Choosing to walk
AI writing tools, labour, and my only writing advice
i get a lot of messages asking me for writing advice, which i’m often hesitant or unable to give because i don’t think i’m particularly good at the actual task of writing: i don’t have a writing schedule, a daily output, good desk posture (or a desk at all), a spiritual artistic practice, or any of that other stuff that people tend to want to hear about. with all that being said, this is something like writing advice. it’s also about AI, and the future.
also: i was trying to write something a bit more accessible and more conversational here but it ended up being way too long so i cut half of it out and might publish the rest as a part two soon. xoxo
My first real writing job was an editor position at a Canadian music magazine that I landed when I was 19, through a combination of luck and salesmanship and a very trusting publisher (thank you, Michael!!!). I’d done some freelance work before that — some reporting for Vice, a piece for Bitch Magazine that went low-grade viral — and managed to parlay that (along with a very idiosyncratic cover letter) into a convincing argument that I would be able to do something cool for them if they gave me a chance. As a broke college dropout with no idea what I wanted to do with my life and no material hope for the future, it was a true heaven-sent miracle. I got to write about music and movies and art, and I was 19, so most of it wasn’t very good — in fact, I wish I could wipe a lot of it from the internet1 — but I got better and better all the time. I made very little money, but I survived. It was an opportunity I was extremely lucky to receive, and it certainly changed my life, for two main reasons: it made me believe I could be a writer, and then it forced me to actually write.
I wrote a few big passion pieces per issue, but beyond that, a lot of my job was about producing capital-C Content: trend roundups, event postings, listicles, movie synopses, 100-word album reviews. To be honest, I often found this work mind-numbing and pointless, and because I’ve never been good at doing anything I don’t want to do, I was bad at it. Every sentence felt like a struggle so intense it frequently made me physically exhausted; I felt like I was pulling the words out of my brain by brute force and shoving them mangled onto the page, ugly and misshapen and never sounding quite how I wanted them to. When I got tired enough, the words got wavy and blurred and I would start messing up the details: I’d mix up dates, forget the formatting rules, make typos. I often cried with frustration. Even the smallest pieces could take hours, and when I finished, I got neither a private sense of creative satisfaction nor a pat on the back (I was, of course, doing the bare minimum poorly).
It goes without saying that this job was in many ways far easier and more enjoyable than any of the other minimum-wage jobs I’d worked in my life, and the last thing I want to do is complain about it. But it was hard. It forced me to write in a way I hadn’t written since I’d dropped out of school: not out of passion or interest, not for artistic pride or creative fulfillment, not in search of meaning or narrative — just writing sentences on a page, word after weary word, signifying almost nothing, being read by almost no-one, for no reason other than because I had to. It was writing-labour in its most unromantic form.
That was only four years ago, but the job I got when I was 19 would almost certainly not exist today (or won’t for long). A great deal of the work I was doing was almost tailor-made to be done by AI: it was mostly information aggregation, impersonal, neutral, ideally produced as quickly and constantly as humanly possible. ChatGPT would have done the job very, very well. It’s possible that if it had existed at the time, I might have used it to do the most tedious parts of my job for me. If I had, I would have had easier days and more free time; I may have gotten more sleep; my boss certainly would have been happier with me. I would also be an incalculably weaker writer and perhaps a worse person, both then and now and at every point in between.
I think about that job often when I’m confronted by ChatGPT’s new role in the world, its growing ubiquity in schools and homes and offices. I’ve been even more alarmed by tools like Sudowrite, a fiction-writing AI program described as “a creative companion that can help unlock your writing potential”. Sudowrite can analyze your writing style, generate sensory descriptions and metaphor, rewrite paragraphs to make them more descriptive or to contain more inner conflict, suggest plot twists, brainstorm writing beats, weave the beats together into a chapter. One user said, “I often struggle to paint vivid pictures with my words, but this mode changed the game for me… By inputting a few keywords or a simple sentence, Sudowrite generated detailed descriptions that added depth and richness to my writing.” This phenomenal article in The Verge follows the inner conflict of a writer who started using Sudowrite to help pump out several of her self-published fantasy novels per year: “It’s just words, she thought. It’s my story, my characters, my world. I came up with it. So what if a computer wrote them?”
This is a great question, not a stupid one; she has arrived at a conclusion that makes perfect sense within the logics of modern labour and production. This author views her writing as a product, and her audience views her writing as a product, and most products are manufactured in line with this logic: someone comes up with an idea, does a bit of the labour in the beginning, and then outsources most of the labour involved in manufacturing the product on a larger scale off to people who will do it for cheap. The mechanisms of capitalism have chugged along for centuries finding cheaper and cheaper ways to outsource labour; ethicists and lawmakers have attempted to push back at various points, discouraging and criminalizing unethical wages, unsafe labour practices, and forced servitude. AI does the job for free, and you can’t illegally exploit it because it isn’t a person. It is, in some sense, a miracle of capitalism.
The real question, then: is writing a product or an art? Obviously, it can be either, and is often both; it’s probably most like a vast product-art spectrum, with most writing falling somewhere in the middle. I’m a firm believer in the idea that labour has value and deserves to be compensated, so it’s not that I think “real” artists should forgo payment or monetization. I just think the job of the artist is in part to resist the demands of the product-side (optimizing their work for the market, maximizing profit, constantly increasing production, prioritizing efficiency at the expense of creativity) as often as they can, for the sake of their own artistic, intellectual, and spiritual health.
I think of writing a lot like walking. It’s rarely the most popular, the most effective, or the most efficient way of getting to your destination. I don’t always want to do it, and it’s not always technically enjoyable; sometimes it’s boring or slow, sometimes it’s tiring and pointless, sometimes it’s cold or wet or windy and I’m retracing the same steps around my neighborhood that I’ve walked a thousand times and it sucks and I’m miserable and wish I’d stayed inside. Nonetheless, I always feel worse in my body and mind when I avoid it for too long, and it’s a loss that feels greater than just the quantifiable enumeration of calories I didn’t burn or sunlight I didn’t see. Of course, walking offers the chance of unmatched material reward: only through walking might you stumble upon a hole-in-the-wall restaurant that isn’t on Google Maps, a rich lady doing a yard sale on her stoop, a garden, a special tree, a cat, a $10 bill on the ground. But you also might get spat on by a pervert. When you choose to walk, you choose not to pursue immediate gratification or even comfort but simply to expand the number of things that might happen to you. Walking invests in the potentiality of your experience with almost no promise of tangible reward at all, which is something like being alive.
Perhaps because of its position as a fundamental tool in the construction and trafficking of narratives, the practice of writing is often lent a sort of grand mythology. Writing is a sacred practice, a spiritual endeavor, an unorthodox path that most normal people could never understand, a lifestyle so saturated with agony and ecstasy that no sane person would ever choose it of their own free will (writing, as they say, chooses you). Writers are monsters, gods, saints, freaks, mystics. Their schedules and rituals and packing lists are fetishized, sacralized, desperately consumed in the hopes of touching the same buzzing amorphous force that they once touched before you. What is often left out, or forgotten, or willfully ignored, is that writing is work. It can be thrilling or painful or exciting or beautiful, sometimes remarkably so, but more often than anything else, it’s boring. It often has no great reward nor any horrible consequence. It’s mediocre more often than it’s especially good or bad. It’s a task. For many people, it’s even a job.
The general promise of this type of AI tool is that it will take care of all this busywork, the part of writing that’s a task rather than an art, so that you can keep your brain pure and untouched by the indignities of labour.2 Then, you can finally focus on generating ideas (intellectual property, even!). Its promotional material frequently describes the difficult aspects of writing, like when you can’t think of the right word or can’t figure out how to join two ideas together, and dangles in front of you a world in which these ego-jabbing roadblocks will be excised from the magical process of bringing your ideas to life. It attempts to separate the mythology of writing from the work of writing, as though the quotidian labour of expression is something that keeps you from your ideas rather than the exact process by which you discover them. In the process, it paradoxically devalues the intangible “magic” of writing by alienating the artist from their own creative process. It ignores the fact that writing is the hard part, and that the struggle for language or structure is a precious and necessary part of forming a real idea. In a way, these AI writing tools feel like the reverse of my first writing job: rather than requiring pure work with no meaningful end product, the tools promise the production of a meaningful product without work.
But a good sentence floating around your head is worth approximately nothing without the ability to write that sentence down, to gird it with other, slightly lesser sentences, to write not just the one sentence that sings to you but all the other ones, too, the ones that are ultimately necessary to turn what we call an “idea” into what we might call an “essay”. An essay is not the process of translating a fully-formed idea into words on a page; it is the process of discovering and testing an idea by challenging it with form, syntax, structure. The friction between idea and ability that AI evangelists promise to eradicate is not a problem suffered by a disadvantaged few. It’s the fundamental condition of the writer, and it is precisely through that friction that we discover what it is we actually have to say.
Expressing ideas effectively through writing is not something that people are born being able to do or not do; it’s a muscle that anyone can develop and that anyone can let atrophy. Even now, when I’m lucky enough to be able to write exactly what I want to write with very few external constraints, writing often feels difficult for me in the same way it did when I was pumping out listicles — slow, sticky, clunky, painful, humiliating. I wonder if this, too, is part of the appeal of AI: it is so often humiliating to push up against the limits of your ability. But I’ve had a lot of practice with it, and it’s that practice — not the books I’ve read or the ideas I’ve had — that has materially made me a writer. I’m no fan of the exploitative, low-paying media industry that forces young writers to churn out content with no job security, to be clear, and I’m no fan of academia, either. It’s strange to balance my distaste for those institutions with a sense of gratitude that I was forced to figure out how to write even when I didn’t want to, even when it didn’t matter, even when it didn’t feel good. Because it really did get easier, and I really did get better.
There are many ways to practice the work of writing. It can come from homework assignments, boring cover letters, copy, filler, blog posts for your dumb job, texts to your friends, school essays you don’t care about, personal essays that you do care about but that no one reads, private journal entries, long emails, discussion questions that only count for completion credit. It’s hard to find the value in any of our work when we’re so deeply alienated from the fruits of so much of our labour, when we’re trained to optimize for the most efficient path to reward or compensation. But I think it’s worth it, particularly in your private, personal life, to try.
AI is a tool, and lots of people need tools at various points for various reasons. I’m not here to call you a bad person or a failure or a fraud if you decide to use this tool on a cover letter or a homework assignment.3 A car is a tool, and a very useful one, but it’s also a tool that has consequences both for the individual and for the society in which it lives — and even with the more serious ones aside, there is a reason why we still might choose to walk.
I know some of you are going to find some of it. All I can say is: if you’re 19, I hope it reassures you that your best work lies ahead.
I’m specifically referring here to the use of AI as a time-saving or production-increasing tool, not the specific and intentional use of AI in experimental, self-referential, or academic internet art & similarly nuanced projects.
To be honest, I personally think everyone should try to avoid doing any and all of that as religiously as one can. But if that offends you at all, who cares what I think!
Thank you for the essay! I'm going into computer science research, so you can imagine how alone I feel on this issue. I've been re-reading Barthes with increasing frequency as my work comes closer to intersecting with LLMs and generative AI (not so much an optional direction of inquiry nowadays as an inevitability). The death of the author, which I imagine OpenAI executives would delight in, although much of his later writing modifies that argument by suggesting that the author is still alive. If not as an extratextual authority then as a body of information that we can parse and interpret as we do a piece.
But also his conception of the "Text" as a field of meaning that the reader creates in their interactions as much as they consume. Reading and writing were not always discrete activities. Our ability to traverse a text, to play with it, and ultimately to procure something of value from it—all tied to our ability to write. It's precisely where they intersect, the describing relationships to other texts and synthesizing context into concrete meaning, that AI intervenes. Offloading all of our intellectual labor onto AI except at the highest level is tantamount to refusing to engage with the text in any capacity, reading or writing. This process debilitates all text and all writers and all readers. I simultaneously do and do not hope that that becomes more obvious with time.
I often think about the Ursula K Le Guin line from a story where they create instantaneous space/time travel and the guys mother asks “where is the dancing? where is the way?”