I’m not against AI tools, despite what a lot of people might say as a defense, but I do believe it’s important to properly separate genuinely human made projects from AI generated code.
For this reason, I genuinely believe it’s a good idea to add an “ai-assisted” or similar tag so that people are required to tag the Resource they post and offer to others as being ai generated / heavily AI assisted.
Having a user AI generate an entire library and claiming that they made it themselves can be damaging to the community, since newcomers might try and use these resources, without realizing that they’re AI generated, possibly buggy, make questionable design and pattern choices as most AIs do, and the creator, if they relied heavily on said tools, might not even be able to provide proper support / updates for it down the line.
Once again, I do not wish for AI submissions to be banned in any way, but I do want a clear seperation of AI generated code / Resources vs. human made submissions.
I’d be in favour for this, but with clear definitions.
For example, if I write the code, but let the AI generate in-file comments for documentation, would I use the tag?
I guess there’s arguments to be made for either side, I’m just trying to vote for clear definitions!
For art / music / sfx - If you use AI in any capacity, mark it. No exceptions there.
For code, if you generated more than a single file worth of AI code and it’s blatantly just copy pasted, mark it as AI.
For documentation I’m split. On one hand, it can help if your primary language isn’t English, and I’m 100% fine using AI for translation, but if you write the addon yourself, you know how it works best, not the AI. And if you can’t even be bothered to write proper documentation for it for the very people you’re creating the addon for, why even make something for others to use? You clearly don’t want to help them, so…
Yeah agreed. The one issue is people who want to pretend they did it all themselves etc. . But it would be an improvement for those people who are honest. There will always be people who lie but still this would be a big improvement.
I think your documentation argument doesn’t really hold up. Again, I’m in favour of this tag, just want to make sure it’s clear for everyone.
But “you don’t want to write documentation so you clearly don’t sant to help them” is quite the opposite. One could argue being able to release more code and tools for the community if you’d speed up doc writing via ai for example. That being said, if any generated content in any form requires the tag, that’s clear and shouldn’t cause confusion.
Yeah I think it’s likely easiest just making any use of AI have to be marked. But something should allow developers to specify where cause generative ai/AI code is a big difference to writing some docs with it.
Why? I’m sorry but if you don’t want to bother explaining how to use the addon / library you have created, then why do you want to make it in the first place?
The whole point of creating these resources is to help out others, make creating games easier for everyone.
How do you achieve that if you just say “ah I don’t care, just generate the docs for me”.
It’s going to be low quality in most cases, and will likely miss important details and “gotchas” if your addon is even slightly more complex than average.
And the mentality of it… if you don’t like the idea of writing useful documentation, or your entire documentation is just AI, then most people like me will also assume your addon was made using AI.
I think use of AI in the docs to speed up the process is fine. But I agree that if someone makes the entire or most of the documentation with a LLM then it as a minimum has to be marked as such.
This is a different issue. As I said, I’m fine with things like Translation from another language, but if you care so little about your own work that you don’t even want to explain it, then it calls into question the quality of your work as a whole.
I think that AI docs should be tagged, even if it’s just a translation. Because it prevents the accusation for trying to “get one over” on people. And TBH, it’s pretty easy to tell. Because AI docs are overly verbose, have lots of icons, and do not tell you anything helpful.
To @tibaverus ’ point, my StateMachine code hit v1.0 this week and I was going to release it and realized the docs were not really helpful enough. So I backburnered that while I think about what to add to the docs. I just can’t imagine how AI could read my mind and decide what was important just by me saying “Generate Docs”.
Fair point. In general I think as previously stated anything using AI should be marked as such. Even just comments cause that’s still editing the code which could mean it edited something else too even if it was told to only do comments.
I would not trust AI to write comments for my code. I’ve already seen it fail when other people have used it to use my code, and I’ve had to fix things for them because the AI doesn’t know what it is doing.
Agreed. For me, once I understood that “AI” doesn’t actually know anything, the magic wore off. It just isn’t human or has the life experience (because it has no life).
I also find AI annoying to work with. If you want to do something in an exact way you want, you have to do more babysitting than your actual work.
It’s nightmare fuel that people still fall for this.
And all the code is just it looking at other peoples code. I know how messy my code is. I don’t want something learning from an entire internet’s worth of horrendous code