5 ms·
It is a complicated subject with no clear resolution, IMO. Train people on a bunch of books from a local library and noone has an issue. Train a generalized AI
by _ea1k 3y ago
It is a complicated subject with no clear resolution, IMO. Train people on a bunch of books from a local library and noone has an issue. Train a generalized AI on the same thing, and the legal result should be the same.
But what about an imaginary specialized AI that really just focuses on reproducing its input training material in a nearly verbatim way?
The current reality is somewhere in between these two extremes, and defining the boundaries sounds difficult.
- soulofmischief 3y agoThat's not an AI, that's a database. The output should certainly not be copyrightable if it does not produce an existing copyrighted work or if fair use is in play, which is why many people seem to be attacking the training set itself, an argument which is also complicated but has little merit.
- _ea1k 3y agoSure, but the line between AI and a database can be blurred, even intentionally. Even with ChatGPT, it can be coerced to deliver parts of its training set nearly verbatim.
- SideQuark 3y agoNewer training techniques likely removed most of that by simply deduplicating training data. All the papers I read on verbatim techniques could only find items that were overpopulated during training.
- bayindirh 3y agoWhat if that output is created by mixing in copyrighted material? e.g.: What happens to your MIT licensed code when an AI assistant generates some code for you which contains some GPL, BUSL or Source Available licensed parts? Moreover, do these code bases scraped with consent of the owners?
- soulofmischief 3y agoWhat is "mixing in" mean in this context? The image itself is copyrighted, or it isn't. Fair use may or may not come into play for things like replication of known copyrighted characters, depending on commercial use. None of this has changed from the traditional copyright process. Even 15 years ago I could find some of my old works rehosted on obscure Chinese websites, and my options for recourse were the same as they are today. Many of my own works were already derivative.
- xigoi 3y ago> The output should certainly not be copyrightable if it does not produce an existing copyrighted work So if I make a program that takes an image and changes a single pixel, I can use it to launder copyrighted images?
- soulofmischief 3y agoGo ahead and try; A lot of this kind of "gotcha" commentary comes from a lack of understanding about existing copyright law. Let me know how far you get.
- xigoi 3y agoI assume from the snarky response that the answer is “no”? In that case, what exactly needs to be changed to not count as the original work?
- soulofmischief 3y agoI'll assume your original question was in good faith because it itself came across as snarky. Here's what you need to know about derivative works in US law: https://www.copyright.gov/circs/circ14.pdf https://www.copyright.gov/circs/circ14.pdf
- preinheimer 3y ago1. The local library at least paid for those books. Which is more than the creators of the current AI generation can say[1]. 2. You can't train people exclusively on books. If you asked someone to explain a concept they learned in those books, sure that knowledge will help, but their response will also affected by every moment in their life so far. They're bringing other unique things to their response. [1] - https://www.nytimes.com/2023/07/10/arts/sarah-silverman-lawsuit-openai-meta.html https://www.nytimes.com/2023/07/10/arts/sarah-silverman-laws...
- tayo42 3y agoIt doesn't seem difficult to me. It seems like the intent is the big difference. Purposefully making a copy of something digitally already has rules around it, which was the second extreme describes