Marc Kohlbry reviews The Eye of the Master

Matteo Pasquinelli. The Eye of the Master: A Social History of Artificial Intelligence. New York: Verso Books, 2023. 272 pp.

Review by Marc Kohlbry

30 May 2024

Behind the forerunners of what has come to be known as critical AI studies, radical publishers from Verso Books to Pluto Press have championed titles dedicated to uncovering how seemingly immaterial technologies are inseparable from the material realities of political economy. Among these is Matteo Pasquinelli’s The Eye of the Master: A Social History of Artificial Intelligence, which holds that previous studies “‘from above’” of AI’s impact on society have failed to account for “the role of collective knowledge and labour as the primary source of the very ‘intelligence’ that AI comes to extract, encode, and commodify” (p. 12). As a corrective to works that would accept AI as a formal imitation of biological processes, The Eye of the Master proposes a materialist turn for critical AI studies by endeavoring to trace how “the design of computation and ‘intelligent machines’ follow the schema of the division of labour” (p. 7).

However, while a cursory reading of the study’s subtitle and these introductory claims might frame The Eye of the Master as a history from below, Pasquinelli’s methodology is not geared toward elevating the voices of those immiserated by these technologies or their predecessors, nor is it ultimately interested in unpacking how AI may be formally modeled on capitalist labor or management. The study instead hinges on intellectual histories and media archaeologies that, among other ends, generatively bring many of the usual suspects from histories of digital technologies (including Charles Babbage, Warren McCulloch, Walter Pitts, Ross Ashby, John von Neumann, and, to a lesser extent, Frank Rosenblatt) together with others from political theory, namely Karl Marx and Friedrich Hayek. This, by the author’s own admission, positions The Eye of the Master less as a people’s history of AI than as an inquiry into the “standpoint—the social classes—from which AI has been pursued as a vision of the world and epistemology” (p. 11).

Even still, without faulting The Eye of the Master in bad faith for not cleaving to some preconceived notion of what a “social history” ought to do, Pasquinelli’s text never quite shakes the tension between its commitment to a materialist historicism and its pursuit of political epistemologies of AI. On the one hand, for instance, the book’s scope initially spans centuries and civilizations, beginning as it does with an analysis of cultural techniques in 800 BCE India, continuing with claims that “algorithmic thinking emerged as a material abstraction” and that “labour [was] the first algorithm,” and expressing a larger goal of exposing how the social intelligence behind the division of labor oriented towards commodity production (and the valorization of value) “shapes the very design of AI algorithms from within” (pp. 16, 12). On the other hand, however, the author maintains that one should read the text as a series of “independent ‘workshops’” that, rather than elucidate “the influence of external factors on science and technology,” provide snapshots of the “dialectical unfolding of social praxis, instruments of labour, and scientific abstractions within a global economic dynamics” (pp. 16, 13).

This tension animates the book’s two major sections, “The Industrial Age” and “The Information Age.” In Part I, Pasquinelli questions how labor functions “as a source of knowledge” by examining “the automation of mental labour during the industrial age in the UK” (p. 16). If “this historical moment is usually studied from the perspective of manual labour, capital accumulation, and fossil energy,” Pasquinelli’s contribution is to argue that the industrial machine was at once “a means for augmenting [manual and mental] labour,” and “an instrument (and implicit metrics) for measuring it,” leading him to the provocative conclusion that “the applied division of labour . . . is the ‘inventor’ of automated computation” (pp. 17, 54). More specifically, here Pasquinelli thinks Babbage’s “difference engine” alongside Marx’s “collective worker” to posit a “labour theory of the machine” (at other times termed a “labour theory of automation” or, in the context of AI, of “machine intelligence”) that would explain how “social relations and in particular labour cooperation” are in fact “the ‘engines’ of technical and political development” (pp. 116, 93).

Where a bottom-up history might next zoom in on how specific figurations of the division of labor (that is, in the factory or workplace) serve as schema for the technical development of modern AI technologies, Part II of The Eye of the Master privileges media histories that chart the rise of connectionism in cognitive science from the 1940s onward with attention to its conceptual resemblance to (and, thanks to Hayek, direct overlap with) cybernetic and neoliberal deployments of self-organization. Here, Pasquinelli somewhat undercuts his introductory critique of histories from above by registering that AI’s form is indeed modeled on (artificial) neural networks, an abstraction that “can best mirror, and therefore best capture, social cooperation” (p. 155). Still, this assertion adds political weight (albeit indirectly) to recent assessments of the “collective intelligence” undergirding large language models by suggesting that connectionist AI also imitates the capitalist division of labor in both society and production insofar as it draws on and obfuscates cooperative work.[1]

Moments like these index Pasquinelli’s overarching concern with historicizing and politicizing the “scientific abstractions” of AI in lieu of detailing how approaches to disciplining or managing workers have developed in lockstep with algorithmic technologies (as Kate Crawford or Lilly Irani have elsewhere shown). For example, though the author follows his political epistemology of neural networks with an interested reconstruction of how Rosenblatt’s 1957 “perceptron” sought to automate the “labour of perception” rather than model the human brain, such insights do not concretely explain how the social relations of production drive the development of the means of production, AI now among them (p. 233). Nevertheless, as a timely riposte in a scholarly dialectic, The Eye of the Master adeptly recontextualizes many of the abstractions that gave rise to AI—all while encouraging more granular study of the modes of labor exploitation that have evolved concomitantly with technological progress.

[1] N. Katherine Hayles, “Literary Cybernetics: The Point (of the Spear),” New Literary History 54 (Spring 2023): 1293.