Kinda? I know what the difference between LLMs and proc gen is. I don't think that it will be clear if a ban comes into place, and people are so knee-jerk at anything that isn't 100% human made that they will care and demand anything that uses any kind of artificial intelligence or non human composed and played music be burned at the stake.

Also, I think that LLM based generation is a subcategory of proc gen. I think LLM based stuff *can* be set up in non slop ways to create non slop work, and I want the "experimental" high effort LLM stuff to be treated as such.

[–] rudyharrelson@lemmy.radio 2 points 3 hours ago (1 children)

people are so knee-jerk at anything that isn’t 100% human made that they will care and demand anything that uses any kind of artificial intelligence or non human composed and played music be burned at the stake.

Maybe let's not presume how the community will behave given a new rule regarding generative AI. Music featuring digital instrumentation has certainly not been "burned at the stake" despite it not being "100% human made".

Music composed and performed by generative AI (which was trained on the works of other artists) would understandably be met with criticism within the music community, and is a far cry from "100% human made" in that it is closer to "0% human made".

[–] Skullgrid@lemmy.world 2 points 3 hours ago

I think once someone mentions "algorithm" or "I generated" it's over. People won't care about the details, because they're people.

[–] Hawke@lemmy.world 1 points 3 hours ago* (last edited 3 hours ago) (1 children)

Also, I think that LLM based generation is a subcategory of proc gen.

Hmm… I would usually think of proc gen as being more deterministic than what LLMs do (even if they use random numbers there’s usually a seed value to get consistent random numbers)

I guess in some ways the important part is how much work the purported artist put in to the output.

[–] Skullgrid@lemmy.world 1 points 3 hours ago (1 children)

Hmm… I would usually think of proc gen as being more deterministic than what LLMs do (even if they use random numbers there’s usually a seed value to get consistent random numbers)

I haven't re-installed my local copy of stable diffusion yet, but I'm pretty sure same seed + same prompt = same output. If you have one to hand, feel free to try... Thinking about this a bit more, I guess not, since you can run batches from the same seed, so I guess it's not exactly the same?

Either way, running the same seed + prompt at least gives you similar outputs. It belongs in the same type of... "tool"?

Maybe if the guidelines are to ban LLMs? But even then, I want the weird , difficult to create LLM stuff to be posted, as long as it's not suffocating everything else.

I guess in some ways the important part is how much work the purported artist put in to the output.

Yeah, I think this is something even the most hardcore AI guys will agree with. No one wants the community to be overrun by drivel.

[–] Hawke@lemmy.world 2 points 3 hours ago (1 children)

Haha… I don’t have stablediffusion because !fuck_ai@lemmy.world but I didn’t realize they were that close to procedural generation techniques.

The problem with “effort” as a metric is that there’s high effort shit as well as low-effort good stuff. But I suppose usually the high-effort shit at least has a good story behind it.

[–] Skullgrid@lemmy.world 2 points 3 hours ago

yep. again, restriction instead of outright bans means that it's up to us to give a shit or not when someone actually tries properly instead of just pulling something out of their ass.