‘Death of an Author’ Prophesies the Future of AI Novels

Literary supercuts have appeared as recently as Graham Rawle’s Woman’s World (made entirely out of language from 1950s women’s magazines) and Kathryn Scanlan’s Kick the Latch (a collage of interviews with a female horse trainer). Many of these supercuts, like LLMs, look for patterns in found language and use them to cluster fragments of text into narratives that reveal something new and often startling about their sources. 

I did something like this in my recent novel, The Nature Book, which is made entirely of found nature descriptions from 300 novels. To write it, I operated like a data scientist or human algorithm, studying thousands of pages of nature descriptions, looking for patterns, and then using them to guide the narrative. 

Death of an Author seems to function in a similar fashion. After studying patterns in how crime fiction works, Marche gathered them into a novel outline. He then took the LLM’s outputs (themselves the product of machines studying patterns in writing) and pieced them together into a final form. This process is not so far removed from a Burroughsian cutup, but it veers from Burroughs’ opaque writing and toward the seamlessness of the best genre fiction. 

Like supercuts, what gives Death of an Author its conceptual rigor are the rules and guidelines Marche outlined for the text. He decided that 95 percent of the novel would be computer-generated, with only minor derivations from the LLMs’s output; that it would be novella-length; and that the text had to be “compulsively readable, a page-turner.” 

This methodology aligns Marche’s book with the constraint-based work of the Oulipo—works like Anne Garréta’s Sphinx, which features a genderless protagonist and is written in French, a language where every noun is assigned a gender, and George Perec’s La Disparition, a detective novel about a missing person written entirely without the letter “e,” the most common vowel in French. Marche’s similarly elegant blending of form and concept becomes a potent meditation on the life (and death) of authorship in the age of LLMs.

We are now a few months into the widespread popularity of tools like ChatGPT. As Marche notes, there are already hundreds of bad books written with LLMs available on Amazon. This number will increase, as will the amount of text that uses these machines in fascinating ways. Like any other literary tool, from constraints to supercuts to collaborative writing, LLMs offer writers a way to challenge their subjectivity and guide their narratives in surprising directions. 

We can expect more machine-driven, constraint-based works, hybrid blends of original and machine writing, and forms we can’t yet predict. Of course, these tools are not for all writers, and not every project would benefit from them (I, for one, felt no need to use one in this article). But even for writers allergic to LLMs, a new challenge emerges: At a time when voice and style can so easily be mimicked, why not try to write something a machine could never do?