‘Impossible’ to create AI tools like ChatGPT without copyrighted material, OpenAI says::Pressure grows on artificial intelligence firms over the content used to train their products
‘Impossible’ to create AI tools like ChatGPT without copyrighted material, OpenAI says::Pressure grows on artificial intelligence firms over the content used to train their products
I don’t think it’s accurate to call the work of AI the same as the human brain, but most importantly, the difference is that humans and tools have and should have different rights. Someone can’t simply point a camera at a picture and say “I can look at it with my eye and keep it in my memory, so why can’t the camera?”
Because we ensure the right of learning for people. That doesn’t mean it’s a free pass to technologically process works however one sees fit.
Nevermind that the more people prodded AIs, the more they have found that the reproductions are much more identical than simply vaguely replicating style from them. People have managed to get whole sentences from books and obvious copies of real artwork, copyrighted characters and celebrities by prompting AI in specific ways.
To be fair, I think your analogy falls apart a bit because you can in fact take a picture of pretty much any art you want to, legally speaking.
You can’t go sell it or anything, but you are definitely not in breach of copyright just by taking the picture.
That’s a rebuttal on the level of “if a tree falls in the forest and nobody is there to hear it”. Legally, theoretically, you should need permission just as much, but nobody is going to sue you over something nobody else sees.
Copyright addresses reproduction and distribution, paid or not, including derivative works. There are exemptions for journalism and education, AI advanced a lot by using copyrighted materials under the reasoning that it was technological research, but as it spun off into commercial use, its reliance on copyrighted materials for training has become much more questionable.
That’s not a thing. There is a right to an education, but that is not about copyright (though it may imply the necessity of fair use exceptions in certain contexts).
Also, you are confused about AI output. It’s possible to make the AI spit out training data, but it takes, indeed, prodding. It’s unlikely to matter by US law.