掲示板 Forums - 2026 Kao Coin Contest - Rule preparation

It does seem hopeless. Maybe it would help to have separate categories for physical work and computer-created work. I don’t think it’s possible to reliably detect faked evidence of manual effort anymore, but an honor system with no penalty for telling the truth would make it easier to comply than cheat.

I think something that would relatively easy for genuine artists to provide is a construction drawing. Most people (me included) draw some sort of basic structure that acts as a guide for their finished rendering; I'm sure it's not impossible, but it certainly seems like AI would have a harder time creating a construction drawing that appears logical to an experienced artist. It would probably be the easiest stage to detect fake submissions.

I took a quick look at last year's entries and I think there are some in there that became kao coins that lack enough proof, and even some that could be AI generated because the proof is bad. I think more checkpoints is definitely necessary. A single checkpoint is easier to fake. The logical flow between multiple checkpoints would be harder to fake.

To take some of the burden off of you, perhaps you could add a little bit of voting? Not sure how it would look, but it could also increase interest. One of the voting options would be "I think this could be AI" or something like that.

Something I heavily leaned towards last year was showing the planning phase and many failed sketches/ ditched ideas that contributed to my final piece before any real drawings existed alongside a 60-70% done drawing.

Using the ideas above, such as having more checkpoints and showing some planning (e.g., sketches, refrence images etc.) would make cheating harder.

However, I can understand some people are not the planning type making this meaning it may be hard for creative people to prove their entries are not AI.

Due to this, if showing planning is not possible for the rules, the next likely thing to implement would be writing a short brief statement accompanied by checkpoints similar to how they force students studying art in schools and universities to write words in the form of an artist's but simplified to at least 2 sentences instead of 300 words.

If these were rules, the first one with planning would be on the lines of "Some form of planning must be submitted with [a certain number most liekly 3] progress checks" while the one leaning towards writing an artist's statement would be written similar to "You must justify your process of making the art such as stating the materials used and mediums involved"

Yes, I know these rules sound boring and may render fewer entries due to a tedious process, but this can force people to justify how they drew their works, what medium was used so no one can generate a water color painting and say it is acrylic and increase the number of progress checks to ensure no AI is used.

I would also think that a select panel of some previous renshuu kao coin artists could help at looking at descriptions and help determine if the artwork is human made based on statements and if it matches to ensure the "security" can increase to prevent AI slipping in. I wouldn't call this voting but getting a few select people that know art to help verify

(I am thinking people who have had many verified entries in the past or have been in older contests before AI happened in the renshuu competitions)

[Absolute not even 1000 people can tell the difference with a real and AI entry] Worst case scenario, force all future entries to be hand drawn or have a hand drawn copy converted/ remastered with evidence it was done by hand. (But this is only a drastic measure)

(the main points bolded)

Edit: ban all renshuu social privileges to cheaters such as use of forums, haiku, word gardens, public shiritoris, crowdsourcing projects or strip their garden kao coins to zero/ factory default coins

My faveroite rule from last year that looks like something that definitely works is obiously stripping away privileges (and feeling like Satan/ a school teacher or even a 100% anti AI person is lurking in October ready to punish those who swear their lives by AI generated content)

Maybe we could send a letter with our work in it???

For digital work, providing images of the multiple layers should work relatively well. I don't think AI diffusion can fakes those yet.

For physical art, having an early and a late version should be enough in my opinion.

maybe for digital art, show each layer in groups so that ai wouldn't be able to recreate like "make the first three layers of this" ,usally the hair and body and stuff would be on different layers,and so the ai would put things that isn't supposed to be there

I think it is hard to foolproof completely for AI without increasing the workload of judging tremendously. I think having process images and hoping that people have a sense of honour and pride in themselves. It should be for fun after all.

Also maybe you can try to appoint some people to filter entries for you (if you are not doing so yet)

I respectfully disagree with the notion that "increasing the workload of judging tremendously" is necessary. The point of requiring additional layers of proof is to filter out entries before they reach the judging phase. Effectively shifting a lot of the workload away from Michael and onto the participants themselves.

I'd argue that, if done correctly, these additional layers wouldn't add more than a few seconds to the initial screening of each submission. They aren't meant to be highly scrutinized, their mere presence should be enough in 95% of cases.

You'd likely filter out the majority of AI submissions simply by making the process of generating all the required proof unappealing. The main draw of AI art is how easy it is to produce. By removing that convenience, you're effectively weeding out the casuals.

I personally think there's a clear choice to be made: do you make Michael's life significantly easier, at the cost of introducing more friction for participants? Or do you keep the process more inclusive and frictionless, at the cost of making Michael's job harder?

You don't need to catch everyone. Making AI submissions sufficiently inconvenient that a substantial number of people won't bother submitting them at all is good enough. The rest you can handle in the judging phase.

You can also, as mentioned by other users, potentially outsource the initial screening of submissions to trusted members of the community, but I suspect Micheal isn't too keen on that idea.

PS: If possible, I'd like to avoid making a traditional version of digital artworks a requirement. A lot of digital art is very difficult to adequately recreate in a traditional medium. There are also certain styles where the digital medium is fundamentally part of the process.

Of course, if the idea of the contest is specifically to showcase freehand art, that's totally fine. It's just that I haven't seen any indication that that's the intention. Even if that were the case, recreating a complex digital piece traditionally could take a huge amount of time and effort, even if the artist is perfectly capable of doing it.

Edit: You could even check exactly how difficult it would be to fake certain layers of evidence yourself.

I agree, timelapses are something that AI would be hard-pressed to replicate. I don't know if all digital programs have that feature, but for those that do, it'd be relatively convenient proof to provide.