While no songs posted to date have been AI-generated, I feel this is a necessary post to make considering my personal experience trying to find music to post here.

I have been trying to find country music tracks to post here since I’m trying to avoid over-saturating the community with electronic music despite that being my personal go-to genre. In trying to find tracks from SoundCloud, I’ve discovered a lot of tracks licensed as BY-NC-SA there that are AI-generated.

Now thankfully majority of these tracks the musicians are transparent about their use of AI, though one egregious example is giving the impression that they’re locking comments if it’s pointed out since it’s not showcased anywhere on their profile. Putting my personal thoughts on AI aside, the past few days I’ve been wondering if in the future, AI-generated tracks should be allowed to be posted in this community.

I came to the conclusion that, no, AI-generated tracks should not be allowed in the community for the following reasons:

1. We have no way of telling the copyright status of the dataset used for the song’s generation. The song could have been generated using data from traditionally copyrighted tracks, meaning that the generated content would be in violation of copyright.

2. Assuming it was possible to tell that the generated track’s dataset was entirely based on BY-NC-SA music, the current status of Creative Commons licenses does not include options to protect musicians from having their music used for AI content generation. In my mind, once again putting personal thoughts on AI aside, considering that AI-generated music is a relatively new development, it is likely that a large amount of musicians would not be comfortable with their music being used for training AI even if the license they released their music under technically allows for that (the BY-NC-SA license coming before the advent of AI-generated music ofc).

3. The question of one being able to tell what an AI-generated song is derivative of comes into question. Assuming we could tell the dataset was entirely BY-NC-SA music, and that all the musicians involved somehow were comfortable with their music being used for AI music generation, how would a listener be able to tell what the song is derivative of? Perhaps this might be a moot point with it already being a hassle to determine the sources of samples in some music or if samples were used at all, but it could also exacerbate this problem that’s currently often faced.

Please feel free to let me know your thoughts and whether you agree or disagree with this new rule. I would love to hear what the community in general thinks, and will consider all feedback given.