Ah there it is. Capitalism seizing the means of generation.
Training is transformative use.
The dumbest thing we could do about spicy autocomplete is make copyright worse.
Not sure about LLM but for image generators training on copyrighted material should definitely not be allowed. So the AI act should not water down those rules
There is no difference. A chatbot that’s read every book in the library is the same as a denoiser that’s seen Zootopia. Published contents are public… that’s what the word means. It’s not theft for the robot to know what Superman looks like, and the only way it’d know that is by lookin’.
None of these models are supposed to reproduce a complete work. That’s a failure to generalize. It means they could be smaller, and still change new footage to fit a novel description. If you film yourself in a bathrobe and motorcycle helmet, and say “Darth Vader,” you won’t get a single from of A New Hope. It only removes the fine details that don’t match the concept. If you’re skateboarding, so is Vader.
Yeah, but the robot can be set to recognize only based on the training data (majority of people would be ok if their work is used when training only for recognition purposes) the problem is the production when the same data set is used to produce images/video. Here it matters what the robot has been trained on. Model that never “seen” picture of pikachu will not be able to reproduce pikachu (For example Firefly by Adobe) Model that has pikachu in its training data will reproduce pikachu even if you don’t use the name pikachu but just describe what pikachu looks like. And that is only part of the problem with original characters (many being done by small authors who don’t have the means to sue companies behind those models) Second issue I see is the AI companies using the training data as a free buffet, even though out of all the sources - electricity, compute, training data the training data are the most valuable - no matter how much electricity and compute power you have, your model cannot exist without training data. Yet no one is paying for those training data anything (again, unless you are big publisher that is able to threat the companies and make a licence deal). All the companies behind those models are for profit companies yet they are using any loophole in the system to pay as little as possible and exploit work of others.
Generators are recognizers. Like how speakers and microphones are the same mechanism, backwards. ImageNet was only a classifier for a couple thousand single-word labels. Dall-E used the same approach to walk an image toward those labels. That’s how you get “avocado chair.”
Yet no one is paying for those training data
They don’t have to, because training is transformative. It’s fair use. If permission is not a factor, why would money be required? Your rights are not a loophole. This is fine for the same reason you’re free to reference, parody, or quote commercial works guarded by flesh-eating lawyers.
Even a vegan model trained on bespoke data could reproduce trademark-infringing characters. If you can describe what Naruto looks like, the model only needs to know what anime means.