this post was submitted on 28 Dec 2024
636 points (93.1% liked)
196
16747 readers
2425 users here now
Be sure to follow the rule before you head out.
Rule: You must post before you leave.
founded 2 years ago
MODERATORS
you are viewing a single comment's thread
view the rest of the comments
view the rest of the comments
AI generated art doesn't meet the definition of plagiarism though?
plagiarize: : to steal and pass off (the ideas or words of another) as one's own : use (another's production) without crediting the source.
Since almost no one actually consented to having their images used as training data for generative art, and since it never credits the training data that was referenced to train the nodes used for any given generation; it is using another persons production without crediting the source, and thus is text book plagiarism.
Every living artist uses other people's art as training data without their consent. That's the way art works and it's ok. Please let's not consider every artist has to pay for every piece of art they ever layed their eyes on to be allowed to create art themselves.
So does that mean that any artist which has viewed another piece of art and learned from it, and used that knowledge in their own works, has therefore committed plagiarism by not asking for permission or crediting every work they’ve ever seen?
I’m an author and one of the most common pieces of advice for authors is to read more. Reading other authors’ works teaches a lot about word choice, character development, world building, etc. How is that any different from an AI model learning from art pieces to make its own?
AI systems like generative art models are trained on large datasets to recognize patterns, styles, and structures, but the output they create does not directly copy or reproduce the original data. Instead, the AI generates new works by synthesizing learned features. This is more akin to how a human artist might create something inspired by various influences. If the generated image does not directly replicate any specific piece of the training data, it cannot be considered "using another’s production without crediting the source."
Also AI platforms like Midjourney do not “reference” specific works in a way that can be credited. The training process distills millions of examples into mathematical representations, not a library of individual artworks. Crediting every source is not only infeasible and impractical, it is also not analogous to failing to attribute a specific inspiration or idea, which is a cornerstone of plagiarism.
Plagiarism is defined in academic settings very precisely. Getting ideas and structure from others rarely meets the standard. Why? Because we do this all the time. Also, plagiarism is 100% legal, because of course it is! Imitation is often a good thing.