I haven’t looked into it too much, because I don’t bother to use these things myself. But if I remember, there are some systems that are open-source, can be run locally, and then a person could train those systems on only public domain and freely licensed works. That is the kind I’m talking about, so bringing up the systems I’m not talking about is just a strawman.
When you say, “However what you’re hypothesizing cannot exist, due to the nature of a diffusion model. The source that the images are generated from, all of that was stolen in 2021. It was illegally scraped.”, are you referring to Laion-5B?
Yeah, I’m sorry but none of this seems to add up. I read the book you linked, the entire thing, and I didn’t really learn anything new from it. In my own searches I’ve found a number of competing image generation systems, most of them just use Stable Diffusion - including Horde, which makes sense as it is open-source and readily available. But others have their own solutions, and some even use their own datasets. I’ve also done other searches and found a variety of instructions on how people can put together their own datasets. Between all of the options, it seems pretty clear to me that even now it is entirely possible for people to use these AI tools without ever needing to come into any contact with “stolen” (ie., copyright infringed - it’s not theft) artwork.
But I’m not going to link to any of these things because I don’t want to endorse or promote anything without taking sufficient time to be really sure that they are in fact ethical, which honestly I’m probably not going to do at least for the time being, since I have no need to use any of these anyway. But yeah, ethical ai art is entirely possible, no copyright ever needs to be infringed, and this is something anyone can do today if they’re inclined to make the searches and assemble free source images themselves.
I understand what you’re saying the point to me that sticks out was the usage of “hypothetical”, which to me would mean things that don’t exist, which wouldn’t be a strawman as i understand it. With that being said I’m genuinely not trying to be an ass, just explaining the way I perceived the comments, I hope it’s not needed to say, but I meant no offense in my reply of your comment.
No offense taken. It is still a strawman though, because regardless of the realness of what I’m talking about, they’re still referring to something different and then attacking that as if it is the thing I’m talking about.
The present models are, is the point i think the other person is making.
I haven’t looked into it too much, because I don’t bother to use these things myself. But if I remember, there are some systems that are open-source, can be run locally, and then a person could train those systems on only public domain and freely licensed works. That is the kind I’m talking about, so bringing up the systems I’m not talking about is just a strawman.
deleted by creator
When you say, “However what you’re hypothesizing cannot exist, due to the nature of a diffusion model. The source that the images are generated from, all of that was stolen in 2021. It was illegally scraped.”, are you referring to Laion-5B?
deleted by creator
Yeah, I’m sorry but none of this seems to add up. I read the book you linked, the entire thing, and I didn’t really learn anything new from it. In my own searches I’ve found a number of competing image generation systems, most of them just use Stable Diffusion - including Horde, which makes sense as it is open-source and readily available. But others have their own solutions, and some even use their own datasets. I’ve also done other searches and found a variety of instructions on how people can put together their own datasets. Between all of the options, it seems pretty clear to me that even now it is entirely possible for people to use these AI tools without ever needing to come into any contact with “stolen” (ie., copyright infringed - it’s not theft) artwork.
But I’m not going to link to any of these things because I don’t want to endorse or promote anything without taking sufficient time to be really sure that they are in fact ethical, which honestly I’m probably not going to do at least for the time being, since I have no need to use any of these anyway. But yeah, ethical ai art is entirely possible, no copyright ever needs to be infringed, and this is something anyone can do today if they’re inclined to make the searches and assemble free source images themselves.
I understand what you’re saying the point to me that sticks out was the usage of “hypothetical”, which to me would mean things that don’t exist, which wouldn’t be a strawman as i understand it. With that being said I’m genuinely not trying to be an ass, just explaining the way I perceived the comments, I hope it’s not needed to say, but I meant no offense in my reply of your comment.
deleted by creator
No offense taken. It is still a strawman though, because regardless of the realness of what I’m talking about, they’re still referring to something different and then attacking that as if it is the thing I’m talking about.
deleted by creator