On February 21, 2023, Portal announced that Amazon’s Kindle store – dedicated to buying e-books for its e-reader of the same name – contains at least 200 books for which ChatGPT is listed as author or co-author. The number that represents only a fraction of reality, since the policy of the American giant does not require the authors to mention the use of such artificial intelligence systems.
The origin of the phenomenon is simple: opportunists trying to make a quick buck by asking OpenAI’s text generator to create e-books and then publishing them through Kindle Direct Publishing, Kindle’s self-publishing arm.
A new business… not really profitable
Illustrated children’s books are the most targeted. For example, in the Kindle Store, ChatGPT is behind the fairy tale The Power of Homework, a collection of poems titled Echoes of the Universe, or The Wise Little Squirrel: A Tale of Saving and Investing, a story for children about a squirrel finds one gold coin and discovers an interest in saving and investing. The thirty-page book, written and illustrated entirely using automatic image or text generation tools, will retail for $2.99 for the digital version and $9.99 for the printed version.
Since ChatGPT’s critically acclaimed release last November, hundreds of YouTube tutorials have posted “business ideas” on how to make money with the system, including “How to Write and Publish a Book” in a matter of hours. . But as always, the authors of these tutorials earn more from the audience for their content (and the paid services they promote) than what their ideas bring to their audiences. The author of the book in question has received less than $100 since it was published in January.
“Many authors are becoming unemployed”?
This doesn’t allay the fears of some industry players: “We have to be really worried, these books are going to flood the market and a lot of authors will be out of work,” said Mary Rasenberger, executive director of the Authors Guild, America’s largest professional organization for writers.
The message that ChatGPT will disrupt the book industry is perhaps a bit strong. It “works” in self-publishing, but this system already has many authors who only have moderate success. The majority of readers still prefer to stick to the recommendations of the major publishers. The classic and increasingly widespread pattern is as follows: an author submits his text to the publisher’s reading committee, which decides whether or not to include it in the (often very closed) circle of published authors.
“Authors and platforms need transparency about how these books are made,” continues Mary Rasenberger. On this point she is not wrong, and it is certain that in this case the readers will prefer to buy works that come from the work of a human being.
Clarkesworld had to freeze all new text input
Meanwhile, Clarkesworld, an American magazine specializing in publishing collections of sci-fi short stories, has been inundated with AI-authored content (the publication this month banned more than 500 users from submitting content suspected of being… stand to be supported by the AI) and had to freeze all new submissions of text on its website. Clarkesworld is one of the few publishers willing to publish and pay for texts by as-yet-unknown authors (12 cents a word), making it a prime target.
Its editor, Neil Clarke, is considering several ways to counteract the problem, such as introducing paid submissions, although he assures that he would regret to see these “crook-imposed changes not result in an increasing number of obstacles.” for new authors”.
A difficult plagiarism problem to regulate
Using ChatGPT, Google Bard and other AI-based text generators for commercial purposes is primarily a plagiarism problem. Without imagination, these systems only use content created by people without the knowledge of the original authors, or at least without their permission.
The most practicable solution then remains to prescribe a clear indication of the source. Also, note that Article 3 of OpenAI’s Terms of Use specifically mentions the transfer of all rights to the user (ChatGPT texts therefore belong to the person who created them), but that “they prohibit presenting the content as if it would be created by a human being. Which is not respected a priori and in any case very difficult to see since they overlap, combine and even paraphrase multiple sources …
Selected to you
