Not that I’m aware of.Are there any rules about AI companies mining online information for training their software? Do they need permission?
Not that I’m aware of.Are there any rules about AI companies mining online information for training their software? Do they need permission?
Based on the uproar over Ai "art" generators, i would say no.Are there any rules about AI companies mining online information for training their software? Do they need permission?
Well that's wierd.
Well the last time they downloaded the Urban dictionary it didn't turn out wellEhh, not really. Forum discussions are widely used to train language models; they provide a gargantuan supply of conversational text. If you sit down to study natural language processing, the first exercise generally involves downloading a bunch of Reddit.
Woah. EN World is in the top 0.1%!Well that's wierd.
The jury is still out. Literally. There's a class-action suit pending right now over the use of artists' work without permission. Note that the suit is not just about the training process; the plaintiffs claim it is possible to get Stable Diffusion-type generators to actually reproduce their copyrighted work -- not perfectly, but recognizably.Are there any rules about AI companies mining online information for training their software? Do they need permission?
I think the plaintiffs have a decent shot of having that hold up. I’ve seen AI generated speeches get delivered; essays answered. And AFAIK, it’s not part of the suit (yet), but things like “AISIS” are damn close to what they purport to emulate.Note that the suit is not just about the training process; the plaintiffs claim it is possible to get Stable Diffusion-type generators to actually reproduce their copyrighted work -- not perfectly, but recognizably.