@MisterBoring That’s an interesting point. There do exist models for SD that are fed on images in public domain, but I’m curious how well they really hold up because of the sorts of examples you mention.
@bear_necessities said in AI PBs:
I’m thoroughly confused by what the argument is at this point. Is it ok to use AI as long as i acknowledge it’s harmful to artists?
It’s okay to use AI. Maybe it’s not perfect, but fuck it. I use AI, but I acknowledge the technology has some real flaws, and I don’t try to pretend that it is ethically better than alternatives. I’ve used AI for images and AI for code. I’ve even used AI to help me figure out why an update failed for BMD, so abandon ship if that’s a problem!!
I use it. I do sometimes think about whether I should buy carbon credits or something to feel like I use it ethically, but on the other hand, I don’t worry about the carbon credits I burn playing video games. I don’t know. On my fucks given scale, it doesn’t really rate, but it does sometimes itch.
In this discussion, I find the approach that AI PBs are ethically preferable to using PBs of existing persons hard to swallow.
What is the line of thinking? Many of these models have used those very same images in their training data. Like, you’re just using the exact same images with an extra layer of ‘and also other copyrighted works’, in a way that is still very much under debate for how much actual harm it causes.
Then there are some harmful beliefs out there which make people blind to potential issues:
@STD said in AI PBs:
Secondly, if the model is made for a for-profit system like Midjourney, then they already have the requisite rights and permissions. That’s part of what you’re paying for when you buy a license for Midjourney.
That’s just incorrect. Make your judgments on the matter based on fact, at least.