Fairy Note

From the Death of the Author to the Birth of the Navigator

AI, Machine Reading, Latent Space, and the Reorganisation of Cultural Cognition

The current panic surrounding artificial intelligence and authorship is fascinating because it presents itself as though something entirely unprecedented has suddenly occurred. Newspapers, universities, publishers, prize committees, and artists increasingly speak as though generative AI has arrived from outside culture and ruptured an otherwise stable understanding of creativity, originality, and intellectual labour. Yet the closer one looks, the stranger this reaction becomes. Long before machine-learning systems emerged, thinkers such as Walter Benjamin, Roland Barthes, Michel Foucault, and later Victor Burgin had already dismantled the fantasy of the sovereign author. What AI has done is not simply threaten authorship from outside. It has exposed the instability that was already latent within modernity itself. The current controversy therefore says less about machines than it does about the fragility of humanism’s assumptions concerning cognition, symbolic labour, and creativity.

This contradiction became newly visible in two very different public discussions around AI. In one article, Jeanette Winterson described artificial intelligence as “alternative intelligence,” instinctively moving away from the increasingly exhausted debate over whether machines are “really conscious” or “truly creative.” Instead, Winterson gestures toward a different possibility: that AI represents another mode of symbolic organisation altogether, one capable of entering into relation with human cognition and reshaping it. Her position belongs, knowingly or not, to a broader posthuman lineage in which thought no longer belongs exclusively to the isolated individual subject but emerges relationally across humans, technologies, archives, systems, and media infrastructures.

The second discussion emerged around accusations that a prize-winning short story may have been generated using AI. The atmosphere surrounding this controversy was strikingly different. Critics immediately began searching for suspicious traces within the prose itself, attempting to identify “obvious markers” of machinic writing through stylistic smoothness, familiar rhetorical structures, and strangely fluent language. What emerged was not merely suspicion of automation but a profound anxiety concerning authorship itself. The panic was institutional as much as aesthetic. If AI systems can produce culturally convincing symbolic work, then where does authorship reside? How can ownership, labour, legitimacy, and intellectual value continue functioning socially and economically?

What makes this anxiety so revealing is that it assumes the author was once stable to begin with.

Yet the intellectual history of the twentieth century repeatedly argued precisely the opposite.

One of the most fascinating moments in Victor Burgin’s symposium discussion occurs when he reflects upon the 1970s and 1980s, recalling how he spent years trying to bring thinkers such as Barthes, Benjamin, Althusser, and ideology critique into photography and arts curricula. There is something deeply ironic in this recollection because the theorists once treated by many as excessively abstract or intellectually outdated suddenly become urgently contemporary again in the age of AI. Burgin almost positions himself historically as someone whose generation diagnosed a set of unresolved problems that culture never fully absorbed. AI has simply forced those buried contradictions back into visibility.

This return feels profoundly Benjaminian in structure. Benjamin’s own writing repeatedly concerns the reappearance of unresolved historical fragments that become suddenly legible under conditions of crisis. The current AI panic has precisely this atmosphere. Ideas once treated as historical suddenly feel newly alive because the technological conditions they anticipated have now materialised. It is not that Benjamin or Barthes “predicted AI” in any simplistic sense. Rather, the tensions they identified concerning reproduction, symbolic labour, authorship, and technological mediation were never resolved. They merely disappeared beneath periods of relative cultural stability and have now resurfaced under new computational conditions.

Roland Barthes perhaps articulated this instability most clearly in “The Death of the Author.” Barthes dismantles the romantic fantasy of the author-God whose inner consciousness guarantees the meaning of the text. Instead, he writes that:

“the text is a tissue of quotations drawn from the innumerable centres of culture.”

This sentence acquires startling new force once read in relation to generative AI. What Barthes once proposed metaphorically has now become computationally operational. Generative systems function through immense symbolic archives assembled from language, images, rhetoric, style, narrative structures, and cultural repetition. Latent space itself begins to resemble a machinic tissue of quotations.

It is important here to pause and explain what latent space actually is, because the term is frequently mystified in contemporary AI discourse. Joanna Zylinska’s account is especially useful precisely because she resists treating generative systems as magical machine imagination. In Image Thinking after Artificial Intelligence, she describes latent space as a multidimensional probabilistic field constructed from relationships between images, words, and symbolic forms. The AI does not store complete images or fixed meanings waiting to be retrieved. Instead, it navigates statistical proximities between fragments of culture. The system learns relational tendencies:
how certain visual forms cluster around certain words,
how symbolic structures recur,
how aesthetic patterns become statistically associated.

Latent space therefore does not function like memory in the human sense. It functions relationally. Meaning emerges through probabilistic association rather than conscious intention.

This distinction matters enormously because it reframes the entire discussion around AI creativity. The machine is not secretly imagining, nor is it simply reproducing isolated copies of existing works. Instead, it moves through a vast statistical field composed of accumulated symbolic traces. Barthes’ “tissue of quotations” suddenly becomes uncannily literal. AI systems operationalise what structuralism had already recognised: that language and meaning always emerge through inherited relations rather than sovereign originality.

Victor Burgin extends this argument in an extraordinarily important direction during the symposium when he revisits Barthes explicitly. Reflecting on “The Death of the Author,” Burgin notes that Barthes displaced meaning from the author onto the reader. But Burgin then introduces a decisive contemporary mutation:

“we do have such an abstract, virtual, mechanical reader today, in the form of AI.”

This sentence quietly reorganises the entire debate.

Most public discourse surrounding AI remains trapped at the level of production:
Who wrote the sentence?
Who generated the image?
Who owns the output?

Burgin shifts attention elsewhere entirely. The real transformation lies not simply in machine writing but in machine reading. AI systems ingest and process culture itself. They read language statistically. They read images relationally. They read patterns, structures, and symbolic repetition across enormous archives of accumulated human expression. The AI therefore becomes not primarily a writer but a machinic reader of culture.

This distinction is philosophically decisive because it reframes AI not as an autonomous creative consciousness but as a recursive interpretive infrastructure operating upon the symbolic field itself. AI systems statistically read culture and then return transformed symbolic configurations back into circulation. The consequences of this are enormous because machine reading begins reorganising the conditions under which human symbolic production itself occurs.

This is also why the panic surrounding “AI style” acquires such a strangely psychoanalytic atmosphere. Critics increasingly claim they can “feel” AI within prose, identifying suspiciously smooth rhetoric or familiar sentence patterns. Yet these accusations often reveal something much deeper and more uncomfortable. What appears “AI-like” is frequently simply the statistical amplification of tendencies already present within contemporary language itself. AI systems privilege dominant symbolic forms because they are trained upon large-scale cultural repetition. But human writing also emerges through inherited rhetoric, cliché, institutional language, and symbolic habit.

The machine therefore exposes something culture would prefer not to confront:
that fluency itself may already be statistical.

Readers begin treating texts symptomatically, searching for hidden machinic traces beneath apparently human prose. What is feared is not simply automation but the possibility that the distinction between originality and recombination was never entirely stable to begin with.

It is at precisely this point that Michel Foucault becomes essential. Barthes may have declared the death of the author philosophically, but Foucault asks a different and much more institutional question:
if the sovereign author is dead, why do institutions continue behaving as though authors matter so intensely?

In “What Is an Author?”, Foucault argues that the author is not merely a person but a function operating within systems of law, ownership, discipline, classification, and legitimacy. The “author-function” organises discourse socially. It determines who is responsible, whose words count, which texts become legitimate knowledge, and who possesses symbolic authority.

This becomes extraordinarily contemporary in relation to AI.

The panic surrounding AI-generated writing is not merely aesthetic. It is institutional panic. Universities, publishers, copyright systems, and prize committees all depend upon stable author-functions in order to regulate symbolic labour. The anxiety is therefore not simply that machines may write, but that distributed symbolic production destabilises the structures through which ownership, legitimacy, and labour are organised.

The symposium discussion around copyright law reveals this perfectly. Filmmakers repeatedly explain that legal systems are increasingly no longer asking simply whether AI was used. Instead, the question becomes:
Where was human agency located?
Who intervened?
Who selected?
Who transformed?
Who directed the process?

In other words, institutions desperately attempt to relocate the author-function somewhere stable.

This is why AI generates such intense anxiety around artistic labour itself. If symbolic production becomes distributed, then how do artists continue being paid? How does ownership function economically? How does creative labour remain legible within capitalism?

The “author” is therefore not merely a philosophical category.
The author is economic infrastructure.

This becomes particularly important in relation to Phantom Mirror because the project occupies a very different position from generic image synthesis. In much commercial or abstract generative production, authorship genuinely becomes muddy. But Phantom Mirror remains phenomenologically anchored in lived experience. The symbolic material emerges directly from caregiving, grief, dementia, autobiographical memory, recursive affect, and psychic specificity.

The work carries the pressure of a lived life-world within it.

This distinction matters enormously because it introduces something that purely statistical generation cannot fully stabilise:
a phenomenological footprint.

The AI does not originate the psychic material. Rather, it participates in its symbolic externalisation. AI phototherapy therefore cannot be reduced either to autonomous machine creativity or to passive technological tooling. The methodology depends fundamentally upon recursive refusal and recognition. Most outputs are rejected. Meaning emerges only when an image acquires affective force through sustained navigation of symbolic space.

This process resembles less the production of a conventional artwork than a psychoanalytic encounter with symbolic material under pressure. The image functions not simply representationally but psychically. Burgin’s concept of the psychical object becomes decisive here. Psychical objects are not merely signs or metaphors. They are structures that organise fantasy, mourning, projection, desire, and affective recognition. They are real because they produce psychic effects.

The significance of the AI-generated image therefore lies not in whether the machine “authored” it. The image matters because it acts upon the psyche.

At this point, the older categories begin to fail. “Author” no longer fully works. Neither does “reader,” “producer,” “user,” or the increasingly inadequate term “prompt engineer.” A different figure slowly begins to emerge historically.

The navigator does not originate meaning from nowhere, nor is the navigator passive. Instead, the navigator moves through latent symbolic fields, selecting, rejecting, stabilising, and recognising meaningful configurations within probabilistic cultural space. Meaning emerges relationally through recursive interaction between human affect, machine reading, symbolic memory, and recognition.

One of the most unexpected consequences of this condition may be the return of modernism. This term also requires careful clarification because modernism here refers not simply to a historical art movement but to a broader cultural response to crisis in representation itself. Writers such as Joyce, Woolf, Kafka, and Beckett increasingly felt that smooth realism could no longer adequately represent modern experience under conditions of industrialisation, technological acceleration, war, psychoanalysis, and fragmented subjectivity. Modernism therefore privileged rupture, opacity, interruption, montage, and symbolic instability.

This becomes extraordinarily relevant to AI because generative systems optimise for precisely the opposite:
statistical coherence,
readability,
symbolic familiarity,
and stylistic smoothness.

The likely cultural response may therefore involve renewed resistance through fragmentation, opacity, and anti-smoothness. AI may unintentionally recreate the very historical conditions that once produced modernism.

This possibility becomes even more striking once considered through Joanna Zylinska’s account of “black-box anxiety.” She notes that both human and synthetic cognition remain fundamentally opaque. Human introspection never fully reveals the processes through which thought emerges, while diffusion systems encode billions of inaccessible statistical relations. This “reciprocal blindness,” she argues, fuels contemporary anxiety surrounding AI outputs.

Modernism historically emerged from precisely such crises of perception and representation. AI now introduces a new version of that instability:
epistemological opacity.

The contemporary panic around AI therefore reveals something much larger than concern over plagiarism. It reveals the collapse of a historically specific understanding of authorship, cognition, symbolic production, and creative labour.

Benjamin destabilised the author technologically. Barthes destabilised the author semiotically. Foucault destabilised the author institutionally. Burgin destabilises authorship through machine reading and psychical objects. Zylinska destabilises cognition itself through recursive image-thinking and infrastructural culture.

Phantom Mirror exists precisely within the convergence of these historical shifts. The project became the site through which these wider transformations became perceptible. It demonstrates that meaning increasingly emerges through distributed symbolic systems involving human memory, machine reading, probabilistic recombination, affective recognition, and recursive navigation.

The sovereign author does not disappear entirely. But neither can the author remain untouched by these transformations.

The author mutates.

And from within the instability of machine-readable culture, another figure slowly begins to emerge:
not the sovereign creator,
nor the passive user,
but the navigator moving through the tissue of computationally organised symbolic space, stabilising meaning temporarily within an increasingly distributed cognitive world.