The Pitfalls of AI from a Writer’s Perspective

by Margery Cuyler

I have written children’s books—more than 60--which fortunately traditional, respected publishers have released in hardcover, paperback, and e-book formats. But with the advent of artificial intelligence (AI), will a machine learn to invent an original, highly individualized children’s story that evolves from a creative spark in the imagination? There are more than 100 YouTube videos that offer step-by-step instructions on how to “create” children’s stories using AI. But what does “create using AI” mean?

Artificial General Intelligence (AGI) can engage in a wide range of tasks using what scientists call “deep thinking.” It can make schedules, organize recipes, write essays, and diagnose medical conditions—all performed more rapidly than a human could manage. As a result, mind-numbing tasks are accomplished with increased accuracy and efficiency. Writers find AGI especially helpful when creating bibliographies, summarizing research, and making an outline. But despite these strengths, AGI has a problem related to copyright that I find gravely concerning. And this is why. To perform varied and sophisticated tasks, AGI amasses vast amounts of text data from the Internet. This data is stored in “Large Language Models (LLMs),” machines trained to organize, categorize, and regurgitate written content when prompted. While collecting written data, AGI has pirated millions of books from hidden online libraries without an author’s knowledge or permission. It has then used these texts to train LLM models. This is a violation of copyright law! In July, a federal judge in the Northern District of California ruled that Anthropic, a tech company that trains LLMs using authors’ works, is doing something “exceedingly transformative” and is therefore protected by Fair Use. Fair Use, as defined by the Copyright Act, focuses on four factors: the purpose of the use, what kind of copyrighted work is used, how much of the work was used, and whether the use hurts the market value of the original work. I recently discovered that four of my books were recruited for LLM training purposes.  As a result, a would-be author could prompt an AGI site like Claude or ChatGPT to produce a children’s Christmas story similar to my book, The Christmas Snowman. AGI could simply amalgamate various pieces of text data to produce a book that seems original but actually relies on copywritten material from the past. For this reason, authors need legal guardrails to be put in place to get copyright protection. Unfortunately, the law tends to lag behind innovation.

Another problem: Authors whose stories are “original” will find their books dwarfed by AGI generated books flooding the marketplace. These books are mediocre at best and do not represent the way a human creates a work from the imagination. So far, AGI has not succeeded in being authentically creative. Instead, it manipulates written data to give the impression is has produced something original. Don’t we owe our children something better than this? They deserve the highest quality that’s available in children’s literature.

Just send them to the children’s section of a library or bookstore!