5

I've read that the training sets and outputs of some generative AI services include copyrighted art. I'm interested in trying out image generators while not accidentally copying someone's art who hasn't intended for it to be used this way.

How do I figure out how to use image generation services without abusing the rights of others? For example, I've tried to figure out which services make sure their training sets include images who's creators contributed the images intentionally.

1

2 Answers 2

3

Link: https://genai.stackexchange.com/a/344/12

I think that "copy someone's art" cannot be determined in such a way that people who aren't art experts might determine.

On one side, the legal and ethical aspects are handled through intellectual property and authorship rights. On the other is the determination that copyright laws can limit what can be used and who and how to train an artificial intelligence entity.

It might help to look at the generative artificial intelligence operator documentation about their stand regarding intellectual property and authorship rights. Some organizations use a structure of vision, mission and values or creed statements, long-term strategies and programs; others might have a "manifesto".

In the coming years, it is possible that the artificial intelligence industry will develop specific guidelines and practices around intellectual property and authorship rights. Simply stating that they operate within the law may not be enough.

Related

3

How do I figure out how to use image generation services without abusing the rights of others?

I'll interpret "abusing" as illegal to dodge the debate around ethics.

There are 2 possible illegal uses of a training sample: 1) the license of the training sample doesn't allow AI training 2) the AI model plagiarizes the training sample (and/or doesn't provide proper attribution in case of plagiarism).

If the image generation service didn't indicate what their training set is, it's very difficult to confidently infer if a sample was part of the training set, unless AI model plagiarizes it.

Also, note that the law itself is unclear on what can be used to train commercial AI models, e.g. see Is it illegal for a firm to train an AI model on a CC BY-SA 4.0 corpus and make a commercial use of it without distributing the model under CC BY-SA?.

Note for dataset providers that researchers have tried to develop some methods to alter a dataset so that one can detect whether a model was trained on it, e.g. see https://arxiv.org/abs/2002.00937


Example of a detection service: https://www.makeuseof.com/how-to-know-images-trained-ai-art-generator/:

Set up by a group of artists, Spawning is a collective whose aim is to help people find out whether their images are on datasets like LAION-5B, used to train AI models. Their web search engine called Have I Been Trained? lets you easily search keywords such as your artist name.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.