2023-09-21 17:33:42
Famous writers have filed a lawsuit against the developer company ChatGPT. Among them is Game of Thrones author George R.R. Martin. They claim that Open AI trained its language models on pirated versions of books and now plans to make billions of dollars. According to the plaintiffs, this practice massively violates copyrights. 17 writers have joined the class action against the ChatGPT developer, including, in addition to Martin, Jonathan Franzen and John Grisham, as well as the American Guild of Authors.
In addition, some plaintiffs found generated books on marketplaces that were signed with their real names, CNN columnist Claire Duffy reported:
“The scale of the problem will grow. Nowadays you can generate text very cheaply and easily. For example, in the style of a specific author or book. Moreover, some people may deliberately sell works created by artificial intelligence, but signed with the names of real writers. This is the situation Jane Friedman found herself in. She recently discovered several books purporting to be hers on Amazon that she actually has nothing to do with. Someone was just trying to make money off her name and reputation.”
Does such a claim have any prospects? The CEO of the developer of artificial intelligence systems “A-Z Expert” Roman Dushkin believes that it will not be easy to prove that language models learned from pirated copies of novels:
“We need to prove what is written in this lawsuit. On what basis is it submitted? Was an illegal copy actually used? If they have clear evidence, then they are probably right and can recover something from OpenAI. But if they file a lawsuit based only on some fabrication of their own, that the company used pirated versions of their books, good luck to them in court.
How the model was trained, what weighting coefficients appeared on the synapses between its neurons, is absolutely impossible to understand. No one can prove whether she learned from the pirated version or not. And even if you ask a model whether she was trained on a pirated version of George Martin’s books or not, the answer cannot be trusted. The AI will invent it.”
As the lawsuit emphasizes, the writers do not want to stop the development of artificial intelligence. They believe that they should have the right to prohibit the use of their work. Or oblige the company to pay royalties.
The fears of American authors cannot be called groundless, but for now, training language models using such examples does more good than harm, says writer Dmitry Zakharov:
“When I read a book by Vyacheslav Krapivin or Jules Verne, do I borrow something from there? I’ll probably borrow it. Somewhere it’s free, but somewhere it’s not. Here, it seems to me, the situation is similar. A neural network cannot create original works in the form in which we understand it. So far, anything that helps educate people and train neural networks falls into this category seems positive to me. We’ll see how the situation develops further. And colleagues are more likely to blow on water where no one is expecting milk yet.”
Open AI admitted that it used copyrighted content, but did not specify what kind of content it was.
Everything is clear with us – Telegram channel “Э FM”.
#Writers #trial #Kommersant