[ 3 / biz / cgl / ck / diy / fa / ic / jp / lit / sci / vr / vt ] [ index / top / reports ] [ become a patron ] [ status ]
2023-11: Warosu is now out of extended maintenance.

/ic/ - Artwork/Critique

Search:


View post   

>> No.6450127 [View]
File: 343 KB, 512x1814, fair use.png [View same] [iqdb] [saucenao] [google]
6450127

>> No.6449959 [View]
File: 343 KB, 512x1814, fair use.png [View same] [iqdb] [saucenao] [google]
6449959

>> No.6448423 [View]
File: 343 KB, 512x1814, fair use.png [View same] [iqdb] [saucenao] [google]
6448423

>>6448252
Here, let me give you guys an example.
Now, if you apply the core concept "the latent space of a dataset is a static archive of images, containing the images originally trained on + interpolations between them", it becomes very easy to see why any claims of "fair use" will never fly. It might be one thing if the algorithm learned general principles and created entirely new images, but what it does is essentially photobash/animorph from its existing database. It is an interpolating machine.
So let's see what happens when we apply this to fair use laws-
1. They took copyrighted data and then used it for profit. Furthermore, they took data that was gathered under the pretense of being non-profit and research only and used it for profit in a clear attempt at creating a legal loophole.
2. The data that they took is comprised of people's own personal and professional photography and artworks, things that are expressly made as creative expression.
3. The dataset and everything derived from it is 100% dependent on all of this copyrighted material, and could not exist without it.
4. Image generation algorithms are expressly meant to replace artists and artistic expression as a whole. There are multiple cases of people using them to impersonate artists, run other peoples' work through img2img and then claim it as their own work, and art communities being flooded with these interpolated derivations.

On the macro level, the dataset itself is a massive copyright infringement case. On the micro level, the users using the image generators did NOT create anything output by said image generators.
Even with inpainting, every bit of output is still something drawn from within the dataset, the only creative input on the users part telling the generator to place the generated content into a specific place.

>> No.6443507 [View]
File: 343 KB, 512x1814, fair use.png [View same] [iqdb] [saucenao] [google]
6443507

>>6443502
If anyone is wondering about the fair use laws, here's the relevant info.

>> No.6441889 [View]
File: 343 KB, 512x1814, 1671924840437640.png [View same] [iqdb] [saucenao] [google]
6441889

>>6441878

>> No.6438027 [View]
File: 343 KB, 512x1814, fair use.png [View same] [iqdb] [saucenao] [google]
6438027

>>6438024
AI shills like to make the argument that since their meme machine draws from so many data points all at once that it automatically constitutes fair use and there is nothing that can be done about it, but it's not so clear cut as that. Not only is the 'fair use' of the output in question, the very fact that all of this copyrighted content is being used in a program for profit is also it's own bundle of problems. I leave you all to draw your own conclusions from this information.

>> No.6436565 [View]
File: 343 KB, 512x1814, fair use.png [View same] [iqdb] [saucenao] [google]
6436565

>>6436558
Here, I'll post a screenie for everyone's benefit.

Navigation
View posts[+24][+48][+96]