🍀 Dating Your (Potential) Executioner
A Conversation with Sean Michaels, author of Do You Remember Being Born?
Kernel Magazine Issue 4 is out now! In case you missed it, you can read the editor’s note here. We’re releasing just a few pieces this week; to read all of our released content so far, visit kernelmag.io. For access to all of our essays sooner, please purchase a copy of the magazine — on sale for a limited time for launch week!
Shira Abramovich is a researcher and translator based in Montreal, QC, Canada.
Sean Michaels is the author of Do You Remember Being Born?, a story of AI and poetry. He is also the author of the novels Us Conductors and The Wagers, and founder of the pioneering music blog Said the Gramophone. His non-fiction has appeared in The New Yorker, The Guardian, McSweeney’s, and Pitchfork. Sean is a recipient of the Scotiabank Giller Prize, the QWF Paragraphe Hugh MacLennan Prize, the Grand Prix Numix, the Prix Nouvelles Écritures, and he has been nominated for the Dublin Literary Award, the Kirkus Prize, the Peabody Awards, and the Prix des libraires du Quebec. Born in Stirling, Scotland, Sean lives in Montreal, Canada.
This interview has been edited for clarity and length.
Dating Your (Potential) Executioner
Interview and introduction by Shira Abramovich
I first encountered Sean Michaels in a rock climbing gym—not climbing, but launching his newly-released novel, Do You Remember Being Born?. Sipping seltzer on crashpads, watching a few younger attendees climb up the overhangs, I listened to Michaels read a conversation between the book’s main character, an elderly poet named Marian Ffarmer, and the book’s AI, Charlotte—and had the eerie feeling that such a conversation, conducted through text chat, could be happening now, or perhaps even tomorrow morning. Whether the parallels to our current technical reality are by luck or by foresight, the novel has struck a chord—Wired recently called Do You Remember Being Born “The definitive novel about art in the age of AI,” naming it one of the best books of 2023.
Fictional depictions of AI have ranged from helpful to malicious, but one of the striking things about Charlotte is that she seems gentle, tender even—fitting for an AI that writes poetry, and even more so for a novel that winds up being, at its heart, about art, labor, family, and relationships. Marian Ffarmer, a 75-year-old poet, has dedicated her life to her poetry—a choice which has resulted in critical acclaim, if perhaps not financial stability, and which has required the sacrifice of most personal relationships. Now, with her son struggling to buy a house, she accepts a lucrative offer from The Tech Company, an unnamed hotshot in the AI space, to write a collaborative poem with Charlotte, their newest AI which specializes in writing poetry. To Marian, it feels like selling out, sacrificing the solo poetic voice she has worked so long to craft—but it is also a chance, perhaps her only chance, to give something tangible and lasting to her son.
I sat down with Michaels on a snowy Montréal afternoon to ask him more about his process, about technology, art, and labor, and what he hopes technologists and artists might take from the book.
Shira Abramovich: When I first heard about Do You Remember Being Born?, I was struck by how Charlotte, the AI interface in the book, is so similar to the LLMs we’re seeing now—I thought, “wow, when did he start writing this?” How did the idea arise?
Sean Michaels: The first seeds were in 2019, when I strayed across the website TalkToTransformer.com. GPT2 wasn’t really open to the public, but this was a little web app—it was a text box you could write a few sentences in and then it would continue your text for a sentence or two. And I was pretty staggered by it. It wasn’t consistently coherent. It wasn’t smart in the way that LLMs in 2023 are, where their quality is measured on accuracy and knowledge. It was really just, on a prose level, able to continue a train of thought or take it twisting into grotesque and interesting ways. And it was really the first time I had encountered a text bot that felt like anything other than a random word generator. Here was something that seemed kind of magical and bewildering and better than I could understand how it might work, better than I could imagine it being, and I was really provoked by it.
That ended up mixing with the idea of a poet in the near future working with some later generation of such a thing. But the book was definitely always imagined to take place in the future. In fact, early drafts had a first page that said “five years from now.” It went from five years to one year and then at some point, I said, let’s just cut that and be vague, because the book does take place in the future. Charlotte, the AI of the novel, is kind of, in a hand-wavey way, a more general AI. She’s not, strictly speaking, a large language model. But still, I thought, by the time publication comes, this might have happened the week before.
Yeah—it was also interesting that it’s a conversational chatbot in the book. That’s what we've been seeing this year.
Imagine that, right?
In a previous interview, you talked about the story of the poet Marianne Moore that led you to choose a Moore-like figure as your poet. I was wondering if you’d retell it.
Sure, I’ll tell you. When I played around with these things, it felt a bit like going on a date, going on a walk with my potential executioner. Here’s this cool technology that might also spell the undoing of an area of human endeavor that I treasure. And so there was something kind of dirty about it.
Those thoughts ended up getting into an interesting dialogue with the story of Marianne Moore that I had read around the same time. She was a great 20th century poet, a contemporary of T.S. Eliot. She was a celebrated poet from a young age, worked for decades, but only became famous in her older age, partly because she was a character. She was full of bon mots and wore a tricorn hat and cape and would go on late night talk shows and throw the first pitch at baseball games and write liner notes for Muhammad Ali. And as a result of that, she was approached in 1955 by Ford, who asked her help naming their new car. To me, it was a perfect encapsulation of that thing. Almost all of us play along with and are complicit with capitalism and industry, which we don’t necessarily ethically or even aesthetically like, but the opportunities can be tantalizing and curiosity-provoking—much like working with that AI.
It’s like, should I? Is this good for humanity, for me to do this? Yeah, maybe. Maybe not. But it seems like a fun project. So Marianne Moore leaps at the opportunity to name a new car—who wouldn’t want to name a major pop cultural institution? And yet, it’s not like we feel like Ford is a good on earth. And so I thought there was something similar in the way that a poet would say, oh, of course I’ll work with this AI. It’s too tantalizing a prospect to refuse, especially if it comes with a check. That, I felt, was an interesting parallel there—how readily Marianne Moore sat down at the table with the man from Ford and how easily my fictional Marian might sit down at a table with the software.
I really like the phrasing “date with your executioner,” and I did think it was interesting that you didn’t really go into some of the moral or ethical issues we’ve seen crop up this year—for example, the copyright stuff around whether artists should be compensated for models that are trained on their work. Was it a conscious choice to not address it?
It wasn’t a conscious choice. I mean, it just shows to what extent I was unconscious of the moral issue of it. My opinion on this is that the moral repugnance, the kind of instinctive recoil that many artists have, has less to do with the idea of AI somehow subsuming their work and more about a giant rich company secretly scraping their work in order to create a product that earns hundreds of millions of dollars for them and their shareholders, and not for the people whose work is adopted. It feels like a separate question from the technological one, so when I started working on this, I had no moral qualms about the idea of feeding it in my work, certainly in my software.
In a world without money, where everyone is okay, would it be wrong to create software that could write poetry that’s nourished on published human thought? In my opinion, no.
Yeah, that reminds me of a quotation by Ted Chiang—“Most of our fears or anxieties about technology are best understood as fears or anxiety about how capitalism will use technology against us.”
Yes. And in the book, that was one of the reasons I wanted to make it an issue with her son and housing, to make this really manifest. You have this manifest inequality and precarity in life that’s implicated in any question about large-scale technology.
Yeah, I found that part of the book really tender, and it was so interesting to me that a book that was about this powerful AI would lead to a book that is so much about caring for one another.
I wanted to write a book that was fundamentally about labor and inheritance and parenting and poetry and art, using AI as a metaphor for some of those topics rather than the other way around. At its heart, Marian is afraid of, and has resisted, collaboration all through her life, and I thought her being confronted with an AI was an interesting way to manifest the idea of opening her life up to disruption and collaboration.
I actually wanted to ask you about collaboration, because it winds up being a really important thread in the book. I was curious what it was like, collaborating with the AI as you were writing, and how you work with collaborations in your writing practice generally.
I don’t collaborate much. I’m a solitary writer. This book is about a collaborative poem, which itself isn’t that common as a form. But this book is, in a certain way, a collaborative novel because of the technology use, which is even rarer, and I understand why. In our culture, there’s a tension between that cultural desire for Capital-G Capital-W Great Works of art and what kind of artistic practice is more likely to lead to a happy life. There’s a healthier mode of life that is about documenting your life of art through products—let’s say, works—trying to make works that are imperfect, but that show the work that you’ve done. Making more good things, I think, leads to a happier life and a life in more equilibrium than where you hide away in search of greatness.
This is a totally lofty conversation but I think in fact, art is improved by interference, by all the noisy things that come into life. Meeting people of all ages, older people, younger people, less experienced, more experienced, artists—that really nourishes your work in surprising ways.
I hadn’t planned out the whole arc of the novel. I was writing, and I realized—I don’t know how this ends, nor how it should end. Should my poet succeed at creating a masterpiece with an AI? Should she fail? Is that more important? What’s the out? What’s the exit? What do I believe? And without wanting to give spoilers for their readers who haven’t read it, I found myself thinking more and more about how the thing I was sure of is the idea that bringing other people into exploring a choral version of art or of human progress—that, I feel, can be generative and nourishing—the other kind of generative.
I came to your book launch, which was in a rock climbing gym, and I am told that this isn’t the first time that you’ve hosted a literary event in “unconventional” spaces. You also had the audience participate in interesting ways—I was curious how and why that became a thing that you did.
The community I merged into in Montreal when I was a younger man was the indie rock music world and it was, you know, a little bit of a scene. I saw the way that being part of a healthy scene was this engine for creativity and for human flourishing. It’s really important to me to feel embedded in a community and then for people to feel embedded in a community with me in that reciprocal way. I don’t want to be dislocated or removed from context. I want to be part of the geographical scene that I am in.
Then I thought, what kind of event, what kind of book launches, could I hold where my community is going to be interested in manifesting? Because something at a bookshop draws on a certain kind of person who will show up, and maybe not a wider audience.
So why the climbing gym?
There was no direct reason. I was trying to think of a space that felt like it could hold many types of people, that could host an encounter that would feel memorable.
I thought it was funny, personally, because there’s the tech bro archetype. A lot of tech bros really like bouldering.
Yeah, right, exactly. Well, at my book launch, the only tech bros bouldering were under the age of seven, I think.
One other thing that I wanted to ask you about was the process of creating that AI character in the book, learning about AI and so forth. What was it like for you and how did you go about it?
It's been a kind of odd process as someone who is outside of that world. Particularly over the past few years. When I came into it in 2019, I was like, what is this weird thing? I'm a very amateur coder, but I know the Internet, I’m very online, I know how to follow instructions. But there was no centralized place to go to understand how to do these things.
I ended up getting a few different grants to get help to come up to speed on some of these issues from the Canada Council for the Arts and from the Quebec equivalent, the CALQ. I was able to hire Jasmine1 to consult, to teach me how to get access to GPT, and then how to use it intelligently. Later, another grant allowed me to find engineer Katie O’Nell to actually build a poetry bot. I discovered that GPT2 and especially GPT3 were really pretty good at literary creation and lyricism, copying my writing style, by giving it as large a prompt as possible, in a way that ChatGPT is terrible. GPT2 and GPT3 were good at this lyrical style mimicry, but cannot do poetry. None of these things can. They’ve just been trained wrong into conceiving of poetry as this rhyming doggerel.2 And anything that’s not rhyming doggerel, it doesn’t understand. You say in the style of ee cummings, it still gives you There once was a woman from Panama… You say, No, no, like this, and give an example. And it still says Ohh. Gotcha. There once was a woman from Panama…
So I found it was impossible at writing in this poetic voice that I had in mind for Charlotte. And so Katie and I decided to make this janky thing. What we ended up producing is pretty slow, not public facing, and it loses the plot really, really quickly. But what it’s able to do is that you give it a short prompt to a few lines and as it starts to riff in such a way that frequently generates interesting phrases, interesting little twists of mind, which are the things that got me excited about AI in the first place—not the ways that it gives me a predictable accurate answer, but instead gives me a surprising imaginative one with speed and limitless inexhaustible energy.
In the past year and a half there’s been such a push that what we need from artificial intelligence is reliability, when what I want from artificial intelligence is a panoply of voices. Can each be different? You want your weird little creature that will perk you up in the morning. You want the less-weird creature that corrects your mistakes. You want the one that will book your plane tickets that’s never weird at all. You want this strange thing that rereads your work, and in the voice of a depressive Virginia Woolf tells you how to make it better. And then the one that reads your work in the voice of a manic, I don’t know, Bukowski or something tells you what to make it better. I think that variety would take away a lot of the fears that we have of blandification of writing and replace it with something else, even as I still feel the same worries about the negative impacts AI would have even in that world I’m describing.
Is there anything in particular that you would hope a technologist would take from the book, and anything in particular that you might hope that writers or creatives take from it—are those different things?
I have two thoughts initially. One isn’t exactly what you asked, but I would love for technologists to take from my book that artists really don’t make much money. I really get the impression as I’ve touched more into the AI world [that] even creative people in AI really have no understanding of the economics of almost all the art that they consume in the world. Some of people’s favorite writers and poets and painters and musicians definitely live on less than $100,000 a year, and may live on less than $50,000 a year, and small amounts of money can really have transformative effects on people's careers. You know, Marian Ffarmer, the poet in my book, is willing to basically mortgage her soul for $70,000 or $80,000 because, for a woman who’s in her 70s, it’s a life changing amount of money.
Also, there are certain kinds of entertainment products that can be engineered and optimized, but there’s a lot of art that deliberately or inherently resists that optimization. It’s not solvable. There isn’t a correct answer. I like that it’s a problem where you can’t just take the measurement or poll people or train and do a probabilistic thing and say, well, in 85 of 100 universes, this is chosen as the best answer, thus it is the best. It’s like, no. And I think that's an important lesson to learn.
Besides that, I would ask [technologists] to play around and experiment and consider the intersection of the human and creative and the weird machine tools. That intersection is really interesting and in the history of humanity has resulted in so much interesting stuff—from the invention of cement and the way that allowed different kinds of architectural creations to photography, to pigments and dyes and paints. The history of humanity is a history of humans using technology in really exciting ways. My most beloved technologies are, you know, the ones that keep people from dying, but also the ones that humans can use in extraordinary and generative ways. And I would like technologists to think not about how they can shut down, have less and less human input, but how they can help humans be more expressive. Similarly, I think that artists should be curious about any tools that make those promises. Because who knows what’s around the corner?
Thanks for reading — if you enjoyed this interview, we’d really appreciate it if you shared or forwarded to a friend!
Launch party links once again: DC, SF, and NYC. Hope to see you there!
Jasmine Wang, editor-in-chief of Kernel 1.
"Rhyming doggerel" is the rhythm and rhyme that is commonly seen in nursery rhymes and children's songs.
 | A guest post by
|