If you’ve ever read a book with an ounce of romance and happily ever after, this story’s ending should feel familiar to you. Who are the characters? It doesn’t matter. Who is the author? We’ll never know! The book, when posted on Amazon, would say I am the author—but I’m not, not really. That entire paragraph was written by ChatGPT after I prompted it to “Write a short paragraph as if it were the last paragraph of a book where the two love interests end happily ever after and the reader closes it and smiles.” Granted, I edited it down and added a comma here or there, but my brain made up exactly zero sentence structure or word choice in creating the lovers’ walk into the sunset. And that’s a problem.

Hollywood Creatives Battle Artificial Intelligence

Two major Hollywood unions are currently on strike, with a portion of their demands revolving around the use of artificial intelligence (AI) in the creative arts. The Writers Guild of America (WGA) went on strike against the Alliance of Motion Picture and Television Producers (AMPTP) on May 2, 2023, with unmet demands including increased pay, viewership-based streaming residuals, staffing requirements for writers’ rooms, shortened exclusivity deals, and safeguards for the use of AI. Similarly, the Screen Actors Guild-American Federation of Television and Radio Artists (SAG-AFTRA) began its strike against the AMPTP on July 14, 2023, after studios, including the likes of Netflix, Disney, Amazon, Apple, Warner Bros., and others refused to budge on performers’ share of streaming revenue, mandatory minimum rates, pension and healthcare contributions, and control of AI.

Though wages, healthcare, and the implications of streaming on the fair sharing of revenue are undoubtedly important, the issue catching the eye of many both inside and outside the entertainment field is the polar opposite sides businesses and creators have staked out on the use of AI in creative work. It seems that for many, this is the hill they are willing to die on. Will AI soon be dancing on the graves of human creatives? Or will people win the robot war? The answer has implications reaching much further than just Hollywood.

As demonstrated by the opening to this piece, AI can be trained to write emotionally moving sentences in the style of almost any current or past writer. As someone who has not actively sought the help of AI in my professional or personal life (I don’t even use Siri, and those home assistants you talk to freak me out), I truly wasn’t sure how good of a job it would do when prompted. So, I sought to find out what the fuss was about.

ChatGPT from OpenAI is a user-friendly, artificially intelligent software with an automated brain of information available up through early fall 2021. I asked it to provide me with a list of its sources, but it was unable to, stating that it had no specific list of sources as it was “trained on a mixture of licensed data, data created by human trainers, and publicly available data from the internet … [by] crawling the web, accessing books and articles, and using a wide range of texts to create a comprehensive language model. However, since ChatGPT doesn’t have direct access to its training data or know the specifics of which documents were included, it can’t provide a list of sources it was trained on.” How convenient.

Because this software is artificially intelligent, it cannot tell truth from a lie. A disclaimer is given about inaccurate or misleading information, which could create offensive or biased answers, but that is for the user to parse. For the software, everything on the internet is true. And not only is it all true—it’s free for the taking.

As part of its requested safeguards for AI, the WGA demands included that AI “can’t write or rewrite literary material; can’t be used as source material; [and] can’t be used to train AI.” For SAG-AFTRA, the AI issue swirls around performers’ consent to having their likenesses used to generate new performances without compensation and the potential elimination of the need for background actors, since digital replicas could be added faster and cheaper in post-production all while the person whose face is in the background can’t pay their rent. The AMPTP rejected the WGA’s demand and is quibbling over the accuracy of SAG-AFTRA’s interpretation of its offer. Despite the muddiness, it remains perfectly clear that it will be extremely difficult to constrain a technological tool that is evolving more rapidly than people can write legalese to control it.

Publishing AI

Hollywood is not the only place AI is being implemented in the arts. The book publishing industry has been besieged with AI use for months, and concern about what that means for authors and illustrators is newly illuminated by its spread into other high-profile creative mediums.

In December 2022, Ammaar Reshi self-published a picture book on Amazon, which is not out of the ordinary. What was different about Reshi’s book was that he used ChatGPT to write the narrative and Midjourney, an AI art engine, to generate the illustrations. Even though the story was average and the character’s appendages more closely resembled claws than fingers, the book has sold copies, and Reshi stated to TIME that he planned to donate copies to his local library.

Although Reshi credited ChatGPT and Midjourney as author and illustrator, respectively, on the Amazon listing, there is no current requirement to do so. As of this writing, there are more than 1,000 books for sale on Amazon that credit ChatGPT as a co-author.

More commonly, ChatGPT and other AI software are used by authors to come up with writing prompts, brainstorm, organize their thoughts, outline narratives, or edit existing work. However, it begs the question whether there is a difference between using AI to get the creative juices flowing and ripping off another person’s work when the information drawn upon to create the prompt was likely used without the original creator’s knowledge or consent—since anything uploaded to the internet could potentially be used to train AI.

In July 2023, The Authors Guild created a petition, now signed by more than 10,000 authors, vocalizing the importance of gaining an author’s consent before their material is used to train AI, and if used in such a way, the author must be compensated for it. AI businesses argue that all material is derivative because “to speak is to imitate,” and therefore feeding the machine does not equate to copyright infringement. The Authors Guild maintains that AI businesses are unfairly anthropomorphizing a machine that does not read—it reproduces.

The MARC 100 Field Says ChatGPT?

The ethics are murky at best, but what does it have to do with libraries? At their core, it is the mission of public libraries to provide open, equitable access to materials. Libraries also have a duty to provide accurate, quality information to the best of their ability. Since AI has no regard for accuracy or bias, how does a ChatGPT-authored or co-authored book fare against a collection development policy without the book being read cover to cover? In a public library landscape often dominated by patron-driven/demand-driven acquisition, if libraries refuse to purchase AI-authored material, is that censorship? Furthermore, AI-written material is not copyrightable under current U.S. copyright law, so the reproduction of these books is unlimited.

The vast majority of authors have historically been some of the biggest champions of public libraries. Librarians have gone to bat for users’ rights to access materials, usually against corporate rightsholders, while the authors themselves are firmly allied with the library. Is it an unwritten duty of librarians and libraries to now go to bat for authors and the rights of creators to create? Even though we are in the era of libraries without borders, in the end, what is a library without quality materials to circulate?

An apt phrase in political psychology states that an entity probably doesn’t tell you what to think, it tells you what to think about. And I can’t help but believe that libraries are the same. We don’t tell patrons what to read or how to think, but our purchasing decisions, especially in small and rural libraries with limited budgets, do tell them what to think about—or at least what we are thinking about. If public libraries use limited financial resources to purchase materials written or illustrated by AI with unclear attribution, we might not be making a statement about their content, but we certainly are making a statement about their legitimacy on the library shelf.

The U.S. Constitution and its protections are foundational for libraries. The First Amendment protects the “right to speak and publish as well as the right to read and hear,” and the Copyright Clause (Article 1, Section 8, Clause 8) explains its ultimate goal of advancing knowledge and creativity to promote the progress of the arts and give creators the exclusive right to their creations. Without exclusive rights to our human ideas, there is no incentive to make our creations public and allow others to benefit from our unique, sometimes even revelatory, thoughts. As AI-generated works multiply, libraries should proactively take a position instead of idling in neutral.

The modern world is a content culture. We now consume content like we live and breathe; it’s a part of being alive. The assumption is, for better or worse, that there is a person behind that content. We take for granted that creators are people. Soon, even if we wanted to find out the origins of a book, article, movie, or song, we may not even be able to tell whether it was created by human or machine. Constraints on AI assume people are reasonable and diligent, and that is likely not a fair assumption given how the humans running movie studios won’t even negotiate the point.

It is of paramount importance that we continue to care about the person behind the creation if we wish to live in a world where truly new, inventive ideas exist. After all, even Frankenstein’s monster was named after the doctor himself.