Gregory's Server@mahmoudajawad Hey, it's https://stablediffusionweb.com/#demo , and it's trained on https://laion.ai/blog/laion-5b/ dataset (same for DallE and Midjourney). This set contains 20 years of all my art online, so hundreds (thousands?) of artworks of me. Open-movies concept art, all pepper&Carrot, all my illustrations, etc... It's possible to find them in the dataset directly (link on the LAION page, or https://haveibeentrained.com/ . It's against my consent, of course. Here is a screenshot sample:
|
6 comments
 @mahmoudajawad Hi, Yes, I have nothing personal against AI, or Image resynthesis software based on neural networks. As a geek and enthusiast for visual art, it's amazing to see this tech. But it should just be trained on Public Domain material. There is enough of them on Wikimedia and other websites for a good starter set. That's why I decided to not use it in my work, and added a "🚫 No A.I." on my bio for now. @davidrevoy @mahmoudajawad So AI generation would even violate your Creative Commons licence, because the generator does not give any kind of credit. Right?  @davidrevoy @mahmoudajawad @lonelyowl Yes, good idea. Maybe just ask the one who want this to upload on Wikmedia Commons under Public Domain. |
@davidrevoy this bothers me. I'm all for #AI advancement, but also with total respect to everyone's work. For me we will be there when we end-up with an AI that is able to tell which works derived the end-results from, which would contribute to ending this dilemma.