In September 2023, the nearly five-month strike held by the Writers Guild of America came to an end. Among the new contract’s agreements was the requirement for studios and production companies to inform writers if materials assigned for rewrites had been artificially generated. Going forward, those materials could not count as having been “written” by a “writer.” After all, WGA rates are lower for rewrites than they are for original work, so the agreement effectively protected writers (human writers, that is) from being significantly undercut by claims they weren’t the first people to work on something. However, the door to AI-generated writing remained ajar, with the backstop provision that writers could use such technologies with the consent of companies.
While studios cannot force writers to use AI under the new contract, that doesn’t exactly stop those in charge from having a preference for writers who partake. It’s a slippery slope, to say the least, but what does such a process entail? And is it worth the time and money saved?
Conversations surrounding generative AI (GAI) have engulfed artistic industries, from the aforementioned Hollywood negotiations to music to graphic design. And while the WGA may have scored a major victory (one that workers have the potential to apply across many industries), the technology still looms over the art world, and continues to grow more sophisticated. Soon, spotting fake images, songs, or scripts won’t be so easy. And if Open AI’s ChatGPT can spit out a borderline functional screenplay, the qualitative argument may not be enough to deter widespread usage—even if what an artificial script brings to the table differs from what a human script puts forward.
In October 2024, the staff at filmmaking website No Film School conducted a comparative experiment where readers could judge two competing scripts without knowing which was AI-generated and which was au naturel. The site subsequently published the results, which were closer than expected. With a shared logline about a disillusioned bodyguard and a young pop star, No Film School had its resident WGA screenwriter Jason Hellerman develop the first 10 pages of a screenplay, solo and over the course of a week, while WGA writing duo Matt Allen and Krista Suh completed the assignment over a matter of hours with help from generative software ChatGPT Canvas and Claude 3.5 Sonnet. Hellerman expands on his process in the results article, while Allen and Suh demonstrate their approach of using GAI as a sort of third writing partner in a YouTube video.
To my eyes, one of two samples clearly has more soul (this turns out to have been Hellerman’s), but whether or not you can readily tell the difference at first glance is—or rather, ought to be—besides the point. If it resembles a screenplay, it functionally is one from an industry standpoint. AI-generated writing can therefore be bought and produced by studios increasingly concerned with cutting corners, even if it means shelving nearly completed or wholly completed projects.
The tech certainly hastens the writing process, and as Allen explains in his video, “Writing is rewriting”—the human act of polishing an existing idea is bound to give it at least some semblance of ethos. However, Allen’s repetition of the old adage misses the forest for the trees. Rewriting is a major part of the creative process, yes, but you know what else is writing? Writing.
GAI at its core is inherently at odds with artistry itself, given its inability to produce anything truly new. One look at how it works lays the gambit bare, in that it trains on existing datasets, but it has no perspective of its own, nor does it have personal experiences on which to draw. These are the most fundamental tools writers can use to imbue their work with humanity, rather than a facsimile of it.
However, this distinction alone is unlikely to stop studios—and a tech industry increasingly taken by the ease of GAI—from pushing it as the new frontier. Allen and Suh are, in the most technical sense, correct in their assertions that such time-saving measures are alluring to creatives and executives alike (Netflix CEO Ted Sarandos believes screenwriters who use AI are the future; filmmaker James Cameron, however, rejects the notion).
For personal insights into the original No Film School challenge—and to satiate my own curiosity—I took only a few seconds to generate a similar script on ChatGPT, and all I needed for a second draft was the input “make it funnier.” But when I asked ChatGPT a question I felt was relevant to the bodyguard premise, “Have you ever cared for someone?”—a theme any human writer would hopefully consider—I received a revelatory answer: “Please upgrade to enjoy unlimited access.”
Fair enough.
In truth, I’m not really interested in the answer. The responses these softwares randomize are based on existing materials, drawn from the thoughts and feelings of other people, which are turned into ones and zeros, and reassembled into what a computer believes a person might say (“believes,” of course, is a misnomer, but you understand). The answer therein would not involve a person introspecting in a way that informs their work. The AI would not even draw the connection between the question and the story. Maybe some screenwriters wouldn’t either—at least not consciously—but it would get their wheels turning, perhaps accessing something in their emotional memory.
AI doesn’t have that. It cannot yearn. It cannot doubt. It cannot fear. And it can only mirror these things if specifically asked, because it has no impulses of its own. No wants or desires, except to imitate and approximate. While emotional impetus can certainly be added manually in second or third drafts, the very baseline of a GAI script is that of something aggregated and compiled. It does not have ideas, inspiration, or curiosity, and it does not perform the very labor of artistry that the WGA sought to safeguard with its strikes.
Enshrining protections in a way that puts the ball in the writers’ court is arguably preferable to letting studios dictate the technology’s use. However, the battle between AI and creative endeavors is only in its first act. Entire features are already being produced based on AI-generated scripts, though these have mostly been experiments thus far, and their reception has been poor. When it comes to using the tech for more traditional screenwriting, the temptation is understandable—why agonize over the frustrating part of the process when you can skip to the rewards?—but the time gained comes at the cost of something true.
Whether or not you consider using ChatGPT and the like to be dishonest, the honesty of the art itself becomes compromised, if not outright corrupted. It plays to the tune of an unfeeling machine that, though it might help you organize your thoughts, cannot tell you what or how to think. And it cannot help you learn from mistakes you never make in the first place—a vital step for any artist. None of us enters the world fully formed. We are not data, after all, and the forces that mold us are the very same ones that inform the screenwriting process, arduous though it may be. The result is worthwhile.
