of the AI-generated images of the big LLM models are plainly plagiarized in this sense.

For example, I myself remember being shocked seeing an "original" AI generated image that looked almost 100% identical to a painting someone I sat in the same classroom with in college made long ago! The models scrap lots of old art like that. People seriously underestimate how huge a proportion of AI generated results are so close in resemblance to copyrighted material that they could very easily be sued for it in the event that the right person becomes aware of it. 

So yes, the mods/admins of this site are indeed wise to cover the legally dangerous prospect of hosting any of this kind of material.

It is very legally unsafe to use any web data scrapping based AI-generated content as anything other than a passing curiosity that you make sure never to redistribute in any form, in whole or in part. The fact that the big tech investors want people to ignore that doesn't diminish the inherently legally dubious and more importantly ethically bankrupt nature of it.

Anyway though, that's my personal thoughts on the matter, for what it's worth.

I wish you all a wonderful night and the best in your creative endeavors!

I don't want even one single grain of sand of the games and assets I support and buy to have ever been touched by any generative AI

If this is about supporting human artists and their craft, very nice of you. Thumbs up.

Unfortunately, there is a high chance for every game engine and library, that it contains code that was made with help of a gen ai. 

There is also a high chance for images created by professional digital artists to contain some gen ai pixels. There are photoshop filters that use gen ai for filler and probably filter effects and there are even outright prompt modifications possible.

Meanwhile, actual professional coders and artists use the state of the art tools of their trade. If there are ethical problems by the training set used, those can be overcome. Artists could even train a model exclusivly on their own data. What cannot be overcome is the seemingly unfair advantage of using tech to do what might be done by hand.

One of the best examples in history is usage of photography to do a portrait in seconds, instead of having a painter paint it for days. But from an art perspective, both are different and customers usually notice the difference. And my impression is, that gen ai games are not really popular, and I suspect this is not because of any ethics aspects. As with photography, it is not enough to push a button, you need to be good at using the tech to create good results. In other words: most gen ai games look not really good.

(+1)

All of the major engines in current use (Unity, Unreal Engine, Godot, Defold, etc) were created mostly before the advent of theft-based generative AI and had the vast majority of their features already complete before it. Most major libraries and frameworks are the same (feature complete long before the advent of theft-based AI). It is also extremely doubtful that theft-based AI use is as ubiquitous as you are suggestion. I don't see much evidence of that, though I do see lots of evidence of AI companies constantly trying to force AI down my throat against my will by adding it to countless things it has no business ever being added to.

Regardless, it does not really matter whether they were or weren't, ultimately, morally. My point is still the same, as is my stance: I don't want any AI to have touched any part of any creative work that I support. Circumstances could force me use something that has been tainted in some way by theft-based AI, conceivably, or I could be unaware of it, but the principle remains the same and always will: I avoid supporting any of it as much as practical. It is automated theft, regardless of anyone ever trying to rationalize the contrary. Training solely on one's one work (or one's team's work) could be the exception I suppose (though I would never do it nor allow it on any team under my own control if that day ever comes), but even then you have to keep in mind that the baseline for the AI models may already contain partial pre-training on IP infringing work even prior to the point where someone trains it on their own work, and so even then it may in essence still contain plagiarism.

I have worked in the AAA game dev industry before and I have seen what artists have sacrificed to get there and it is disgusting to see so many people making up excuses for systematically stealing the work of these people. Concept artists were hit especially hard.

The AI situation reminds me somewhat of the Trail of Tears and other historical events where settlers repeatedly made up excuses to just keep stealing the Native Americans' land based on nothing but their own desire for more wealth and land. That is very similar to the same kind of energy and rationalization I sense every time I see someone going into contortions (like you are here, frankly) to rationalize the systematic theft of other people's work or to conveniently buy into the misleading rhetoric and propaganda the AI companies are delibrately spreading to make people think that it is all just merely technological progress in action or that it isn't hurting people or that the AI is just taking "inspiration", etc. That is all just AI propaganda designed to open up an avenue for the AI tech companies to essentially steal 95%+ of the rest of humanity's digital work without any need to compensate them, causing wealth inequality to potentially become worse than ever before in history (de facto digital slavery essentially... it is the most dystopian thing I have ever witnessed).

Above all, I do not believe in defeatism with regards to ethics.

I can only control what is in my power to control and in that regard I will always do what is in my power to minimize the extent to which anything touched by theft-based AI has any presence in my life. Indeed, with the pervasive and unethical and non-consensual pushing of AI by these numerous immoral companies it may indeed become impossible to avoid partial contact with it, especially with having one's own data unwilling stolen by the companies to be "trained" (in reality to be integrated into an automated randomized plagiarism engine that is called "training" just as part of a way of covering up the systematic theft and reducing the chances of being held accountable for it).

Theft-based AI is the most unethical thing I have ever witnessed the tech industry do in my life. It is almost mindbogglingly evil and greedy and will likely be highly corrosive to the integrity and meaningful survival of the internet if current trends continue.

It is essentially the largest act of intellectual property theft in the entire history of humankind and the people responsible for it should receive punishment proportional to the magnitude of that unfathomably large crime against real artists (and all other creators) everywhere.

Your rhetoric about theft-based AI merely being "progress" akin to photography is complete disingenuous sophistry. Photography didn't involve stealing from painters. There is nothing about this theft-based AI that is inevitable. It is simply being forced upon everyone else by greedy and unethical people for who no amount of wealth or power ever seems to be enough for. It is absurd.

Also, theft that "looks bad" also does literally nothing to justify or excuse the act of stealing. It is still theft, just as much.

Your whole argument amounts to nothing but excuse-making and rationalization and red herrings and strawman fallacies, just like every other "argument" in favor of theft-based AI I've ever seen over these past few years.

When it comes down to it, you are just making up excuses to rationalize your own desire to cut corners by systematically robbing the creative work and livelihoods of others from them. There can never be a genuine justification for such unethical treatment of other human beings. Never.

Wanting something can cause people to make up contortions and rationalizations to "justify" it but that does not ever justify it in reality.

As I like to say recently: Good is a journey where every step is treated as an end of itself, whereas evil is a journey where only the outcome is treated as mattering. The rationalizations of theft-based AI developers and users all fit the later. 

Joy can be found in any amount of anything, if one has a wholesome and benevolent enough perspective. 

Thus there can never be real justification for harming others for the sake of needless "gains".

One square mile of forest holds more true wealth than any wealthy person's wallet ever can.

Theft-based AI users are thus accomplishing nothing but inflicting suffering upon others, just like any other rationalized evil throughout human history, whether it be the Trail of Tears or (now) theft-based AI.

Goodnight...

I do not want to hear a reply of any kind if it contains even one whiff of rationalization of theft-based AI.

I may block any such respondent next time, potentially, but not this time. This kind of stuff ruins people's lives

One cannot be certain of what is or isn't AI, but that never justifies defeatism. 

There is never a good reason to hold the door open for evil.

Please do not misrepresent what I wrote. Because you did that. A lot. I did not argue for gen ai. I also did not argue against it - which might have mislead you to the belief that everyhing I wrote must somehow be an argument for gen ai. 

It is also extremely doubtful that theft-based AI use is as ubiquitous as you are suggestion.

New software code by google is 25% ai generated to put out a known number of the industry. And for your tained touch thinking it does not take that much. A single function would be enough.

Your rhetoric about theft-based AI merely being "progress" akin to photography is complete disingenuous sophistry

You accuse me of a fallacy by constructing a strawman. 

Usage of a term like "theft ai" begs questions about "non-theft-ai". So I described a scenario where gen ai would be trained ethical - the non-theft-ai -  and the consequences. Namely that the advantage of that tech would still exist.

Your whole argument amounts to nothing but excuse-making and rationalization and red herrings and strawman fallacies

My what?! What argument are you even talking about? I like semantics and to talk about fallacies. So kindly point out what you think the fallacies are. I am especially interested in the strawman fallacy you claim I used.

I was giving you my opinion about the matter. My point of view, my observations. I made a lot of claims, not arguments. And the whole post itself was also not an argument. I mostly wanted to point out to you how impractical your tainted touch approach is and how shortsighted your theft rhetoric is.

I did not even give my stance about gen ai! If you wanna know, I see it very critical, but also pragmatical. Without a Butlerian Jihad ai will stay. And I care more about the usage of ai then how it was made. Struggling artists will still be out of work when ethical ais do their job. We need future proof solutions. Not shortsighted method based solutions, like banning "theft" ai.

Moderator moved this topic to Ideas & Feedback
Moderator(+5)

Our admins explained from the beginning that first creators will get a grace period to label their projects correctly. Visible labeling is coming.

Good to know! Also, thanks for moving my post to the right category.

Do you happen to know when, roughtly, they plan to turn on the label visibility?

Moderator

No idea, sorry.

That's okay. But you have heard them say that visible labels are coming? I just want to be super clear on this, so I can let the matter rest.

Moderator(+1)

Yes, like I said, this has been discussed since the new feature was first announced.

Thanks, just wanted to be super-sure!

For assets

https://itch.io/game-assets/tag-no-ai

For games, ai tagging is not required, so the various ai tags will be not be accurate - like all tags. Especially for older games.

But I think disclosure of ai was in the guidelines before all that new gen ai tagging. "If your project involves automatic or AI generation, make sure it’s clearly stated in your project description". But there is much confusion about terminology. Maybe it meant live gen ai and not ai generated assets.

I would wish the formal disclosure to be a bit more elaborate then a yes/no filter. Or visibility of a yes/no tag.

(+1)

If I search for the "No AI" tag, click on any result, then click on "more information", I can see "No AI" among the list of tags.  So you can see if something is marked as "No AI".

However, the converse for the "AI Generated" tag is not true.  If I search for the "AI Generated" tag, click on any result, then click on "more information", I can't see "AI Generated" in the list of tags.

I don't know why it works this way, but if it's AI-safe, then it should have the "No AI" tag which is visible, and if it doesn't have a visible "No AI" tag, then it's not AI-safe.

Interestingly, I can get some search results by searching for both "No AI" and "AI Generated" at the same time.

(+1)

"No AI" is just a tag someone made. Just like the Horny-Eyeballs tag and the thousands of other tags. This has nothing to do with the AI generation disclosure section of the project page, which is still unavailable in search.

(+1)

Horny Eyeballs. Hah. ;-)

But seriously, the way the ai disclosure currently works for users to filter with, is with tags. Read the announcement. https://itch.io/t/4309690/generative-ai-disclosure-tagging

This here is an automatic tag: https://itch.io/games/tag-ai-generated-graphics It does not appear on the tag list on such games. Unless someone chose it manually.

So yes, no-ai was used before and very likely currently by developers. But it is also given automatically by the disclosure. And the way I understand it, it should only be visible if manually chosen.

(+2)

The problem is, that the new meta tag is not a true meta tag. The ai tags existed before as regular tags.

What I call meta tag is everything not added to the tag list by selecting tags and genres. You can have 1 main genre and 10 freely chosen tags.

The ai disclosure should not appear in tags at all in that info box. There should be a separate entry. Like "made with". And that should appear only when there is information available. Like "made with". You do not inform players that a game was not made with Unity. For assets that might be different, but it was said that assets without disclosure would be delisted after a grace period.

We shall see how this works out and gets changed over time to be a usefull feature. You currently cannot filter for no-ai-assets, only positive for graphics, sound, story, code. Some people would like a general no-ai, but others's might want a no-ai-assets filter, since there is mandatory ai disclosure for assets. Story and code are not assets. And code is not content. And some people might want to avoid ai narrative, but would not care for the other things either way.