Roald Dahl has been making headlines of late, after a series of sensitivity edits were made to the author’s famously macabre children’s stories. Dahl’s adult short stories aren’t as beloved, but are very much worth reading; one story in particular appears to have predicted the rise of generative AI, particularly ChatGPT, with uncanny accuracy.
At this point, many have experimented with some form of generative AI (reportedly, ChatGPT is the fastest-growing consumer application in history), and any writer tinkering with ChatGPT might confess to feeling a little nauseous; the AI is far from perfect, but it is coherent, and lightning fast.
ChatGPT spits out cliched phrases and confidently spews misinformation, sure, but it might just be good enough to make a quick buck; there is a real concern that the internet will be spammed with millions of AI-generated articles, and that emerging fiction writers might just be drowned out by the volume of content produced by ChatGPT.
Roald Dahl’s 1953 short story, “The Great Automatic Grammatizator,” sees two characters, Adolph Knipe and John Bohlen, “disrupt” the publishing industry with a similar device. Knipe is a tech genius and aspiring fiction writer who is deeply frustrated by his own artistic limitations. Bohlen is a businessman who doesn’t even like Knipe, but recognizes his potential.
After a fruitless writing session, Knipe is suddenly overcome with inspiration, and decides to create a machine that can write for him, better than he ever could. Knipe understands that “a machine, however ingenious, is incapable of original thought.” This truth is still just as relevant today, as it was back in Dahl’s time; ChatGPT and Midjourney are simply remixing and regurgitating a gargantuan volume of original work created by artists.
Knipe comes to the conclusion that “an engine built along the lines of the electric computer could be adjusted to arrange words (instead of numbers) in their right order according to the rules of grammar. Give it the verbs, the nouns, the adjectives, the pronouns, store them in the memory section as a vocabulary, and arrange for them to be extracted as required. Then feed it with plots and leave it to write the sentences.”
Knipe’s motivation for creating this machine seems to be based on a desire to be recognized as an artist, but he sells the idea to Bohlen by emphasizing how profitable it could be, and reminding him that handmade items, once crafted by skilled artisans, are now overwhelmingly made with machinery: “The quality may be inferior, but that doesn’t matter. It’s the cost of production that counts. And stories—well—they’re just another product.”
The concerns expressed by today’s working artists — that the rise of generative AI will devalue their labor — are perfectly expressed by Dahl’s story, which builds to a terrifying conclusion.
Bohlen is initially skeptical of Knipe’s ability to build such a device, but he can get behind the logic, and a few months later, the machine is ready. Of course, Dahl imagined his device in the form of the technology of his time; it is a vast, unwieldy machine, filled with whirring cogs, rods and levers.
Much like how one can feed prompts into Midjourney, or instruct ChatGPT, Knipe has an array of buttons that control the tone, the theme, and the literary style of the machine’s stories, taken from the words of great writers such as Hemingway, whose life’s work has been flattened into content to feed into the machine.
Today, one can ask ChatGPT to write a story in the style of Hemingway, and countless others; the output might not be great art, but it is impressively fast, and the AI is growing more sophisticated by the day.
At first, Knipe and Bohlen are confounded by the technology’s teething problems. Much like the early days of generative AI, the machine’s initial output is riddled with errors; at one point, the two are horrified after overusing the “passion” prompt, causing the machine to generate smut.
ChatGPT also suffered similar issues, spitting out uncomfortable content and hateful ideas that it had absorbed from its training data, which was solved by hiring an army of Kenyan workers who trained the AI to avoid toxicity, many of whom suffered trauma after reading the AI’s most vile output.
After patience and perseverance, the machine is perfected; both Knipe and Bohlen have attached their names to its stories, and gained reputations as prolific writers. But critical acclaim is not enough, and Knipe sets out to buy the “brands” of famous writers, seeking permission to use their name and style, and replace them entirely by the machine.
Knipe is largely met with aggression from writers who find the idea repulsive, just like many artists today who find generative AI threatening. To many, the act of creation is the point, and the idea of simply editing an AI’s output is profoundly dispiriting.
Hence, Knipe makes the decision to target “mediocre” writers, understanding that they will be more receptive to his offer. It doesn’t take long for Knipe’s machine to flood the market, as more and more writers choose the money over their craft.
Dahl’s story ends on a tragic note, describing an author holding out against the technology, refusing to sign the “golden contract,” but unable to feed his family. The machine, it seems, has permanently displaced the creatives.
The story was always chilling, but today, it reads like a disturbing prophecy. Generative AI is still in its infancy, and it is unclear how it will develop, or if the public will accept the uncanny mediocrity of AI-generated art and writing as “good enough.”
Countless artists have spoken out against the technology (although, some view it as a tool), and many have expressed anger that their work has been fed into the AI without their permission. Cartoonist Sarah Andersen wrote an insightful piece for the NYTimes that outlines the many dangers of AI, and notes that her own art style had been absorbed and spat out by the machine; casual fans might not even notice the difference.
There is a glimmer of hope on the horizon; recently the U.S. Copyright Office ruled that AI-generated images “are not the product of human authorship” and therefore cannot be copyrighted. Artists are building tools to protect their work from being stolen for training data, but the battle is still ongoing, and AI might well adapt to such protective measures.
AI-generated creative writing already threatens to drown out human writers; recently, sci-fi and fantasy magazine Clarkesworld was forced to temporarily close submissions after receiving an influx of machine-written stories.
In a detailed Twitter thread, the official Clarkesworld account lamented that they “have no solution to the problem,” only “ideas for minimizing it.” Putting up paywalls and restricting submissions to previously published authors would be an obstacle for emerging authors, and would risk shutting many out entirely.
There are several out there, it seems, who share the same perspective as Knipe, and view stories as “just another product.”