Copyright questions loom as ChatGPT's Ghibli-style images go viral

They fed Ghibli movies, which are copyrighted, into their AI to allow it to create similar images. How is that not a violation of that copyright?

They fed Ghibli movies, which are copyrighted, into their AI to allow it to create similar images. How is that not a violation of that copyright?

And since Hayao Miyazaki deeply hates AI produced images this feels like a personal attack to him, hopefully this will have serious repercussions.

They fed Ghibli movies, which are copyrighted, into their AI to allow it to create similar images. How is that not a violation of that copyright?

And yet when a human looks at many Ghibli movies and draws something in similar style, we do not call it a violation of copyright. The problem is that AI does it much faster and much better. So it’s like complaining that someone is better at something than you.

Copyright issues only make sense with industry products, machines, technology and such. In the area of art or artwork in the wider sense it doesn't. Anyone can become an artist and often art is based on similar art of other similar artists. In fact, if the intention of an artist is, that his works remain unique and uncopied, then he simply doesn't publish it and keeps it all a hidden private secret. That is real art, the big rest is for sale or publicity, commerce, and such it cries indirectly for being widespread and copied.

This seems very complex. Is any comic who does a schlurry impersonation of Sean Connery as James Bond infringing his copyright? I don't think it is.

It also opens the door to any anime or manga artist whose work predates Miyazaki and looks similar to sue Miyazaki himself for a vast sum of money. It is very naive to regard Ghibli's drawing style as 100% original. He is standing on the shoulders of many who came before him. It is also incorrect and superficial to think the magic of Ghibli films lies in that drawing style. This would be a huge insult to the other people working on the films, coming up with the stories, doing the voices, doing the music etc.

If ChatGPT has seen every film and heard every song ever made, I am tempted to sign up and ask it some questions. There are loads of videos on Youtube analyzing films in various ways, historical accuracy of battle scenes and the like. Some of them are quite interesting.

This will never be resolved.

Direct copying - pixel by pixel - can be deemed a copyright infringement/theft, but anything loosely based on a "style" will not be able to prove theft of an image.

For better or worse, AI production will increase, there's no stopping it's progression.

Safeguards are needed, but animation characters have become so ubiquitous that it's inevitable that crossovers occur.

If I drew a new super hero that looked just like Superman, then I'd expect I'd be in trouble.

If I drew a new super hero that sort of had some aspects of Superman - ie physique and cape - well I wouldn't expect to be in trouble.

And yet when a human looks at many Ghibli movies and draws something in similar style, we do not call it a violation of copyright.

But this is not a human. That's the whole point. This is a company that used copyrighted material without any authorization to develop a product it is now making a lot of money on.

I'veSeenFootageToday  01:13 pm JST

But this is not a human. That's the whole point. This is a company that used copyrighted material without any authorization to develop a product it is now making a lot of money on.

OpenAI has never made any money and never will, the business case just isn’t there. They are burning through billions of dollars of investor money.

If you understand how LLMs work, you know it’s blatant copyright infringement on all levels. It can mimic anything you feed it with during training.

OpenAI has never made any money and never will, the business case just isn’t there. They are burning through billions of dollars of investor money.

True, but that's a little bit besides my point. As stated in the article, the Ghibli ripoff is limited to paid users only, meaning they are making money off it. Which is legally what matters.

OpenAI is facing a barrage of lawsuits over copyright infringements...

But still they can freely operate without facing any issues. Why? Because of this:

OpenAI has projected its annual revenue could exceed $12.7 billion in 2025, up from $3.7 billion expected this year.

Sure, AI art imitates Ghibli style, but it's obvious it's not Ghibli. ...right?

ibb.co/YFdY1rXs

Styles like Ghibli’s can’t be copyrighted. Copyright protects specific works, not general aesthetics. If someone generates an image that directly copies Ghibli content, that’s on the user, not the tool.

They trained the AI model on publicly available data to learn patterns, not to memorize or store copyrighted images. It doesn’t contain or reproduce exact works - just like a human artist learning by using mood boards.

It would be like suing Adobe because someone made a forgery in Photoshop. If we start treating style as property, we risk restricting human and machine 'creativity.'