How does the fansites protect their contests from AI generated submissions?

Question

I was wondering if there is any policies in place or has fansites established any specific way of determinating if the submitted work (for example: poem, article, or even drawing) is generated by AI not by the player itself.

Comments

I wonder this too and if there's a way to answer it. Honestly, I've been seeing some answers that made me suspicious already but is there any proof you know

---

To be honest this question was more as a "poke" that some things should be consider for future contests. The AI chat is not real ideal to provide information about game as is providing misleading information  ( just ask "how to gate access to boss oberon in game tibia?" and see what it says- it told me following: Access to Boss Oberon in Tibia requires completing a series of quests known as the "Forgotten Knowledge Quest." This quest can be started by speaking with NPC Zushuka in the Zao Palace, located in the Zao Steppe.


the saddest thing which I saw lately was people using Chat GPT to resolve 469 language. At first exciting- yes; but ultimately...  years of mystery and trying to resolve it by many people, simplified and degraded.

I hope AI will continue trolling about game knowledge as it does now.  ;-)

---

These 469 theories generated by chatGPT are terribly annoying. On our TibiaSecrets discord we have been spammed with just such garbage posts :/ Every few days someone comes to our channel and writes that he has solved 469 :)

Answer 1

Well, that is an answer that may vary from fansite to fansite. TibiaLight, the fansite I manage, has not yet created any contests in which AI tools could be used, such as texts, poems or songs, but certainly the caution about submissions made without proper work (and only using ChatGPT and other similar AIs) should be a new priority.

To this end, one of the ways we can use is another tool that allows us to check whether the generated text is obtained through AI or a legitimate human creation:

https://platform.openai.com/ai-text-classifier

Obviously this tool is not perfect, but it allows us to get an idea, I think it is too early to eliminate submissions just by checking one tool, but maybe several. The reader can see that the work for admins and fansite staff has only increased, checking each submission in one tool can be time consuming and hard work, even more so considering that we do all this in the few spare hours we have.

Answer 2

Unfortunately, there are no official regulations that address these issues.

I must admit that I myself am wondering how to deal with this problem in the future and have no idea. In the context of AI-generated images, it could be considered that the solution could be to show the stages of the work (screenshots from a graphics program showing sketching, line art, coloring, etc.), to bet on work done manually, etc. But, in my opinion, this does not solve the problem either, as it is trivial to simply redraw the graphics by taking the generated image as a base (model). I am also assessing this problem on my own abilities as a contestant. Hypothetical situation: I generate a graphic, create a sketch based on it from scratch, and present it in the steps of creating a drawing. Is it possible to prove such cheating? No.
The only thing I can think of is to choose artwork that depicts drawings of Tibian characters/creatures so that they look like they are from Tibia, rather than referencing common images from the Internet.

A good example of this was the recent TibiaDraptor.com competition, where there were about a hundred entries posted, some of which were generated and disqualified: they all had a common denominator - the draptor more closely resembled a raptor from Jurassic Park, rather than a creature known from Tibia.

Comments

In fact, it will be a big problem for such contests and a challenge for admins. I myself play with tools such as MidJourney, StableDiffusion and Leonardo.ai. I am terrified by the capabilities of these tools....