In early July 2026, on the occasion of publishing a new novel, Haruki Murakami gave an exclusive interview to Kyodo News in which he drew a clear line between his own writing and the fiction that artificial intelligence can now produce.<\/p>\n\n\n\n
“AI synthesizes and infers from things that have already happened. But what I do when I write a novel is something entirely different.”<\/p>\n\n\n\n
The novelist’s task, he said, is “to seize hold of something new that suddenly surfaces.” That surfacing, he insisted, is not the product of inference but something that “truly comes from thin air.” “AI probably can’t do that.” The word probably<\/em> is worth pausing on \u2014 it suggests that even Murakami himself senses some fragility in the claim.<\/p>\n\n\n\n Starting from this remark, I want to open up a few questions.<\/p>\n\n\n\n The first thing to notice is that the contrast Murakami draws \u2014 AI works by analogy, I pull things from thin air \u2014 is, in principle, unverifiable.<\/p>\n\n\n\n It resembles a variant of the philosophical problem of other minds. When someone says “it hurts,” there is no way for an outside observer to know whether they are truly experiencing pain or simply producing the appropriate utterance for the situation. In the same way, Murakami’s introspective report \u2014 his testimony that something “comes from thin air” \u2014 is a first-person account accessible only to him. Not just third parties, but Murakami himself, has no means of distinguishing whether the experience is genuinely creation from nothing, or the fast, unconscious recombination of an enormous store of patterns accumulated over decades of reading and writing.<\/p>\n\n\n\n But unfalsifiability does not mean the claim is meaningless. It should be read not as a scientific proposition but as a creator’s self-understanding \u2014 and, in the line “that’s not the kind of novel I want to write,” as a declaration of intent.<\/p>\n\n\n\n Murakami’s position, however, runs against the grain of the dominant currents in literary theory and philosophy since the mid-twentieth century.<\/p>\n\n\n\n Roland Barthes’s “The Death of the Author” (1967) decentered the author, treating them not as an “origin” but as a site where existing linguistic and cultural codes are recombined. Kristeva’s notion of intertextuality holds that every text is a mosaic of quotations and transformations of other texts. As early as 1919, T. S. Eliot argued that the most original poets are precisely those who have most deeply internalized tradition. Harold Bloom described poets as perpetually caught in an “anxiety of influence,” generating work out of struggle and misreading of their precursors. Propp’s morphology of the folktale reduced an entire genre of narrative to a finite set of combinable functions.<\/p>\n\n\n\n Within theory, the idea of originality springing from nothing has been more or less abandoned; the dominant view is closer to “everything is repetition with difference.” And yet Murakami \u2014 a writer who has openly acknowledged debts to Kenzaburo Oe, Raymond Carver, American literature, and jazz \u2014 reaches for a non-genealogical, discontinuous vocabulary, “from thin air,” when describing the moment of creation itself.<\/p>\n\n\n\n This is less a contradiction than a sign that theoretical description and the phenomenological experience of the creator belong, from the outset, to different language games. Literary theory can explain, after the fact and structurally, why a given pattern emerged. But that is not the same as describing what the writer is experiencing in the moment of writing.<\/p>\n\n\n\n There is another axis to this that is easy to overlook: the physical means of writing itself.<\/p>\n\n\n\n The visual arts have already circled this question once. In the early days of digital art, there was real debate over whether such work could be called painting at all. Today, using technical aids like layers or copy-and-paste is barely considered grounds to doubt a work’s originality. Generative artists go further still \u2014 feeding an algorithm a set of generative rules and adopting the “unexpected result” as the finished work. This is strikingly close to Murakami’s “thin air”: the artist is not in full control, but waiting for the surprise the system produces.<\/p>\n\n\n\n Literature cannot escape the same question. In the era of writing by hand on manuscript paper, one wrote character by character, at a high cost for going back and revising; correction meant physical resistance \u2014 crossing out, starting again. With the word processor and the PC, cutting, copying, and reordering became instantaneous, and text came to be treated less as something generated linearly than as an editable mass of material.<\/p>\n\n\n\n If so, it becomes possible that Murakami’s own notion of “pulling something in from thin air” was itself formed within \u2014 and presupposes \u2014 a digital writing environment of writing, erasing, rearranging, and recombining fragments after the fact. The era of handwritten manuscript paper may, ironically, have been the more linear, less reversible, more singular creative process. There is a paradox here: the persuasiveness of the claim that AI and human creativity differ turns out to depend on which era of writing technology one has in mind.<\/p>\n\n\n\n What is the authority \u2014 the instance of judgment \u2014 that decides a combination is “good,” that says “this is finished”? The question becomes concrete the moment we imagine a novel co-written by a human and an AI.<\/p>\n\n\n\n Several models of collaboration are possible. In one, AI generates candidates and the human selects among them; formally, judgment remains with the human, but since AI itself defines the population of options, the human is not choosing from infinite possibility but from a finite set of candidates the AI has already shaped. In another, a dialogic back-and-forth, judgment shifts continuously between the two parties, making it genuinely difficult, after the fact, to say who “decided” what.<\/p>\n\n\n\n Worth noting here is that the word “collaboration” can create the false impression of two equal partners at work. In reality, AI has no will to “put this before the world,” bears no responsibility of the kind that comes from “this work carries my name,” and has no subject position from which to experience how the work is received once it exists. Before we even ask where the power to decide resides, there is a prior asymmetry: only the human carries the stake<\/em> that the concept of judgment presupposes.<\/p>\n\n\n\n And if some future writer were to compose a novel in dialogue with AI and still claim, “in the end, I was the one who chose” \u2014 the further question remains whether the very criteria of that choice, having been formed through dialogue with the AI, can still be called “a judgment that fell purely from thin air.” The moment the standard of judgment itself is collaboratively constructed, the pure independence of that authority is already unsettled.<\/p>\n\n\n\n Interrogating what “thin air” actually names is the crux of the whole discussion.<\/p>\n\n\n\n Read through a Freudian or Jungian lens, it might be the personal or collective unconscious \u2014 and the wells, undergrounds, and passages to other worlds that recur throughout Murakami’s fiction have long staged precisely this descent into a subconscious reservoir. On this reading, “thin air” is really “underground” turned inside out, a metaphorical inversion. Read structurally, “thin air” need not be located inside an individual mind at all; it could be the shared system of signs belonging to a language community \u2014 something that precedes any individual speaker’s intention, much as Saussure’s langue<\/em> does.<\/p>\n\n\n\n What is striking here is the proximity to how generative AI actually works. Large language models represent the meaning of words and sentences as coordinates in a high-dimensional embedding space; generation is the process of probabilistically tracing a “next point” through that vast, multidimensional space, given a context. This is not a metaphor \u2014 it is a technically real space.<\/p>\n\n\n\n The intuitive, ill-defined place Murakami calls “thin air” and the high-dimensional vector space AI actually computes over turn out to be, as structural metaphors, strikingly homologous. Both amount to selecting a single point from a vast field of possibility that lies outside conscious, causal reasoning.<\/p>\n\n\n\n If so, who is it that stands in that airspace? Is it the writer, who first receives whatever descends? Or is it the reader, who ultimately receives what the writer has given form to? If we suppose the writer is merely a medium that gives shape to what it receives, what, then, distinguishes that medium from AI?<\/p>\n\n\n\n This reframing is not as radical as it sounds. Reception theorists like Iser and Jauss held that a work’s meaning is not complete at the moment of the author’s intention but is completed only in the act of the reader’s reception. What Barthes announced alongside “the death of the author” was, in the same breath, “the birth of the reader.” Further back still, when Homer opens his epic with “Sing, O Muse, through me,” the oldest layer of Western literature already contains the self-understanding of the author as a conduit \u2014 receiving something divine and giving it form. Murakami’s “from thin air” can be read as a contemporary version of this ancient image of the author as a possessed medium.<\/p>\n\n\n\n If the site of creation is not the author’s interior but something that arises in the between<\/em> \u2014 between author and reader \u2014 then the point of contrast with AI shifts. It is no longer “who generated this” but “between whom does the event occur.” Even a text generated by AI, once read by a human reader, may produce the same event at the site of reception \u2014 a question that only pushes Murakami’s position further into a corner.<\/p>\n\n\n\n Pressed this far, almost nothing substantive is left except the presence or absence of a body. This is not a defeat of the argument but, if anything, an honest place to arrive at theoretically.<\/p>\n\n\n\n Give the body some content, and a few things surface. As Heidegger’s “being-toward-death” suggests, the urgency with which a human strings words together is conditioned by the fact that it will end. AI has no such thing as running out of time. The vulnerability of the body that Levinas emphasized \u2014 the fact that what one receives can change or wound a person \u2014 is a risk borne only by a receiver with a body. Hunger, fatigue, and aging are not mere constraints but the very source of the urgency behind “why this must be written now.” The body, in other words, is the condition under which what one receives becomes something that weighs<\/em> on the receiver.<\/p>\n\n\n\n But honestly, even after stacking up this much theory, the claim that “the body makes the weight different” ultimately collapses back into the same kind of unfalsifiable introspection as Murakami’s “because that’s how it feels.” It hasn’t been proven that the presence or absence of a body constitutes a substantive difference \u2014 only that more words have been piled up to explain why it might matter.<\/p>\n\n\n\n Push further still, and we arrive at the place where explanation in language runs out entirely \u2014 something knowable only to those who have a body. Ironically, that may be the other face of what Murakami calls “thin air.” The question of human creativity has no answer. And yet it may be only by facing that lack of an answer directly, without flinching, that the outline of the word “writer” gets redrawn for this era.<\/p>\n","protected":false},"excerpt":{"rendered":" The Occasion In early July 2026, on the occasion of publishing a new novel, Haruki Murakami gave an exclusive […]<\/p>\n","protected":false},"author":1,"featured_media":527,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"footnotes":""},"categories":[1],"tags":[],"class_list":["post-526","post","type-post","status-publish","format-standard","has-post-thumbnail","hentry","category-uncategorized"],"_links":{"self":[{"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/posts\/526","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcomments&post=526"}],"version-history":[{"count":1,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/posts\/526\/revisions"}],"predecessor-version":[{"id":528,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/posts\/526\/revisions\/528"}],"wp:featuredmedia":[{"embeddable":true,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=\/wp\/v2\/media\/527"}],"wp:attachment":[{"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=%2Fwp%2Fv2%2Fmedia&parent=526"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=%2Fwp%2Fv2%2Fcategories&post=526"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.junkatanuma.com\/index.php?rest_route=%2Fwp%2Fv2%2Ftags&post=526"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}An Unfalsifiable Claim<\/h2>\n\n\n\n
The Myth of Originality \u2014 What Literary Theory Already Knows<\/h2>\n\n\n\n
When the Tool Itself Unsettles the Definition of Creativity<\/h2>\n\n\n\n
Where Does the Authority to Judge Reside? \u2014 Collaboration with AI as a Test Case<\/h2>\n\n\n\n
Where Is the Air?<\/h2>\n\n\n\n
Is the Writer a Medium? \u2014 Reframing the Author as Receiver<\/h2>\n\n\n\n
What Difference, If Any, Remains<\/h2>\n\n\n\n