Author’s Note: This was written for a college assignment. I’m pretty happy with it, though frankly this feels unfinished to me from where I want to be. I’d like to return to it someday and finish/expand it, but that’s a project for future me. Also I personally think it reads worse in this format than in distinct pages but oh well not much I can really do about substack formatting.
The first step one has to take to answer whether or not a proposed solution would fix or improve a given challenge or problem or issue, I would argue, is to understand the supposed issue first, well before giving any solution the light of day. Such is the basis of the oft-touted scientific method that’s ubiquitous in western sciences: understand the question before searching for answers. In the case of so-called Artificial Intelligences within the arts, that means attempting to understand what art is. In this case, that’s writing, as I am a writer well before a painter or sculptor or photographer or—perhaps to the surprise of those who know me better—a woman of science. Exploring the use of AI in writing then necessitates understanding both the process behind writing and the intentions with which it is done to guide a reader along a given path. Ursula K. Le Guin perhaps puts it best, “I would go so far as to say that the natural, proper, fitting shape of the novel might be that of a sack, a bag. A book holds words. Words hold things. They bear meanings. A novel is a medicine bundle, holding things in a particular, powerful relation to one another and to us,”1 though I would go further and expand that definition beyond just that of the novel and to all writing, and that sometimes bags do not hold things: sometimes they are empty or only partially full—words left unsaid, both purposeful and not, too are a part of writing. Perhaps those blanks are the most important part of a piece, perhaps they are an oversight that leaves the reader unfulfilled or dissatisfied or feeling lesser. To illustrate this, I’ll turn to the idea of maps. No matter how detailed or precise it tries to be, things will be left blank; either unknown or deemed unimportant. This does not, however, inherently make a map worse. “Blanks can also represent what is known, but deemed unimportant in a particular context, for a particular map. While it’s useful to know where public libraries are, and while some bicyclists no doubt make use of public libraries, we wouldn’t expect to find libraries on [a] map of county bike trails.”2 This is to say that writing is a collection of words and blank spaces bound by both their author’s intent and the ability for others to interpret those words for themselves. Artificial Intelligence, then must seek to ‘fix’ this, by way of stripping words or spaces or ideas of any intentionality. What’s worse is the underlying implication of this: that creation is something to be stifled and reduced in favor of more. Maybe this is an artifact of societal pressures—people seeking a quick influx of cash by churning out novels or a fear of perceived failure in school. Mostly, I’d argue that it is a result of our own fixation on value and the ways in which our definitions of the word have shifted: time could be better spent earning or spending money rather than creating something without an inherent use or physical and tangible monetary value to society. After all, the implication that writing is something to be ‘fixed’ or ‘solved,’ to be made more efficient, is a rather capitalistic view of the world, is it not? Incidentally, that capitalistic mindset is exactly that which is used in the tech industry to continue making their advancements on large language models:[Open AI’s] belief in scaling was once viewed as extreme. Now scaling is seen across the tech industry as doctrine. And should the industry’s adherence to that doctrine continue unabated, future deep learning models will make the once-unfathomable size of generative AI models today look paltry. [. . .] The scaling doctrine has become so ingrained that some are even beginning to view it as something of a natural phenomenon. Scaling compute is the way, not just a way, to reach more advanced AI capabilities.3 This is identical to the capitalist growth mindset, “capitalism cannot exist without constantly expanding the scale of production: any interruption of this process will take the form of an economic crisis.”4 Of course, infinite growth is simply not possible in the long term as “we continue to embrace economic systems that prescribe infinite growth on a finite planet, as if somehow the universe had repealed the laws of thermodynamics on our behalf. Perpetual growth is simply not compatible with natural law.”5 Thus I claim the following: the Artificial Intelligence frenzy—specifically that of Large Language Models—will, out of necessity, die. Indeed, many economists agree with this conclusion. First, it must be stated that companies like OpenAI are facing huge economic losses: “The most hyped AI companies in the space are taking huge losses, and profitability is still far away for most of them. OpenAI and others will need to keep raising money to stay afloat. To attract future investors, they will have to be able to continue growing their valuations to unprecedented levels.”6 To cover for this, these companies must take on huge quantities of debt. “[CoreWeave] expects to bring in $5 billion in revenue this year while spending roughly $20 billion. To cover that gap, the company has taken on $14 billion in debt, nearly a third of which comes due in the next year. [. . .] CoreWeave also faces $34 billion in scheduled lease payments.”7 It is with this information I present the crux of Hyman Minsky’s Financial Instability Hypothesis:Over a protracted period of good times, capitalist economies tend to move from a financial structure dominated by hedge finance units to a structure in which there is large weight to units engaged in speculative and Ponzi finance. Furthermore, if an economy with a sizeable body of speculative financial units is in an inflationary state, and the authorities attempt to exorcise inflation by monetary constraint, then speculative units will become Ponzi units and the net worth of previously Ponzi units will quickly evaporate. Consequently, units with cash flow shortfalls will be forced to try to make position by selling out position. This is likely to lead to a collapse of asset values.8 Of course, prior to the rapid rise in Artificial Intelligence, the tech industry was in the midst of a ‘protracted period of good times,’ with companies like Nvidia comfortably bringing in a billion in revenue quarterly since at least 2012.9 Today, there has been a mass influx in private investment within the tech industry—like all of CoreWeave’s aforementioned obligations. In other words, a shift to a reliance on speculative finance; thus leading to Minsky’s hypothesized collapse of asset values. And thus I’ve walked us headfirst into the trap I’ve set: I’ve fixated on the economic value of Artificial Intelligence—our proposed ‘solution’—and rather specifically haven’t understood the question beyond that. Sure, I’ve ‘defined’ writing and posited a solution to the encroachment of AI in the creative spaces of simply waiting it out, but that rather misses the point, doesn’t it? Is there value to be had, on an individual and specific level, to using AI as a tool in the process of creation or learning? I simply ask: what is the point of using AI in the process of creation or learning or simple exploration; what ownership of creation can you claim if your bag of words is stripped of the intent with which you put each piece into it? The answer, I think, must be built up not from what writing may or may not be, but the process in which it is created and read; the why behind an author’s intention.We might set out intending simply to describe what we see — to open the curtains beyond our desk and report on the landscape outside our window — but even then we describe what we see, the way we see it. We know the names of the trees and birds and grasses, or we don’t. We’re aware of the different types and formations of clouds, or we aren’t. Even if we could know it all, at any given moment we would have to choose the evocative description or the scientific fact. No matter how hard we work to be ‘objective’ or ‘faithful.’ we create. That isn’t to say we get things wrong, but that, from the first word we write — even by choosing the language in which we will write, and by choosing to write rather than to paint or sing — we are defining, delineating, the world that is coming into being. ‘Intention’ is a useful term when more broadly defined. Intention might be meaning, but for another writer, or another poem or story, the intention might be to depict a particular emotional state, or to explore an ethical dilemma in its complexity, or to understand how a particular character could commit a particular act — or, for that matter, to test the limits of associative movement, or collage structure. Certainly we have some intention(s) for each piece, or we wouldn’t be writing. [. . .] The plan that guides our exploration may or may not be the structure-defining intention of the map, the document that leads the reader; experience reminds us that there is often a world of difference between what we hope to find or think we might find, and what we discover. Goals of our exploration, then, include refining our intention and determining the best way to present it.10 Thus I claim that the process of writing is that of exploration. Regardless of a piece’s intention, the work is exploration, too—of a new place or feeling or time or scientific knowledge—for both the author and the reader. A piece of writing then, the bag of words rather than the process behind them, is then akin to a map, guiding the reader along the author’s chosen path to the destination of her intent, rough edges smoothed away and pitfalls markedly avoided. To write then is to explore, then refining the path so that it may be a guide to others. Historically, cartographers have had a disproportionate influence on their society’s perception of land and the world. Take, for instance, cartographers during the colonization of the Americas. Beyond the colonies, many of these maps were simply blank. “Blanks within the borders of a map can represent many things, among them the deliberately withheld [. . . or] what is known, but deemed unimportant. [. . .] More ominously, Native American tribal regions on early European maps of the Americas, giving readers of most of those maps the impression that no one lived there — at least, no one of consequence. No landowners.”11 “A blank on a map became a symbol of rigorous standards; the presence of absences lent authority to all on the map that was unblank. The logical end to such a scientific approach would be a comprehensive map of a verified world.”12 This leads to a notion that the blanks then, must be empty, reducing what nature or hidden places, or, in more nefarious cases, people simply do not exist. Over these blanks, colonial expansion would, over time, march ever wast-ward, filling in these supposed blanks with something new. “Renaming was one the principal instruments with which colonists erased the prior meanings of conquered landscapes. [. . .] In such acts of renaming, the adjective ‘New’ comes to be invested with an extraordinary semantic and symbolic violence. Not only does it create a tabula rasa, erasing the past, but it also invests a place with meanings derived from faraway places, ‘our dear native country.’”13 This is not to say that each and every blank within a given piece of writing must be agonized over, but rather that blanks are powerful—that intentionally saying things in a given way, or perhaps not at all—can shape meaning and perception; changing the relation from one word to another shifts the meaning of our bundle of words. The most basic of blanks in writing are perhaps those which denote the end of an idea and the transition to the next, those that are considered to be grammatically correct. In literature, however, blanks can often instead coincide with the passage of time. Artificial Intelligence removes this from writing perhaps most of all. The networks of data that form a given model’s training is mathematically transformed in a way that gets from one word to its most likely successor, with no regard for what is or could be between and even less for what could be around—where else the map could meander. Whereas writers avoid recreating the same thing as another, from the same base of shared knowledge, through creativity and their own perception and understanding of world and words, LLM models avoid recreating identical content from identical prompts via different words being assigned as ‘more probable’ than another based entirely on the occurrence of that word in a given database next to words associated with the prompt given. With enough permeations, enough different layers of compute, the likelihood that a prompt will receive the same output as the exact same prompt is approaching zero, giving an illusion of newness built on probabilities of words derived from authors past. Within the model of our map, this allows AI to evoke a vague shape of something that we could reasonably recognize as such. Borders will separate countries; oceans and mountains will separate land. A prompt given to AI may be able to influence those basic shapes—make the countries have harsh, square borders or make islands more or less common. Telling it to make a map of the Earth may make the shape vaguely familiar. It does not know what a map of the Earth would look like, but can impart the shapes it has associated with the word “earth” onto the map, giving it an uncanny appearance, but far from accurate. Importantly, this also means that its tendencies and biases will echo that of its database. Today, we live in a world divided into nation-states whose borders were the result of colonial expansion; our mapping AI’s database is similarly skewed by modern perceptions of the world (the parts of it that have internet presence enough to be scraped for data, anyway), as there simply exists more modern works than surviving historical ones. Our mapping AI then must be prone to leaving blanks where it perhaps should not, whether for making the piece worse or for whatever implicit implications may be present.But simply discovering [the rules] isn’t enough. We need to devote ourselves to the ongoing practice of questioning the rules we have found most useful (including those we hear ourselves offering as advice) and the fundamental assumptions of our work, constantly checking for empty routine, thoughtless employment.14 AI models, by their very nature, are unable to question the rules of its creation; cannot decide to stray from its algorithm outside of randomness determined by yet another algorithm; an ‘empty routine, thoughtless employment,’ of mathematics to recreate the illusion of human writing. This leads to a misplaced attribution, removing the data from which the words were pulled from and imposing a new sort of tabula rasa on them: this is new, this has no other authors. Writing like this leads to a sort of formulaic reproduction—redundant and reductive facsimiles of literature.When we grow frustrated with received literary forms, it is because we feel the forms are reductive, placing artificial constraints on what can be represented on the page. But no matter whether we work within those forms or try to make them more elastic, the challenge is to find ways to express, not everything in the world, but some part of the world in its complexity. The tension between our vision for the work and the form we choose mirrors the tension between the world and its incomprehensible vastness and our attempts to make sense of it.15 Most notably, AI is incapable of any worldly experience; has no attempts to make sense of the world. Sure, perhaps it was trained on several thousand different recounts of walks through different forests along different trails, but it has not walked them; it has no experience to make sense of. Its reductive recreations not only lack originality in thought, but in experience. This necessitates that whatever use one could conceive for AI it has no place within writing. Perhaps it could help one jump the process; to create a meaningless product or—in the case of many students—a product whose sole meaning is to check off an assignment they have no interest in. But to create something new requires creativity and the ability to have thoughts that are uniquely the writer’s and have been shaped by the world and their understanding of it from every moment of their life until a given piece is finished. AI, by its very nature, lacks both. Its “creativity” is randomness weighted by the frequency of words strung together across the internet and its “thoughts” are a mindless echo of a congealed average of everyone on the internet. The words it spits out have no relation to another in its sentence but a shared, algorithmically defined, relation to every word in the training data, resulting in the inability to impart real meaning into whatever it supposedly creates. Words bear no meaning when they are not held in that “particular, powerful relation to one another and to us.”16
Le Guin, Ursula K. “The Carrier Bag Theory of Fiction.” In Dancing at the Edge of the World: Thoughts on Words, Women, Places. Grove Press, 1989.
Turchi, Peter. Maps of the Imagination: The Writer as a Cartographer. Trinity University Press, 2004. (p. 33)
Hao, Karen. Empire of AI: Dreams and nightmares in Sam Altman’s OpenAI. Penguin Press, 2025. (p. 89)
Foster, John Bellamy. The Vulnerable Planet: A Short Economic History of the Environment. Monthly Review Press, 1999. (p. 124)
Kimmerer, Robin Wall. Braiding Sweetgrass: Indigenous Wisdom, Scientific Knowledge, and The Teachings of Plants. Milkweed Editions, 2013. (p. 308)
Kost, Danielle. “AI Companies Don’t Have a Profitable Business Model. Does That Matter?” Harvard Business Review, November 12, 2025. https://hbr.org/2025/11/ai-companies-dont-have-a-profitable-business-model-does-that-matter
Karma, Rogé. “Something Ominous Is Happening in the AI Economy.” The Atlantic, December 10, 2025. https://www.theatlantic.com/economy/2025/12/nvidia-ai-financing-deals/685197/
Minsky, Hyman P. “The Financial Instability Hypothesis” Levy Economics Institute Working Papers Collection, no. 74. https://www.levyinstitute.org/pubs/wp74.pdf
Macrotrends. “NVIDIA Revenue 2012-2026.” Accessed February 21, 2026. https://www.macrotrends.net/stocks/charts/NVDA/nvidia/revenue
Turchi, 14
Turchi, 33
Turchi, 37
Ghosh, Amitav. The Nutmeg’s Curse: Parables for a Planet in Crisis. University of Chicago Press, 2021. (p. 49)
Turchi, 102
Turchi, 157
Le Guin