While I am no fan of ‘merica, In many areas the UK is definitively worse.

Disney goes hard after uses of their art and characters, but doesn’t seem to be as DCMA-happy about other minor projects.

The problem is also that nearly all of these AI nets are trained on stolen art.

The gameplay loop of Horizon *really* wouldn’t work for a Monster Hunter style multiplayer game without some massive overhauls.

You’re a day late to the conversation, keep scrolling.

The biggest problem is that in the grim dark future, there are only assholes. There’s no heroes, so that rather limits potential tones for a show.

Then I do apologize for painting you with the broad brush. If datasets were generally sourced ethically, I’d have little problem with the technology overall. But that’s not where we live right now.

Yeah, it’s that ‘publicly available’ dataset that I’d be suspicious of. If you started entirely from scratch, feeding your own dataset completely on public domain (not ‘publicly available’, which is just any gallery a web-crawler can reach) and licensed images, and then ran or purchased the extensive server time to…

- and there it is. The loudest voices supporting AI are always those who exploit the stolen content to avoid paying actual artists.

Sounds like me in my early 20s. Gods, that sucked. Hope the job hunt goes well.

I am baffled at the amount of time you have spent in the past week defending people you now admit are jackasses. Like, this wasn’t just a 2-cents situation, this was investment. I am invested in this because I have a number of professional artist friends who are directly affected by this. What’s your deal?

One last note on the topic. I would point out that there are Stability-based diffusion AI for music as well, but they are specifically trained on “datasets composed entirely of copyright-free and voluntarily provided music and audio samples”. So what’s different here? Simple. Musicians have major litigious labels…

And my argument is that the bits of images stored in a trained neural net are still pieces of the images that were fed into it, not something that was transformed into something else via.. I dunno, digital alchemy or something. So if the software was trained on copyrighted content, the output is also infringing.

“Theft” is a generalized term. Feel free to substitute “copyright infringement” in all cases in this discussion and actually address the issue. “publicly facing” is a meaningless term in that context. AI art is just a fancy twist on napster/limewire. Imagine if someone built a music-remixing database that ran off of…

The fact that it’s a more innovative form of copying doesn’t mean it’s not copying. There’s no argument claiming lack of infringement that doesn’t rely on humanizing the algorithm. You, Sylvers, censure, and the other less-prolific AI-champions in these comments keep repeatedly making the comparison to human artists…

They’ve been paid.

Last paragraph, “moral” compasses, not morale. (though imagining the latter as a game-stat, my morale compass drops notably any time this game shows up in the news)

Are you arguing that you’re infringing on the dead animal’s copyright? Or perhaps implying that the copyright is owned by the evolutionary process? Perhaps more whimsically held by God or Mother Earth?

This is how insidious capitalism is. You can’t imagine a system where decisions are made on some criteria other than “number go up”, just varying who gets the profit. Decisions can be made on non-financial criteria too.

RE: web scraping, URLs do not fall under copyright, art does. Search engines scraping page text are not fully reproducing it somewhere else, they are linking to the original so it remains fair use.