Sean Michaels’s remarkable new book Do You Remember Being Born? sits squarely at the intersection of technology and literature. After reading this beautifully realized and moving novel, I interviewed Michaels about the book and how he is thinking about the new frontier readers and writers are exploring with the advent of generative artificial intelligence.
Michaels also appeared, with author Greg Hurwitz, on a panel about the new threats writers face during PW’s half-day conference Artificial Intelligence: Revolution and Opportunity in Trade Publishing, held on September 27. Both spoke eloquently about the shifting economics of writing, the need for transparency in labeling products created with AI, the dangers of monolithic large language models controlled by tech giants, and the desire for human excellence and community. In our interview, Michaels explained that he has been thinking about the implications of generative AI since at least 2019, when he discovered Talk to Transformer—an early autocomplete program based on GPT-2—and started experimenting with it.
Do You Remember Being Born? was inspired in part by a real incident with the poet Marianne Moore, whom the Ford Motor Company asked to help name its new car model, and by Michael’s experiences with generative AI. In the book, a world-renowned but penniless poet, Marian Ffarmer, agrees to take on a writing assignment proposed by a large tech company: to write a “long poem” with a poetry bot called Charlotte, an AI fine-tuned on poetry. For a week she struggles with the process and questions the mission and her motives, as well as the results of the collaboration. The novel offers a unique perspective on generative AI from an author who has studied its capabilities, and has incorporated what he’s learned into both the form and the content of his work.
As part of the writing process, Michaels used Moore’s poetry to feed a generative AI called Moorebot that he created with researcher Katie O’Nell. Moorebot is the source of the verse in the novel, which Michaels describes as “fake poetry.” Michaels and O’Nell tuned early language models with both human and robotic poetry, but still the results were often nonsense. Michaels recalls the effort and “meticulous crafting” it took to generate passable verse from Moorebot, comparing the process to “writing with a slot machine.”
In the novel, the prose is also “infiltrated”—Michaels’s term—by generative AI. The passages he includes (highlighted for the reader) are intended to show the range of what GPT can generate, from uncannily clever to blandly ordinary, and to demonstrate the “internal biases” of generative AI.
At the heart of the novel are questions about what writing is and how it is produced by humans and machines. His protagonist thinks about the difference between herself—composed of what Michaels calls her “biography, biology, and dream”—and Charlotte, which has access to every modern poem ever written.
“My Marian-ness is in the slender sample of the world that I am able to bring to my work,” Marian reflects. “If we did not have this smallness, these limits, there would be no way to tell Ffarmer from Sappho, or Eliot, or anybody. So what was I to make of Charlotte—not small but all-devouring, ubiquitous, remembering? Anointed, in a way, by her magnitude. And at the same time, I am certain, diminished by it.”
A foundational model built with billions of “tokens” (what developers call words or pieces of words) has no such constraints, which can make its voice, paradoxically, banal and anodyne. Exacerbating this problem are the parameters (or guardrails), designed to reduce randomness and ambiguity, that GPTs have increasingly been assigned.
In Do You Remember Being Born?, Michaels explores another unfortunate corollary of AI-generated output: the tendency of humans to imbue words with meaning that they do not have. Humans as pattern-finding beings are notoriously prone to conspiracy theories and other irrational beliefs.
Michaels writes that the problem is “not merely the emptiness of these emissions, but the boundlessness of human beings’ capacity to interpret, to make meaning from. I could draw substance from any line I read, no matter how hollow its intention.”
In our interview, Michaels made it clear that he thinks we have not yet started asking the right questions about generative AI. What happens when generative AI is able to “write” a moving short story? What are the consequences for the creative professions when the output becomes good enough? When audiences can’t tell if writing is an artful human creation or an industrial product? “The questions that AI brings up for me—like, What does it mean to be fooled by fake art?—are actually exciting to engage with,” Michaels remarked.
The enjoyment readers will experience reading Do You Remember Being Born? is the result of a supremely intelligent writer’s deep engagement with this dizzying new frontier.
Maja Thomas is the chief innovation officer at Hachette Livre and the director of the Hachette Business Development and Innovation Department.