News for hackers. Stuff that matters.
Re:OpenAI's justification for why training data is fair use,... by wordpad25 on Thursday October 05, 2023 @12:18PM (#37780518)
One could easily argue derived work is sufficiently different from the original to avoid a violation.

ChatGPT doesn't just steal and slightly modify the original. It fully copies the style and feel, which aren't copyright protected.

You can't just change the color of Mikey Mouse, call him AI Mouse and be good, but you can certainly create a new Disney-style character without triggering copyright violations.

Re:OpenAI's justification for why training data is fair use,... by sleepybrett on Thursday October 05, 2023 @12:34PM (#37780737)
https://creativecommons.org/2023/03/23/the-complex-world-of-....

"Style and artistic expression are certainly linked, but they are not the same thing. In some instances, mimicking an artist’s style may constitute copyright infringement, but when this is true will necessarily require a case-by-case analysis. Simply working in a style that is similar to another artist does not automatically mean the work is a copy. Moreover, there may be legitimate reasons to produce works in another artist’s style. The law should allow artists to be inspired by each other, even if that means having similar styles. Doing otherwise would chill expression, ultimately harming human creativity and the purposes of copyright law itself."

Re:OpenAI's justification for why training data is fair use,... by cj on Thursday October 05, 2023 @12:25PM (#37780614)
> It fully copies the style and feel, which aren't copyright protected.

Is that for certain? I have a feeling this point could be debated.

Re:OpenAI's justification for why training data is fair use,... by Sohcahtoa82 on Thursday October 05, 2023 @12:39PM (#37780797)
"style and feel" are entirely subjective.

If I give DALL-E or MidJourney a picture of myself that I took (and therefore I own the copyright), and ask it to "Recreate this picture in the style of Pixar", am I violating Pixar's copyright?

What if I decide to model it myself and deliberately imitate Pixar. Have I violated copyright?

What if I decide to model it myself, and it just happens to look similar to Pixar, because I've watched a lot of their movies and it has influenced how I model people in 3D animation. Have I violated copyright?

Re:OpenAI's justification for why training data is fair use,... by cj on Thursday October 05, 2023 @12:55PM (#37781010)
What if an AI model took the Barbie movie and added a filter to change the hue of every pixel and altered the voices to have Australian accents instead of American accents?

I wonder if Disney would sue if you tried to sell and distribute despite 100% of the pixels and audio file not exactly copying the original.

(Rhetorical question, Disney would definitely sue)

Re:OpenAI's justification for why training data is fair use,... by greggsy on Thursday October 05, 2023 @01:21PM (#37781308)
I think early YouTube movie uploaders tried this angle but ultimately failed.
Re:OpenAI's justification for why training data is fair use,... by nickrubin on Thursday October 05, 2023 @01:20PM (#37781291)
The answer to all of your questions is no
Re:OpenAI's justification for why training data is fair use,... by agloe_dreams on Thursday October 05, 2023 @01:29PM (#37781404)
Question - Does it knowing what the style of Pixar is and knowing to read it constitute trademark infringement instead?
Re:OpenAI's justification for why training data is fair use,... by dcow on Thursday October 05, 2023 @12:31PM (#37780699)
Exactly. The court would consider other things like the impact of your modified copy and whether it competes financially with the original.