The Automation of Antihumanism

On Leif Weatherby’s ‘Language Machines’

Last week’s post on gay frogs and cosmic horror was the first of what will be several follow-ups to my modestly viral August thread on X arguing that the “obligatory dogma of humanities academia is antihumanism.” I pointed this out because the unifying function of antihumanism is something right-wing critics tend to miss. In academe, “you [can] be a feminist, Marxist, postcolonialist, etc,” as I wrote, “but the thing you [aren’t] supposed to be was a humanist.” The fact that this is the unifying credo of the humanities is part of the point: Their guiding ethos long ago became all-consuming internal critique. Two examples I like to cite are the 1992 book Against Literature, written by—you guessed it—a literature professor, and the 2020 essay “The Case for Letting Anthropology Burn,” published in—of course—anthropology’s top journal.

Often, antihumanism was more of an undercurrent, but in recent decades, having decentered every other “centrism,” scholars in various humanities fields set their sights on the most centric of them all: anthropocentrism. Accordingly, they questioned the category of the “human,” and thus “humanism” on both intellectual and moral grounds. Privileging the human, the standard argument goes, provides a basis for exclusionary and violent ideas and practices. The posthuman, the inhuman, the “animal turn,” and other mots du jour all evidenced a deep philosophical discomfort with the very category that historically grounded the humanities.

But what happens when academic antihumanists turn out to be dispensable to the realization of antihumanism? This is one way to understand the crisis generative AI has created for the already beleaguered humanities. The most widely discussed aspect of this crisis is practical: What is the continued value of the humanities when many of the core competencies it claims to impart to students are easily outsourced to technology? But the philosophical conundrum it represents for these disciplines is, if anything, more daunting. Like so many humanities scholars before them, the entrepreneurs and engineers pursuing AGI openly seek to “decenter the human,” disaggregate the bundle of qualities associated with it, and explore the possibilities of nonhuman agency. Needless to say, they have considerably more resources at their disposal than humanities departments.

Put differently, antihumanists have spent half a century deconstructing “Man,” but now face steep competition in that effort from tech companies with trillion-plus dollar valuations. The most common response seems to be the defensive crouch that NYU professor Leif Weatherby calls “remainder humanism” in his interesting recent book Language Machines. This a “humanism without a theory or doctrine of what is human, in which humanity is remaindered, like a book past salability. The human here is defined by technology’s creep, but only negatively.” Having repudiated all positive definitions of humanity, academic humanists can only “[shout] more loudly each time something is quantified that the ‘essence of the human’ can never be quantified”—and on this basis, presumably, demand that university administrators continue to fund their departments.

Weatherby, conversely, makes a strong case that the rise of generative AI presents an opportunity to revive a particular strain of antihumanism: the structuralism of Ferdinand de Saussure, Claude Lévi-Strauss, and Roman Jakobson, which articulated a “nonreferential theory of language” as a system of “differences without positive terms.” “For machines to communicate,” Weatherby writes, “they need to use language as it is internally structured … Once we see this, we can see that ‘meaning’ resides not in one or the other part but in the relationship of part and whole.” In other words, machines have been able to master language because the structuralists were right when they argued that language itself is a kind of machine, and not primarily a vehicle for expressing human meanings and intentions.

Resuscitating this approach, Weatherby argues, can offer a new vocation to the humanities. To study the “encounter between the generativity of computation and the generativity of language,” he says, “we need a general poetics”—a revival of the project pursued by Jakobson and others in the midcentury heyday of structuralism. “Literary scholars,” he says, are “the only professionals positioned to do this work.” Here he converges with his nemeses, the “remainder humanists,” in being attuned to the practical problem of justifying the budget lines of humanities departments. As he puts it elsewhere: “[W]e will surely have to train our citizens to understand how to use language in computational settings in new ways.”

The potential problem here is that LLMs threaten to automate the labor of literary scholars and literature instructors no less than the production of the textual artifacts they might scrutinize. Weatherby recognizes this, to an extent. He recalls Frederic Jameson’s proposed vocation for humanistic study in postmodernity: the “invention and projection of a global cognitive mapping.” Today’s generative AI, he says, “fulfills exactly this function” because it shows us the totality of our culture in the form of algorithmically processed statistical aggregates. The twist is that instead of demystifying culture by revealing how it works as a linguistic system—as, for instance, Roland Barthes attempted in S/Z, his analysis of Balzac’s “Sarrasine”—AI’s output is itself “ideology,” and thus in need of demystification. This is where Weatherby thinks literary scholars come into the picture, to perform the debunking operations of Ideologiekritik.

But this project isn’t immune to the threat of automation either. The anthropologist Bruno Latour already anticipated this in his 2004 essay “Why Has Critique Run Out of Steam?” Latour argued there, in short, that the hermeneutic of suspicion that long defined humanistic inquiry had been banalized in a culture permeated by postmodern cynicism toward the establishment, institutions, and official truths. Latour argued that the radical skepticism of the academy had been—metaphorically—automated and mass produced. “[C]ritique,” he wrote, “has been miniaturized like computers have. I have always fancied that what took great effort, occupied huge rooms, cost a lot of sweat and money, for people like Nietzsche and Benjamin, can be had for nothing, much like the supercomputers of the 1950s, which used to fill large halls and expend a vast amount of electricity and heat, but now are accessible for a dime and no bigger than a fingernail.”

Two decades after Latour wrote this, his metaphor has been strangely literalized. Today, LLMs can be prompted to churn out instant structuralist analyses and Frankfurt-School style critiques of ideological artifacts. Weatherby castigates his fellow scholars for “retreating into [the] remaining corners untouched by computation.” But when he invokes structural linguistic analysis and ideological critique as non-automatable tasks that can only be performed by literary scholars, he seems to be retreating into such a corner himself. At the end of his essay “What Is an Author?”—one of the foundational statements of antihumanist theory—Michel Foucault asked: “What difference does it make who is speaking?” For Foucault, the question was rhetorical: His point was to herald “the stirring of an indifference” as to the answer. But Weatherby and others who wish to salvage the humanities from the AI-occasioned impasse of antihumanism will need to reverse the valence of Foucault’s question and try to answer it literally and concretely.

This week in Compact:

• Leila Mechoui on Trump, Mamdani, and the centralization of power

• Juan Rojas on the rise and fall of Bolivia’s MAS

• Ted Reese on why America may be in worse shape than the late USSR

• Dan Hitchens reviews Cory Doctorow’s Enshittification

• Gabriel Rossman on AI and the last literate generation