Forums / Discussion / Meme Research

"Ooooh boy, that topic again"
Yes, yes, but I think that for the sake of this database, we need to consider the impact of AI-generated pictures and videos online. I haven't seen any official announcement, nor any prior thred on this theme, but if there is, please feel free to close this one.
See, I have kept in touch with several other pictures databases, some of… more or less savory repute HEM and a common occurence is that without clear guidelines against AI-generated pictures, they tend to be flooded in concerning amounts. It's my opinion that we should survey our options so that we avoid such spam.
-
What makes me particularly wary of AI-generated content, in the case of KYM, is that the site's role is of archiving, preserving pieces of online media for the future. But such a role seems trivial when it comes to AI content, especially pictures. They were created with literally minimal effort, just a prompt fed to a software, and therefore don't have the same historic value than art from actual artists; and the fact that they are so easily created makes their preservation virtually useless. One such picture disappeared from the Internet, so what? Type the prompt again, and voilà. On top of that, one prompt can potentially lead to hundreds of variants of the same picture. On all those grounds, I think that AI-generated pictures should have a minimal, if not null, priority of archiving, unless they have become so viral that they became memes in their own right.
-
Here are a few possibilities, on top of my head:
- Making sure that each and every AI-generated picture is uploaded in one or several of relevant entries, so that they can then be filtered by e's script. This would be the simplest in theory to implement, but it would be putting the burden of filtering on the user's shoulders, and necessitate to make sure that no AI-generated picture gets uploaded in another gallery, which sounds as tedious as it is. Not to mention, it would't stop any spamming, if spam were to happen.
- More unlikely to happen, we could write in our guidelines that no AI-generated content must be uploaded from now on, unless it is absolutely necessary to the documentation of a specific meme. Basically, the same rule we're using for NSFW pictures showing F̴̠͍̲̮͙̀̽̾̄̈́̚e̷̛̮͖͎͔̟̎̄̌̿͠m̷̢̛̠̯͓̘̔̿̂̇â̵̱̝͛̏̐̕͠ĺ̵͓̿e̴̟̺̍̇͑͒͘-̵̰̳͔̓͜͝͝p̵̤̝̈́̊̄̀̄̚ȓ̵͍͇̙͊̒̓̐͗e̵͖̪̖͚̓͑̉̍̓s̵̡̒̿͠ȩ̶̬͙̘͉́̈́̄n̵͙̿͂̑ẗ̴̡̡i̵͔̠̱͇͖̔̅̎͜ṅ̸̝͙͕̞̠͋̈́̈́͠g̷̟͇̭̩͚̈́̃͛̾̊͝ͅ ̵̫̥͉̌͑͜͜n̵͍̙̋̐ỉ̵̛̳͓̯͕̂̊p̴̧̺̻̓̀̿̅̚ͅp̴̦̈l̸̛̫̲͓̭̼̣̾͐͐͂̉e̴̟̊́̓́͜͠s̶͖͗̈́ and whatnot. It would need to be an explicit and well-known rule, and it would probably require some more work from the mods in case of spam.
- In a worst-case scenario, we have the option to lock certain AI-related galleries from any new uploads. This would however bring two obvious and crucial issues: AI pictures could just be uploaded in another relevant gallery, if in any gallery at all; and pictures such as text screenshots and actual drawings which are relevant to those galleries' themes would get blocked too.
-
Now then, what's your opinion? And do you have any suggestions on how to manage them?

I vote none unless it's absolutely necessary
or at least put up a barrier like the spoiler/nsfw

I agree with this.

The barrier should say "AI" and have a picture of a hand with between 8 to 11 fingers

See this?

This should be an actual thing.

Inspired by martinprince12345 and blank profile pic, I present a "non-copyrighted" AI spoiler:

As for my opinion on the matter, I agree with you Triplem on their priority in archiving, and I think that the best solution is to only allow AI uploads if they started or indicated a pivot in the development of a meme (option 2). However, I don't know the feasability of moderating something like this because I've never been a moderator.

From a logical POV, moderating AI-spam will always be more difficult since they're harder to distinguish from genuinely high-effort pieces. We'd also run the risk of banning users who may rapidly upload real pieces, or allow users spamming AI to go free because of that confusion. Really, there is no perfect solution considering that we're recording the spread of memes, not archiving artists' art, and AI generated images can fall just about anywhere from irrelevant to critically important.
While coming up with a strategy is important, until we're able to figure out a practical order of operations, I say we just take advantage of the current user-base's good will, since I haven't really seen any AI spam on this site outside of the designated AI gallery.

One thing's for certain in my mind…
We should turn off that Reddit Moments bot because whether or not it posts AI stuff, it still clogs up the "New Images" tab.

Just have an intentionally poorly drawn “AI” image of KYM-tan as a spoiler similar to the NSFW/Spoiler tag we have.

AI generated media should only be allowed for media in the body of a meme's entry. Anything more than that is subject to potential user abuse/spam. Just thinking of what rules-skirting wojak/pepespammers would do with that deeply concerns me.

I think allowing broader ai generated content submissions even under a banner adds unnecessary validation requirements by mods/staff. Limiting it to entries seems more professional.

I'd honestly prefer that over using an real AI image like the one in @Checkpoint_Flag's post (no disrespect to them, just so we're clear).

You aren't in favor of using AI to mock AI?

Implement an AI that can detect and hide AI art.

Meh. It doesn't really hit home if you ask me.
The reason I even proposed this idea is because I'm at a point where I'd just rather not have to look at any AI stuff period, regardless of the intent behind it. I inherently find it all to be a visually unappealing eyesore and even occasionally wish it simply didn't exist.

Than have an AI which keeps an eye on the AI detecting AI.

The problem is that AI-Genrated media has been in KYM for several years, dating back and before Deep Dream and many of these are their own memes as well.

Another thing to consider that advice animals and other template memes, while not AI-geneerated, take very little effort as well. And those too can be reproduced easily.

I have no idea sadly. Wish I could say it.

This is definitely something to consider. We may have some problems tagging images.