Reveries of the Solitary Author

AI writing tutors mean you'll never have to write alone again

Welcome to Ever Not Quite—essays about technology and the human.

I think of each essay I post here as a sort of first draft. Nobody else sees anything before it’s published, and whenever I return to an older piece with fresh eyes, purified by the passage of time, I can often see more clearly what I was getting at than I could when it was first posted—a common experience in writing which I will have occasion to mention below. It was nearly two years ago that I released an earlier version of this essay, but recently I found myself returning to some of these same themes in my last post, which, in case you missed it, can be found here:

A Portrait of the Artist as an LLM

As something of a companion to that one, and because the original went out to only a few subscribers, I’ve re-worked this one in order to share it again now.

The greatest challenge of publishing writing to the internet is getting it in front of the people who will value it. If you’d like to support my work, consider subscribing and sharing this essay around within your circles. It’s the best way to bring it to new readers.

“So now I am alone in the world, with no brother, neighbor, or friend, nor any company left me but my own.” Jean-Jacques Rousseau, Reveries of the Solitary Walker

There’s a famous line which is often attributed to Dorothy Parker: “I hate to write, but I love having written”. Whatever its provenance, this remark has always captured something true to my own experience, and I suspect the same of many others: writing is a difficult business, but while the road may be arduous, the destination is rewarding. It can be painful, at times even excruciating, but its travails ultimately give way to a distinct sense of clarity, of having surveyed the contours of an issue, mapped some of its invisible pathways, its false leads, its trap-doors, and discovered a way through. In a word, writing is a process—perhaps the best we have—by which confusion may be transformed into something resembling understanding.

But here’s a claim I’m willing to defend which has apparently become controversial in certain quarters: the fact that writing is such hard work and exacts such a price on its practitioners is no argument that it ought to be made any easier, nor that we should consider its struggles to be problems in need of solutions.

Let me be clear right away: I’m not talking about all of the textual proliferations that can be found online which barrage us with meaningless ‘content’—a perfect word, really, for the co-incidence of abundance and emptiness—click-bait, advertising, listicles, and so on: all that is now meant by the term ‘human slop.’ You’ll have no trouble thinking of your own examples. All this, it hardly needs to be said, is the lexical equivalent of the widget punched out on the assembly line, and the reason it could be offloaded so readily onto machines in the first place is that the techniques by which it has always been produced were already mechanical in nature. In fact, I’m not really interested in any text that can be said to perform a useful function—giving clear instructions, say, or summarizing the body of academic literature on a certain topic. I’m interested instead in that much smaller share of writing, whether it’s found online or anywhere else, which proceeds from an open attunement to reality and which is the fruit of the faculty of judgment in its vital pursuit of understanding. The process by which we understand and become reconciled to our world, Hannah Arendt once wrote, must engage the human mind “in nothing less than an interminable dialogue between itself and the essence of everything that is.”¹ This is what interests me.

Although it constitutes only a modest portion of the text we might come across—and notwithstanding the many doubts which have been newly raised—a capacity for this sort of writing continues to be an essential skill which the students at any serious educational institution are expected to master. And the continued existence of this kind of education is what warrants all the recent hand-wringing—both within and outside of these institutions—about the place of LLMs in their endeavor to fulfill their mandate to educate their students. It has been proposed, for instance, that AI software be introduced not just to play the role of conversation partners and writing assistants for students who are writing essays, but also as all-purpose tutors for those who are studying just about anything else. Many history and language arts teachers are now evaluating, with all apparent seriousness, the allure of AI-generated resurrections of historical figures and impersonations of characters from literary works, which their advocates insist can help breathe life into the otherwise dull classrooms which already compete with social media for the attention of modern students. What were mere proposals just a few years ago have already found their way into actual pedagogy in classrooms both public and private, upper and lower, all over the world.²

Setting aside my profound skepticism of all this, I won’t comment here on the dubious educational value of these gimmicks, nor will I decry the pressure that is placed on educators to run these sorts of high-tech experiments on their students. I simply want to point out what I doubt anybody who is paying attention would deny: that the very shape of education itself, and not only the methods by which students are trained to write well, is in the process of being redesigned in ways that will have unforeseeable results.³ Because it seems so often that these initiatives are pursued in order to better accommodate the needs of technology to extend its reach and to make itself indispensable, these developments in the way students are taught to write asks us to consider what it is exactly that makes writing the sort of activity that it is.

One way to define the purpose of a humanistic education is to say that it seeks to cultivate in students a sensitivity to the whole of reality—to prepare them, in the words of the late educator Eva Brann, “to go about interpreting all the world”. The world doesn’t need another 5-page essay on the issue of, say, political authority in Antigone, but the student, according to a humanistic pedagogy, does need to undergo the piercing experience of reading, thinking, and writing about the questions raised by Sophoclean tragedy, and to wrestle into being a perceptive and coherent analysis of some of its more serious issues. Simply put, the purpose of learning to write is to render the entire world—including, to be sure, questions about what we are doing with technology itself—accessible to the inquiring mind. The result is twofold: it’s the mind’s deeper relationship with that totality, and it’s also the new product which is introduced into the world—a written work capable of revealing these insights to a reader. The act of writing thus initiates a dual motion, one which directs the writer outwards, into the world and towards the reader, just as surely as it draws the writer inwards, into a conversation with oneself in which we strive to take the measure of the truth.

---

But the argument is now made anew every day: LLMs can teach students to write, and also aid anyone else who finds themselves, for whatever reason, in the position of having to produce a written piece.⁴ The algorithm functions as a partner in writing, one that can inform, goad, challenge, and raise Socratic questions that push both the writer and the work into new and fruitful directions. This technology, it is alleged, enables students to learn better and also enhances the final product, without enabling unethical shortcuts that might falsify its authorship; in short, it can make writers better able to achieve the understanding that the effort of writing has always furnished to those willing to undertake its rigors.

In March of 2023, the online education nonprofit Khan Academy, which boasts of more than 169 million users worldwide, introduced Khanmigo, a chatbot that runs on OpenAI’s GPT-4 to assist students taking its courses. In his 2024 book, Brave New Words: How AI Will Revolutionize Education (and Why That’s a Good Thing), Khan Academy’s founder and CEO, Sal Khan, discussed Khanmigo’s promise in working not just with students who are writing school papers, but also with their teachers whose duties include assigning and grading them. Here’s Khan describing the working relationship between Khanmigo and the student:

With Khanmigo [...] we are creating a middle path for teachers, intending to balance the need for students to learn how to write with the utility of generative AI for support. In this scenario Khanmigo acts as a powerful guide, but the student does the bulk of the work. The student can ask Khanmigo: “What points should I consider as I write an essay about The Great Gatsby?” It will suggest themes including the American dream, social class and inequality, the Jazz Age and the Roaring Twenties, and key symbols in the novel, such as the green light, the eyes of Dr. T. J. Eckleburg, and the valley of ashes. The technology works well when a student takes a couple of paragraphs of something they’ve written and asks the large language model to read it and critique its strengths and weaknesses. It can help students with counter-arguments and make what they’ve written more compelling, encouraging them to think about the essay as a good writing coach might do. Again, the AI tutor does not do the work for the students. Rather, it works alongside them. Within seconds, the artificial intelligence will provide feedback, highlight areas for improvement, and offer suggestions for how to revise and refine the work. It helps craft tone, voice, and perspective even as students develop their own. At its best, an education-based AI platform can be the world’s finest assistant and co-collaborator, objective in its assessments and thorough in its analytics, designed to do one thing and one thing only: to sharpen a student’s skills.⁵

I’ll leave it to you to ponder what are, to my mind, a number of hair-raising specifics in this description—to say nothing of the mystery as to what the neologism “co-collaborator” might possibly mean. I’ll merely point out that writing with Khanmigo, as it is described here, involves more than simply coaching students to become better writers: it also means tacitly, if inadvertently, to alter fundamentally what it means to write in the first place.

Part of what makes writing so difficult, and certainly one reason it might seem to require this sort of high-tech remedy to begin with, is the fact that, in the absence of any input from an LLM like Khanmigo, it calls upon the writer to play a number of distinct roles: a writer, of course, writes.⁶ But a writer also edits what has already been written; a writer must fashion the tone and voice of the work, taking into account its audience, and the existing discourse to which the final product is intended to contribute; and perhaps above all, a writer must conjure that critical inner voice which objects to every argument, takes issue with every characterization, demands greater clarity from every explanation: “he who knows only his own side of the case,” the liberal piety has it, “knows little of that,” and few activities demand the same dutiful observation of this precept that writing does. What’s more, any good writer knows how to step away from the work in progress, letting it murmur indistinctly in the background of thought and imagination, returning after some time with fresh eyes now better attuned to blemishes in pacing, diction, or logic.

This is all to say that the writer engages in an ongoing self-dialogue, the participants of which cannot be pulled apart without compromising the integrity of the writer’s total possession of the competencies demanded by the very nature of the craft. Proponents of software like Khanmigo and other AI-based writing assistants, which deeply insinuate themselves among the writer’s native faculties, suggest that these myriad functions—which any writer properly so-called must command and balance—should be classified and ranked, with some being deemed essential and therefore inviolable, while others which are deemed dispensable are outsourced to the effort-saving tool. They cast a pall of suspicion over the writer’s need ever to struggle, and implicitly pathologize writing’s exertions as senseless agony and wasted time. Against all this, I continue to suspect that the best way for anyone to become a better writer—properly speaking—is to write, which means by definition to exercise all of the faculties which the activity requires. Unfortunately for the solutionist ethos which perceives the need for time- and effort-saving tools wherever it discovers the pains of human striving, this means that writing must continue to be hard.

Leonid Pasternak, The Passion of Creation (1892). Oil on canvas.

---

Now, perhaps my appraisal here is overly harsh, and maybe there’s something to Khan’s vision after all; maybe you’ll object that LLMs can further advance the pursuit of the special sort of understanding to which writing offers unique access. Karl Popper attributed to Einstein the remark that “my pencil and I are much cleverer than I am,” and I too sense that I am well served by the keyboard and word-processing software with which I write these words. Why shouldn’t we expect Khanmigo or any of the myriad other chatbots designed for this purpose to similarly serve future writers who have learned to make the best use of them?

I’d answer this with an observation, which can be stated simply: to precisely the degree that writing is difficult, the practice not only produces a written work, but it also inevitably shapes the writer. Conversely, and perhaps more relevantly in the present context, to the degree that writing is not difficult, the practice does not shape the writer. Of course, just as word processing software once rendered typos and simple grammatical errors inexcusable, students who are being trained to write using Khanmigo might be expected to clear a much higher threshold of formal quality in order for their teachers to consider their work satisfactory. My point is that the conditions under which we write determines what the undertaking makes us into: what it requires us to do, to know, and to attend to.

All of the roles which have been marshaled into the automated persona of Khanmigo—creative consultant, co-writer, all-around editor—have of course always been available to writers in one form or another, but at a distance, and it usually required real effort to access them: I could seek the conversation of a thoughtful interlocutor to help bring shape to my thinking, or the counsel of a trusted editor, or a proofreader’s second pair of eyes. But for the most part, the basic condition of the writer has always been to be alone: not the desperate isolation of loneliness, but the dynamic inner dialogue of solitude—what Arendt later called the “two-in-one”, that conversation with oneself which constitutes all real thinking. Not for nothing does the scenery which has surrounded writers of all ages tend to conjure images of the secluded forest hut, the rustic cabin on the shores of a remote pond, or the barren retreat in the wilds of the desert. Even the modern icon of a figure like Jean-Paul Sartre urgently converting cigarettes and coffee into upwards of twenty pages per day of existential philosophy on the patios of crowded Paris cafés depicts, for all the infernal churn of the multitude, the writer working alone, and in dialogue only with himself: L’enfer, c’est les autres.⁷

But if writing has always involved a measure of solitude, Khanmigo is the product of an entirely different, and decidedly contemporary, impulse. Indeed, even the name “Khanmigo” plays on the Spanish “conmigo”, meaning “with me”. It was created to be, in the words of Khan Academy’s website, “your always-available writing coach”. Consider some of the antecedents here: Facebook’s defining prerogative was to connect everyone with their family and friends—and ultimately with the rest of humanity; as Mark Zuckerberg put it in a 2017 statement, Facebook’s “whole mission” is to “bring the world closer together”. This sort of connectivity-at-a-distance, of course, was all the more necessary after nineteenth and twentieth century technologies of transportation, together with various other forms of social fragmentation, had rendered physical separateness more practical than in any prior era. Like countless purveyors of digital media before him, for Sal Khan, there is a democratizing and equalizing potential behind the endeavor to connect each and every student to an always-available artificial tutor capable of performing functions that may otherwise remain unreachable for most of them; but whatever else the proliferation of AI tutors is likely to bring about, it’s important to recognize that it represents only the latest episode in a much longer-standing trajectory which bends towards rendering everyone and everything available to everybody, its supreme commandment: only connect.

An earlier phase of this transformation was explored in an influential 2009 essay, “The End of Solitude”, by William Deresiewicz . Many of the distinctive characteristics of our modern sense of self had been made possible by a unique technological configuration which, beginning in the seventeenth century, was marked by the widespread availability of texts which had been touched off by the appearance of the printing press two centuries prior. The press not only created the modern printing industry, with the books, newspapers, and scientific and literary journals that sustained a new culture of reading and writing; it produced such notional associations as the new Republic of Letters, which later established what we have come to call more broadly the ‘public sphere,’ in which events of public interest could be debated, opinions aired, and a kind of dynamic consensus reached, which ultimately lent political legitimacy to the state which purported to act on the public’s behalf. But the wide circulation of newspapers—and, later, radio and television broadcasts—also created a new kind of mass society, one which could be measured and polled, and the results would come to represent what we now call ‘public opinion.’

But in a complimentary development, this new relationship to the printed word also made possible our distinctly modern sense of ourselves as individuals—as direct correspondents with God through an inner religious conscience; as rational minds in possession of an interior source of self-certainty; as political subjects whose inalienable rights must be defended against the trespasses of the state. And with these newfound dimensions of subjectivity came a corresponding need for the forms of solitude and introspection necessary to sustain and defend them. In its Romantic manifestation, it’s the writer—and above all the poet—whose singular genius lies in a gift for channeling nature’s awesome forces into verse and giving written expression to the visions that can be wrested from that numinous totality. For these voices of creative intensity, as well as for their nineteenth and twentieth century cultural descendants, the greatest danger was submersion beneath the currents of mass society: conformity, inauthenticity, assimilation to popular culture, and any other failure to live deliberately.

But today, Deresiewicz suggests, “we no longer believe in the solitary mind”; “[O]ur great fear is not submersion by the mass but isolation from the herd.” The socially-networked mind enjoys our allegiance, and it’s impossible to overlook the role of era-defining technologies in making this so. Telephone, radio, television, internet-enabled social media: each of these, in its own way, is a technology of connection, of making contact at a distance. (“Reach out and touch someone”, Deresiewicz reminds us, was AT&T’s ubiquitous slogan in the 1970s and ’80s.) But every benefaction also harbors a corresponding flipside: in addition to the entertainment television can deliver, its absence now subjects us to a new experience of boredom. “Television, by eliminating the need to learn to make use of one’s lack of occupation, prevents one from discovering how to enjoy it. In fact, it renders that condition fearsome, its prospect intolerable.” And if social media connects us with others, its absence creates novel forms of loneliness—of isolation as it had never been experienced before: in its absence, as an affliction which none should be expected to endure, and in its presence, as contact without substance.

A self-avowed product of the 1970s TV generation, Deresiewicz observes:

[T]he Internet is as powerful a machine for the production of loneliness as television is for the manufacture of boredom. If six hours of television a day creates the aptitude for boredom, the inability to sit still, a hundred text messages a day creates the aptitude for loneliness, the inability to be by yourself. [...] If boredom is the great emotion of the TV generation, loneliness is the great emotion of the Web generation. We lost the ability to be still, our capacity for idleness. They have lost the ability to be alone, their capacity for solitude.

And, to extend the pattern into the chatbot phase of media-enabled connectivity, if perhaps the great emotion of the AI-generation may be helplessness, then they are at risk of losing the ability to depend on themselves for knowledge, skill, and proficiency—their capacity for self-reliance. It is above all this Emersonian virtue, I want to suggest, which the peddlers of ubiquitous AI-powered writing assistants appear to have set their efforts against: the solitary writer, thinking, possessed of the full range of the competencies demanded of them by their craft, and pursuing a truth or an artistry which must take a form not answerable to the crowdsourced verities of algorithmic consensus.

As a cultural value, self-reliance—long so deeply rooted, especially in North America—appears to be coming due for reappraisal: by suggesting themes for the student to write about, Khanmigo mocks the student’s capacity to discern the meaning of a text in a way that is not pre-packaged and spoon-fed; by providing counter-arguments to a student’s reasoning, it diminishes the need for the writer’s own intellect to discover how an argument will sound to those who are not already convinced of its truth; its advice about tone, voice and perspective communicate to the student that it is unnecessary for a writer to develop their own ear for these qualities; and finally—and, indeed, above all—its constant presence and watchful gaze means that the writer, once an icon of solitude, now need never be alone. It supplies a powerful and perpetual reminder that the many struggles that necessarily accompany writing—and which alone can turn a writer, properly speaking, into one-who-has-written—are needless burdens which are borne only in vain. The message is implicitly addressed to the writer: this difficult business is beneath you, and ought instead to be turned over to the machine that can do it better.

If, as L. M. Sacasas has argued, the message of the medium of artificial intelligence in its many guises is that faculties like thinking, responsibility, and agency can and should be outsourced rather than continuing to be exercised by human beings, then its use should be expected to encourage us gradually to become de-skilled, intellectually disempowered, and divested of our responsibility for the fate of the world. And in one of the many ironies that have characterized much of the history of communications technology, it seems that this may occur at precisely the moment—and, indeed, for precisely the same reason—that we will also convince ourselves that we have gained rather than lost capacities, made ourselves more rather than less powerful. Beyond the individual’s experience of intellectual helplessness which Khanmigo seems almost to have been designed to cultivate, the more of our competencies we cede to machines, the narrower the share of the world we will have preserved for ourselves, and the more we will have consented to become a burden, merely to be tolerated, supported, and looked after, on the otherwise efficient systems that increasingly manage our fate.

---

If the act of writing, whether the fingers make contact with a pen or a keyboard, shapes the writer by offering an incomparable gateway to understanding, then it goes without saying that whoever renounces the practice also surrenders this possibility. The emergence of the modern self out of, among other things, mechanical technologies of textual reproduction opened unrealized possibilities not previously explored; beyond its capacity to establish readerships and to help found public institutions, the printing press—with the essays, poems, and psychological novels it made newly reproducible and profitable—helped to reveal a subtler and more articulate sense of the unseen spaces we harbor within ourselves. In one influential framing of the resulting social transformation, a culture of sincerity, in which we derive our sense of identity primarily through the sincere enactment of social roles which are largely fixed and externally dependent, yielded to a culture of authenticity, one which is more interested in the possibilities that await discovery within the interior of the individual. This inward turn meant that the authentic self now needed to be defended against the influences and injustices imposed upon it by external social realities. But a different constitution of the self, encouraged by technologies that offer collaboration and partnership, is now coming into view. The human-AI hybrid that emerges from this alliance, as it has been described and defended by entrepreneurs like Sal Khan, declares an outward-facing orientation, one that looks askance at solitude and has little need for self-reliance—once the cherished virtues of a now fading literary culture.

And it will also inform the kinds of writing that will be produced: indeed, every technology influences the stories we tell, just as surely as it shapes which ideas which we are inclined to take most seriously. As Karl Marx once wondered:

[I]s Achilles possible when powder and shot have been invented? And is the Iliad possible at all when the printing press and even printing machines exist? Is it not inevitable that with the emergence of the press, the singing and the telling and the muse cease, that is, the conditions necessary for epic poetry disappear?⁸

It isn’t that the poet is irrevocably deprived of some essential faculty with the appearance of the press; it’s that these technologies don’t just change our material conditions, but also reshape our world: our estimation of what is beautiful or excellent, our sense of what is interesting, of which stories need telling, of what is worthy of inquiry and exploration: the world begins to speak to us differently. I won’t be so bold as to venture a prediction about which forms of the written word might flourish as algorithms become more deeply insinuated into their composition, or about which are most likely to wither from neglect. But I do expect that the technologies that define the process by which we interpret the world will in turn shape what and how we write.

1. Hannah Arendt, “Understanding and Politics” in Essays in Understanding, pg. 322.

2. Mary Harrington recently wrote, for instance, about Alpha, which calls itself an “AI Powered Private School.”

3. If you’re interested in the topic of AI in education more broadly, check out Stephen Fitzpatrick’s Teaching in the Age of AI, which engages thoughtfully with these questions every week.

4. It may seem as though this attitude is currently in some disrepute, given the backlash which AI has lately received; AI may be less popular now than it has ever been, but the pedagogical aspirations which I describe in this section are far from dead.

5. Later in the same paragraph, Khan even looks forward to a world where Khanmigo recommends grades to teachers for the papers it has helped the student to write. He also reassures the reader that:

It would also be difficult for a student to cheat using ChatGPT in this context. If they use ChatGPT to write the essay but simply copy and paste text into their assignment, Khanmigo will tell the teacher: “We didn’t work on this essay together, it just showed up, so we should be suspicious.

In other words, students who have chosen not to use ChatGPT in the process of writing their paper will be accused—by ChatGPT itself—of cheating on their assignments.

6. See this recent post by John Warner. Or, better still, see his new book, More Than Words: How to Think About Writing in the Age of AI, which didn’t come out in time to influence the present essay, but which I highly recommend if you’d like to do some deeper thinking about writing and AI.

7. Indeed, it was Charles Baudelaire who wrote: “Multitude, solitude: identical terms, and interchangeable by the active and fertile poet. The man who is unable to people his solitude is equally unable to be alone in a bustling crowd.” (From “Crowds”.)

8. The passage is from Karl Marx, “A Contribution to the Critique of Political Economy”.