There is the new "AI generation disclosure" option, but what is exactly covered by this? For some cases this is clear to me (AI generated images & text), but what is with cases like:
Is there an extensive list what needs to be tagged as "generative AI used"?
To quote the opinion of leafo about this in general
Are you making a game? It’s just a tag, use your best judgment. If the output of Gen AI is something you put into your project, then tag it. If you are bundling a dependency that you chose because of its use of Gen AI, then I think it’s fair to tag it. Otherwise, I wouldn’t really worry about it.
It is about content, and important for assets. You use them in other projects, so information should be acurate and given.
My opinion on the mattter, if you used a prompt to create it, it probably is generative ai.
machine translations (nowadays they are based on LLMs)
Why should you care how your free translation service black box does it? The actual content is your original text. Applying a computer generated filter to your images in your phone is probably also ai assisted these days.
text to speech (based on AI)
You should not promote tts in your game as a voice over anyway. Computer generated voices were called such long before the new gen ai thing. Also, this usually happens on the user's machine. You probably want to look at the licensing of your computer generated voice, what it actually is and how you can use it and how you should or must attribute it.
AI opponents based on neural networks
That's ... not content. That's how opponents were always called: ai. I think using actual gen ai like methods in your game as a way to create game play interaction is too meta to be covered by the current tagging situation. I know games like that, that even connect to one of the llm services. They usually do use some ai tags, as advertising, because the promise is that the opponent is smarter and not railroaded in conversation. But I just do not think the recent ai tags would fit.
spell/grammar checking by LLMs for in game texts
Same as translations. Even less so. Why would you even know how your word processor does it. Or care.
use of copilot coding assistents based on LLMs (Github copilot etc.)
Subject to debate. Personally, I would exclude code ai tagging from that tagging business as a whole. Normal users do not understand the difference and probably would not want to filter out such games, if they knew how software development works. In a way, if you can formulate a prompt so exactly that the outcome is a working code segment, you basically did what all programmes do when using a library function call and higher programming languages to generate assembler code under the hood. But a level or two higher up. When pseudo code in natural language becomes the code.
Of course, if you just prompted the ai to give you a match three game, that's a bit much. But those assisted things are often possible with procedural assitants as well and simple function generating is not really different from using a template, imho.
It gets complicated, if you want to use a function, or even the operating system, and you do not know, if those libraries or the operating system did have ai generated code in them or not. But as said above, for games it is just tags, do not worry too much.
My personal interpretation:
I am curious. Why would machine translation count as ai generated content? For one, you do not even know if it is "ai" or not. Machine translations are older than the current things that are misleadingly/confusingly called "ai".
There is no artificial creativity for lack of a better expression. A translation is usually 1:1, if you use a translation site. You always get the same result for the same input. If you use an "ai" to generate an image or text from a prompt, this is not so. And even if a llm is used under the hood to translate something, there still is little to no content created. The content is translated, not created.
I am especially curious about your reasoning, since you would count grammer correction not as ai gen. As you wrote, the content is still created by the human. This is also true for a translation. One can define grammar correction as translation from "broken English" to "formal correct English".
Regarding tts, why would it matter, if the text was ai gen or not? If the text is ai gen, it stays ai gen, no matter how you deliver the information. Be it by written word or by spoken word - even if a human would read it out loud.
Also why would the intent matter? If I generate an image but do not "intend" to replace human generated images, it also does not change the nature of the content creation process. tts is tricky, because the content would be a speech pattern. The llm ai gen stuff works different for that, if at all. And there are professional solutions where they just pay a voice actor to record samples to generate a computer voice. This is also possible without modern llm tech, but I assume it is an easier task with state of the art tech.
For me, the defining attribute to recognise the modern ai stuff, or large language model generative systems is the prompt. You tell the system what to create. Like, "smoky voice of an elderly woman with a southerner dialect". Or "80's robot voice". Then the llm digs in it's training data to create something that would fit to that prompt according to it's understanding of the language. That the systems kinda understand the language in the prompt and can match it to the data is the big breakthrough in my opinion.
I don’t think you understand the process of translation or creative writing very well. For simple texts, machine translation can offer some context through translating dictionary meanings and matching patterns, much is lost because of the interpretation of the translation and the word choice that the individual translator chooses.
When a translator translates, there are often phrases that are impossible to directly translate. The translator requires some creative input to figure out what is the best representation of the expression - which extends beyond just word choice and arrangements. These arrangements are formulated between the translator’s biases, what parts of the text interests them, the expectations of translating similar texts from other writers etc.
Generative AI like ChatGPT don’t have these capacities, it simply will use a model trained on general use between these languages, so it has a real potential impact to the artistic quality of the work. I think that if you are using a translation engine to translate the content of your game, you should inform your audience because of that distinct difference.
Similarly, copiloting assistance in programming does not have the same kind of interaction or experience that an individual developer can provide. In fact, copiloting assistance has generally been correlated with decreased quality of code. A big part of this is because the creator is no longer really creating a large part of their own work.
Compare this with using AI assisted grammar assistance. The writer still has full control over the language of the document. They can ignore suggestions made by the editor. And unlike something like a copilot development mode, grammar can be checked independently of the production of the content itself. While its nice to have correct grammar, it’s not essential to the delivery of the product, unlike code, translation etc. I also think it can just be very difficult for creators to trace the influence of AI generated content through something like grammar checks, while using a translation service, ChatGPT generated text, AI generated images etc. are a lot more obvious because you are trying to replace the production of a critical part of game creation.
Ultimately there is some level of all of this that relies on some level of honesty. After all some cases of AI generation can be hard for audiences to tell apart. However, personally, and I think most people would agree with me on this, I much more value my integrity as a developer and a creator to not lie about the technology that I used to create my game, especially if its one as controversial and impactful to the labor force of game production as AI.
I think you overestimate the compexity of language translations. A creative process can help, but is not necessary. And if you use a translation website you actually do not know the tech used. Unless of course you use a site like chatgpt - and I am not even sure, if they use the actual llm capacity for translation. It is very likely, but it might also be, that they use the llm to understand what you want from it and use a different engine underneath for the actual task of mere translation to save computing power.
In it's core, a translation is not new content. The computer system does not know how freely it is allowed to translate. What you describe would be more accurately be called localisation and not translation.
If I upscale an image using an ai, or use a filter, that is not new content either. But all the new pixels were created using "ai". Maybe even the kind of ai that was trained on data. But I would not consider it "content produced" by the ai. And neither would I consider a grammar check or a translation as such.
Now, if you were to tell the ai prompt to generate a translation of your text and use the writing style of a certain author in that new language - or rewrite it in your own language in a different writing style - that would be new content.
In the context of the topic, it is about generated content. So I try to apply both words. It should not only be generated but it should also be content. A gen ai translation might be ai generated, but I do not consider it (new) content.
to not lie about the technology that I used to create my game
Fair point, but it is actually not about the technology used to create the game, but about the content and who initially created it.
