Kabury wrote: Sat Aug 15, 2026 11:48 pm
Yeah it's not ideal but it's the best we can have I think. I think the mod author should be the one placing the tag on their own mod (as with all other tags). If they block other people so other people can't comment it's AI and refuse to label their stuff as AI, I think it's better to leave it like that than to "report as AI" because inevitably some people will report stuff that isn't AI.
I understand your proposition isn't the worst can of worms
, though i think it still is one or worse for a few reasons. I'm assuming your post may have been moved to this discussion due to the topic/request, and i see the poll is much older, it could very well have happened that someone voted on it and then went to open this discussion in the meantime
There were couple topics like this opened on the forum.
Given your imo reasonnable assessment of the limitation of the proposition and the general presentation of your argument, i think i recognize you from discord, so i'm going to try a science based approach ^^
There is something unfortunate i see in the wording of the poll that asks : What's your reaction to a factorio mod having an AI generated thumbnail? and currently the most picked answer is : Dislike, but may still download it around 36 %, the second is Neutral / I don't care 34 %, and the last one is "Not downloading it at all" at around 29%.
The (recent) scientific litterature disagree with the assumption that human can correctly identify AI pictures by looking at them [
https://arxiv.org/abs/2512.22236 ,
https://link.springer.com/article/10.11 ... 25-00683-w], sure the context and types of pictures matters but what i take from this is that unless it's blatantly AI, people don't have a clue, it's the same for music, and for small snippets code currently, but in 5 years, it may very well be the case for a whole lot more things. AI won't get "worse" at making indistinguishible stuff from non-AI to human senses. They will develop and have already a way to make things that are "obviously AI" but they will be increasingly able to do both distinctively, i feel. The "deep-fake" direction, and the ""AI-art"" are different.
The poll has the merit to exist and materialize a concern, but a wording such as "when i think it is AI" would have been accurate.
29% is about the percentage of people who rely on paper for their password give or take, it's wrong and most of them know it, yet still do ^^ To me that would be like judging a book by its cover, or a mod by its thumbnail (only). Hopefully half of the time they are incorrect and would miss an AI image and still get to enjoy the good mod, but the other half they are comiting a mistake !
I'm not talking about the people who are "philosophically allergical to AI" as someone explained to me once, to me this is a lost battle, because there is no way to guarantee a perfect curated non-AI safe place AND with public/semi public submission without constant surveillance/monitoring ( the worst can of worms).
What i expect is a more moderate approach of " i don't want ugly mass dumped AI slop all the time", the "bad AI" , a system that discourage low quality stuff, but that doesn't mean the tag is appropriate for this purpose, consider someone who spend a lot of time coding or doing the assets, and using AI for the part they are unfamiliar with but spend so much time and effort that it doesn't look AI anymore, it's indistinguishible from non-AI, they label it as AI, because of their good ethics, only to learn 29 % people will straight ignore it ? When they would not have been able to tell otherwise ? Why would you label it ? If you spent 1000 hours making your mod, and 2 hours correcting some grammar for some tooltip with an AI or made sure it's not ambiguous , would you disclose it ? This is slippery if you are unfamiliar with code or modeling in 1000 hours you may not get very far and your mod would look like essentially a lot of AI-generated tooltip to someone else.
Fundamentally it would be the one tag that you use not to help people finding your mod but on the contrary, only a negative filter thought to make your mod less visible. In a context with monetization, the incentives to "not disclose" would be higher, when currently you can read it in the description of the mods, some write they use AI, for which parts, some write they didn't use AI, a "NO-AI" tag would be a different approach, where you would have to actively mislabel your mod instead for a "bad behavior". It's like the "home-made-meal" or the "cratman's work", that new label in a red circle you can see on youtube for people with strong opinions on the matter because seemingly the self applied label isn't enough, here you don't expect the "bad actors" to label themselves (that's irrealistic expectation imo), and you account for all the cases where people "dont know" if instead you make more visible the "good actor". It would then sound more legitimate to ask for a workflow explanations , and/or holding those using the no-AI tag more responsible for the mod using the tag, if there is a need for "monitoring" the curated content, it's not the whole mod portal, but only those mods. You don't comit a "lie by omission by default" if you don't know and allow like minded people to find themselves or mod that suit their request, which is more the purpose of the tag ^^
I don't think the AI tag is the best we can have, i can think of less problematic, but even then it's superficial and/or potentially misleading imo, i only agree that that there is no ideal solution to what appears to be to me a "taste" issue. Currently abuses are dealt with on a case per case basis, it's not ideal because it may cause devs a lot of work, but i imagine if a person behavior was out of bounds the devs would give a warning and explain before considering removing the mods or the authorization for the author to publish. If someone is particularly triggered and report everything, their report i imagine are ignored at the discretion of the devs. It appears better ^^
