Post

fargo
@fargo.dev

Would you be willing to apply the same standard you pick for users to yourselves? I don’t condone your use of AI art, even internally. You host a community of artists. I don’t mind LLM coding, that’s distinct from genAI art, but apply it fairly. Allow categories 1-5 for creators and VRChat equally.

scaine.bsky.social
@scaine.bsky.social

Curious why coding gets a pass? I see this all the time - people want to protect artists, but developers are thrown under the bus. LLMs can code for the same reason that they can create pictures - they've stolen billions of lines of code to learn, with no attribution or compensation.

fargo
@fargo.dev

AI art fully replaces the artist. If you generate a fursona instead of a commission, that’s direct job displacement. Code is not like that. Software is never finished, tools that accelerate development don’t replace programmers, instead they improve their speed and capabilities.

fargo
@fargo.dev

As for the training data argument, largely it is the same story. Artists almost universally do not condone training on their art, while in programming, open source and the free sharing of ideas is commonplace. There is of course complications and nuance, like licensing concerns, but that’s the gist

0 reposts0 likes

asie
@asie.pl

That's really not true these days. I have seen many websites and tools created from a single prompt, maybe with some back-and-forth for reporting bugs. I have been commissioned to make small things like that in the past, I probably won't be anymore. Also, a lot of software *can* be finished.

asie
@asie.pl

I also know of about as many artists who condone training on their art as I know programmers who condone training on their code. Free sharing of ideas is commonplace, but open source was never about just taking a dev's work without even attribution (even if it did often happen, but so did tracing).

fargo
@fargo.dev

That’s genuinely surprising to me. It seems like there’s such a cultural difference: Even if art is posted publicly (like FA) it still is like, a PART of that person, deeply connected to our identities often. But an open source GitHub repo was always intended to be shared knowledge for public good.

asie
@asie.pl

Some artwork is utilitarian - think store banners or municipal posters; some code is artistic - think the demoscene or code golfing. I like to believe that my programming work is very much an expression of my identity, of my preferences and perspective.

Johan Schuijt
@resolve.works

I fully condone training models on my code just as I condone others learning from my code. I would argue that art and code are two very different things. Code is a means to an end, if you can make my life easier by improving my tooling to get to that end, I'm all for it. That does not apply to art.

scaine.bsky.social
@scaine.bsky.social

It doesn't matter if you condone it or not. The genAI companies will take it anyway. Your opinions don't matter to them. Of course code and art are different, but attribution, compensation and the license under which both code and art is created is the same. Again, the companies don't care.

scaine.bsky.social
@scaine.bsky.social

Sure, if the coding license is honoured. But it's not. They grabbed millions of lines of GPL code and shoved it in there. No copyleft was honoured. No share and share-alike. No "must not be used for commercial purposes". That £200/month Claude license was built on "protected" code.

fargo
@fargo.dev

Yes this is a real complication, that the models might output some protected code snippets without the accompanying required license. Honestly, I have a lot of problems with how these companies trained their models. I’m not condoning their behavior, we ought to sue and regulate the hell out of them.