SO WHAT? GENERATIVE AI AND THE HISTORICAL REDEFINITION OF COGNITIVE COMPETENCE
Generative AI has entered intellectual practice with a quality best described as irresistible. Much of the intellectual response in research and education has been organized around an alarm: that these systems induce a surrender of thinking, an atrophy of cognitive capacity, an accumulation of cognitive debt. This essay argues that the alarm conflates two claims that must be told apart. The first is categorical — that generative AI corrupts a cognitive interior whose integrity consists in operating unaided. Its governing metaphors, debt and atrophy and renunciation, presuppose a prior solvent interior against which the deficit is scored, and no such interior is to be found at any documented moment of the historical record: the Greek transition from orality to literacy shows a comparable cognitive figure lost without loss of the human. The second claim is empirical and concerns the ecology in which competences are formed. It survives the separation, and the essay relocates rather than refutes it. Drawing on Latour’s argument that we have never been modern, and on convergent descriptions of cognition as coupled practice (Hutchins, Stiegler, Malafouris, Barad, Goody, Ong, Yuk Hui), the essay reconstructs cognition as a configuration of brains, bodies, tools, environments, and other people — a configuration whose history has repeatedly redefined what counts as cognitive competence. Recent measures of cognitive loss under generative AI assess competences with the ruler of the configuration those competences are leaving behind. What survives the critique is a systemic question, left open here: whether the emerging configuration sustains the practice it draws upon or depletes it. The burden of showing that it does not falls on those who raise the alarm. Concerns that are political-economic, environmental, labor-related, and about opacity and cultural standardization survive examination and deserve deliberation. The categorical alarm about cognition does not.