Reply to thread

Message

<blockquote data-quote="Jfdlsjfd" data-source="post: 9237898" data-attributes="member: 42856"><p>Copyright is a monopoly that is granted by the State in order to allow the original author to profit reasonably from his work, in order to incentivize him to show his work to the public and to keep creating without having to get a paid position (like it used to be before, but tended to create official art et religious art mostly), with exceptions so that the public benefit of having these works created still remain (like, how you can copy it for your own private use). The details were mostly voted before AI became a thing, but there is a strong case to be made that the public interest leans toward allowing training. Look at the EU's text and datamining exception, it's quite broad and allow cultural institutions to do exactly that: as an exception to copyright, they can scrape freey and train a model. Sure, Stability AI trained its model with a subset of images linked to LAION-5B, and made their model available for free ; and Stability AI isn't a cultural institution, but I am not sure everyone would be cool if it was the Rijksmuseum that was publishing the model instead of SAI... Much like I think a lot of people (not you, in general) aren't happy with database trained on Adobe's images or public domain image (like pixart-alpha), which I think is evidenced by using the term AI broadly and not specific models they have a problem with.</p></blockquote><p></p>

[QUOTE="Jfdlsjfd, post: 9237898, member: 42856"] Copyright is a monopoly that is granted by the State in order to allow the original author to profit reasonably from his work, in order to incentivize him to show his work to the public and to keep creating without having to get a paid position (like it used to be before, but tended to create official art et religious art mostly), with exceptions so that the public benefit of having these works created still remain (like, how you can copy it for your own private use). The details were mostly voted before AI became a thing, but there is a strong case to be made that the public interest leans toward allowing training. Look at the EU's text and datamining exception, it's quite broad and allow cultural institutions to do exactly that: as an exception to copyright, they can scrape freey and train a model. Sure, Stability AI trained its model with a subset of images linked to LAION-5B, and made their model available for free ; and Stability AI isn't a cultural institution, but I am not sure everyone would be cool if it was the Rijksmuseum that was publishing the model instead of SAI... Much like I think a lot of people (not you, in general) aren't happy with database trained on Adobe's images or public domain image (like pixart-alpha), which I think is evidenced by using the term AI broadly and not specific models they have a problem with. [/QUOTE]

Verification