Isn't the narrative that AI does not infringe on copyright literally, demonstrably false? One of the 'goodness' measures of AI training is the loss function, meaning its ability to reproduce the training set. If it has low loss, and some copyrighted content is included in its training set, like the Harry Potter novels, it is by definition able to reproduce the novels contents more or less verbatim, perhaps with some minor inconsistencies. This is like storing a compressed version of a movie on my server, and claiming that it doesn't infringe copyright because the copy is lossy.
The fair use argument is retarded.
Memory is an infringement
You shouldn't be able to memorize and repeat any copyrighted work no matter how small without paying a fee
Yes exactly. Storing a copy of a copyrighted movie/book is infringement, even if I zip/recompress the movie.
/this: if I quote piece of text or prose verbatim from memory am I infringing the authors copyright by keeping those words in muh head & then saying them out loud?
what if I skip a word - can I then claim the work as muh own?
People trying to bring copyrights into AI are literal fucking garden gnomes. It's a slippery slope and they want to monopolize something that should be free for all of humanity. There are 2 outcomes to this basically.
1. You can train AI on anything and use it to create anything. Basically it's PS on steroids, easy to use for anyone with an internet connection. This is where we're at now and it makes perfect sense since anyone can use it, we're all equals. Artists HATE this because they want to stand above everyone and make as much money as possible on their art.
2. FOSS models are deleted from the internet and the remaining models are cucked to hell so grifting artists can safely be garden gnomes now and forever. Ideally they want to have rights to their own styles which would eventually mean they'll just start suing other artists for copying them. Entire anime, cartoon and comic industries will have people assmad that other people, not just AI users, are getting inspired by their styles and making similar things. Art will die because it's always been a game of inspirations and copy cats, a global iterative stream of creativity. But now you can't get inspired and make transformative work, AI was bad because umm.. so yeah.
Artists actually want the latter. It's crazy.
copyroght laws simply shouldn't exist
Post proof of it writing Harry Potter verbatim
if i personally can recount harry potter verbatim, am i infringing copyright by "storing" it in "memory" ?
or is it only an infringement when someone pays me to read them harry potter and i read it entirely from memory?
the anti-ai arguments are so retarded
What sort of false equivalency is this? AI is not a person, its a sequence of bytes on your disk. If there is a sequence of bytes called harrypotter.zip containing all the books you are infringing copyright.
The method of encoding/decoding said bytes does not matter.
and there is a sequence of electrochemicals in my brain storing the same data
if the method of encoding/decoding the data doesn't matter, surely the storage medium doesn't either.
digital copyright is a fucking joke
by this logic, a drawing with some lines and numbers on it counts as a firearm and you should be arrested for owning it without a license
the method of encoding and decoding doesn't matter, so long as the end result is the thing? well if I encode my AR-15 into a technical drawing and then decode it using a block of steel and a CNC mill, that drawing is considered a firearm weather or not I reproduce it?
you are fucking retarded
At some point an AI will have to have copyrighted stuff in its training data. I don't think an AI without that data could pass a Turing test where any of the questions related to copyright data. Like you said though it should never present copyrighted data verbatim to anyone interacting with the AI, it should be able to describe the data without just reciting it, within reason, as a human would
>, it is by definition able to reproduce the novels contents more or less verbatim
It isn't "by definition" anything because you didn't define the loss function, retard. You can train the model to output whatever. You can train it to specifically output something that looks like random garbage and never lines up with the original.
Stop running your mouth about shit you don't understand.
either it is you who doesn't know shit, or you are just nit-picking with hypotheticals and arguing in bad faith. In supervised learning, the loss function always measures how similar the output is to the training data.
Copyrights need to be abolished, it's currently destroying media preservation efforts, reducing innovation with patent trolls and only enforceable if you're a super large corporation.
doesnt it get old to call people trannies and to tell people to kill themselves? like i've seen it a thousand times, its just boring, its not even funny
You will never truly be accepted. People who pretend to like you are secretly repulsed when being in your presence.
The best Cortana.