No offense intended, but I can't help wondering how AI is any different in principle from a young author learning to write by reading other authors. Face it, copyright is broken as a means of earning a living. I don't know what the fix is, but I don't think railing against AI is helpful. - ThreadSky

ronwebb3.bsky.social • 19 days ago

No offense intended, but I can't help wondering how AI is any different in principle from a young author learning to write by reading other authors.
Face it, copyright is broken as a means of earning a living. I don't know what the fix is, but I don't think railing against AI is helpful.

Comments

lpersonius.bsky.social•19 days ago

One problem, the AI did not buy or check the book out from the library, they stole it from authors. No consent or compensation.

Another problem is, with Citizen United, we've seen what happens when we give a non-human personhood rights. It doesn't end well for the humans, except for the uber-rich.

ronwebb3.bsky.social•18 days ago

2. Regarding rights, I hope we’re not giving rights to machines (yet, anyway). I would say the developers of the AI model are the ones whose rights are in question. If there were a lawsuit over copyright, I’m sure they would be the defendants.

ronwebb3.bsky.social•18 days ago

1. Good point. AI was originally supposed to be trained on public domain sources, but I’ve heard that other data was used, perhaps illegally. I’m not sure how much additional compensation each author would have received if the books had been checked out from a library though.

lpersonius.bsky.social•18 days ago

They are compensated somehow for library inclusion, and also for subscription models like Kindle Unlimited, around <$2 per read depending on length which isn't much. So, I think a couple bucks per ebook is not outrageous to compensate authors who knowingly sign up for that, which isn't happening.

anna-louise-ellis.bsky.social•19 days ago

Further it’s taken the Machine Learning papers we ML scientists write and has used them to advance its architecture. Meta have also refused to sign up to the AI Security Institute meaning, unlike the 60+ countries that did (including UK - bar one part about sharing resource), there are 0 guardrails

ronwebb3.bsky.social•18 days ago

Sounds like you know more about this than me. 😄 Is it not usual for knowledge to build on previous work? Why would they need your consent to use your ML papers?

I know nothing about the AI Security Institute, but that sounds like a separate issue.

Comments

Posting Rules

Reply