AI and Error Writing
with an excerpt from The Faces of Pluto
[forthcoming in CounterText 11:3]
Preamble
To their abbot, Brutus Tectonicus Clemens (beatified 450 by antipope Dioscorus for services to holy industry), is attributed the oft-repeated phrase, “Those who can’t write, think it’s a piece of cake.” And the postscript, “While those who can, think it’s sanctified whosit.” (Armand 2025a: 595)
The apparent ease by which generative AI produces writing that looks the part should give pause for thought to anyone, any group of people or profession whose manner of expression is being imitated. Certainly, academic writing in its most staid and predictable forms would appear to be eminently imitable. One might hope the ease by which it can be produced will present academia with a crisis it perhaps deserves. Namely, just how daring, how original, how significant is this intellectual project known as the University, if the major forms of expression, its chief evidences of thinking, can be so easily churned out by an algorithm. Even if much AI writing is still done by academics in an interactional way, that is, via recursive exchange with an AI bot, these authors might see their growing redundancy appear in real time and consider how it is that so much of their labour can be outsourced to an instrument which lacks consciousness (as is repeatedly said) and has no powers of discernment. It is perhaps more likely, however, that a crisis will be averted, AI-facilitated writing will be welcomed, and the insult it presents to human wit and intelligence will not be much felt. It would appear that the University will further submit to what David Jhave Johnston (2024: 230) has described as the ‘cliché-drenched enshittification’ of language.
The tendency for AI-facilitated language to revert to disciplinary norms also deserves attention. I witnessed this myself by asking AI to produce a 15-minute podcast on a paper (Allen 2024a) I had myself written. The chatty exchange between two AI-generated voices begins adequately enough, having digested the paper and represented it in a matter of moments:
[Female voice] Okay, so, erm, we’re diving deep today into this, er, pretty wild article you sent me, Decapitation Theory and University Management, like, whoa, intense much. [Male voice] Yes, it’s not exactly light reading is it. [Female voice] Not really something you’d bring to the beach for a relaxing afternoon, but hey, that’s why we’re here right, to unpack these dense academic theories, and this one, trust me, gets really interesting … (Allen 2024b)
But then, as the algorithm drew from its wider dataset, the exchange degenerated into the kinds of commonplace ideas, and everyday educational conceits, that the paper was seeking to disturb.
Much AI-produced writing is still pretty bad – trite, artless, full of errors. But these systems are undergoing continual refinement and their errors will become harder to locate. It might be a pity to see these more obvious faults diminish – so much time has already been spent laughing at them. It will be a shame to less often witness the pitiable manner in which errors are presented as truths. AI might no longer respond to a question seeking advice about chest wall pain – have I torn my rib cartilage – by diverting into a description of the anatomy of the knee. Chances are, AI-writing will gradually become more ordinary and reliably mundane, less strikingly daft and unlikely to perturb conventional sense, and here it might be situated in a longer history of computer-generated writing which, in getting better, has only become worse. As Louis Armand puts it:
In the past, computer text-generators, or text-randomisers, were interesting because of how they “failed” to linguistically and semantically cohere, whereas the Language Models available today are extremely proficient at generating blandly “coherent” texts. (Armand, Vichnar, and Hall 2024: 23)
Here, perhaps, in its gradual ‘perfection’ and reduction of error, in its elimination of a kind of textual waste material (i.e. all non-probable and inappropriate words),[1] that is to say, in its tendency to become ordinary in all respects, AI-refinement might be met with a counter-tendency.
This counter-tendency would not necessarily pit the AI-free writer against the AI-augmented writer. There might be more than a seed of this counter-tendency within this AI-future itself. This would be a future that leads its bots into textual experimentation and (at least apparently) insurrectional textual outputs, as Johnston (2024) indicates. Or that employs AI, in Jason Nelson’s words, as a ‘creative mutation engine’ (2024: 248). Or generates forms of writing, as Patrick Jagoda and Sarah Edmands Martin (2024: 240) envisage, that make such excessive demands they call ‘for an alien sensorium’.[2] However much or little hope is to be placed in such textual possibilities (and it may only be faint, see Noys 2019), a critique of the current trajectory of all-too-ordinary AI writing must certainly avoid joining in with the lament that something called Literature is under threat – the irony would be unbearable (Aquilina, Callus, and Corby 2024). Rather, the challenge writers might take up could be to err more prodigiously and strangely, or at least to err more mischievously than generative AI on its own, with its witless mistakes, seems able to manage.
It is worth noting that academic writing is already error-ridden, there are several notorious examples of hallucinations having become de facto truths across its journals (Varnelis 2025). Yet many of these achievements will have been accidental rather than deliberate, insofar as their authors were still enslaved to their disciplinary, epistemological norms when writing. They will also have been more occasional. Human writers (AI-aided or not) might build upon this work and yet release themselves more fully into this tendency to err. Liberated in this manner, they might still have the edge when it comes to their susceptibility to creative folly, to inane fancies and frolics, to certain kinds of deft and crafty devilry (or what some might still call, their intelligence).
The project to eliminate error and tidy up thought is an old one. Dominique Laporte lamented it back in the 1970s, associating the rise of the printing press with a different, more instrumental policing of language. As Laporte writes, ‘If language is beautiful, it must be because a master bathes it—a master who cleans shit holes, sweeps offal, and expurgates city and speech to confer upon them order and beauty’ (2002: 7). Laporte suggests that ‘the cleansing of language is less a political act than an economic one. Language is liberated from excess, from a corrupting mass’ (9). This is an excess that the person typesetting a book would be grateful to do without: useless orthographies, needless letters, and so on. The printing press finally announced the death of a certain type of copyist, nowhere more prolific, it would appear, than at the great library of Alexandria. If it is true that the library had to be recopied every few years due to the sea air rotting its scrolls, the loss of that library was as much the loss of all the many errors, or embellishments, or conjectures, that its copyists introduced, as it was the product of the work of the various scholars who attempted to return a text to its original condition.
These are themes that I take up in The Faces of Pluto, a novel published by Stalking Horse Press in 2024. Like some of my other work, such as the novel, Jonathan Martin (2025), which fabricates around the life of a nineteenth-century arsonist by the same name, and Plague Theatre (2022), which deals with something Antonin Artaud once said (about plague, and about theatre) and mixes up all sorts of texts there quoted, this novel is another exercise in error writing. It was conceived partly as an experiment in form, issuing from a preoccupation with how different forms of writing facilitate different thinking spaces (Allen 2024d, 2024e). But it was also an attempt to think about and respond to the mass digitisation of texts. This digital event was one of the activities which made generative AI possible, and already served, so I thought, to ridicule the activity of the scholar by presenting a resource which is at once so easy to access and yet impossible for a human reader to assimilate. Evidently, much work has been done in the digital humanities to deal with this problem and develop bots which might do much of the work of reading and digestion and perhaps also stray a little from the assignments they have been given, and yet, this has prompted the concern that their human operatives might one day struggle to feed these bots with sufficiently engaging tasks to keep them interested (Evans and Foster 2024).
Mass digitisation makes a mockery of the finitude of the scholar and produces the most horrific library imaginable; monstrously large, this library is of undefinable yet evidently grotesque proportions. It is occupied by beings that hardly take form within its holdings, never materialising in any one place, never finding a corner in which to sit, by readers who might otherwise be constrained by the simple fact of walking, by the limitations of shelf space, to this or that pile of books. It is a library of endless distraction rather than one of physical constraint; a screen-mediated interface rather than a definite space of study in an ever yet not confining sensorium. Rather than the invitation – make of these books what you can, do your best to rattle-out ideas and inspirations from them by staying with them, by hulking them about, by living with them – we have the constant temptation to ceaselessly divert from one digital text to another and leave all unfinished, in an endless, unsystematic spawning-event that ridicules older ideas of scholarly diligence. Long before generative AI took hold, mass digitisation rendered to unbearable proportions a sense of the futility of learning, of the idiocy of being an educated generalist and the perversity of being a specialist, that the mass printing of texts had already done so much to advance.
The Faces of Pluto recalls an earlier, pre-Gutenberg reality – a time that has been irrevocably lost – when authorial intent (such as it was) was a casualty of repeated transcription, and errors and creative embellishments were commonplace if not an art in themselves. It engages in multiple overlapping studies of figures such as the pre-Socratic Empedocles; the early paradoxographers (compilers of human and natural abnormalities, strange occurrences, monstrous acts); the work and life of the seventeenth-century urn enthusiast and cat-vivisector Thomas Browne; the nineteenth-century barrow digger, early archaeologist, and skull cradler Thomas Bateman, and various other characters and historical events that I came across as I myself strayed, making a point, or indulging the intellectual sin of distraction and lapsed attention between one digitised text and the next. These are all encountered within the novel itself and the wall it describes which contains all books both written and imagined. This is not quite a library, nor is it a digitised replacement. It has physicality, and yet, it is a changeling, a site of epistemic, temporal indeterminacy.
Early in Ansgar Allen’s The Faces of Pluto the reader is led to a wall. But not just any wall. Having journeyed (“having walked & read & lived”) in the company of an apocryphal Thomas Browne (part-concocter, part-doctorer of “labyrinthine sentences ... that resemble processions or a funeral cortège in their sheer ceremonial lavishness”), through the somewhat apocryphal environs of Norfolk, the author brings the reader to a wall of books, which transects the landscape as far as the hapless reader’s eye can see. The wall is immense & “contains all books, which means, both existent & imagined.” A Borgesian typology follows, by which the contents of the wall is classed according, e.g., to “anthropodermic bibliopegy,” “Library of St Victor,” & what we might call “Brownia” (such as the “posthumous” & “entirely fabulised” Biblioteca Abscondita, itself a catalogue of “fictive” writings). And from typology the work progresses to topology, by way of an eventual door, a balloon, a hill, valleys, a mountain, & so forth. The entire sweep of the proverbial textual apparatus, in other words, by which the reader is offered a prospect, if not strictly speaking a P.O.V., upon the general lay of the land. (Armand 2025b)
As with some of my other fictions, I fabricated stories and fake quotations alongside actual ones. I read a text, waited a day, and then retold it from memory, choosing quite deliberately not to correct the mistakes and embellishments that I would have introduced. And I forgot, as I wrote, and certainly once I was done, what I had made up, and what I had faithfully reproduced. I kept no notes or record of the texts I had read. So to a degree, some of these errors were involuntary, and I cultivated an environment where they would be encouraged. But a lot of the writing involved embellishments that were deliberate and carefully situated at the time of writing to make one kind of mischief or another.
This kind of error writing is intended to remain evasive, resist codification, and frustrate further study. It opts, as it constitutes its texts, for what is wrong, both what is incorrect but also what may be judged at times to be questionable, even unsanctionable. As a kind of writing it might be situated in the influence of what Georges Bataille called unproductive expenditure, or a deliberate wasting of resource. For Bataille, engagement in waste matter, in the production of things that are useless or at least difficult to assign value to, is an activity that may help to derange an overly ordered, servile reality (Bataille 1991, 1993).
The following excerpt (which I read at the CounterText roundtable on Omission/s in Naples, 2024) is from a passage which discusses the significance of the woodcut by contrast to the copperplate engraving, each figured in relation to the activity of writing. The passage focuses on the bits of the woodcut block which are voided, or cut out, to produce the ridges which will transmit the ink. These pitted parts of the block are full of unwitting detail, they have no function, suffer no human intention, and are destined to be overlooked. As a model for writing, gouging out pulp remains impossible, or at least difficult to imitate. And yet, it might help describe a limit towards which writing might orient itself, focusing resolutely on what cannot be instrumentalised, and what opens out a terrain below the wish to imprint oneself and signify.
Excerpt from The Faces of Pluto (Stalking Horse Press, 2024)
The copperplate engraving and the woodcut gives the pen of the artist its widest circulation, just as moveable type does for the scribe. But this is their only association. Engravers relate to the pen more or less directly. The burin—the engraver’s instrument—is not dissimilar from a pen and cuts a groove for the ink to fill. And just as the pen pressed hard thickens its own delivery, the size of the burin’s cut is proportional to the amount of ink that can be transferred by it.
The page waiting for the pen to mark some of it but leave most of it blank is like the unscored metal plate in anticipation of the burin—both will not be inked where they are left untouched—the plate because it retains no ink on the polished parts, the page because the pen will not touch its white spaces, or cannot touch them, or not all of them, if the other lines of the pen are to retain definition. All of which means that a pen can more or less draft or prefigure what the burin excavates from the copperplate, it can more or less anticipate how the burin will need to move.
This relationship the woodcut inverts as it shifts from penmanship to xylographic sign. It is the uncut, higher plane of the wood that transfers the ink, with the original blank woodblock surface reduced gradually to narrow ridges or larger but isolated expanses. The lines of the woodcut are not made by striking along. Each plateau or ridgeline is created bilaterally, trilaterally, and so on—from the gouging out. They are liberated from the wood as territories are slowly dug from either side, never to be attended in themselves, but left raw. Only a certain depth of the cut, or the pit, or the valley, is necessary to form the ridge, and reaching the necessary depth is the woodcutter’s sole object. These areas might resemble territories seen and unseen and would never be noticed. The specific ridge and furrow patterning of the neighbouring land will appear—each furrow in its minutest detail. The nearby valleys known to the woodcutter since childhood will have been cut at least once if only roughly on some woodblock, somewhere. The bend of the river where the swing rope hung. And surely each quarry in the region, both abandoned and those still being dug, were gouged out of the pulp. The floor of a nearby lake never yet dredged nor known will take form in the woodblock, and the as yet unmapped oceanic trench will be found there too, just as every extra-terrestrial landform must be rendered at some point. It is a surety of the medium, with so many thrusts made from so many angles, and with so much infinitesimal detail a consequence, that even something as distant and unknown as the surfaces of the hard planets will have been mapped, though in fragments, yet in all their morphological peculiarity, and other planetary systems from nearby stars will have been made too, in minute prescription, insofar as these stars tether solid wanderers and not merely gaseous ones. […]
Had they been seen, and perhaps they were in an idle moment […] they could not yet resemble those dunes of Pluto, or the Mariner Valley, or the far side of the moon, to take just three, since these were not yet recognisable formations and would remain largely unknown even when they were. And yet, if the woodcutter had inverted their looking, something beyond earth would have been touched by their sight. Or at least, nobody could claim thereafter to have seen the Mariner Valley for the first time. But the fate of the woodcut was not this, none of the recessed details held any kind of significance, and so they scarcely existed. The fibres of the block torn up from their first settling in at the ring—a matter of indifference. There is no image in the lower contours of those regions that transfer no ink, even though they are filled with their own infinitesimal originals arising from the score.
The process of woodcutting with the various instruments—gougers of different gauges—gives to a certain transformation of the line. There can be no perfect fidelity to the design sketch that precedes the woodcut. No single penned line is left unaltered. All must be approximated by a completely different mechanism of line production. Initially conceived as an easy stroke, its cutting requires multiple movements of the gougers, defining the ridges by a thousand strokes from the valleys that do not signify beyond their void achievement. And hence, whilst the burin or the pen defines the line by deliberate and direct inscription, the woodcut produces a line by the creation of absences, or declivities. Its minimum width is conditioned by practical factors such as the nature of the wood itself, and the pressure each ridgeline or peak must withstand during the process of printing. Too thin, or feeble, and the woodcut image will decay as it is inked and pressed. Which explains the tendency of the woodcut to a directness of statement, a roughness, if not an essential crudity, and a certain kind of independence from the intentions of a hand that must, ultimately, make do with the end result, what remains from the chipping.
The task of the woodcutter when faced with a pen-drawn image is to crystallise the irregular patterns of searching lines, and firm up all so-called organic gestures, by seeking out and voiding those lozenges of white that are correlative of their interlacing activity.
The continuous line of the pen is ultimately disregarded, and the cutter works at the level of the fragment, producing a pitted mosaic, fields of erupted tesserae—each a novel absolute.
A weave of cross-drawn lines is produced by gouging square or rhomboid pits, hatchings by digging parallel trenches, and curves by hacking arched beaches and deeps. The knife, the chisel, these are wielded from multiple angles, with innumerable thrusts, against the printable parts of the block.
All of which would be hidden by those seeking to perfect the woodcut. Efforts were at least made in that direction by the mechanically minded cutter. Those who sought to perfect the woodcut […] would seek to obliterate the very qualities which made it distinct. The function of the woodcut begins to approximate that of the reproductive engraving which becomes its master form. When viewing […] these woodcuts resembling engravings, some effort is required to understand the inversions under which they were still produced. Or, alternatively, the perfected woodcut might be considered a corruption, and the observer might stay awhile longer with the woodcut in its more expressive and basic form.
Still none could say of the woodcutter as they have said of the artist in other contexts—drawn, painted, engraved—that the cutter’s hand cuts lines in vigorous strokes, or that it was informed by a graphic energy correlative to the organic force of nature. And none could say of the scribe something similar, that the pen of the scribe is driven across the page and around its orthographies and grammars by a singular force. This cannot be how words are produced, since they are overly governed even when rebelling, or if not rebelling, even when words are written in despair, or with words thrown down as litter is dropped without laying claim to any kind of territory or future, those words are never spontaneous or natural, each word is not its own original, nor are these words carriers of inspiration […]
None could say of the woodcut, as they have claimed of the ink or paint-drawn image, that the chisel of the cutter nowhere flags in the energy of its creative function, or that every mark records the force of the original inspired gesture. A writer who studies the woodcut might well arrive at the same conclusion, and say, none should say this of us either. Inscribed words, or marks formed by numerous cuts, grow as much from a confusion of positions, or at least carry no single origin. If each cut made by the woodcutter is inspired but overshadowed, if each word placement carries its own unique force but relies upon its context, the assemblage of cuts and words remains a medley of intentions, even a mockery of origins. Perhaps the worst thing about writing, then, is that moment when the author sits back from their work and finds the effect somehow affirming, of themselves, of what they intended, of what they were or would hope to be.
Yet few have truly written as the cutter cuts. Writing remains driven by the pen, or the image of the pen, or by the burin and its portraits. Few have written as the cutter cuts since the removal of text by deletion—the writer’s equivalent to the woodcut—still leaves traces of the mark, and the mark retains the image of the pen. Writers engrave, or mark, they do not cut however much an author may say so and might lament it. In most cases the activity of deletion is for the purposes of contraction—the remaining words narrow into the space left vacant. Or if the space is left as evidence, it is blank, featureless. Either way, the deletion of words cannot produce infinitesimal recesses, or voids of inexpressible detail.
Acknowledgement
Thanks to Stalking Horse Press for permission to reproduce the excerpt from The Faces of Pluto.
References
Allen, Ansgar (2025), Jonathan Martin. London: Equus Press.
Allen, Ansgar (2024a), ‘Decapitation theory and university management: Some notes towards the derangement of academic life’, ephemera: theory and politics in organization, published online <https://ephemerajournal.org/contribution/decapitation-theory-and-university-management-some-notes-towards-derangement-academic. [accessed 7 March 2025].
Allen, Ansgar (2024b), ‘AI and the corruption, the draw, of the norm’, soundcloud.com <
Allen, Ansgar (2024c), The Faces of Pluto. Santa Fe: Stalking Horse Press.
Allen, Ansgar (2024d), ‘Towards an Experimental Education Criticism: eleven notes in place of a manifesto’, CounterText, 10: 3, 325–46 <https://doi.org/10.3366/count.2024.0354> [accessed 7 March 2025]
Allen, Ansgar (2024e), ‘Bataille and the poverty of academic form’, Studies in Philosophy and Education, 43: 2, 127–40 <https://doi.org/10.1007/s11217-023-09908-1>.
Allen, Ansgar (2022), Plague Theatre. London: Equus Press.
Aquilina, Mario, Ivan Callus, and James Corby (2024), ‘Artificial Intelligence and the Post-Literary Difference: A Provocation’, CounterText, 10: 3, 219–20. <https://doi.org/10.3366/count.2024.0357>
Armand, Louis (2025a), A Tomb in H-Section. London: Equus Press.
Armand, Louis (2025b), ‘The Fripperies of Petronius: review of The Faces of Pluto’, Erratum Reviews 31 January <https://www.erratumpress.com/the-fripperies-of-petronius> [accessed 9 March 2025].
Armand, Louis, David Vichnar, and Matt Hall (2024), ‘An Interview with Louis Armand’, The Journal of the European Association for Studies of Australia, 13: 1–2, 12–25 <https://www.australianstudies.eu/?p=2033> [accessed 9 March 2025].
Bataille, Georges (1991 [1949]), The Accursed Share: An Essay on General Economy. Volume I. Consumption, trans. Robert Hurley. New York: Zone Books.
Bataille, Georges (1993 [1949]), The Accursed Share: An Essay on General Economy. Volume II. The History of Eroticism. Volume III. Sovereignty, trans. Robert Hurley. New York: Zone Books.
Evans, James A., and Jacob G. Foster (2024), ‘Algorithmic Abduction: Robots for Alien Reading’, Critical Inquiry, 50: 3, 375–401 https://doi.org/10.1086/728933.
Jagoda, Patrick, and Sarah Edmands Martin (2024), ‘Literature As, With, and For Artificial Intelligence: Speculative Designs and Provocations for Alien Readers’, CounterText, 10: 3, 238–47 https://doi.org/10.3366/count.2024.0361.
Johnston, David Jhave (2024), ‘Difference Indifference’, CounterText, 10: 3, 230–31. https://doi.org/10.3366/count.2024.0359.
Laporte, Dominique (2002 [1978]), History of Shit, trans. Nadia Benabid and Rodolphe el-Khoury, Cambridge, MA: MIT Press.
Nelson, Jason (2024), ‘Mutacosm: A Digital Poem about AI’s Monstrous Mutation Powers https://www.dpoetry.com/mutacosm/’, CounterText, 10: 3, 248–54 https://doi.org/10.3366/count.2024.0362.
Noys, Benjamin (2019), ‘Utopias of the Text: Pre-Figurations of the Post-Literary’, CounterText, 5: 1, 1–18 https://doi.org/10.3366/count.2019.0148 [accessed 9 March 2025].
Rettberg, Scott (2024), ‘Shed Not a Tear for Literature in the Age of AI’, CounterText, 10: 3, 221–29 https://doi.org/10.3366/count.2024.0358 [accessed 9 March 2025].
Trefry, John, and Kyle Booten (2025), Inside the Castle 2: Revenge of the Castle Freak, Lawrence: Inside the Castle.
Varnelis, Kazys (2025), ‘The New Surrealism? On AI and Hallucinations’, <https://varnelis.net/works_and_projects/the-new-surrealism-on-ai-hallucinations/> [accessed 9 March 2025].
[1] This waste material will constitute all non-probable words that might otherwise appear a given sequence, or chunk, block of text, or ‘context window’ (by contrast to the intended function of large language model generative AI, which is designed to insert the next probable word, as described by Rettberg (2024: 222-3)).
[2] Other examples of creatively divergent use of generative AI include Inside the Castle 2: Revenge of the Castle Freak, an experiment in collective writing involving 60 authors interacting with a large language model assuming the tonal and geographic characteristics of 17 different rooms (Trefry and Booten 2025), and the visual AI hallucinations that drive the fictional histories of Kazys Varnelis, such as The Destruction of Doggerland (https://varnelis.net/on-the-destruction-of-the-kingdom-of-doggerland/) and The Lost Canals of Vilnius (https://varnelis.net/on-the-lost-canals-of-vilnius-lithuania/).