Authors demand credit and compensation from AI companies using their work without permission | OpenAI, Alphabet, and Meta have been called out::The letter, published by professional writers’ organization The Authors Guild, is addressed to the bosses of OpenAI, Alphabet, Meta, Stability AI, IBM, and Microsoft. It calls out…
To date, it remains legal for humans to borrow a book from the library, read it, learn skills & knowledge from it, and apply what they’ve learned to make money — without ever paying the author or publisher.
Copyright does not, in general, grant control over the ideas in a work; only their specific expression. It deals with copying the text; it is not a tax on the information or knowledge contained in that text.
It also does not assure the author or publisher of a share of all revenues that anyone is ever able to make using the knowledge recorded in a work.
Yes, but it isn’t legal to download a repository of pirated books to read and learn from.
Did OpenAI check out the books they trained their model on from the library one at a time?
I’m generally very much against the copyright creep that’s being advocated by some trying to get training to be infringement.
But at the same time, OpenAI should have at least needed to buy retail copies of books they were using to train the AI on, or getting access through legal means, and one of the chief allegations against them was that they effectively built the AI on the open seas of piracy by using a data set that contained copyrighted content illegally distributed.
So the overreach by copyright holders to claim rights to training is BS, but there may well still be a valid claim against OpenAI in how they went about it.
Even if I illegally download a book about carpentry, and learn how to build a doghouse from it, and go into the business of building doghouses, the extent of my liability to the copyright holder does not include my entire doghouse-building revenue.
Even if I subsequently teach other people to build doghouses, if I’m not further copying the actual contents of the book, I am not further liable for copyright infringement.
Copyright is actually pretty narrow, and should not be construed to give authors or publishing companies unbridled control over the ideas or knowledge contained in works.
Yes, but you are liable for damages in having pirated it.
Where did I say anything about being liable for future revenue?
But it’s a special level of dumb to build a billion dollar company on material that you pirated and can be confirmed to have possessed and used by your end product.
Suits trend to have multiple claims trying to get the plaintiffs as much compensation as possible. Even if all the crap about training as infringement gets thrown out (as it should), claims OpenAI committed one of the largest copyright infringements in recent history by obtaining and using pirated material in violation of copyright law is likely going to have hefty damages attached if it can be proved (which it will be if it happened).
If you downloaded music from Napster and got caught in the early 2000s, did the MPAA fine you got only the retail price of the song?
If you illegally downloaded a book about carpentry, and get caught, do you think you don’t have to pay anything for having illegally downloaded it?
Yes, but you are liable for damages in having pirated it.
Sure, if someone can show that you did.
Based on my own experimentation, ChatGPT knows facts about the Harry Potter novels, but it does not recite the text of them when asked to do so. Does it contain a pirated copy of them? I can’t tell. Maybe it just reads a lot of open-source fanfic off AO3.
You are not logged in. However you can subscribe from another Fediverse account, for example Lemmy or Mastodon. To do this, paste the following into the search field of your instance: !technology@lemmy.world
This is a most excellent place for technology news and articles.
To date, it remains legal for humans to borrow a book from the library, read it, learn skills & knowledge from it, and apply what they’ve learned to make money — without ever paying the author or publisher.
Copyright does not, in general, grant control over the ideas in a work; only their specific expression. It deals with copying the text; it is not a tax on the information or knowledge contained in that text.
It also does not assure the author or publisher of a share of all revenues that anyone is ever able to make using the knowledge recorded in a work.
The lines between a specific expression and the idea behind them are very blurred with AI…
Yes, but it isn’t legal to download a repository of pirated books to read and learn from.
Did OpenAI check out the books they trained their model on from the library one at a time?
I’m generally very much against the copyright creep that’s being advocated by some trying to get training to be infringement.
But at the same time, OpenAI should have at least needed to buy retail copies of books they were using to train the AI on, or getting access through legal means, and one of the chief allegations against them was that they effectively built the AI on the open seas of piracy by using a data set that contained copyrighted content illegally distributed.
So the overreach by copyright holders to claim rights to training is BS, but there may well still be a valid claim against OpenAI in how they went about it.
Even if I illegally download a book about carpentry, and learn how to build a doghouse from it, and go into the business of building doghouses, the extent of my liability to the copyright holder does not include my entire doghouse-building revenue.
Even if I subsequently teach other people to build doghouses, if I’m not further copying the actual contents of the book, I am not further liable for copyright infringement.
Copyright is actually pretty narrow, and should not be construed to give authors or publishing companies unbridled control over the ideas or knowledge contained in works.
Yes, but you are liable for damages in having pirated it.
Where did I say anything about being liable for future revenue?
But it’s a special level of dumb to build a billion dollar company on material that you pirated and can be confirmed to have possessed and used by your end product.
Suits trend to have multiple claims trying to get the plaintiffs as much compensation as possible. Even if all the crap about training as infringement gets thrown out (as it should), claims OpenAI committed one of the largest copyright infringements in recent history by obtaining and using pirated material in violation of copyright law is likely going to have hefty damages attached if it can be proved (which it will be if it happened).
If you downloaded music from Napster and got caught in the early 2000s, did the MPAA fine you got only the retail price of the song?
If you illegally downloaded a book about carpentry, and get caught, do you think you don’t have to pay anything for having illegally downloaded it?
Sure, if someone can show that you did.
Based on my own experimentation, ChatGPT knows facts about the Harry Potter novels, but it does not recite the text of them when asked to do so. Does it contain a pirated copy of them? I can’t tell. Maybe it just reads a lot of open-source fanfic off AO3.
I’m starting to realize several people in this thread don’t understand how subpoenas work.