3 ms·
Apparently current Japanese law allows models to be trained on copyrighted content. https://www.deeplearning.ai/the-batch/japan-ai-data-laws-explained/ https:/
by pnw 2y ago
Apparently current Japanese law allows models to be trained on copyrighted content.
https://www.deeplearning.ai/the-batch/japan-ai-data-laws-explained/ https://www.deeplearning.ai/the-batch/japan-ai-data-laws-exp...
- lxgr 2y agoSuch a law doesn't automatically imply that everything a model outputs is automatically copyright free though, does it? Otherwise: Meet my new and surprisingly performant new model, IdentityFunctionGPT! Oh, the weights? Sorry, trade secret.
- Cthulhu_ 2y agoI don't know, I'm no lawyer. However, if I make a drawing in the style of Ghibli, I own the copyright because I made it. I don't know if in Japan, art style is copyrightable. I don't know if copyright applies to generated content (of any sort) though. It probably does.
- lxgr 2y agoThe important question here isn't who owns the copyright in the output, but rather whether the output also infringes on somebody else's, and if so, who's liable for that: The company that trained the AI, the one that hosts it as a service, or the user that prompted it.