4 ms·

OpenAI is pretty clearly using their work to make derivative content that in certain cases (CNET) is a direct competitor. Honestly, this seems open and shut

nomel 3y ago
> to make derivative content How would they prove this? Is it safe to say that each article used has a nearly meaningless influence on the weights? Could this be used as a defense? Perhaps train a (smaller) model, remove a single article, and show how it doesn't influence performance?
prepend 3y ago
It’s not making a derivative product though. The fact that they compete or not doesn’t matter for copyright. It’s not like it’s ok to violate copyright if you don’t compete. It’s still illegal to take a NYT article and print it on a t-shirt. The issue is that copyright law doesn’t prevent the kind of model training as there’s no clearly derived work. I don’t think that’s been tested in courts yet, but I expect it won’t be found to be copyright because there’s other precedent that influenced is not infringement.
CuriouslyC 3y ago
No more open and shut than the NYT having the right to sue people writing editorial news stories with no new content based on reading their news (along with many other sources). This issue is ultimately going to come down to the transformative clause of fair use. The fact is that the _model_ is unquestionably a transformative product of the inputs, and a judge ruling otherwise is going to cause a cascading shitstorm of litigation and put a chill through the creative economy. The outputs of the model under certain conditions can be guided towards copyright infringement, and any sane ruling will focus on protecting rightsholders from overly derivative model outputs. In all likelihood the precedent will be that the standard for being transformative will be raised for "algorithmically generated" content, and the people who distribute that content will still be fully liable in the event of infringement, with "I didn't know, the AI did it" not being an acceptable defense.
flir 3y ago
> derivative content If I read five calculus textbooks and write a new one, I don't think that's derivative content (or maybe it is?) Seems like that's what an LLM does - read many works, write a new work.
bandrami 3y ago
It's not derivative though. For derivation you have to literally point to sequences of words in the original that are also in the alleged infringer, and those sequences have to be long or unique enough to not be able to come from somewhere else or just common English usage.
- 8note 3y ago
  You can make a derivative work by using the character of Harry Potter in your own book. Fan art and fan fiction are derivative without copying sequences of words

nomel 3y ago

> to make derivative content How would they prove this? Is it safe to say that each article used has a nearly meaningless influence on the weights? Could this be used as a defense? Perhaps train a (smaller) model, remove a single article, and show how it doesn't influence performance?

prepend 3y ago

It’s not making a derivative product though. The fact that they compete or not doesn’t matter for copyright. It’s not like it’s ok to violate copyright if you don’t compete. It’s still illegal to take a NYT article and print it on a t-shirt. The issue is that copyright law doesn’t prevent the kind of model training as there’s no clearly derived work. I don’t think that’s been tested in courts yet, but I expect it won’t be found to be copyright because there’s other precedent that influenced is not infringement.

CuriouslyC 3y ago

No more open and shut than the NYT having the right to sue people writing editorial news stories with no new content based on reading their news (along with many other sources). This issue is ultimately going to come down to the transformative clause of fair use. The fact is that the _model_ is unquestionably a transformative product of the inputs, and a judge ruling otherwise is going to cause a cascading shitstorm of litigation and put a chill through the creative economy. The outputs of the model under certain conditions can be guided towards copyright infringement, and any sane ruling will focus on protecting rightsholders from overly derivative model outputs. In all likelihood the precedent will be that the standard for being transformative will be raised for "algorithmically generated" content, and the people who distribute that content will still be fully liable in the event of infringement, with "I didn't know, the AI did it" not being an acceptable defense.

flir 3y ago

> derivative content If I read five calculus textbooks and write a new one, I don't think that's derivative content (or maybe it is?) Seems like that's what an LLM does - read many works, write a new work.

bandrami 3y ago

It's not derivative though. For derivation you have to literally point to sequences of words in the original that are also in the alleged infringer, and those sequences have to be long or unique enough to not be able to come from somewhere else or just common English usage.

8note 3y ago
You can make a derivative work by using the character of Harry Potter in your own book. Fan art and fan fiction are derivative without copying sequences of words