Published 300 years ago, Jonathan Swift's “Gulliver’s Travels” (1726) described a “word combination machine” which foresaw the dangers today's AI writing poses to knowledge, authorship and human judgment.
An illustration of Gulliver visiting the academy of Lagado, from an 1850s French edition of “Gulliver's Travels” (1726). The illustration's caption reads: “Je passai dans une autre chambre, l'odeur en ètait repoussant” (“I went into another room, the smell was so disgusting”).WIKIPEDIA
Noh Myung-woo
The author is a professor of the Division of Economics, Politics and Society at Ajou University.
There is a joke about classics: They are books everyone knows but no one has read to the end.
If we use the excuse that everyone else has skipped them too, however, we can miss their remarkable ability to speak across centuries. In that vein, Jonathan Swift’s “Gulliver’s Travels” (1726) still gives us plenty to think about in 2026.
Among the strange lands Gulliver visits is Laputa, an island floating in the sky, a name familiar to many through Hayao Miyazaki’s “Castle in the Sky” (1986). Below Laputa lies Lagado, the capital of Balnibarbi. Radical intellectuals there, contemptuous of traditional farming and architecture, establish a vast academy under a royal patent. Some scholars try to extract sunbeams from cucumbers, while others study how to turn human excrement back into food.
One professor explains to Gulliver an invention that might be called a “word-combination machine.” With it, he says, even an ignorant person can write books on philosophy, poetry, politics, law, mathematics and theology with modest expense and a little physical labor, without the slightest help from genius or study.
To prove his claim, the professor demonstrates the machine. The roughly six-meter-square (64.5-square-foot) device is packed with dice-sized wooden blocks, each bearing a word. Forty students turn its handles, randomly rearranging the words. Thirty-six boys scan the resulting combinations and, whenever chance produces a meaningful phrase, read it aloud for the remaining four to write down.
The professor believes the machine can eventually produce a complete body of writing on the arts and sciences. He boasts that he has included the entire vocabulary and carefully calculated from numerous books the frequency with which participles, nouns, verbs and other words occur.
The academy’s absurd experiments are funny because Swift pushes the worship of technique to its logical extreme: A process can look systematic, elaborate and learned while losing sight of the human judgment that gives knowledge meaning.
By now, however, the resemblance to today’s generative AI is hard to miss.
The irony is sharper three centuries later. The machine in Lagado required forty students to turn handles and four to record its accidental sentences. Today, one user can summon prose in seconds. The technology has changed, but Swift’s question about knowledge, authorship and thought remains unsettlingly familiar.
The number of papers submitted to academic journals has reportedly surged in recent years. It is hardly because researchers everywhere suddenly stopped eating and sleeping to devote themselves to scholarship. Rather, it is part of the farce that has unfolded since generative AI, the 21st-century word-combination machine, became an indispensable research tool.
Journal reviewers now face a new task: determining which submissions were written by humans and which are products of automated writing. And as the volume of papers requiring review grows, reviewers themselves are reportedly turning to generative AI. AI-generated papers are submitted under human names, then human reviewers ask AI to assess them. What appears on the surface to be communication between humans may in fact be communication between machines.
The momentum behind so-called one-click publishing — entering a prompt and publishing what the machine produces — shows no sign of slowing. One publisher recently released more than 9,000 e-books spanning economics, classics, humanities, self-help, fashion, food and drink. In some cases, more than 100 books appeared under a single author’s name.
Publishing editors are stunned to receive manuscripts that plainly bear the stylistic fingerprints of AI. Meanwhile, some writers are furious when publishers use AI to polish their work. Students ask generative AI to summarize lecture notes rather than read books or papers themselves, and request model answers and study texts produced by machines. Professors, whenever they read student submissions, must first wonder whether a human or a machine wrote them.
The world is noisy with the confusion and mistrust brought by automated writing. In 2026, when the value of human-written books risks being reduced to their usefulness as training data for large language models, we have postponed the most fundamental questions: Why are humans giving up writing for themselves and asking machines to do it instead? And, by entrusting writing to machines, are we voluntarily choosing ignorance?
This article was originally written in Korean and translated by a bilingual reporter with the help of generative AI tools. It was then edited by a native English-speaking editor. All AI-assisted translations are reviewed and refined by our newsroom.
