@j what a nonsense argument lol, you're copyright infringing if you re-share or reuse the data. seeding is de-facto copyright infringement. but if you turn around and use the data you downloaded in a product, you're back to infringing again, just in a different way
for the record I don't give a shit about copyright but their argument is silly
@sun@vokainen099@j it’s not obvious in a legal sense, the output of a training session isn’t copyrightable (because it’s the output of an algorithm) so using copyrighted material as input might not be a copyright violation in itself. (This is the angels on a pin stuff that IP law leads to)
@vokainen099@j yeah I am saying, seeding is what got you caught. Meta is saying they didn't seed (or not much). but they are releasing a model based on that data so obviously they are still infringing even if they didn't seed
@lain@vokainen099@j that's fair, that is downstream of the terrible seeding argument though. They're lawyers though, obv they are going to try any argument they can