The debate over generative AI doesn’t look to be settled anytime soon but the specific contours of what it should and should not do are starting to form, and fast.
One case is over in Japan where the recent spate of ChatGPT-generated Ghibli-esque images has garnered attention and possibly future regulation.
Ostensibly to protect creators and their work, such legislation targets this specific implementation of generative AI because it cannot exist without having, in some way, imitated the style directly.
How it learned this is more of a question for the courts to decide should ChatGPT need to explain itself but the debate over how this material is obtained and then used is salient enough for our community to discuss.
There is also the broader debate over whether a “style” can really be copyrighted or if it should be. After all, many of the masters and even those today teach themselves through imitation of those that have come before. Yet this sequence ignores the scale at which AI can create images and teach itself. It’s one thing to endeavor to become an artist in the style of Picasso and a whole other entirely to quantify and condense that style into a readily available, generalizable and replicable algorithm of some kind.
To be clear, the illegality here would stem from how the training materials were obtained. If copyrighted images were used then that is a problem. Simply copying a style is not but, again, maybe it is if it is AI? There’s a lot more nuance here than there would be if it was a human creator behind the wheel.
Let us know if you think that this is a good step forward for protecting creators or a bit misguided in the comments.
Check out some other news at this link.
