Dear Editor
Artificial intelligence has not simply enlarged my library. It has shortened the hallways.
I am told, repeatedly, that the great intellectual danger of generative AI is that people will let machines think for them. Perhaps. But I suspect a subtler danger is already upon us: AI can make a thought look finished before the human being has actually finished thinking.
That is worse than a hallucinated citation. A false citation can be checked. A beautifully phrased half-thought may never be checked at all.
Using AI can feel like being a tenured professor surrounded by thirty exceptionally capable doctoral students. I arrive with an intuition—stubborn, half-formed, linguistically awkward—and they immediately begin work. One offers a term. Another restructures the argument. A third says, “Perhaps what you mean is this.” Others run toward Peirce, Wittgenstein, Confucius, or contemporary sociolinguistics and return carrying notes.
It is exhilarating.
It is also a fraud, if I forget one detail: thirty real doctoral students are thirty centers of resistance. One will dislike my argument. Another will say I misunderstood Peirce. A third will accuse me of romanticizing East Asia. Someone, sooner or later, will say what every scholar needs to hear: “That is beautiful. It is also wrong.”
AI can imitate such objections. But imitation of plurality is not plurality. Several AI systems inhabit overlapping textual worlds and habits of exposition. They are gifted at coherence. Give them two distant ideas and they often find a bridge before asking whether one ought to exist.
This may be generative AI’s characteristic epistemic danger: the texture of consensus without the evidential weight of consensus.
Five machines admiring my idea do not constitute peer review.
More troublingly, five machines can help my idea put on a suit and tie by afternoon. Every analogy can acquire a literature. Every intuition can return with an abstract, six headings, and the confidence of a tenured concept. The old difficulty was that a living question might die before a scholar could pursue it. The new difficulty is that a foolish question can become a polished manuscript before dinner.
We keep speaking as though AI has democratized intelligence. I think it has done something more unsettling.
It has subsidized the appearance of intelligence.
Eloquence can now be rented. Erudition can be rented. Structure can be rented. Even the external appearance of judgment can be rented by the hour—or by the token.
This does not mean the renter is stupid. I use AI constantly, and gratefully. At seventy-six, formed between Korea and America, Christian theology and Chinese thought, retirement and books, I have found an extraordinary workshop. A question once requiring weeks of preparation can now be tested while warm. I can put Confucian zhengming (正名, the rectification of misalignment between name and reality) beside Peircean semiotics, ask Wittgenstein to disturb the arrangement, then ask what a sociolinguist would object to. The library is illuminated at three in the morning. The translator is awake. The seminar reconvenes in seconds.
Speed preserves the temperature of a question. That is no small gift.
But speed also creates a new temptation: linguistic completion before intellectual completion. The machine closes sentences, fills conceptual gaps, smooths awkward transitions, and reconciles tensions with extraordinary politeness. What once remained visibly unresolved can now be made to look settled.
Fluent closure has become almost free.
The institutional consequence is not difficult to imagine. When half-formed thoughts can be dressed for submission in an afternoon, journals inherit part of the cost AI has removed from authors. Editors and peer reviewers become unpaid inspectors of premature coherence. The manuscript becomes cheaper to produce precisely when judgment becomes more expensive to supply.
My own response has been to cultivate a peculiar intellectual virtue: the ability to say “almost.”
Almost.
Too smooth.
You have Westernized the problem.
Do not reconcile them yet.
That word is technically correct, but existentially wrong.
You improved my English and weakened my thought.
The concept is right, but the center of gravity is wrong.
Something is absent. Begin again.
I increasingly suspect the ability to say “almost” will matter more than the ability to prompt brilliantly. Prompt engineering is becoming a minor priesthood, complete with incantations and claims to esoteric skill. Fine. Learn the spells. But the deeper human competence is recognizing when an elegant answer has betrayed the question.
This matters especially to those of us who think across languages and civilizations. A thought may arrive with a Korean emotional temperature, a Chinese conceptual skeleton, a Christian theological horizon, and a Western philosophical question. Academic English asks it to pass through customs.
AI is a magnificent customs broker. It can move the cargo quickly.
It can also quietly replace the cargo.
I have, therefore, adopted an almost embarrassingly primitive discipline. Before inviting AI into an important thought, I try to preserve two or three ugly sentences of my own. Sometimes in Korean. Sometimes grammatically incomplete. I want the pre-AI intuition on the page before fluent language improves it.
Why? Not to establish legal ownership. The mythology of the solitary author was always exaggerated. We think with teachers, enemies, books, languages, lovers, traditions, and ghosts.
I preserve the ugly sentence for intellectual provenance.
Months later, when a phrase, such as “indexicality without occupancy”, appears in my writing, I need to remember the crude perception from which it grew: AI가 권위 있는 사람처럼 말은 하는데 그 말 뒤에 책임질 사람이 없다. 표지는 있는데 사람이 없다. (AI speaks like a person of authority, but there is no one behind the words to bear responsibility. There is a sign, but no person there.)
Perhaps the machine offered the elegant phrase. Perhaps I refined it through forty exchanges. Fine. The important question is whether the language clarified my intuition or replaced it.
Academia will need disclosure rules for AI use. Publishers will invent policies. Universities will produce forms. These are necessary and will predictably become bureaucratic.
The more intimate problem will remain unregulated: Will scholars still recognize their own thoughts after the machines have improved them?
This is why I am less worried that AI will become intelligent than that human beings will become satisfied with the signs of intelligence.
We already reward the signs. A polished abstract. Correct vocabulary. A symmetrical framework. Six boxes in a diagram. A confident conclusion beginning, “This requires a fundamental reconceptualization of…”.
AI did not invent academic vanity. It simply industrialized some of its preferred costumes.
The antidote is not nostalgia. I have no wish to return to card catalogs as a spiritual discipline. Nor will I dismiss a technology that has liberated what I call speculative adjacency: bringing distant bodies of thought into rapid, exploratory contact. When the hallways become short enough, the architecture of thinking changes.
But the cheaper inquiry becomes, the more expensive discernment must become.
Not every connection deserves development. Not every curiosity deserves a paper. Not every publishable paragraph contains a thought worth publishing. In the old economy of scholarship, scarcity imposed crude discipline. We had only so many books, research trips, conversations, and years. AI has removed many of those restraints.
Now we need intellectual fasting.
At seventy-six, I feel this urgently. My problem is not whether I can generate more material. I plainly can. My problem is deciding which questions deserve the hours I have left.
AI can help articulate a question. It can search, compare, synthesize, and object. It may even help me discover that my favorite sentence is nonsense.
But the final act of judgment remains stubbornly human: not merely saying “yes” or “no,” but recognizing when the truthful answer is “almost.”
Almost right.
Almost clear.
Almost mine.
Almost worthy of publication.
So perhaps we should stop asking whether AI will make us more intelligent. The question flatters machine and human alike. Ask instead whether, surrounded by nearly free fluency, we are becoming more discriminating.
Are our questions improving?
Can we recognize a false harmony?
Can we preserve an inelegant intuition long enough to discover what it means?
Can we throw a curveball that breaks the window of premature coherence—and then inspect the glass?
Above all, can we still say “almost” when the machine gives us exactly the beautiful answer we hoped to hear?
That small word may become one of the last defenses of formed intellect.
AI has made fluency cheap.
Judgment is now expensive.
We should spend accordingly.
Data availability
No datasets were generated or analysed during the current study.
Curmudgeon Corner
Curmudgeon Corner is a short opinionated column on trends in technology, arts, science and society, commenting on issues of concern to the research community and wider society. Whilst the drive for super-human intelligence promotes potential benefits to wider society, it also raises deep concerns of existential risk, thereby highlighting the need for an ongoing conversation between technology and society. At the core of Curmudgeon concern is the question: What is it to be human in the age of the AI machine? -Editor.
Author information
Authors and Affiliations
Contributions
S.K. is the sole author.
Corresponding author
Ethics declarations
Conflict of interest
The authors declare no competing interests.
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Rights and permissions
About this article
Cite this article
Kim, S. The ability to say “almost”. AI & Soc (2026). https://doi.org/10.1007/s00146-026-03297-y
Received:
Accepted:
Published:
Version of record:
DOI: https://doi.org/10.1007/s00146-026-03297-y