Title: “OK Computer”
Author: Dawn Chan
Publisher: Bookforum
Date: April 8, 2021
Form: Book Review
Keywords: Artificial intelligence, algorithms, authorship, experimental literature
Editorial Letter is framed as a newsletter about making arguments. It tends to proceed from the assumption that effective arguments consist of lucidly articulated claims clearly linked to well-chosen evidence and organized into a transparent structure. In other words, it usually implicitly puts a high value on certainty, and on the apparatus that makes that certainty communicable.
What follows is a post about uncertainty: a reminder that the field of “effective arguments” is capacious enough to encompass not just tight, polished artefacts of certainty with perfectly honed thesis statements but also writing that’s more speculative and open, that runs up against what can’t be easily grasped, and that tentatively begins the process of mapping out new terrain. This kind of writing still produces new knowledge, but it does so as much by attending to what can’t be known as by perceiving what can.
This post grew out of reading the following one-word sentence in Dawn Chan’s Bookforum review of Pharmako-AI—a book by K Allado McDowell cowritten with an AI language model called GPT-3:
Hmm.
I really like this move. Beyond simply signifying uncertainty, it does something important to the pacing and perspective of the piece: it’s a brief pause—taking a moment to deliberate—and it marks a shift in point of view: what we’re reading is no longer the words of the writer after the fact (the writer who has gathered their thoughts and turned to setting them down in prose) but rather the words of the writer reading and thinking, in the midst of engagement with their object. Here’s the sentence in its original context:
What’s strange is that when GPT-3’s musings in Pharmako-AI leave you flummoxed, you don’t know who or what to blame; just as when its insights feel startling and wakeful, you don’t know who to thank. Take, as an example, the machine-generated line appearing in the book’s twelfth chapter: “A cybernetic poetics would have to recognize the ways that the Western medium of consciousness, the modernist Umwelt, perpetuates the unsustainable reality that we are experiencing.” Hmm. Is Western consciousness perpetuating an unsustainable reality? Does Western consciousness constitute a “modernist Umwelt”? Do you agree? And does it matter? What do we make of our own grappling with the meaning of these words, when they’re not even understood by the algorithm stringing them together?
The experience Chan is describing here is one that’s fundamentally structured by uncertainty: the book she’s reviewing is partially written by an entity of an entirely different constitution than a flesh-and-blood author. Reading words written by another human is an experience that’s already opaque and precarious enough; the words of an AI-language model are murkier yet.
Consequently, the review is dotted with markers of uncertainty. Here are a few:
Consecutive sentences beginning “Perhaps”
Consecutive sentences ending “sometimes”
Adverbs like “arguably”
Verbs like “seems”
Qualifiers like “a sort of”
Judgments made explicitly subjective, e.g., “I enjoyed,” “For me”
Often, though, these moments of unknowing are followed by a sentence like this one, which grounds the reader in what can, after all, be known:
But the project never leaves us guessing as to which passages are written by human and which by machine.
These sentences mark out the boundary between the territory that’s safely mapped and the places where the dragons are.
The review ends with the claim that “we’ll all have to get used to this way of reading soon.” For now, what Chan is charting is the experience of a text that, though not without precedents, asks to be read in a way that nobody is really used to. For now, we’re still in the “Hmm.”