A_Very_Big_Fan comments on In Cringe Video, OpenAI CTO Says She Doesn’t Know Where Sora’s Training Data Came From

[–] A_Very_Big_Fan@lemmy.world 0 points 2 years ago (2 children)

It copied all of its text from the article, and it allows me to get all the information from it I want without providing that publisher with traffic or ad revenue. That's not fair use.

I do like the bot, and personally I'd rather it stay, but no matter how you look at it this isn't "fair use" of the article.

permalink

fedilink

source

parent

hideshow 2 child comments

[–] Guntrigger@feddit.ch 0 points 2 years ago (1 child)

Interesting take. In all of the defences of LLMs using copyrighted material it's very often highlighted that "fair use" allows exactly such summaries of larger texts.

In reality, "fair use" is ruled on a case by case basis, so it's impossible to judge whether something is or not without it going to court.

permalink

fedilink

source

parent

hideshow 1 child comment

[–] A_Very_Big_Fan@lemmy.world 0 points 2 years ago

We're not making legislation here, so we don't have that level of burden of proof. But either way, when it comes to factors of fair use that every authority on the matter will list, it violates almost all of them.

It's non-commercial, and it's using facts rather than using a more creative work, so it's got that going for it... But it's

composed of 100% copied material
it's not transformative
it's substituting the original work
it uses officially published work
it specifically copies the "heart" of the work
it bypasses all of the ads and impacts their traffic/metrics so it has a financial impact on them.

It's pretty obvious that there is no argument here. The factors that are violated the hardest and most undisputably are the ones that most authorities on the matter (including the one I linked) agree are the most important.

permalink

fedilink

source

parent