Writing Machines: LLMs and the Artifice of Intelligence

by Jonathan Basile

Review: Leif Weatherby, Language Machines: Cultural AI and the End of Remainder Humanism (Minneapolis, University of Minnesota Press, 2025), vii + 267 pp.

The debate over ‘artificial intelligence’ often proceeds as if there is, on one side, self-knowing human intelligence and, on the other, a machine that either does or does not adequate or correspond to it. Both sides of this debate rest on a misrecognition. If we were to turn to the apparently natural pole of this correspondence-structure, we would nonetheless find that there is no way of coming to any knowledge or self-knowledge of it without first positing some concept or precomprehension, some operationalization, or perhaps developing a model, program, or machine. Indeed, the study of human intelligence is no less controversial than the study of its ‘artificial’ offshoots—countless metrics and tests have been developed to measure the ‘intelligence quotient’, for instance, each of which is frequently criticized precisely for failing to capture something essential about what we recognize or ought to recognize about ourselves, about our own ‘intelligence’. That is to say, these intelligence metrics are criticized for their artificiality.

Rather than turning back from every discursive, quantitative, scientific, and technological representation of intelligence, to guard an ineffable essence that would count as truly our own, we should recognize that this supposed origin always appears as the after-effect of its representations or models. It is a moving target, displaced by precisely what is claimed as inessential to it. The very structure of failed or failing imitation or repetition, where we construct a mirror in which we see reflected something other than ourselves, should be understood as itself the intrinsic or essential structure of ‘intelligence’, and related terms or stand-ins such as subjectivity, consciousness or self-consciousness, etc. In other words, ‘intelligence’ is artificial.

We can see already that we have shifted the ground on which the customary debate takes place, over whether contemporary technologies referred to as ‘AI’, foremost among them the ‘large language models’ (LLMs), are truly doing what we humans do when we think. Do they have a ‘I’? It is no longer as simple as choosing among two opposed alternatives: thinking subjects or mere imitators, the latter accused of a bad repetition lacking the originary insight or lumen rationale that the former preserve. We too merely imitate thinking, which certainly does not mean that it’s all the same (say, whether ‘thinking’ or writing is done by human or machine, in the narrow senses), but rather that the differences we can intuit or measure among different instantiations of what we call by the same name—intelligence—must be thought as something other than the difference between autonomy and heteronomy, freedom and program, intuitive response and reaction, finalism and mechanism, or self-consciousness and parroting.

Such a reconfiguration of the scenes of intelligence certainly does not require us to accept the inflated claims made for ‘AI’ today. Even if we may have more in common with these models than we like, anyone who understands the first thing about how they work can see that attributing to them freedom, self-consciousness, or other powers of the humanist subject is a simple misunderstanding—all the more so if we do not possess such powers in any purity ourselves. Moreover, it is or should be—to any thinking person—obvious that the claims of efficacy, efficiency, and superior insight attributed to these technologies are bloated. The hype surrounding their spread is not a function of their capacities, neither their vaunted subjectivity nor their benefit to anyone’s bottom line, but is an end in itself in an economy that runs on speculation (and perpetually seeks out new ways to undermine the power of labor). It also serves further ideological purposes in the justification of military and governmental uses of such technologies.

It is not just because of their false promise, of subjectivity or profitability, that an overarching artificiality attaches to ‘AI’. It is not just that no one can agree on a definition or metric of intelligence, whether artificial or natural/human. If we examine the various instantiations said to represent this category, ‘AI’, today, we find irreducible heterogeneity. While certain machine learning algorithms can be said to be responsible for the popularity or success (however limited and desultory) of ‘AI’, much work that does not strictly follow from such algorithms themselves goes into the curation of data sets, the ‘fine-tuning’ of models for specific tasks or users, the filtering of inputs and outputs (with ‘guardrails’), and so on. In many cases, low-wage workers in the Global South intervene to produce outputs that users are told are AI-generated; at the extreme, there may be no machine learning at all involved in a process nonetheless promoted under this name. Despite this technology’s trumpeted novelty, the figure that best represents these hidden operators comes from the 18^th century: the Mechanical Turk.

In focusing on this non-self-identity of the category ‘AI’, I am not simply trying to burst its bubble. In fact, if we hoped to regulate or outlaw ‘AI’, to protect against its social, economic, and ecological harms, or even just to critique it, it is all the more difficult given that we do not strictly know what that is. There is no one essence or identity underneath the name.^[1] The modeled is an after-effect of the model. I start here in order to emphasize this parallel; the ‘same’ non-origin allows us to declare ‘AI’ a free and self-conscious agent despite it not being anything in particular, and to inflate a massive investment bubble on the basis of empty promises, hot air. Metaphysical and economic speculation have the same origin, the lack of origins.

Structuralism, Poststructuralism and Deconstruction, Then and Now

I draw this understanding of the relationship between modeling and modeled from Derrida, and from his own writing on thinking machines (as I’ll show below). The most common critiques of ‘AI’ today invoke the supposed self-consciousness and intentions of the human-natural pole of this structure, to argue that learning or writing machines are a failed imitation (‘stochastic parrots’).^[2] I hope to make the critique or rather deconstruction of ‘AI’ more robust by not making it depend on this origin that never was.

In pursuit of this end, this essay’s trajectory will be as follows: 1) an examination of narratives of the humanities in the 20^th and 21^st centuries, focusing on how structuralism, poststructuralism, deconstruction, new materialism, post-critique, and the digital humanities are variously characterized and periodized. Structuralism has been declared newly relevant because LLMs are said to reveal that the agency of language is internal to its system or structure, a notion of systematic immanence that Derrida can help us to puncture; 2) a critique of what I argue are Leif Weatherby’s misreadings of Saussure and Derrida—these misreadings serve to uphold conceptual and disciplinary demarcations between language and thought, human and machine, or humanities and sciences that Derrida can help us to question; 3) a deeper exploration of Derrida’s writings on cybernetics, in order to dislodge both the humanist and the structuralist readings of ‘AI’ and machine/technology/scientific modeling in general; and 4) a reflection on how this reading of Derrida can displace several common debates, within and beyond academia, over the nature and use of these technologies.

A handful of scholars in the humanities have recently made a salutary turn or return in the direction of the subversion of the subject by drawing continuities between LLMs and structuralist models of language. Ted Underwood notes that structuralism dislodges the fundamental presupposition shared by AI boosters and their humanist detractors: that meaning depends on the intentions of a thinking subject. For structuralism, Meaning or rather signification emerges from the relations of elements internal to the language-system, of which the speaking subject is an effect. Ted Underwood broke ground in reintroducing this perspective to the theorization of LLMs with a blog post in 2023, and it is the focus of Leif Weatherby’s 2025 monograph Language Machines.^[3]

Underwood’s faithful though brief development of a structuralist perspective introduces a complication from at least one author who might be considered ‘poststructuralist’ (Foucault). Underwood is silent on Derrida—not an inherent fault in a short and evocative blog post—but a noteworthy omission given that Derrida’s intervention in ‘structuralist’ and ‘poststructuralist’ theory was precisely to displace these presuppositions about the autonomy or agency of language as system. Weatherby, on the other hand, aims to shore up the structuralist autonomy of language by attacking a caricatural version of ‘poststructuralism’ exemplified by Derrida. Two recent essays in The OLR Supplement, by Víctor Betriu Yáñez and Justin Joque, have broken this silence or challenged the dismissiveness toward Derrida in this field.^[4]

For Weatherby, dismissing Derrida serves both a theoretical and narrative function. As Weatherby tells it, following Saussurean structuralism’s too brief heyday in the 1960s, the humanities ‘lost language’ because Derrida dislodged it from its privileged place as a ‘concrete object’ at the center of humanistic inquiry (73).^[5] Thus, Weatherby declares that only cognitive and computer science have advanced our understanding of language (language itself, language as such); moreover, only by undoing the destabilizing influence of Derrida does Weatherby believe the humanities can earn their rightful place in a conversation about language whose tone is now set by language’s technoscientific implementation in LLMs. Language as pure theoretical object can only be regained by righting the historical errancy of our discipline—both theory and history are set on course by undoing the influence of (Weatherby’s peculiar version of) Derrida.

I have several misgivings about this narrative, some of which I’ll return to in conclusion. I’ll note for now just what is idiosyncratic about its genealogy: where Weatherby sees ‘the humanities’ as following in the footsteps of Derrida by turning away from technoscientific understandings of language, some degree of hegemony has gathered in the 21^st century around narratives (undoubtedly contested) of the academic humanities according to which they have become deferential to technoscience by repudiating deconstruction, poststructuralism, and the ‘linguistic turn’. This would include ‘post-critique’, new materialism and the ‘material turn’, and certain approaches to the digital humanities (such as ‘distant reading’).^[6] At the same time, we can note one parallel: for all of these narratives including Weatherby’s, reanimating and slaying—at last, at long last—Derrida’s ghost is understood as necessary in order for the humanities to achieve a desired scientificity.

A frequently invoked pretense for ‘post-critique’ and self-styled materialism has been the claim that a skeptical attitude toward science gave rise to climate denialism, resistance to public health measures, and a broad distrust of expertise (a more historical materialist account might look to the neoliberal degradation of any function official institutions had apart from naked exploitation).^[7] Ironically, this skepticism is typically blamed on a ‘humanism’ (attention focused on ‘human’ language rather than ‘nonhuman’ nature/matter) erroneously attributed to structuralism and poststructuralism, because it is all but forgotten that these movements of thought unmoored language from the humanist subject.^[8]

There is a similarly curious misremembering undergirding certain uses of computational literary analysis. Certainly, not every practitioner appeals to such justifications, but at least some have claimed (whether out of conviction or for grant funding) that they were making the study of literature newly scientific. The structuralists made the same claim for their approach, so in reality whatever shift has taken place is not from nonscience to science, but from theory-driven science to data-driven science, a turn paralleled in this period in the natural sciences.^[9] Data-driven science is often cast as ‘theory-free’, with the pretense that the knowledge emerging from it is unencumbered by human finitude.^[10] In reality, whether one looks for patterns in data from nature or from literature, everything from how data is collected, to what correlations are deemed significant, to how those results are interpreted, requires the whim of the interpreter. Here as well, the claim of newfound scientificity required the repression of this same claim within the structuralist generation.

The Structure or Structuralism of LLMs (in Deconstruction)

Weatherby demonstrates, thoroughly and adeptly, that the form of machine learning that has produced the present iteration of LLMs reconfigures many of the models and debates that shaped cognitive science and linguistics. Earlier theories grounded language either in the evolved capacities of our cognition or in the cultural contexts of its usage; its true nature was said to lie in its syntax or in semantics; and various attempts to model language or reproduce it programmatically set out on the basis of these opposing theories.

Philosopher Juan Luis Gastaldi gives a lucid illustration of how a model called word2vec transformed these premises.^[11] This algorithm was an important milestone in the development that led to contemporary LLMs; it sets out from no more than examples of language use and trains itself by attempting to predict what element will follow a given element in a text. It treats each unit in the corpus as text and context; so, one might characterize each word in terms of the five words immediately following it. Originally, this algorithm focused on the level of the individual word, but subsequent algorithms make use of sub- and supra-lexical elements. Each element is represented as a vector, which means that the resulting matrix can be treated as an ‘embedding space’ in which each element appears a certain distance from all the others. A striking discovery that followed from this model was that the relationships among words in this embedding space faithfully reconstructed some syntactic and semantic lexical relationships.

In other words: with no information beyond language usage—no cognitive structure guiding learning, no access to the world beyond language, no predigestion of syntactic structures or semantic hierarchies—syntax and semantics can be reconstructed, at least to a degree, from statistics. Of course, we are all familiar enough with similar algorithms by now to know that these reconstructions are imperfect. Regardless, this has been taken, by Gastaldi, Weatherby, and others, as vindicating a structuralist view of language, according to which language’s essence is defined by the differential relations of its signs. (While this is a step in the right direction compared to the subjectivist/intentionalist interpretations, it seems to me that significant differences may remain. Saussure’s structuralism was not statisticalist in any legible way—the structure of language was made up of signs, with each word or ‘acoustic image’ seeming to be equally weighted. This langue or language structure was said to exist in the brain of speakers and listeners—it removed itself completely from speech or parole.^[12] On the other hand, these statisticalist algorithms digest language usage into a functional model, the matrix model representing words as predictive vectors, and thus are not simply equivalent to language usage, to speech or parole, themselves.)

All the effects that have been observed to follow from the use of LLMs—which, even if we remain thoroughly skeptical of the inflated claims of their powers, are certainly not nothing, and remain worthy of analysis—start from here. (The generation of LLMs that gave rise to the present investment bubble were based on a more sophisticated machine learning algorithm than word2vec, but one that also digests statistical relationships of elements within language usage.) In a way, it is more striking to recognize what is really happening: the impression received by many, interacting with such algorithms (through the many layers of mediation noted above), not just of receiving expressions of knowledge or sympathy but of receiving them from a knowing and feeling subject, are effects that can be reproduced merely from the statistical relationships of words in a representative corpus.

Against both the boosters and detractors of AI who insist on the irreplaceability of the knowing subject, Underwood writes, ‘It is hard to imagine a clearer vindication of a thesis that linguists, critics, and anthropologists spent much of the twentieth century advancing—the thesis that language is not an inert medium used by individuals to express their thoughts but a system that actively determines the contours of the thinkable’.^[13] This systematicity of language is precisely what a deconstruction of structuralism was meant to puncture; if language is an active agent, this implies outside it some passive substrate (here it is thought or thinkable meaning), and so one feigns knowing where these apparently different systems begin and end. Thus, this very claim that thought or the thinkable is unthinkable outside of language must grant itself the pretense of thought as a virgin substrate, receiving the impression of language. Though he is frequently misrepresented as a philosopher of language and associated with something like a Sapir-Whorf hypothesis or ‘linguistic turn’, Derrida was in fact deconstructing the autonomy of language as much as anything else. Deconstruction sets out from there never having been a simple origin, of language or anything else.

In ‘The Supplement of Copula: Philosophy Before Linguistics’, Derrida wrote about an almost identical claim—regarding the determination of thought by language—in the work of structuralist linguist Émile Benveniste.^[14] Benveniste sought to demystify Aristotle’s categories, by arguing that what philosophers take to be categories of thought and being are really categories of language. ‘Substance’ is really the substantive, etc. Yet, this very gesture must grant itself precisely what it claims to be impossible. In order to claim that language and thought or philosophy are one, it distinguishes and compares language and philosophy: ‘Benveniste himself can only repeat the operation he imputes to Aristotle: to distinguish saying from thinking (these are his words), and to consider that there is only an empirical relation between them’. For Aristotle, this empiricity means treating language as the external and contingent clothing of thought; for Benveniste, it means that a particular, historical, and natural (concepts that are not independent from the philosophical tradition) language determines thought in this instance. In other words, this particular language is not thought itself, can be recognized as an imposition on thought to the extent that it can be recognized as particular, empirical, contingent, etc.

We see the scope of the problem when we recognize that even opposing Benveniste’s thesis depends on the same premises. A philosopher, Jules Vuillemin, attempted such a reversal, by pointing out that Aristotle selects and organizes linguistic givens when forming the categories, which thus are not mere reflections of their language. In Vuillemin’s words, ‘this philosophical language separates itself as much as necessary from the suggestions primarily imposed by the Greek language’ (qtd. in ‘SC’, 194n24). With regard to this separation of philosophical and natural language, Derrida writes:

The general presupposition of [Vuillemin’s] discourse seems to be the—symmetrical—opposite of that which supports Benveniste’s analyses […]: the contents of thought are essentially, principially, and structurally independent of language, despite the ‘borrowings’ and ‘suggestions’. […] The specular symmetry of the present theses, their profound resemblance in in(de)terminable opposition, from the outset and by itself would invite a reelaboration of the problem; a reelaboration in which one would not in advance take as given, and as if it went without saying, with a feeling of familiarity, mastery, and ‘knowledge’, the access to the ‘essence’ of ‘thought’ and of ‘language’, to their opposition or identity. (‘SC’, 194n24)

Thus, the structuralist declaration of the agency of language rests on the same premises as the philosophical gesture that declares thought’s independence of language: the pretense both philosopher and linguist grant themselves of being able to distinguish pure thought from its empirical clothing. Even in order to claim that thought has been limited, distorted, or created by language, one must grant oneself the pretense of knowing thought and language as originally independent and purely self-identical systems.

Thus, the linguist remains a philosopher. Derrida would not want, by pointing out this complicity, to pretend that philosophy was somehow sovereign over linguistics; he does not endorse ‘the most traditional philosophical objection’ (‘SC’, 193), found already in Benveniste’s text, according to which recognizing the empiricity of the empirical (such as a particular language) requires knowing already the universal and necessary, and thus elevates philosophy above all other knowledge domains. Rather, Derrida recognizes such philosophical questioning as necessary and impossible at once—in an impossible simultaneity—which destabilizes both linguistics and philosophy, the regional and the general (this is what he means by describing philosophy before linguistics; philosophy is both prior to and before-the-law of linguistics). The linguist acts as if he knew already the philosophy he claims to overshadow, while philosophy necessarily depends on all the ‘particularities’, including those of ‘natural language’, that it would hope were its subdomains. A break with this specular mirroring, a deconstruction-event, could come to pass only where that shared ground is questioned, where true thinking is no longer represented as a return to an original and lost authenticity. (Derrida concludes this essay by noting that even Heidegger, who aimed at a certain deconstruction of metaphysics, still retains this presupposition of a fall from an original authenticity).

Misreading and Rereading Saussure and Derrida

If it is true that LLMs have renewed the relevance of structuralism, then Derrida’s recognition of constitutive aporias within structuralist thought, indeed the very aporias of the philosophical tradition certain structuralists claimed to surpass, might have some pertinence as well. It would not be possible to stage that conversation on the basis of Weatherby’s tendentious reading of Derrida. Rather than trying to correct it in every particular, the most revealing approach is to show in an example what reanimating and slaying Derrida’s ghost does for Weatherby. In a word, it is the prop supporting his claim over ‘language as such’.

Weatherby gives a slightly different analysis than Gastaldi does of how to situate structuralism within existing approaches to language from cognitive science and natural language processing. For Weatherby, structuralism provides a third way, navigating an impasse that he sketches out between ‘syntax’ and ‘statistics’ as approaches to understanding language. The syntax view, derived from Chomsky’s work, reduces existing languages to a set of generative operations said to be innate in human minds and capable of learning and producing any language whatsoever. While this provides an explanatory account of language, it does so by distilling its manifestations into an essence that has left many researchers dissatisfied. In particular, it seems to leave an explanatory gap between these essential principles or operations and the actually existing languages they are said to govern.

While the statistical analysis of language certainly works more closely with language-existence (‘external’ or ‘E-language’), Weatherby sees it as ‘recreating the problem that the syntax view produces from the other side’:

If the syntax view cannot fit its rules to the artifacts [i.e., existing languages] those rules must somehow have produced, the statistics view of linguistic meaning has to leave the rules in the black box which leaves the normative element of all linguistic meaning floundering in approximations that fail to account for abstraction. (62)

In short, while statistical analysis has proven quite successful at the reproduction of any given language, its critics nonetheless feel that it provides no explanatory understanding of language whatsoever. Syntax would account for essence without existence, statistics for existence without essence, with the possible implication that neither explains much at all.

According to Weatherby’s idiosyncratic reading of Saussurean structuralism, it is uniquely situated to ‘dialectically’ bridge this divide (68–69). If language’s structure is composed of the differential relations of signs, Weatherby argues that this establishes its empirical grounding in an actual language while nonetheless providing explanatory value: ‘[T]he empiricism of the structure view splits the middle between syntax and statistics in constructing an empirical domain. It does this by way of semiotics, by insisting that language is a concrete object only insofar as it is a series of signs. This will mean that language is modellable enough for there to be computational language, but not in such a way as to be reduced to number’ (75). This attempt to construct an impasse then overcome it may be a bit too neat: Weatherby seems to be collapsing the distinction between langue and parole—language structure and speech (language use)—which is foundational to structuralism. Moreover, trying to import speech and other manifestations of language into langue or structure itself plays a curious trick, attributing bits of Derrida’s deconstruction of Saussure to Saussure himself; Weatherby performs a similar operation when collapsing Saussure’s distinction between static and evolutionary (synchronic and diachronic) linguistics, even describing the language system as the ‘drift’ of the values of its signs (65). Nonetheless, in order to secure his grasp of the structure of language, ‘language as such’, Weatherby feels that he must exclude Derrida’s reading.

In broad strokes, Weatherby argues that Derrida went looking for a ‘metaphysical’ premise in Saussure’s linguistics, the premise of the privilege of speech over writing. Because of Derrida’s fixation on this ‘external’ discourse of philosophy (‘a metaphysics outside of semiology’), he failed to recognize what Saussure had really discovered, language as such, language as structure (71). We have just seen that this premise of a philosophy external to linguistics, of an original purity of these domains, is precisely what Derrida places in question when he reads structuralists.

In order to banish philosophy from the scene, from any claim over a language restored to its purity and thus to the pure authority of the structuralist linguist, Weatherby denies that Saussure has any real interest in what he (Weatherby) deems a ‘metaphysical concept’: the priority of speech over writing. ‘The model does not require speech to be privileged’ (72).^[15] Weatherby is engaged in an odd—but not atypical—pattern of argumentation here. He has, apparently unwittingly, taken a premise from Derrida’s text and treated it as a critique of Derrida. In fact, such gestures were so common in attempted criticisms of Derrida that Derrida gave them a nickname in his famous debate with John Searle.^[16] He referred to Searle’s argumentative technique, borrowing ideas from Derrida’s text and then pretending they critique that very text, as the objection à Sec (‘LI’, 47),^[17] ‘Sec’ being an abbreviation for the title of the essay Searle was critiquing, ‘Signature Event Context’, as well as the French adjective meaning dry. The English translator, Samuel Weber, sometimes renders the phrase à Sec as ‘from/to Sec’, given that these ‘objections to Sec’ are nonetheless derived from it (à has senses that move in both directions). Weber also suggests we hear these as ‘Dried-out-objections’, emphasizing their shopworn, twice-used, and hackneyed quality. Indeed, à sec can also refer to being broke, and Derrida might mean to suggest Searle is so hard up for ideas he’s reduced to begging, borrowing, or stealing. À sec has the further sense of ‘without lubrication’ which adds to the sexual subtext running throughout Derrida’s response—a text which is, as he later admitted, ‘not devoid of aggressivity’.^[18]

There is undoubtedly a reason for the recurrence of this argumentative strategy. Deconstruction is not a positional discourse. So, in the present case—Derrida’s reading of Saussure in Of Grammatology—he is not arguing simply that Saussure privileged speech over writing, nor is he arguing the opposite (that linguistic structure has no special relationship to speech).^[19] He is recognizing the undecidability of these theses or positions, and thus that Saussure’s self-contradictions are symptomatic of that undecidability, and betray an attempt to repress or control it. So, Derrida himself argues that Saussure’s structuralism makes the privilege of speech over writing untenable, that Saussure recognizes this and makes it explicit—yet, nonetheless, Saussure himself still argues for that privilege.

When an interlocutor such as Searle or Weatherby hopes to reduce such undecidability, to occupy one position against the other, the easiest way is to pretend that Derrida maintains the opposite. Then, one can argue as if positionality and oppositionality have been restored.

So, Saussure maintains both that ‘language is a form and not a substance’ and that it finds its privileged representative in a phonic substance that excludes systems of writing.^[20] In Of Grammatology, Derrida notes the paradoxical effects that these dual tendencies have had for the reception of Saussure: if Saussure argues, at the same time, that language both is and is not uniquely embodied in speech, it is possible to take either position oneself while claiming to either agree or disagree with Saussure.

And, in fact, this is just what happened. Derrida gives a brief overview of Saussure’s successors to draw out this pattern:^[21] Roman Jakobson claimed that he agreed with Saussure by privileging phonology within linguistics. Martinet, on the other hand, claimed that he was breaking with Saussure by doing the same thing as Jakobson (privileging phonology). Only Hjelmslev and Uldall recognize the competing tendencies within Saussure. Thus, they recognize that they are at one and the same time breaking with Saussure and building upon his legacy by purifying linguistics of phonologism. Though it is not a perfectly tidy taxonomy, we can formalize this classification:

	+Saussure	−Saussure
+Phonologism	Jakobson	Martinet
−Phonologism	Hjelmslev/Uldall	Hjelmslev/Uldall

Weatherby is yet another Saussurean acolyte who has slotted himself into this grid, with the coordinates (+Saussure, −phonologism).

The strain of Saussure’s text that banishes writing serves the same function for him that banishing Derrida does for Weatherby. (We might instead say conjuring Derrida, needing to revive him and bring him near in order to exorcise or distance him again and again.^[22]) Structuralism’s claim to scientificity is anchored in its representation of the structure of the language-system, embodied in the differential relations of its signs. As Derrida writes, ‘the condition for the scientificity of linguistics is that the field of linguistics have hard and fast frontiers, that it be a system regulated by an internal necessity, and that in a certain way its structure be closed. The representativist concept of writing facilitates things’ (OG, 33). By ‘representativist’, Derrida means writing is treated as external and contingent in its relation to speech and thus to language proper, precisely those properties traditionally attributed to the signifier (the contingent clothing of an idea) and elsewhere displaced by Saussure.

Similar to Saussure, Weatherby would like to argue that structuralism, LLMs, and he himself have access to ‘language itself’ or ‘language as such’. Though he has otherwise misread Derrida, he has correctly perceived that Derrida poses a threat to this integrity of language. If something external to the system of language so defined, such as properties of writing not reflected in speech (like the a of différance), nonetheless displace its ‘internal’ structure, there is no pure inside of language or structure in general, any more than there is a pure externality. Far from an agency of language, what-acts-on-what is undecidable once language and everything else have no rigorous boundaries. This does not mean that it is dissolved in a soupy indistinction where we wallow in indifference—it is precisely because of this undecidability, the insecurity of borders, that we make a genuine decision every time we intervene. We will see in a moment what effects this has for the discourse today surrounding ‘AI’.

Derrida on Cybernetics: The Writing Machine as Relay

Derrida certainly will not return us to the anchoring of language or meaning in the intentions of a human subject (though, as we will see, he would not exclude altogether the pertinence of intention). Nor will he ground meaning or the subjectivity-effect in linguistic structure as Saussure understood it. To understand how Derrida deconstructs our models for the relationship of thought and machine, we can turn to his own writing on the thinking machines of his day.

In ‘Freud and the Scene of Writing’, Derrida examines the many metaphors and models of the psyche found in Freud’s texts that picture the psyche as some sort of writing process or technology (as we will see, he is aware of and aims his analysis at the cybernetic servomechanism as well).^[23] On the one hand, he is demonstrating (despite Lacan’s frequent claims of making a ‘return to Freud’) that Freud’s psyche-as-writing models are more advanced than Lacan’s epochal declaration that the unconscious is structured as a language, which continues like other structuralist models to privilege speech. On the other hand, Derrida shows that Freud still tries to control or repress certain of the effects of self-exteriority that are legible within his own scriptural models.

The last of these writing-thinking-machines Freud has recourse to is the ‘Mystic Writing-Pad’, composed of a wax base, a two-layer ‘transparent sheet’, and a stylus, an apparatus he takes as an image of the psyche. We must consider this model on two levels to understand what it can say to us or rather what is inscribed in it or on its surface as our psyche. I’ll follow a long paragraph where Derrida’s analysis reaches its conclusion:

Far from the machine being a pure absence of spontaneity, its resemblance to the psychical apparatus, its existence and its necessity bear witness to the finitude of the mnemic spontaneity which is thus supplemented. The machine—and, consequently, representation—is death and finitude within the psyche. (‘FSW’, 228)

Let’s start with what might be the traditional understanding of such models or modeling in general, a metaphysico-scientific tradition of which remnants remain legible in Freud: psyche or mind (and with it life) are spontaneous and self-identical, they move themselves and know themselves, while a machine or artifact is heteronomous, it must be created and operated by the other, a living other. Furthermore, the very concept of a model mirrors this logic: the model should be an image or representation, an external resemblance outside the thing it models, whose integrity and self-identity are untouched by it (like the Mystic Pad, the model in general is an aide-mémoire, assisting thought supposedly without changing it or its object). Thus, this particular model, writing-machine-as-psyche, is intimately related to modeling in general; it represents, or attempts to represent, pure internality by recourse to self-externality. In the same vein, the model and the machine are metaphors, representing supersensible or intelligible internal qualities by means of sensible externalities.

It is this traditional logic that is breaking down when the psyche, life’s own self-relation, is represented as a writing machine. In order to achieve self-relation or self-reflection, in order for the supposedly purely internal, auto-affecting psyche to become what it is, relating to itself as itself, it depends on recourse to what is supposedly most unlike it, to the externality it would exclude.^[24] This dependence manifests both when psyche relays through marks or traces left in the world, through writing technologies, to accomplish its own thought or memory, and when it takes those writing-machines as models of itself and its inner self-relation. The logic distinguishing being from resemblance by distinguishing psyche from machine trembles or breaks down (illness, disorder, system failure, organic and machinic malfunction): in order for the psyche to be what it is, itself, it must be what it is not, mere resemblance or mere imitation of self-relation, which is to say: model or machine. Thus, as Derrida writes here, what this model and modeling-as-such would hope to quarantine outside the modeled nonetheless persists within it: death and finitude (the absence of spontaneity and autonomy), machine (which requires an external operator or power source), and representation (again, as external to its represented content) are within the psychic structure.

At the same time, if this requires us to picture our (who, we?) psyche as always already in a circuit with the world, the machine, the inorganic and so on, that is not the ‘pure absence of spontaneity’. Agency-effects can emanate from any pole or relay within this structure, even a device or mēkhanē as simple as the Mystic Pad.

Derrida continues this reflection by noting the continuity of the Mystic Writing-Pad with the cybernetic machines of his day:

Nor does Freud examine the possibility of this machine, which, in the world, has at least begun to resemble memory, and increasingly resembles it more closely [et lui ressemble toujours davantage et toujours mieux]. Its resemblance to memory is closer than [Beaucoup mieux que…] that of the innocent Mystic Pad: the latter is no doubt infinitely more complex than slate or paper, less archaic than a palimpsest; but, compared to other machines for storing archives, it is a child’s toy.^[25] (‘FSW’, 228)

Alan Bass’s translation here risks sounding as if Derrida is drawing an opposition between the Mystic Writing-Pad and the memory-machine; as if Derrida would take Freud to task for failing to consider, in 1925, the computer or cybernetics. Rather, the Mystic Pad is a primitive example of ‘this machine, which, in the world, has at least begun to resemble memory’, and it is the shared possibility of it and its successors that Freud is not sufficiently examining or questioning. The following sentence makes this continuity obvious:

This resemblance—i.e., necessarily a certain Being-in-the-world of the psyche—did not happen to memory from without, any more than death surprises life. It founds memory. (‘FSW’, 228)

In other words, it did not wait for the first ‘computers’ (in the typical sense), let alone for LLMs, for machinic self-exteriority to befall thought; these are not historical accidents displacing something that was originally self-identical or self-contained. This alterity ‘does not happen to memory from without’; far from preserving an inner sanctum, this means that the inner only attains its interiority by relay through what it is not, the psyche in the world, the world in the psyche.

Then, whatever machine is chosen as emblematic of thought, its status cannot be reduced to that of the illustrative metaphor or scientific model (again, a would-be relation of exteriority):

Metaphor—in this case the analogy between two apparatuses and the possibility of this representational relation—raises a question which, despite his premises, and for reasons which are no doubt essential, Freud failed to make explicit, at the very moment when he had brought this question to the threshold of being thematic and urgent. Metaphor as a rhetorical or didactic device is possible here only through the solid metaphor, the ‘unnatural’, historical production of a supplementary machine, added to the psychical organization in order to supplement its finitude. The very idea of finitude is derived from the movement of this supplementarity. (‘FSW’, 228)

The ‘solid metaphor’ here is the Mystic Writing-Pad or any other writing-, memory-, or thinking-machine, not sublimated or abstracted as a rhetorical figure, but just as it is, if there is such a thing. Before one can imagine that there is an essence of psyche, and thus a literal meaning that the machine would only supplement rhetorically, the psyche must do any number of things ‘in the world’ to supplement itself, to enable itself to come to know itself through the auto-affecting circuits of thought, language, and writing. External, inessential, mere resemblance achieved at the price of heterogeneity is the ‘nature’ and ‘being’, the ‘internal’ ‘essence’, of what is to be modeled or metaphorized. Freud could never have described the psyche (himself) this way, and we could never have read it, not only if that model-machine did not exist, but also if he did not have the technologies of writing that record it and make it legible beyond what we think of as his living self-consciousness. Not simply because the range of ‘communication’ is being extended (not because we arrived too late to speak with him in person), but because prosthetic supplementation first makes the ‘origin’ possible: first makes the psyche a psyche by making it other than itself, exposing it to alterity.

In fact, when a fragment (e.g. the Mystic Pad) of this historico-technological relay is then treated as a ‘metaphor’ or ‘model’, if these are taken in their traditional sense as mere props to understanding then one has actually obscured precisely what one pretends to be ‘representing’. In this case, the metaphor-model is meant to make intelligible interiority and intelligibility themselves. One treats as external what constituted the inside, the thing to be modeled, and in the process tries to restore the pure, spontaneous sphere of interiority. Thus, Derrida goes on to say, the sort of ‘metaphor’ we are dealing with is categorically different from this traditional notion, what he calls ‘intrapsychical metaphor’, which may not exist in any purity:

The historico-technical production of this metaphor which survives individual (that is, generic) psychical organization, is of an entirely different order than the production of an intrapsychical metaphor, assuming that the latter exists (to speak about it is not enough for that), and whatever bond the two metaphors may maintain between themselves. Here the question of technology (a new name must perhaps be found in order to remove it from its traditional problematic) may not be derived from an assumed opposition between the psychical and the nonpsychical, life and death. (‘FSW’, 228)

This nonoppositionality of psyche and its others or life death entails that countless events or inventions may come along to displace this relay-relationship (changing both how it operates and how it pictures itself to itself, changing the one by changing the other), but certainly not because some special thinking machine would one fine day dislodge the inner essence from itself. Rather, the essence of interiority or thought is already exposed to anything that can come to pass in history or the world, from the ‘beginning’.^[26]

In other words, the question of our relationship to this or any writing machine (indeed, this would be a better name for what we call ‘artificial intelligence’) cannot simply be reduced to how well or poorly it imitates us. Such a question presupposes that we are off to one side, as a self-contained thinker, awaiting our reflection or passing judgment on its failure. A writing machine cannot be reduced to a relationship of either identity or difference with the thinking subject. It is insufficient to either claim that it is a self-relating subject itself, or that it is not, that it fails to imitate us by merely imitating us. Anything that happens or comes to pass in what we call history or the world may change how thought relates to itself, how its circuits of auto-affection are accomplished, by displacing (subtly or massively) both how it pictures itself and how it actualizes those circuits. (So, if an LLM seems to highlight by its shortcomings some creative capacity or thought process that remains uniquely human, that does not mean that we have identified the essence of the human subject, but just the opposite, that in transit or relay—traduction—with a given model, we receive a certain relationship to ourselves. This self-relation inevitably changes with new models and technologies, without that meaning either that the original human essence has maintained itself or that it has been reproduced outside of itself as a machine-subject.)

Even though Weatherby’s structuralism attempts to discard the thinking subject, it retains certain presuppositions of AI-boosterism merely displaced onto ‘language as such’. Weatherby would certainly not say that LLMs are the first example of truly thinking machines. Nonetheless, he does make an epochal or apocalyptic claim for them, that they are the first to truly do language (204). Of course, this requires that language be an essence unto itself, inaccessible to machines until now. It requires ignoring all of the chatbots that since the 1960s have successfully inserted themselves into our language-relays, and more fundamentally, the tele-communications devices that mobilize our thought, memory, speech, and writing from the beginning (Weatherby is undoubtedly aware of such devices, but he treats them as essentially distinct from the LLM, a view we are placing in question). It is to displace each of these presuppositions, language as a distinct power, properly belonging to the human being but wrested from our unique control only recently by a newly empowered machine, thus heralding a conceptually delimited and linear epochal break, that I distinguish the writing machine from what Weatherby calls ‘language machines’. The writing machine would already find its gears turning or its relays in motion wherever a hand wields a pencil or chisel, and even with the hand or body ‘itself’, as well as ‘language itself’. That is certainly not to equate these technologies, prostheses, or supplements or efface their differences, nor is it to create a genealogical continuity; these ‘origins’ can be seen as such only in light of some present or rather its anticipated future.

Because of Weatherby’s epochal claim, it is worthwhile to dismiss a potential misreading of Derrida’s ‘Freud and the Scene of Writing’. If we were to think that Derrida could say such things about the modeling of the psyche and language because he dealt with quite primitive writing machines, and with an old-fashioned (and quite odd) neurologist, but that we had left them behind by entering a new machinic age, we would be mistaken.

Derrida wrote about the advent of cybernetics in more detail than just the passing reference quoted above.^[27] His recently published seminar, Life Death, provides far more context for a Derridean reading of this generation of cybernetic or writing machines. Francesco Vitale has written at length about this text in his Biodeconstruction, and also places it in the context of ‘Freud and the Scene of Writing’.^[28] Derrida gives a lengthy reading in Life Death of geneticist François Jacob, who played a pivotal role in the revolution of molecular biology, which brought cybernetic notions into the life sciences to give us the concept of the genetic program, of life as a language machine.^[29]

The understanding of DNA as text, an alphabet and genetic code, was undoubtedly transformative for the life sciences and for life itself. When Jacob elaborates this view, as what he calls the logic of the living being, he understands himself in a more or less traditionally scientific and metaphysical fashion to be using an external model (text, code, program, the computer or computation as combinatoric language) to allow an object (the living thing) to be understood by the scientific subject (Jacob himself, and all biologists). Knowing subject, known object, and mediating third term would form something like a scientific if not dialectical synthesis. Yet, text, computer, and all technē, machines or artifacts, are themselves creations or extensions of living beings, constitutive of the self-reproduction taken to be the definition of life, and often treated as themselves self-reproducing by cyberneticians including Jacob; these models do not have the relation of exteriority to the modeled that would allow them to merely model, metaphorize, or resemble life (they are members of the set or system they pretend to define). Moreover, the scientific subject, the life scientist, is also textual in nature; not just in the sense that he inherits from the textual tradition of biology or that he composes a text in order to define the object of his study, but he himself is a living thing, a member of the class he studies. And all those living beings (the known objects) have been defined in turn as textual in nature or essence. The consequences of this breakdown of trinitarian mediation are far-reaching, as Derrida observes:

The text is thus this time the dominant model. But can we blithely accept this so seemingly straightforward claim? Jacob speaks often of model, of image, of analogy, of comparison—these are all his words. But before refining my question and coming a bit closer to Jacob’s discourse, let me indicate the general form of the problem I would like to pose: is the text or is textual language—the silent text, since it is always a matter here of a gram (an engram or a program) that is voiceless—something, something determinate that can provide a model for some objective knowledge (so, is it something, on the one hand, and is it something that can provide a model, on the other) without entailing a transformation in the structure of knowledge, in the objectivity of knowledge, in the referentiality of scientific discourse, in the very concept of model, a transformation so complete that the very axiomatic that subtends all these statements would have to be structurally altered? If the object, the referent of a scientific text (and science is a text), if the object, the referent, of a scientific discourse (and science is a discourse), if this object and this referent are no longer meta-­textual or meta-­ discursive realities, if their very reality has a structure that is analogous to or fundamentally homogeneous with the structure of scientific textuality, if the object (the living, which is to say, reproducibility), the model, and scientific subjectivity (the knower, etc.) have an analogous structure, namely, that of the text, one can no longer speak as one does elsewhere of a knowing subject, of a known object, and of an analogical model. What is more, and for the same reason, what we human beings claim to take from culture as a model, namely, discursive texts or computers and everything we believe we know and are familiar with under the name text, what we then claim to take as a model, comparison, or analogy in order to understand the living at its most elementary level is itself a complex product of life, of the living, and the alleged model is external neither to the knowing subject nor to the known object. There is nothing fortuitous or external about the living being that we are producing things (texts) that might seem to serve as a model for the knowledge of the living, nothing fortuitous about this analogy of structure. The text is not a third term in the relation between the biologist and the living; it is the very structure of the living as the structure common to the biologist, as a living being, to science, as a production of life, and to the living itself. Since this cannot be said about every region of knowledge or about every model or every appeal to models, should we not admit that we are no longer dealing with just a model, that the science of the living or the logic of the living is not one region among others in the scientific field, and that there must, therefore, follow from this powers and risks that are absolutely unique?^[30]

When the scientist claims to know life by means of text-as-external-model, what in fact takes place is that text knows text by means of text. The result is not a plane of textual immanence, not self-identity or continuity, and certainly not the predicate Jacob took as definitive of the living: self-reproduction. What takes place, in and as life and life-science, differs from itself in just the fashion described above when the psyche took a writing machine as its supposedly external model. Life achieves its ‘self’-reproduction by relay through its ‘inorganic’ milieu or environment, its technological extensions, its viral supplements and so on; its ‘self’ is always already this self-exteriority. When the scientist pretends to encapsulate this process by finding at hand what just happens to serve as a model and metaphor, he is simply doing what life has always done, accomplishing his ‘self-reproduction’ (creating a text that reproduces life, the logic of life) by means of what is neither self nor other, what Derrida calls life death.

Life and psyche or ‘intelligence’ are not circumstantially modeled or mirrored. They have model or mirror within them, which as we have just seen changes the value of both (life/psyche on the one hand, model/metaphor on the other). Derrida finds an elegant figure for this incorporation of the mirror in Psyche: Inventions of the Other. Psyché happens to be the French name for a kind of standing mirror: ‘But the mirror named psyche does not figure an object like any other. Nor is the gesture that gets caught wanting to show the mirror just one gesture among others. Whether or not it is granted this right, whether or not it makes of the right a duty, it has no choice but to watch itself showing while listening to itself speak. Is that possible?’^[31] By bringing Derrida into this discourse surrounding ‘artificial intelligence’ or the writing machine, we are drawing out the consequences of a psyche that incorporates the mirror, one whose models are neither purely internal nor external to it.

The Always Already of Artifice

If we understand the writing machine in this way, neither as thinking subject nor as a mere imitation of the thinking subject, it can transform certain habitual discourses that have cropped up around ‘AI’ and the ‘LLM’.

The LLM is no more a self-conscious subject than we are, but this removes one commonplace stopgap for critiques of ‘AI’. Its statements can no longer be claimed as meaningless, or its artworks as devoid of aesthetic quality, simply because it lacks the intentions of a subject. Our own language has never depended on our intentions for its meaning (as we can see, machine writing remains legible—so much so that at times it is difficult to distinguish or prove its provenance—even if we may tend to read it differently than human writing, when we know or think we know its origin). A structuralist view might conclude that language, as a system unto itself, not only does not depend on such intentions but is the matrix in which they are generated, that all the effects we attribute to our own subjectivity or initiative are effects of language, misrecognized. Derrida was never satisfied by such notions of the ‘death of the author’:

Rather than oppose citation or iteration to the noniteration of an event, one ought to construct a differential typology of forms of iteration, assuming that such a project is tenable and can result in an exhaustive program, a question I hold in abeyance here. In such a typology, the category of intention will not disappear; it will have its place, but from that place it will no longer be able to govern the entire scene and system of utterance.^[32]

In other words, intention does not disappear from the scene once it is deconstructed—it is just that it, like everything else, is written and read. It cannot govern the meaning of a statement (we argue over presumed intentions or can dispute explicit intention-claims, can dissimulate intentions and even be at odds with ourselves over them, and can see further possibilities of interpretation or reading beyond a given intention, etc.). It follows from deconstruction not that intentions and statements of them are meaningless and not that language or writing univocally masters the scene, creating a meaning of and unto itself that forecloses all others, but that intentions are textual. In other words, we know them by that relay described above, understanding our own through language, the world, and the other. They exist for us or an interlocutor or witness only to the extent that they can be inscribed (implicitly or explicitly), which means that they are not the origin of language or action, but rather a participant in the scene of writing.

Then, our own ‘intentions’ can be transformed by inserting a novel writing machine into our circuits, and that machine may produce intention-effects just as we do, by displacing those relays, by taking part in the self-exteriority of textuality. That certainly does not mean that we ought to treat the writing of LLMs just as we do writing from other sources (except in the sense that both must be read, one way or another), it only means that the linkage we try to form between text and context is insecure in both cases.

Nor do we need to seek out our inner light to dismiss the artistic productions of ‘AI’. While it is often said that some form of genius, sympathy, or suffering sets apart human artistry, we should recognize that whatever self-relation we accomplish through art-making (a) necessarily subjects itself to established and iterable art-forms, genres, etc., even or especially when it attempts to exceed their limits, and (b) first comes to itself in a relay with such formulaic art (life imitates art, etc.). Again, it is what we have in common with writing machines that allows us to read them otherwise. Context plays a nonprogrammable role in how we read anything, including an artwork. The stories we can tell about a work and its author, the history and tradition within which we contextualize it, and the environing scene in which we situate it all play some role in how we interpret and relate to the work. This context does not determine or constitute reading any more than intention does (nor is it a simple given what counts as context); context participates in the scene. So, anyone is free to imagine that the products of ‘AI’ emerge from a thinking and feeling computer consciousness that they then relate to in an identificatory fashion (as they would any other artist—but perhaps not exactly so), or to see such works as the lazy detritus of a human machine-user or prompt writer, worthy of no more of the reader’s effort than was put into their creation.

Finally, even if we share a disdain for the type of poetic or artistic work thus far created with ‘AI’ (I certainly do), we should not try to theorize its deficiency by claiming that all automated or programmatic processes are inimical to creativity. As Derrida often wrote, it is necessary to think ‘the event with the machine’.^[33] Invention or the unanticipatable is not only possible as a repetition that changes contexts, but every invention takes place this way. Long before LLMs, we had poets and artists making use of much simpler algorithms, and even before the computer, combinatoric processes were used to bypass the censors of consciousness;^[34] moreover, processes of recontextualization (such as found art or poetry) demonstrate the interrelation of invention and iterability. Even if such artistic practices remain matters of taste, I would argue against dismissing them on the grounds that art ought to emanate from the mystical font of pure creativity; if ‘AI’ art disgusts us today, if there is a truly visceral reaction against it, that is perhaps not because of its reliance on machine or iterability (always already the case), nor because of its uniquely bad aesthetic quality (the average and majority of human composition has been similarly slop-like, but may evoke a different reaction); rather, the context and story surrounding this technology is part of our aesthetic experience. That it is dismantling the economic possibilities for human artists and many others, forced upon us by a capitalist system fundamentally hostile to life and nature, serving as the fun-and-innocent front of an ideological campaign building infrastructure and ideological support for ever more totalitarian forms of surveillance, warfare, and governance, all for the benefit of eminently mediocre man-children (too lacking in taste to even appreciate what they are destroying, except for perhaps a barely conscious resentment), this shapes aesthetic experience in a manner that a theory of intentionalism can only obscure.

As above, when we were considering the relationship of the psyche to its writing machines, it is certainly not the case that this essential continuity (which is itself discontinuity and the inessential, the essentiality of the inessential) means all art-makers, whether we would think of them as organism, human, or machine, are simply the same. There is enormous heterogeneity, not just between but within each of these categories—it is to multiply or complexify these differences that I place in question both the thesis of dichotomous conceptual boundaries and that of essential identity.

Hopefully, this points the way toward a criticism or deconstruction of claims of the subjectivity of ‘AI’ that do not fall back on the stopgap of the human subject as a ground of meaning. A deconstruction of these scenes would also differ from their structuralist interpretation, as Weatherby can help us to recognize. By securing the notion of language itself or language as such, Weatherby aims to shore up certain disciplinary demarcations.^[35] Just as LLMs recreate language—often though not always in context-appropriate ways—based solely on their internal models of intralinguistic properties (the statistical relations among terms), Weatherby sees literary language as similarly self-referential. He understands literature or the poetic as dispensing with external reference and reflecting upon internal properties of language. Thus, the study of ‘language as such’ requires a dialogue between the two academic specialties Weatherby happens to possess: those disciplines of computer/cognitive science that have developed ‘AI’ and literary criticism:

A general poetics will study language at this point of meaning generation across the difference between its origins—computational and neurocognitive, in the only two examples we have so far. It will involve literary scholars—the only professionals positioned to do this work properly—but it will not take literature as its primary examples. Language as such, in whatever variety it comes to us, will be its object. (171–72)

This image of the literary theorist or critic, turning away from ‘examples’ of literature to focus on ‘language as such’, might remind us of the man described by Hegel, who goes to a fruit vendor seeking fruit; he is offered apples, pears, cherries, and rejects them all. No, he wants fruit. Weatherby pictures the essence of language and literature as the reduction of its external relations: ‘what has long been called the “literary” aspect of language, the quality that language has before it is turned to some external use’ (128). Literature has never done without reference; even where it is uncontroversially fictive, which is only a limited case of its instantiations, it achieves this not by doing without reference but by making use of powers of referral, a self-externality or exposure to the other prior to history, biography, and other discourses of truth.

The specialist in literature has anything but a hermetic knowledge, not because all meaning inhabits ‘language as such’ as a closed system, but because literary language, like everything else, is exposed to its outside. Any discipline of study and any object such disciplines take or create as their own may find themselves mirrored in literature (and in a more rigorous sense, always already do). A given reading of literature may require knowledge thought to belong to any other discipline, and that reading may in turn transform our understanding of this other. This leads to countless interdisciplinary permutations, and should not give any special privilege to the union between ‘AI’ specialists and literary theory, which is one combination among others, and certainly does not contain the program or essence of all the others. In fact, long before the LLM, literature has been able to put such computational thinking on its stage and parody its pretensions of essential completeness, Borges’s ‘Library of Babel’ being only a relatively recent example.^[36]

Weatherby’s vision for literary studies become a ‘general poetics’ of ‘language as such’, studied ‘empirically’ by focusing on LLMs, trades on a common framework for interdisciplinary work that bridges (or claims to bridge) the humanities and the sciences. Despite his overarching narrative that the humanities have turned their backs on science, in fact his own field of digital humanities is one of many in which humanities scholars often try to absorb the cultural and institutional cachet of the sciences by proximity. His vision of ‘general poetics’ accepts a version of the proper object and method of language-study that is, in the present, quite destructive to the humanities and its disciplines. In short, if our proper object of study is ‘language as such’, as a cognitive or computer scientist would define it, an essence shared by all particular languages, what need is there then for multiple language departments? Indeed, this is the precise logic that is dismantling the study of languages today, our language departments being among the most embattled by austerity-minded administrators.

Of course, I think there is good reason for scholars in the humanities, as well as departments of language and literature, to study computation, digital technology, and so on; but such study ought to be careful to avoid framing itself in ways that might suggest the study of ‘particular’ languages has become obsolete. Moreover, Weatherby’s vision of what remains for us to do may not be as inspiring as he thinks; he laments that humanists have not spent more time arguing with Steven Pinker, for example (69); I am unconvinced that this would have deepened our understanding of language or anything else. To stay with the example of Derrida (as I’ve argued, a symptomatic one for Weatherby), his relentless exploration of the singularity and untranslatability of languages in fact provides a response to pretensions of scientificity in this and all domains.^[37]

‘Language as such’, lost and regained, constitutes Weatherby’s epochal claims. I have mentioned one of these epochal narratives already: his intra- and inter-disciplinary story according to which, when literary scholars abandoned structuralism, language as such became the proper object of the sciences:

And this is how the humanities lost language, allowing both cognitive science and [natural language processing] to update analytical and technological approaches that literary theory rarely engaged. By sweeping structuralism’s focus on a concrete object to one side in the name of opposition to metaphysics, poststructuralism fumbled the object itself. (73)

Beyond the many caveats already raised about this narrative, we should recognize that it participates in the perverse economy in which humanities scholars blame themselves for the political and cultural campaign seeking to dismantle our institutions and the support necessary for our research and teaching to continue.^[38] If resources today are funneled toward particular disciplines and their approaches to studying ‘language’—which may be more of a homonym—that is not because the humanities stopped having something relevant to say about language (on the contrary, it is often that all-too-relevant critiques of the failed aspirations of scientificity, including the biases it exhibits and structural violence it furthers, can be found there). Rather, a concerted effort has been made to dismantle educational institutions and their cultural influence, for which a story such as Weatherby’s merely provides ideological cover. Moreover, the massive investments in ‘AI’ and the rush to adopt them in educational settings are themselves a salvo in this neoliberal push to strip our schools for parts. If deconstruction provides some resistance to Weatherby’s narrative and the political forces it obscures, it is by reminding us that the humanities, language, and everything else receive their meaning or their heading only within some context (as here, where Weatherby’s internalizing and ideal or ideological narrative masks the contextual determinants). Rather than a self-determination from their essential interiority, they are exposed to the manipulations of forces that such essentializing may only obscure.

Second narrative: for Weatherby, the isolation of ‘language as such’ leads to his claim that the LLM represents the first encounter between language and computation.^[39] It is ironic for a self-styled structuralist to make this claim, given that structuralism was precisely a combinatoric and differential-digital model of the essence of language.^[40] Nonetheless, he claims that language ‘itself’ is devoid of computation, and thus that it is a historic event if the ‘two’ are brought together. Weatherby may be motivated merely by the appeal of grandiose epochal or apocalyptic pronouncements, such as this opening salvo:

Language has entered the age of its algorithmic reproducibility. The human ability to speak can no longer define our singularity. […] AI has accidentally divorced language from cognition, raising more questions than it answers. The isolation of language as a sign-system from human cognition is an unprecedented experiment about how meaning is made. (1)

I’ve tried to show how and why deconstruction resists picturing differentiality or the machine as a fall from an original plenitude, autonomy, spontaneity, etc. Beyond the rejoinder of always already, I would note just how oft-repeated, how derivative and uninspired this claim of novelty is today. Indeed, already in Derrida’s day he found several occasions to express some exasperation at the countless declarations of the ‘end’ of this and that, a gesture related to the notion of synchronic structure in structuralism, undoubtedly filtered through Foucault’s ‘historical a priori’.^[41] It finds its prevalence today in the felt impetus to speak to the moment, to achieve the kind of relevance that Weatherby worries we sacrificed voluntarily, even as the demand for it is foisted upon us as a cover for decisions taking place according to different financial and political logics. It is so often rehearsed today that someone attempting to take it seriously (maybe that is already to misread it) would be left to think we were living in the most chaotic and rhapsodic of taxonomies or temporalities, in an age or ages whose definitive trait was apparently every single scholar’s favored research subject. Perhaps this is the age of ‘the age of’.

The claim of novelty is itself the most formulaic, repetitious, machinelike or parroting declaration. If I place in question certain of its pretensions, that is certainly not to position myself against novelty. It is only that a truly unanticipatable event happens otherwise, beneath the possible forms of recognition thus far available to the intentional subject and even its ‘language as such’.

1. I am grateful to Dez Miller, Víctor Betriu Yáñez, Timothy Clark, and Junting Huang for helpful suggestions on earlier drafts of this essay.

   The most thorough and insightful debunking of ‘AI’ I have read has been in the work of Kevin Baker. See, e.g., Kevin T. Baker, “AI Got the Blame for the Iran School Bombing. The Truth Is Far More Worrying,” News, The Guardian, March 26, 2026, https://www.theguardian.com/news/2026/mar/26/ai-got-the-blame-for-the-iran-school-bombing-the-truth-is-far-more-worrying. Baker makes the undoubtedly correct observation that when a technology becomes ‘charismatic’ in the fashion that ‘AI’ has, even the impulse to critique it can reinforce the gravitational force it exerts over our understanding of what has agency in the present. In other words, critiques of ‘AI’ may reinforce its supremacy by attributing to it—to something that may not even exist with any definite boundaries—by attributing to its name vast transformations with more complex or perhaps more traditional causes. In many cases, what we despise and what vexes us is not ‘AI’ but capitalism, for example (though this economic designation undoubtedly has its own ‘charismatic’ effects). I acknowledge that there is a risk, by engaging in this discourse over machine learning, ‘AI’, the LLM, and so on, that by responding to the literature taking this as its guiding focus, even in order to deconstruct its foundations, there is an ever present risk of reinforcing this charisma.

   Baker makes an important further point with respect to this gravitational force of the name, one he calls the ‘bureaucratic double bind’. This double bind exists between any organization’s regulations and the judgment necessary to implement them. Formalized and standardized operating procedures allow an organization to put its decisions forward as impersonal and automatic, principled, logical, or at least inevitable, what we sometimes call ‘faceless’ bureaucracy (which has its advantages for the bureaucrat). Nonetheless, every such procedure requires the judgment of individuals at every step of its implementation—even if one has put a computer in place to execute a number of intervening steps, someone has decided that this computer operating according to this program will be utilized and someone decides to act on its evaluation (or program it to execute some action—the example Baker focuses on is the military ‘kill chain’). At most, one can eliminate from the equation the judgment or conscience of certain bureaucrats or underlings, but only by granting greater scope to the judgment or whims of those setting and keeping this system in motion. Ultimately, automatization (long before it took the form of ‘AI’) serves the dual purposes of control and ideological cover (unilateral control is both secured and denied). Baker’s analysis of the ineluctability of judgment—for which there is no law, as Kant explained, because a law of judgment would require judgment to be implemented—is precise and incredibly pertinent. I would only want to add or emphasize the extent to which this is an important theme in Derrida’s work on justice and the law; indeed, Baker may have had this in mind already by referring to it as a ‘double bind’ (though the term originates with Gregory Bateson, perhaps a more proximal source in this context). See Jacques Derrida, ‘Force of Law: The “Mystical Foundation of Authority”’, in Acts of Religion, ed. Gil Anidjar, trans. Mary Quaintance (Routledge, 2002). ↑

2. Emily M. Bender et al., ‘On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? 🦜’, Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency (New York), FAccT ’21, 2021, 610–23, https://doi.org/10.1145/3442188.3445922. ↑

3. Ted Underwood, ‘The Empirical Triumph of Theory’, In the Moment, June 29, 2023, https://critinq.wordpress.com/2023/06/29/the-empirical-triumph-of-theory/. Bernard Dinysius Geoghegan has also done important work on the connections between structuralism and information theory (he makes only glancing reference to Derrida): Bernard Dionysius Geoghegan, Code: From Information Theory to French Theory (Duke UP, 2023). See also, Lydia H. Liu, The Freudian Robot: Digital Media and the Future of the Unconscious (University of Chicago Press, 2011). For an earlier generation of scholarship on this nexus see, e.g., Anthony Wilden, System and Structure: Essays in Communication and Exchange, 2nd ed. (Tavistock, 1980); Christopher Johnson, System and Writing in the Philosophy of Jacques Derrida (Cambridge UP, 1993). For analysis, see my ‘Grafting Disciplines’, forthcoming.

   Lydia Liu’s earlier attempt to connect cybernetics and information theory with deconstruction, in her essay ‘iSpace’, has certain similarities with Weatherby’s treatment of deconstruction. Lydia H. Liu, ‘iSpace: Printed English after Joyce, Shannon, and Derrida’, Critical Inquiry 32, no. 3 (2006): 516–50, JSTOR, https://doi.org/10.1086/505377 (hereafter ‘i’). She forms a series of temporal demarcations and continuities in order to relate Derrida and cybernetics, together with modernists such as Joyce and the much more ancient figure of the ideogram. It is already a sign of the untenability of Liu’s intellectual history that paradoxical relations of haunting must be invoked in order to delineate its periods. She makes information theory both a pure invention and a continuation of tendencies in Joyce and ideogrammatic writing. According to this schema, Derrida is both indebted to the ‘biocybernetic revolution’ and lagging behind it, a ‘belated modernist’ navigating or mourning in the ‘wake’ of Joyce and Mallarmé (‘i’, 518).

   When Liu considers Derrida’s invention or discovery of a ‘hypermnesic machine’ in the text of Joyce, she distorts its functioning to make it appear retrograde relative to the inventions of the information theorists (‘i’, 520; cf. Jacques Derrida, ‘Two Words for Joyce’, in Derrida and Joyce: Texts and Contexts, ed. Andrew J. Mitchell and Sam Slote, trans. Geoffrey Bennington [SUNY Press, 2013]). Derrida describes Joyce’s Ulysses and Finnegans Wake as more powerful than any computer given the manner in which they are able to seemingly potentiate language with every imaginable linguistic and thematic connection and disconnection. In Derrida’s words, ‘For example, at what speed is the Babelian theme or the word “Babel”, in each of their components (but how could you count them?), co-ordinated with all the phonemes, semes, mythemes, etc. of Finnegans Wake?’ (qtd. in ‘i’, 520). Unexpectedly, Liu reads this as Derrida imposing a phonocentrism on Joyce’s text and on the figure of the machine or computer in general, ‘To expect a computer to think in terms of syllables or recognize things like phonemes, semes, mythemes, and so on is not so much to make a demand on its speed and hardware as it is to ask the machine to be a linguist like Roman Jakobson’ (‘i’, 522). According to Liu, Derrida’s suggestion that there are ‘phonemes, semes, mythemes, etc.’ traversing Finnegans Wake dissimulates the operation of computer language (and Joyce’s language), for which there are supposedly only printed characters understood as statistical probabilities. Of course, if one were to ask what dictates the probability of one sequence of letters over another, recourse to the phoneme and semantics (the word), as well as much broader contexts, would be difficult to avoid. More fundamentally, Derrida is bringing to mind that the ‘information’ imparted by Joyce’s writing, even if it must and can only dance across the surface of Joyce’s text, is nowhere present there. No computation of those letters is simply equivalent to what they make possible, even if that possibility belongs to nothing else. It belongs nowhere (else). The speed of this operation, which we could call reading, is not calculable because it is not a present entity or process that can be placed within definite borders and timed.

   Liu’s insistence on the purity of this information theoretical framework severely restricts the textual interpretations her discourse is capable of. About Finnegans Wake, she can say nothing more than that it is an improbable sequence of letters, a conclusion whose merit she defends without justifying, ‘I suppose that Joyce scholars might gain additional insight into his feat of engineering by analyzing the stochastic dimension of his multilingual riddles in Finnegans Wake. I am aware that this idea may strike some literary critics as frivolous, but it is not’ (‘i’, 535). It is noteworthy that this impoverished language of information leads to no conclusion other than the one English teachers admonish their pupils to strike from their essays, the genius of the author (Joyce’s ‘feat of engineering’). This demonstrates how easily the supposedly pure machine-language can be subsumed by an intentionalist (logocentric) ideal of genius and creativity, while also reminding us that we do something other than what the information theorist does when each of us reads.

   The objective of reading, of any and everything that could count as reading, is not to imagine that the insight into meaning exists in a topos hyperouranios, outside of the computer and its operations. Rather, one must see that even though language has never been anything other than a combinatorics of marks, it has never been simply that either. This out-of-syncness with itself is dissimulated when Liu attempts to form a historical narrative of it, in which a statistics of language is ‘invented’ by Claude Shannon one fine day in 1948 (‘i’ 532). Liu’s ultimate aim is to treat information theory as an advance over Derrida in the understanding of Joyce and of language in general, while nonetheless forging an alliance, which may jeopardize the historical claim, between information theory and ideogrammatic writing systems. We could also look toward the Ancient Greek atomists, and to Borges, to discover a long history of language treated as a combinatoric system. The deconstructive foment within all such systems, prior to or after Shannon’s intervention, is the non-closure of their calculative systems. Derrida’s reading of Joyce, for example, is meant to puncture the very possibility of a systematic analysis or calculation of his textual operations. ↑

4. Víctor Betriu Yáñez, ‘Is a Derridean Critique of Generative AI Possible? Writing Machines and Logocentrism without Subject’, The OLR Supplement, January 15, 2026, https://olrsupplement.com/2026/01/15/is-a-derridean-critique-of-generative-ai-possiblewriting-machines-and-logocentrism-without-subject/; Justin Joque, ‘The AI Con, by Emily Bender and Alex Hanna’, The OLR Supplement, February 17, 2026, https://olrsupplement.com/2026/02/17/the-ai-con-by-emily-bender-and-alex-hanna/; see also Justin Joque, Revolutionary Mathematics: Artificial Intelligence, Statistics and the Logic of Capitalism (Verso, 2022). Daniel Bashir, ‘The Third Yes’, Substack newsletter, Sincerely, in Jest, January 27, 2026, https://thejester.substack.com/p/the-third-yes. This essay has benefited from a chapter Justin Joque shared with me of his Artifices of Intelligence: AI and the Dream of an Infinite Capitalism, forthcoming with Verso, as well as forthcoming work on the subject by Dez Miller. ↑

5. ‘Poststructuralism has on the whole tended to be less concretely interested in language in the sense of strings of words on a page, other theories of linguistics [sic] structure, and even—unfortunately—the automatization of language alongside which the statistics view developed […] Saussure in particular is best suited to parse LLMs. Derrida’s reading of Saussure, in particular, seems to obscure this usefulness in the present. (220n71). ↑

6. For deconstructive readings of ‘new materialism’, see Vicki Kirby et al., ‘How Do We Do Biodeconstruction?’, Postmodern Culture 28, no. 3 (2018), https://doi.org/10.1353/pmc.2018.0021; Jonathan Basile, ‘The New Novelty: Corralation as Quarantine in Speculative Realism and New Materialism’, Derrida Today 11, no. 2 (2018): 211–29, https://doi.org/10.3366/drt.2018.0187; Thomas Clément Mercier, ‘Resisting the Present: Biopower in the Face of the Event (Some Notes on Monstrous Lives)’, CR: The New Centennial Review 19, no. 3 (2019): 99–127; Jonathan Basile, ‘Life/Force: Novelty and New Materialism in Jane Bennett’s Vibrant Matter’, SubStance 48, no. 2 (2019): 3–22, https://doi.org/10.1353/sub.2019.0018; Jonathan Basile, ‘Other Matters: Karen Barad’s Two Materialisms and the Science of Undecidability’, Angelaki 25, no. 5 (2020): 3–18, https://doi.org/10.1080/0969725X.2020.1807132; Eszter Timár, ‘Bacterial Sex and Death: Darwirigaray and Life Death’, Paragraph 47, no. 3 (2024): 273–88, https://doi.org/10.3366/para.2024.0470. ↑

7. Bruno Latour, ‘Why Has Critique Run out of Steam? From Matters of Fact to Matters of Concern’, Critical Inquiry 30, no. 2 (2004): 225–48, https://doi.org/10.1086/421123. ↑

8. See Jonathan Basile, ‘Misreading Generalised Writing: From Foucault to Speculative Realism and New Materialism’, Oxford Literary Review 40, no. 1 (2018): 20–37, https://doi.org/10.3366/olr.2018.0236. ↑

9. Though it is no longer so central to computational literary analysis’s self-representations (we are here reconstructing history or histories), it has been frequently repeated that computational ‘reading’ would make literary study more scientific by grasping empirical totality (all novels of a period or genre reduced to a map or data set). Levi-Strauss saw the scientificity of his approach in opposition to precisely this idea of empirical completeness: ‘If critics reproach me with not having carried out an exhaustive inventory of South American myths before analyzing them, they are making a grave mistake about the nature and function of these documents. […] Unless the population dies out physically or morally, this totality is never complete. You might as well criticize a linguist for compiling a grammar of a language without having complete records of the words pronounced since the language came into being, and without knowing what will be said in it during the future part of its existence. Experience proves that a linguist can work out the grammar of a given language from a remarkably small number of sentences, compared to all those he might in theory have collected’. Claude Lévi-Strauss, The Raw and the Cooked, trans. John and Doreen Weightman, (University of Chicago Press, 1983), 7. ↑

10. For critical analysis, see Cristian S. Calude and Giuseppe Longo, “The Deluge of Spurious Correlations in Big Data,” Foundations of Science 22, no. 3 (2017): 595–612, https://doi.org/10.1007/s10699-016-9489-4. ↑

11. Juan Luis Gastaldi, ‘Why Can Computers Understand Natural Language?’, Philosophy & Technology 34, no. 1 (2021): 149–214, https://doi.org/10.1007/s13347-020-00393-9. ↑

12. ‘Whereas speech [le langage, here: language in the broadest sense] is heterogeneous, language [la langue], as defined, is homogeneous. It is a system of signs in which the only essential thing is the union of meanings and sound-images, and in which both parts of the sign are psychological’. Ferdinand de Saussure, Course in General Linguistics, ed. Charles Bally and Albert Sechehaye, trans. Wade Baskin (Philosophical Library, 1959), 15. ↑

13. Underwood, ‘The Empirical Triumph of Theory’. ↑

14. Jacques Derrida, ‘The Supplement of Copula: Philosophy before Linguistics’, in Margins of Philosophy, trans. Alan Bass (University of Chicago Press, 1982). Hereafter ‘SC’. ↑

15. ‘If it has been common to lose track of the contradictory nature of the signifier, and to use that term to designate the physical aspects of language, that is partly because of Derrida’s misreading here. The signifier in ­ Saussure can peregrinate as much as it wants, but it cannot engage in any “play” without taking material form. Taking material form as a signifier in the value-system of language as such is all that Saussure requires of these “acts.” The model does not require speech to be privileged, and it is hard to see how a signifier could derail some presumed systematicity without being strung together in a sequence, written or otherwise’ (71–72).  ↑

16. Jacques Derrida, ‘Limited Inc a b c…’, in Limited Inc, ed. Gerald Graff, trans. Samuel Weber (Northwestern University Press, 1988). Hereafter cited as ‘LI’. ↑

17. Jacques Derrida, ‘Limited inc, a b c’, in Limited Inc. (Éditions Galilée, 1990), 94. ↑

18. Jacques Derrida, Limited Inc, ed. G. Graff, trans. S. Weber and J. Mehlman (Northwestern University Press, 1988), 113. For example, ‘The gayest thing that Sarl has written [Ce que Sarl a écrit de plus gai], in the “never quite takes place”, is “never quite”. For this slightly too scrupulous nuance, if I haven’t misunderstood it, opens a space for the very thing that should not, should never have taken place; thus, I get my foot in the door. Indeed, it has long since slipped in, and at bottom Sarl may not quite want me to pull it back, at least not too quickly. Or rather, Sarl’s wishes in this regard seem rather paradoxical, caught in a kind of double bind, impelled to do everything to keep my foot there, to prolong the scene’ (‘LI’ 36). ↑

19. Jacques Derrida, Of Grammatology, trans. Gayatri Chakravorty Spivak (Johns Hopkins University Press, 1976). Hereafter, OG. ↑

20. de Saussure, Course in General Linguistics, 122. ↑

21. See Derrida, OG 53–59 ↑

22. See Jacques Derrida, Specters of Marx: The State of the Debt, the Work of Mourning, and the New International, trans. Peggy Kamuf, Routledge Classics (Routledge, 1994), 49–60. ↑

23. Jacques Derrida, ‘Freud and the Scene of Writing’, in Writing and Difference, trans. Alan Bass (University of Chicago Press, 1978). Hereafter ‘FSW’. ↑

24. On this logic of modeling or analogy, especially as it functions in the life sciences, see Jonathan Basile, Natural Lection: Cultures of Evolution (U of Minnesota P, 2026), especially its first and fifth chapters. ↑

25. Jacques Derrida, ‘Freud et la scène de l’écriture’, in L’écriture et la différence (Éditions du seuil, 1967), 336–37. ↑

26. As with any deconstruction, the displacement of a traditional opposition or dichotomy (human/machine, nature/technology) is certainly not meant to level differences or put forward an essential identity. When I say that psyche or life and its writing machines are not two separate processes, this is meant to open each to a heterogeneity or heteronomy that cannot be purely appropriated or reincorporated. This alterity is what each ultimately has in common. Víctor Betriu Yáñez alerted me to a passage in Echographies where Derrida discusses his notion of continuity and discontinuity with respect to the specificity of given technological ruptures: see, Jacques Derrida and Bernard Stiegler, Echographies of Television: Filmed Interviews, trans. Jennifer Bajorek (polity, 2002), 37–39. Cf. Jacques Derrida, The Animal That Therefore I Am, ed. Marie-Luise Mallet, trans. David Wills (Fordham UP, 2008), 29–30. ↑

27. There is a passing sentence in Of Grammatology that has been frequently cited by those who write on Derrida and cybernetics or systems theory. It is misconstrued as saying that cybernetics has exceeded the metaphysical closure: ‘If the theory of cybernetics is by itself to oust an metaphysical concepts—including the concepts of soul, of life, of value, of choice, of memory—which until recently served to separate the machine from man, it must conserve the notion of writing, trace, gramme [written mark], or grapheme, until its own historico-metaphysical character is also exposed’ (OG, 9). The publication of Life Death should resolve any doubts about how Derrida viewed this metaphysico-scientific discourse. Jacques Derrida, Life Death, ed. Peggy Kamuf, trans. Pascale-Anne Brault and Michael Naas (University of Chicago Press, 2020). ↑

28. Francesco Vitale, Biodeconstruction: Jacques Derrida and the Life Sciences, trans. Mauro Senatore (SUNY Press, 2018); see also Dawne McCance, The Reproduction of Life Death: Derrida’s La vie la mort (Fordham University Press, 2019). I have written about this text at greater length in Basile, Natural Lection: Cultures of Evolution. ↑

29. François Jacob, The Logic of Life: A History of Heredity, trans. Betty E. Spillmann (Princeton University Press, 1993); Evelyn Fox Keller, Making Sense of Life: Explaining Biological Development with Models, Metaphors, and Machines (Harvard University Press, 2002). ↑

30. Derrida, Life Death, 80–81. ↑

31. Jacques Derrida, ‘Author’s Preface’, in Psyche: Inventions of the Other, I, ed. Peggy Kamuf and Elizabeth Rottenberg, trans. Peggy Kamuf (Stanford University Press, 2007), xiii. ↑

32. Jacques Derrida, ‘Signature Event Context’, in Limited Inc, ed. Gerald Graff, trans. Samuel Weber and Jeffrey Mehlman (Northwestern University Press, 1988), 18. ↑

33. Jacques Derrida, ‘Typewriter Ribbon: Limited Ink (2)’, in Without Alibi, trans. Peggy Kamuf (Stanford University Press, 2002), 136. ↑

34. More fundamentally, all art is technē, which means that it has relied on our models and extensions of a relation with natural possibility, while nonetheless transforming them. ↑

35. Nina Beguš provides an interpretation and periodization of ‘AI’ that stems from quite similar motives as Weatherby’s. She aims to treat contemporary technology as an epochal rupture from what came before it, a novel relationship of machine and language. Her aim of creating an ‘artificial humanities’ stems from this rupture; in other words, she argues that because computers are now engaged in the arts, there is an important space for humanities disciplines and scholars to dialogue with them and their specialists. She harnesses the energy of novelty while preserving disciplinary demarcations, a standard approach in a certain style of ‘interdisciplinary dialogue’. Her engagement with structuralism and deconstruction is much more perfunctory than Weatherby’s, however; e.g., ‘Postmodernists, inspired by Saussure’s work, claim that language is semantically self-contained, i.e. self-referential. For example, the principal tenet of Jacques Derrida’s deconstruction is that language is unstable and ambiguous, simultaneously holding conflicting meanings. Meaning is thus as much at the center as it is marginal’ (150–51). This attempt to equate the ‘self-contained’ and the ‘unstable’ is quite curious. See Nina Beguš, Artificial Humanities: A Fictional Perspective on Language in AI (University of Michigan Press, 2025). ↑

36. For further examples, reaching back as far as Aristotle and the ancient atomists, see Jorge Luis Borges, ‘The Total Library’, in The Total Library: Non-Fiction 1922-1986, trans. Eliot Weinberger (Penguin Books, 1999); for analysis, see Jonathan Basile, Tar for Mortar: “The Library of Babel” and the Dream of Totality (punctum books, 2018), doi.org/10.21983/P3.0196.1.00. ↑

37. See, e.g., Jacques Derrida, Edmund Husserl’s Origin of Geometry: An Introduction, trans. John P. Leavey, Jr. (University of Nebraska Press, 1989); Jacques Derrida, ‘Ulysses Gramophone: Hear Say Yes in Joyce’, in Derrida and Joyce: Texts and Contexts, ed. Andrew J. Mitchell and Sam Slote, trans. Francois Raffoul (SUNY Press, 2013); Derrida, ‘Two Words for Joyce’; Jacques Derrida, ‘What Is a “Relevant” Translation?’, trans. Lawrence Venuti, Critical Inquiry 27, no. 2 (2001): 174–200; Jacques Derrida, H.C. for Life, That Is to Say…, trans. Laurent Milesi and Stefan Herbrechter (Stanford UP, 2006); ↑

38. John Guillory provides another example of this pattern, a humanities scholar internalizing the blame for institutional failure, which I analyze in Jonathan Basile, ‘The Political Climate: Ecocriticism and the Apocalyptic Tone’, ELH 92, no. 4 (2025): 1199–234, https://doi.org/https://dx.doi.org/10.1353/elh.2025.a977609. ↑

39. Weatherby tries to purify structuralist ‘language as such’ of the digital, in order to claim that the ‘two’ are only recently encountering each other for the first time. Once again, Derrida is the scapegoat for such ‘one fine day’ narratives: ‘Perhaps it is no accident that, after literary theory clashed with form and structure, then forgot them, the “post-theory” moment would be forced to confront its own history, paraded in front of it as a digitalization of the language it held so dear. In the face of this problem, poststructuralism has tended to become a humanism, retreating into phenomenological and other remaining corners as yet untouched by computation’ (137). I would argue that this idiosyncratic storytelling has nothing to do with ‘poststructuralism’, or anything else, other than maintaining the narrative form and taxonomic claims of Weatherby’s discourse. Its interest lies rather in comparison with another discourse by Alexander Galloway, with ostensibly similar aims, trying to situate the legacy of French theory relative to contemporary digital culture. Where Weatherby saw Derrida (his privileged representative of poststructuralism) on the side of phenomenology as against computation, Galloway sees the opposite—Derrida as purely digital (as opposed to analog): ‘Consider the age of écriture […] Consider the notion that the unconscious is structured like a language or that there is nothing outside of the text. This, I suspect, represented peak digitality’. Alexander R. Galloway, ‘Golden Age of Analog’, Critical Inquiry 48, no. 2 (2022): 231, https://doi.org/10.1086/717324.

    Correcting the misreadings here would almost be beside the point. Surely, Derrida or deconstruction would start from the undecidability of analog and digital or qualitative and quantitative, perhaps exemplified by the encounter of phenomenology and structuralism in his works. But reading Derrida, and reading in general, is inimical to the sort of position-taking and epoch-making that these discourses take as their task. ↑

40. The generation of structuralists including Jakobson and Lévi-Strauss was explicitly influenced by what they saw as the common ground between their science and cybernetics; see Geoghegan, Code. ↑

41. Jacques Derrida, ‘The Ends of Man’, in Margins of Philosophy, trans. Alan Bass (University of Chicago Press, 1982); Jacques Derrida, ‘Of an Apocalyptic Tone Recently Adopted in Philosophy’, trans. John P. Leavey Jr., Oxford Literary Review 6, no. 2 (1984): 3–37, https://doi.org/10.3366/olr.1984.001. ↑