How one screenwriter finds AI useful (and useless)
My union, the Writers Guild of America (WGA), is on strike over a host of issues, one being a proposed set of rules governing the use of AI. When negotiations fell apart nearly a month ago, here’s where things stood:
WGA proposal: Regulate use of artificial intelligence on MBA-covered projects: AI can’t write or rewrite literary material; can’t be used as source material; and MBA-covered material can’t be used to train AI.
Studios’ response: Reject proposal, counter with offer to hold annual meetings on the state of the technology.
Take a moment to appreciate the studios’ audacity. Holding annual meetings on the state of AI would be like trying to paint a still life of a moving train. They sang the same tune in 2007 when the WGA sought participation in streaming revenue. Then, as now, the studios claimed to have no idea how one might distribute content over the Internet and offered annual meetings on the subject. We struck. Within days of that strike ending (after they conceded to our demands), Hulu went live to the public. Today, their flat rejection to discuss AI reveals the whole game: they intend to employ it without restriction to cut costs and eliminate jobs.
I see their logic. If hiring a writer costs x in dollars and y in time, then using AI would lower both. This is the logic of optimization, the newest euphemism for cannibalism, and if you don’t think it’s coming for you, you’re fooling yourself. I have already shown how well AI can do the job of the very people itching to replace me.
I can also appreciate that many regard the act of writing as either tedious or terrifying, and don’t understand why writers wouldn’t welcome the digital assist. But this is our medium. We sweat and bleed for it. Trust me on this one point: writers don’t write to get rich or become famous (there are easier ways to do both); we write because we have to. It’s what keeps us human. So when our employers telegraph their intent to cut costs at the expense of our humanity, we get… edgy.
To those crying “buggy whip!” I concede the AI genie is out of the bottle. So how might writers learn to cohabitate with AI, or (gasp) use it to their advantage? A 2021 Harvard Business Review article posited that the coming AI revolution would result in what it called “Augmented Intelligence”, a system in which computers would perform the analytic drudgery of a given job, freeing humans to engage in higher-order thinking. It’s the pocket-calculator theory: delegate the arithmetic so the mind can ponder the math.
That theory has its limits when applied to writing. For me, writing is both arithmetic and math. Heading into a first draft, I might know the premise and basic outline of the story, and the themes I hope to explore (i.e. the math). But the process of writing word after word, sentence after sentence (the arithmetic) reveals the emotional arc of the piece in ways that cannot be predicted or denied. If you have ever cried at a movie, or laughed, or felt anything, it was because of the arithmetic. Therefore, the idea of using AI to gin up a first draft which I would then edit not only sounds like soul-sucking torture (I’ve rewritten plenty of other people’s bad scripts — it’s painful), but it would actually take longer than writing from scratch because I’d have to spend additional time deciphering the machine’s arithmetic.
Nor would it be wise for me to feed an AI one of my scripts and ask for notes. First there are the legal issues of handing over copyrighted or proprietary material. If I were hired to write a movie for Marvel, say, and then shared it with ChatGPT, I’d be in breach of contract. Next quandary would be the uncomfortable reality that I’ve helped train the AI to write in my style, which would enable Marvel to “hire” me again without actually paying me. Finally — and this is a big point — I don’t want an AI’s notes. If I want to improve my script, that is, make it more appealing to human beings, then I should probably seek feedback from human beings.
All that said, can tools like Midjourney and ChatGPT be a kind of pocket calculator for a writer? I’ve been playing with both for a few months now, and so far I’ve found two legitimate uses.
Research assistant. For a recent project I needed to identify a number of novels in the public domain. I fed my parameters into ChatGPT and got back as many titles as I could ever desire. No ads, no visual noise, no hunting for nuggets of information while dodging pop-ups. For the novels I was most interested in, I asked ChatGPT to generate 250-word summaries. I have tried to use Google for this exact task in the past, and got nowhere. After thirty minutes, I’d identified a number of relatively obscure titles that fit my needs, which I then sourced from my local library.
Did this replace a human? I suppose I could have hired an assistant to do this, if I had the resources or inclination (I had neither). I could have consulted a librarian, though “consulting a librarian” feels more aspirational than practical. I could have wandered the stacks on my own, letting fate guide the way — a supremely human experience — but I was pressed for time.
As a research assistant, ChatGPT feels like the concierge at Hotel Wikipedia, meaning you can ask it for anything, and it will deliver (mostly) reliable information with minimal fuss and a here-to-help attitude. Since every project requires some amount of research, I can easily imagine myself turning to ChatGPT in the future to at least point me in the right direction.
Visual Aids. When Covid forced pitch meetings onto Zoom, I and many other writers turned to Powerpoint to help sell our ideas. I know some writers who loathed this, preferring the energy of the room and their own ability to talk through a pitch. I welcomed the change. I tend to ramble when I get nervous, and Powerpoint kept me on track. I filled my slides with photos (culled from Pinterest) to communicate mood, scale, and tone, much like a vision board.
The photos I culled weren’t exactly what I had in mind, but they served the purpose. I have since toyed with Midjourney to create more bespoke reference photos, and the results surprised me. The Midjourney pics weren’t necessarily closer to what I’d had in mind, but the positive feedback loop created by describing them to the bot, seeing the results, then iterating on them, ended up sparking new ideas. Some have extolled ChatGPT’s capacity as brainstorm partner, but I find Midjourney to be far more stimulating.
Again, does this replace a human? Yes and no. When I worked at Pixar, I experienced positive feedback loops like this whenever the art department would generate concept paintings based on script pages I’d written, which would in turn give me new ideas for story. But as a freelance writer, I have no art department at my disposal. Nor does any other screenwriter I know. Most studios do not spend a dime on art until they have a green-lit script (which is to say, after the writer has gone home). So I don’t think a writer using Midjourney for inspiration costs anyone a job any more than Pinterest or a Google image search does. The danger is that studios will cut their art department in the mistaken belief that AI-generated art is a viable replacement.
Every writer will decide for themselves how or if to use this technology. As I continue to experiment, I’m trying to keep an eye on one what it’s doing for me, and to me. Is it stretching my creative process in new directions or is it shunting me into a derivative corner? Is it saving me time or is it just another rabbit hole I’ll have to swear off every few months? This is an ongoing process, one that I and every other artist should have the room to let play out, on our own terms. That’s the key.
Let’s look again at the WGA’s proposed regulations, because there’s a larger significance to them, one that pertains us all. By preemptively outlining a set of rules, the WGA is arguing that labor should define the use of this technology, not capital. On this there can be no compromise. For the companies, the measure of AI’s usefulness will be how many people it can replace. As writers, artists, laborers, we need to make sure AI is useful to us, and useless to them.
Only in tech can a company release an untested product to the public without facing legal peril. If AI were an experimental car, people would already be dead. If AI were a drug, the FDA would ban its sale. Yet it is available over-the-counter, without a prescription. Though Sam Altman of OpenAI has called upon Congress to intervene, we cannot afford to wait that long. Nor can we wait for cases to worm their way through our legal system. This is a collective problem that requires collective action. The WGA may be the first union in the nation to strike over this issue, but it must not be the last.
