AI Won’t Make Our Minds Useless According to 400 Years of History

Venkatesh Rao calls AI panic “cognitive mercantilism”: the belief that intelligence is scarce, fixed in supply, and therefore something to hoard or control. Drawing on Joyce Appleby’s history, he contrasts that outlook with capitalism’s contingent emergence from the recognition that wealth could be created rather than only captured. Mercantilist thinking treated wealth as finite and favored elite control; it also assumed most people would remain agricultural laborers, not become significant consumers. Rao argues that similar assumptions about intelligence may be narrowing expectations about what people could do if access to it expanded.

The historical analogy begins with a society in which roughly 80% of people worked in inefficient subsistence farming. As agricultural productivity increased, fewer people were needed to produce food. The consequence was not simply that more people became “useless mouths”: more people could do other things while having access to more food. Rao presents AI’s potential to let most people consume intelligence without producing it as a parallel, not a forecast. The analogy suggests a possibility for changed economic roles, but does not establish how quickly or evenly such a shift would occur.

The discussion challenges job-loss arguments that assume a fixed “lump of labor” and no elasticity in demand. Cheaper intelligence could support demand through six channels: price, access, complexity, continuity, personalization, and relational demand. The point is not that any specific list of new occupations can be predicted, but that lower costs and broader availability may make people seek services and work they could not previously afford, access, or sustain. Those possibilities complicate the assumption that automating existing tasks must simply reduce the total amount of work.

Appleby’s account of Britain illustrates how demand and productivity can reinforce each other. Before the Industrial Revolution, expanding trade helped ordinary Britons develop appetites for goods including tobacco, tea, sugar, and Indian cottons. Annual working days rose from roughly 250 to 300, with a significant portion of the increase voluntary as people sought income to buy goods they wanted. Higher productivity supported more consumption, while that consumption encouraged further productivity gains. The history does not prove that AI will produce the same pattern, but it offers an alternative to treating demand as fixed.

Rao argues that cheaper intelligence could lead people to find more uses for it and pursue projects or curiosities that were previously out of reach. If AI raises the returns on attention, people may have reason to think harder about problems they previously lacked the ability or incentive to tackle. He describes this as a possible “cognitive industrious revolution,” analogous to the increase in British working days. Ben Katesh suggests that AI’s largest consequence may be freeing people from producing intelligence merely to survive, rather than only making intelligence production more efficient. He sketches a possible shift from 80% of human brains engaged in subsistence cognition to 95% meeting previously unfulfilled cognitive demands; the figures express a scenario, not a prediction.

Rao distinguishes cognitive mercantilists, who fear losing scarce intelligence and the expertise, authority, status, or geopolitical leverage attached to it, from cognitive Malthusians, who fear intelligence will exceed people’s capacity to absorb it and make them economically redundant. Both, he argues, assume limited demand. The host’s “infinite backlog” idea sharpens the anxiety: AI agents could replicate work in parallel and continuously, making the backlog appear unlockable and unsettling familiar ideas of what counts as done. The resulting labor politics may concern not just preserving current cognitive jobs, but ensuring fair access to intelligence, the ability to direct it, and the freedom and rights to pursue projects with it. The discussion’s optimism about expanded exploration does not dismiss the costs of AI taking on work previously done by human brains; it separates addressing those costs from trying to stop the transition.