Adrian Howlett
Literary Criticism
There is a direct lineage from Sumerian tablets to the abstractions of the Semantic Web. It can be argued that writing is the ultimate technology, so there is a real irony in realising that most writing from the most technologically fixated society of all time, modern America, is rather technophobic. This attitude often extends from writing to writers, but in the case of William Gaddis, things are not so simple. In his fiction, Gaddis treats technology as another character; just as his characters are almost exclusively seen through dialogue, technology is similarly placed, at once autonomous but still often refracted through the words of others. Gaddis’s letters, edited by the eminent scholar Steven Moore and recently republished in an expanded form by NYRB, do much to explain the complexity of the situation. Having previously used either a mechanical typewriter or displayed his calligraphic hand, in the early 1990s, Gaddis began to use a fax machine. While this switch added little more than novelty (and some grainy images), the relationship between medium and style becomes clearer here.
A Gaddis enthusiast previously unfamiliar with his correspondence will be delighted to find it often revels in the same excesses of language and form as his prose fiction. One particularly amusing example depicts a mundane day in the life of the writer in the style of a classified espionage report (L 344-348). This imaginative use of the epistolary format ultimately explores an alternative use of the typewriter as a professional, even political, tool. The fax, however, takes this aspect – something akin to what the influential Joycean A. Walton Litz termed ‘expressive form’ – to a far greater extreme. In a fax to his partner Muriel Oxenberg Murphy, sent during the collapse of their relationship, Gaddis produces a block of prose ‘IN THE STYLE OF THOMAS BERNHARD’ (L 634-635) to locate a literary analogue for the breakdown of a relationship. What is so remarkable is that this produces a singularity of the writer and the person, considering Gaddis’s late discovery of Bernhard’s work led to the distinctive style of his posthumous novel Agapē Agape.
Like Agapē Agape, these letters often attune the polyphonic noise of Gaddis’s writing, honing it down to an equally noisy solo performance. Remarkably, the novel is the culmination of the longest strand of Gaddis’s creative life: a book on the history of the player piano, first envisioned as ‘an exhaustive history of the Player Piano’ (L 118) in 1948, when Gaddis was only 25; much later, he reflected that this became ‘a casualty of overresearch’ (L 589). These letters demonstrate the persistence of the idea, which ultimately only came to fruition as a work of fiction, and then only because Gaddis had found a technologised style (i.e. that of Bernhard) to match it. Gaddis called Bernhard ‘my Cicero’ (L 634), a measure of the scope of this influence, but also surely said with some irony given Bernhard’s erstwhile status as an outsider. Questions of influence were always vexed for Gaddis – the collection includes numerous letters to Joyce scholars denying he had even read Ulysses or Finnegans Wake in full, let alone been influenced by them – so for him to state an influence so boldly is a very significant thing indeed.
The ’whole pile of books notes pages clippings and God knows what’ (AA 1) that opens Agapē Agape is disarrayed, only structured by the (superficially unstructured) prose framework offered by Bernhard. Here, we are given the raw materials of literature and the technological form (i.e. style) that is able to shape them into a literary form. For Gaddis, these raw materials are reified in a manner that is often very literal. This has an especially amusing resonance in J R, in which noise is physically frozen and dumped in the sea. Naturally, this has consequences for art: in a trial of the process, ‘the shards comprising Beethoven’s Fifth Symphony proved more difficult to handle than had been anticipated, and the sequential thaw technique was not entirely reliable’ (JR 714), leading the technique’s inventor Vogel to become injured, ‘ascribing the damage mainly to the strident quality of the musical work’s opening bars’ (Ibid.). The preposterousness of this scene aside, the essential suggestion here is that art has raw materials that can often be ascribed a physical form, but that while technology is intrinsically related to the process of converting these materials into art, this by no means asserts the supremacy of technology over art.
On a reductive level, it is possible to argue that Agapē Agape is the conversion of this discussion into more traditionally philosophical terms, essentially constituting a debate on whether the creative process is a question of techne or episteme. What immediately complicates this theme is its extremity of style, which marks a major break with Gaddis’s previous four novels, each of which occupied a continuum of polyvocality. While the expansive dialogue of those novels could theoretically contain a (potentially rather abstract) dialogue, even a Platonic symposium of sorts, the reduction to the inner world of the individual here denies this possibility.
Although he died twenty-five years ago, Gaddis’s technological thinking makes him a vital voice in considering the present landscape, particularly as questions of artificial intelligence and its relationship to art are asked with increasing urgency. By turning his work into a machine-like system, Gaddis pre-emptively inverts the mechanised writing of AI, producing a view of what the scholar Joseph Tabbi calls the ‘textualization of reality’ (p. 144). If everything is already textual, where does this leave the act of writing literature? For Gaddis, this makes the role of a writer that of an artificer, a Daedalus-type figure weaving with the threads presented by reality.
One of the most striking aspects of Gaddis’s writing process is the attention he pays to literary criticism, most notably Hugh Kenner’s remarkable book The Counterfeiters, which explores the relationship between art and technology against a backdrop of mimesis. The book may have been the catalyst for Gaddis returning to his player piano study, presenting him with a sense of ‘what it might have become’ (L 548). Around the same time (he was working on A Frolic of His Own), he was introduced to the work of Walter Benjamin by the scholar Gregory Comnes, allowing the technological undercurrents that had long existed in his work to become conscious elements. Two brilliantly perceptive critics, Kenner and Benjamin each discussed the art-technology relationship in ways which have remained remarkably salient; The Counterfeiters has such a broad reach that it can even inform our present discussions around artificial intelligence. The fact that its title (presumably unintentionally) echoes both the name and content of Gaddis’s first novel underlines how close Kenner’s ideas are to those explored by Gaddis throughout his career.
In Kenner’s view, the novel is an inherently technological form, matching the mode of production of its present era (the ‘novel-factory’ of Dickens; Emerson’s prediction that Babbage would ‘invent a Novel-writing machine’) (p. 148); by extension, the abstraction of production will lead to the abstraction of the novel. Just as the absurd exaggerations of Babbage’s ideas invent the possibility of a mechanical novel, the increasingly powerful AI programs of the 2020s bring the notion of a novel written by machine closer and closer to reality. The mere suggestion of this is enough to produce an existential crisis in the arts, but I would argue that Gaddis’s treatment of writing-as-technology offers a readymade antidote to the possible threats of AI. Rather than literally allowing AI to write novels (which would be inevitably terrible anyway), the form of AI provides a new analogue to the writing process, one in which the writer themselves can become a form of AI in a manner which challenges and disrupts the technology’s attempts to intrude into the literary domain.
Before delving deeper into what Gaddis offers us, it is worth reflecting on the present capabilities of AI. In their publicly available forms, current AI has a largely novelty value; the fascination with programs like ChatGPT and DALL-E often comes from the humorous failures and incongruities of these models to grasp the nuances of human language. Even if we presume that giant leaps ahead are imminent (and possibly already in private use), these challenges remain unsolved. These absurd programs could be the invention of Thomas Pynchon or Joseph McElroy, writers obsessed with the possibilities and limitations of technology. In the vast novel Women and Men, McElroy describes the sound made by an analogue television with the volume muted (pp. 102-109), the sort of essential immanence of technology that can only be understood in relation to human perception; no large language model could ever allow this kind of insight to be gleaned. This is not a statement that writing is fundamentally a humanist endeavour, but rather the understanding that writing of this nature is the space where technology and consciousness intersect, allowing the writing to become technological. McElroy’s extraordinary description of the television is made possible by the fact that his writing acquires the perspective of the television, which is then triangulated with the perceptions of the two characters who debate the presence of its sound.
If we accept the argument that the textualization of reality is an extant process, this puts literature and AI in a similar position, albeit with radically different methods of dealing with the material at hand. Large language models are trained on enormous quantities of language and data to produce what is supposed to be a cogent response to reality, but in the process, these models inherit the biases and flaws of human language and society. Discussing the longstanding issues people of colour have faced in dealing with facial recognition software developed using LLMs, James Bridle defines technology as ‘the reification of a particular set of beliefs and desires: the congruent, if unconscious dispositions of its creators […] to continue to assert an objective schism between technology and the world is nonsense; but it has very real outcomes’ (p. 142). In this specific example, the technologies were not intended to be racist, but they reflect ‘the historic prejudices deeply encoded in our data sets, which are the frameworks on which we build contemporary knowledge and decision making’ (p. 144). The moral perspective of the novelist allows these prejudices to be addressed with a degree of mediation, or at least enough self-awareness to make them apparent to a perspicacious reader. There is a debate to be had elsewhere about the extent to which Gaddis’s work had a moral centre, but his belief that Mark Twain was ‘the quintessential writer’ (WG p. 13) would lead many readers to observe an abiding Twainian morality to Gaddis’s writing, particularly in its satirical strains; this affords his work a dimension of moral critique that is inconceivable in AI.
In the guise of satire, literature can become a powerful counter-technology. This is particularly prominent in J R, where the relationship between technology and capitalism is the most frequent target of the caustic satire that pervades the novel. As Moore explains, ‘satire can defuse the threat of mechanisation, to some extent, simply by ridiculing it. But satire is an adult response largely unavailable to the child, the real victim of technology in J R’ (WG p. 95). When literature is weaponised as a counterforce to technology in this way, we may still see the nefarious effects of technology in action, it is just that these effects are nullified for us, the reader; while we can delight in the hilarity of J R, the children of the novel must endure the rigours of an education steeped in mechanical processes which increasingly treat them as objects. The humour here is derived from distance, namely the mismatch between the objectives of technology and its actuality. This is also why ChatGPT and DALL-E are often funny: the lofty claims are often undercut by the silly, incongruous and downright absurd results when the public try to use these programs. It becomes less funny when the distance between us and the technology in question is closed, however, and just as the children of J R are powerless to ridicule educational technologies, those whose jobs are increasingly threatened by AI are unlikely to get the joke.
One of the most enduring themes of Gaddis’s career was that of failure, something he addresses directly in the essay ‘The Rush for Second Place’ and mentions throughout his correspondence. A common theme is for a letter to be greatly delayed, opening with a few apologetic sentences damning his own failure to keep on top of letters. A representative example: ‘I’ve been so inexcusably poor about writing that at this point it would be graceless even for me to apologise; as graceless as it would be for me to in some way thank you for the piece in the Berkeley paper’ (L 287-288). The litany of apologies that runs throughout the volume reflects a failure to write; it takes a remarkable writer to be able to transform this persistent failure into literary substance, almost the kind of alchemy hinted at in The Recognitions. For Gaddis, however, failure is an aesthetic driving force which can even allow for artistic refinement, much as how J R sees its talented composer Bast able to produce a distilled version of his art because the stress of the economic world sabotaged his previous efforts. With technological failure as a subject, this affords the writer a great deal of nuance in being able to develop artistic responses.
When we talk about the threat AI poses to art, it is worth noting that to a significant degree, art is already managed by similar forces. Although created by humans, television, music and even literature are largely subordinated to algorithms. The heaps of low-quality programming made by Netflix, the nondescript music by anonymous artists that many Spotify playlists eventually yield to, the morasses of pointless print-on-demand books that have probably never had a single reader, all these things are already made to fit the quantified systems through which we approach modern culture. Ironically, despite the commonalities between this algorithmic form and AI, the use of actual AI in the arts often provokes a squeamish reaction from the giants of streaming; in 2023, Spotify removed thousands of evidently AI-made songs which racked up plays from fraudulent, AI listeners, music made by AI for an AI audience. Meanwhile, the likes of Google and Universal are eager to license some of the less tangible elements of their artists’ work and images to counter the spread of deepfakes. While these may turn out to be little more than a TikTok novelty, they are being taken increasingly seriously as a threat to copyright law.
Literature is, in some respects, a different case. The economic and technological means required to write and publish a book are considerable, of course, but they pale compared to the requirements for a film. The notion of literary style being copyrighted in the same manner as a singer’s voice is simply amusing; were he still alive, it would be easy to imagine William Gaddis writing an absurd novel where literature professors argue in the courtroom in copyright cases to demarcate the boundaries between legitimate influence and outright theft. In a field which has always used quotation, homage and pastiche, similar rules would be unenforceable. Gaddis’s writing uses all these devices extensively, even in superficially non-literary work. For much of his career, Gaddis produced corporate work, often scripting promotional or technical films for large corporations, smuggling in veiled criticism of business and technology through diverse literary allusions to Thoreau on technology as “an improved means to an unimproved end,” to Max Weber on society’s decline from Status to Contract with the rise of capital, and […] to Gertrude Stein on the circularity of programmed culture, which is only capable of answering questions that it has already formulated (RSP 15).
The Stein quotation concludes a treatment for a short film on software, developed in 1976 to promote IBM; computing technology has advanced beyond recognition in the intervening forty-eight years, but the charge of recursion and self-reference remains.
The role of large language models in present-day AI technologies is intended to counteract this problem, but as we have seen, algorithms then merely end up reflecting real-world problems and attitudes. As these systems grow in scale, the relationship between world and algorithm becomes increasingly unintelligible, creating an opacity which, alarmingly, is sometimes taken as objectivity. Justin Joque writes powerfully about the impact of this:
These machine learning systems function in a way that is analogous—and, as we shall see, metaphysically tied—to capitalism: they move the locus of social domination from the material world into the abstract one of capital and probability, yet they do not oust history. In fact, because their aim is only to predict, they actively reproduce it. While these systems make some categories more fluid and open, they simultaneously work to solidify extant social systems: capitalism, racism, patriarchy and imperialism, among others. And they do so in ways that are potentially more insidious and harder to resist, presenting their outputs as objective facts. It is thus necessary to account for the metaphysical force of this objectification—something we can ascertain only by tracing the ways in which probability and statistics function socially and economically (p. 41).
Reason becomes an obfuscatory technique, in which “the supposed presence of a formal proof that can be understood by a single mathematician is made commensurate with 200 terabytes of data that cannot be reviewed except by another computer” (Ibid, p. 42). This is the Steinian loop taken to absurd ends.
When we apply these models and concepts to culture, it becomes apparent that AI artworks are made not for a human audience but for an AI one. As the example from Spotify has demonstrated, forms of this are already taking place for base economic reasons, but the inherent unintelligibility of these systems does not bode well for artistic potential. If the AI-like commissioning strategies which blight popular culture are anything to go by, such an art would strive for nothing more than profit. This is nothing new, of course, but the lack of a human impulse would likely make this motive far more brazen than any actual artist or writer would dare consider. These motivations are usually felt less at the level of the artist than in other parts of the cultural foodchain, and of course, literature tends to be less lucrative than other forms. If we look at the visual arts, the instrumentalization of profitability has already arguably taken place, with deleterious effects for the wider public. Many works which should be displayed in galleries are now stored in shipping containers in freeports as they are bought and sold, sometimes in incremental shares. Their presence in these private, gated communities confers the status of a financial instrument rather than an actual artwork; just as AI art being ‘viewed’ by other computers raises conceptual questions, we can ask whether an artwork in this state is really art at all, when its latent monetary value stops it from even being on display in private, let alone in a publicly-accessible location. The art and digital culture scholar Rachel O’Dwyer wonders ‘if the industrial economy perfected the material commodity – bringing together materials from distant places to make a toaster, for example – then the economy of finance runs this process in reverse, carving into the heft of things, making them more liquid’ (p. 214). If the solid artwork melts into air, does its cultural value dissipate?
The legalistic satire A Frolic of His Own provides some unlikely guidance. In one of the novel’s key strands, a dog is trapped beneath a sculpture called Cyclone Seven, leading to much legal grappling around the material and culture value of the sculpture as opposed to the tangible and intangible values of the dog’s life (the artist initially obtains a restraining order, preventing the local fire department from damaging the structure to rescue the animal) (FHO pp. 29-30). Once art enters the domain of legal disputes – or indeed finance capitalism – it becomes a teleological organ, undergoing a phase transition from art to techne. This conceptual shift does not necessarily happen at the moment of conception – paintings by old masters have mutated into digital commodities unthinkable in the days of the artists – but it necessarily delimits the art available to us.
Contemporary artists are nevertheless afforded a degree of agency in how they navigate these issues. Just as Gaddis seeded critiques of technology in his writing for IBM, art remains capable of biting that hand that feeds it and developing a degree of social criticism. An unflattering portrait by the Australian painter Vincent Namatjira of the billionaire Gina Rinehart has recently received a significant degree of attention after Rinehart demanded its removal from the National Gallery of Australia in Canberra. This comically infantile response entirely misses the point of Namatjira’s painting, which is part of a satirical, often very funny body of work, but it may lead us to consider what happens when people of Rineheart’s means control not just the purse strings but also the tools of artistic creation and distribution. With his acrylics, Namatjira has managed to create a succès de scandale, spreading his work to a wider audience, but algorithmic platforms may well have stopped him in his tracks. Even if this work could be created with digital media, the inherently financial logic of the algorithm would have probably suppressed it, or at least slowed its dissemination to the point that it would have had no audience at all. Furthermore, as an Indigenous artist depicting a wealthy white subject in an unflattering light, we may reasonably assume the encoded racial biases observed in many algorithmic entities would have impeded Namatjira.
Despite the worrying issues raised in this essay, Gaddis’ approach to technology nevertheless provides the tools required to turn potentially oppressive technologies into a fertile literary substrate. In this case, the dense, virtually incomprehensible network of information that underpins algorithms has the potential to be taken to absurd ends within literature; perhaps the overdetermined language that overwhelms many readers of JR could be taken to an even greater extreme, challenging the noise produced by algorithms. Above all else, the primacy of the human voice in Gaddis’ work provides a challenge to the technologization of the self which sometimes appears to be an inevitable consequence of our current technological path. Voices in Gaddis’ writing are frequently distorted by technology – telephones, televisions and radios provide some of the most extreme challenges to conventional perspective in his work – but, although they means often impede comprehension or recognition, they also allow at least parts of the voice to carry. This focus on the relationship between voice and technology can become a strong challenge to AI too, demonstrating the ridiculous distortions precipitated by AI and allowing a subject existing beyond reason and techne to be glimpsed in between the bursts of static.
Adrian Howlett lives in Dublin, where he teaches English literature at Trinity College. He writes about the intersection between literature, money, and technology, and his work has been published in several Joyce Studies journals.
Photo Credit: Eric Calloway.
L'Esprit Literary Review 