More than 8,000 authors have asked not to train neural networks on their texts without payment

More than 8,000 authors appealed to AI-developers with a request not to use their work for training neural networks without permission, mention and compensation.

In the letter of the Guild of Authors of the USA, it is emphasized that millions of books, articles, essays and poems serve as the basis for language models, including ChatGPT and Bard chatbots. However, the authors of the works did not receive anything for their contribution. Companies like to say that their neural networks simply “read” the texts they are trained on, the letter says. However, this is inaccurate anthropomorphization: the models copy the texts into the software itself, and then reproduce them again and again.

“It will be fair if the authors receive compensation for the fact that they “fed” artificial intelligence and continue to contribute to its evolution,” said the head of the Guild of Authors, writer Maya Lang. She stated that the texts cannot be used without the author’s consent, mention and compensation.

The letter also notes that over the last decade, writers’ incomes have decreased by 40% – and the emergence of neural networks only exacerbates this.

The Authors Guild of the USA sent a letter to Meta, IBM, Microsoft and OpenAI. In the latter, The Wall Street Journal responded that the company respects copyright and that ChatGPT is trained on “licensed and publicly available content, as well as texts created by AI trainers and users.” Other corporations have not yet responded to the WSJ.

• Hollywood actors and screenwriters are also on strike in the USA. They demand guarantees that their work will not be replaced by artificial intelligence, as well as an increase in the basic salary and deduction from repeated showings of films.