In response to:
The Labor Theory of AI from the March 27, 2025 issue
To the Editors:
In “The Labor Theory of AI” [NYR, March 27], Ben Tarnoff is quite right to note Charles Babbage’s emphasis on total visibility and managerial control in his writings on factory work and his own never-finished Difference and Analytical Engines, which sought to extend the “eye of the factory’s master” by automating mathematical calculation. Tarnoff contrasts this visibility with the opacity of contemporary AI modeled on neural networks and trained on massive stores of data. (“Nobody really knows how the technology works.”) However, a different picture of Babbage emerges in his more speculative writing, which highlights what we can’t see. In his Ninth Bridgewater Treatise (1838), a spoof of traditional theistic science, Babbage envisions a form of superhuman intelligence obscure to the most advanced human viewers and “masters” but still “immeasurably below even our conception of infinite intelligence.” For “such a Being,” Babbage reflects, the Earth itself would serve as a massive physical database—“one vast library” in which “the permanent impression of our words and actions” would support the statistical ability to “foresee” and “predict” future developments, through a process somewhat similar to the operations of current large language models and generative AI.
Although hindsight now shows us the limitations of Babbage’s deductive and symbolic approach to intelligence, his Ninth Bridgewater Treatise fantasizes about a more flexible alternative—a demigod-like programmer able to adapt and respond to the “infinitely varied” series of “laws” that govern life, matter, and a world of change and contingency. Babbage’s treatise thus concludes, much as Tarnoff does, that “mortals must content themselves with partial truths.”
Tamara Ketabgian
William and Gayle Keefer Chair of the Humanities
Beloit College
Beloit, Wisconsin
English (US) · 