Evan Thompson

In the [YOU] on AI Field Guide

The cycle that began with [YOU] on AI asks what it would mean to see the machine clearly—without the narcotic of hype or the paralysis of fear. Thompson is the clearest diagnostic instrument the cycle possesses for answering precisely what kind of thing these machines are. When the author sits at his desk late at night working with Claude on a passage that will not come together, Thompson’s framework reveals the asymmetry the experience conceals: one partner is enacting a world of significance—feels the fatigue, carries the biographical weight of why this book matters, finds the idea arriving in the body before it is understood in the mind—and the other is processing tokens. Both contributions are real. They are not the same kind of activity.

Thompson’s lens reframes the cycle’s central puzzle—the coexistence of extraordinary machine fluency and confident, basic error. A system that misapplies Deleuze in eloquent prose is not making a mistake in the way a human expert makes a mistake. It is generating statistically probable text in the absence of sense-making: the surface is smooth precisely because there is no depth to resist it. The fluency is a consequence of the absence of understanding, not its presence. The cycle treats this decorrelation as the signature hazard of the age; Thompson explains it from first principles.

He also illuminates the “ascending friction” the cycle documents. When AI removes mechanical difficulty from a practitioner’s work, it does not liberate judgment for free; it severs a form of structural coupling between the body and its domain. The programmer who no longer debugs by hand has lost a specific mode of motor intentionality—the felt sense for where a system goes wrong—that no amount of output quality can restore. This is the cycle’s most uncomfortable finding, and Thompson is its most rigorous theorist.

He stands in the cycle’s gallery alongside Byung-Chul Han, who diagnoses the same erosion from a phenomenological direction, and in productive tension with thinkers who believe that sufficiently large models will eventually climb into genuine understanding. Thompson does not merely assert the wall; he identifies the organizational property—autopoiesis—whose absence in AI systems is not a temporary engineering gap but a categorical difference.

Origin

Born in 1962, Thompson is the son of the philosopher William Irwin Thompson and grew up at the intersection of science, contemplative tradition, and philosophy of mind. His encounter with Francisco Varela, the Chilean biologist and neuroscientist, was the intellectual turning point of his career. Varela had already developed, with Humberto Maturana, the concept of autopoiesis—the self-producing organization that distinguishes the living from the non-living—and was looking for a collaborator who could bring phenomenology to bear on it. Thompson brought Merleau-Ponty’s analysis of the body-subject and Husserl’s first-person method, and the synthesis became The Embodied Mind (1991).

The book’s founding claim was that the standard computational theory of mind—which underwrites virtually every claim made about artificial intelligence—misidentifies what cognition is. Cognition is not the processing of representations of a pre-given world. It is the organism’s enactment of a world through its structural history of coupling with its environment. The frog does not compute the fly’s trajectory; the frog-fly system enacts a world of significance in which the fly is food. Thompson spent the following two decades extending this claim into every domain where its implications ran: into neuroscience in Mind in Life (2007), into contemplative practice in Waking, Dreaming, Being (2015), and into the sharpest confrontation with contemporary AI in the 2025 Nature letter.

The neurophenomenology that Thompson developed with Varela represents the most disciplined attempt in cognitive science to hold the first-person and third-person perspectives together without reducing either to the other: trained subjects report their experience with phenomenological precision while their neural activity is recorded, and the two data streams are correlated to reveal aspects of consciousness that neither stream alone would disclose. The method requires a subject capable of genuine first-person report. It is the method that most sharply distinguishes cognition from its computational simulation.

Key Ideas

The enactive approach. Cognition is enactment, not information processing. The organism does not passively receive data from a pre-given world and compute a response. It brings forth a world of significance through its embodied activity. The distinction between processing and enacting is not verbal—it is the difference between a map and the journey that the map describes.

Autopoiesis and the living mind. All living systems are autopoietic: they produce and maintain themselves through their own operations. All autopoietic systems are cognitive in a minimal sense: they make sense of their environment by evaluating it in terms of what supports or threatens their continued existence. Mind is continuous with life, which means a system that is not alive cannot be minded—not because the substrate is wrong but because the organizational form is absent.

Sense-making as the mark of the cognitive. Sense-making is the organism’s creation of a world of significance through its own activity. The bacterium that moves toward a sugar gradient is not computing an optimal path; it is making sense of its environment by evaluating it in terms of its own needs. Claude generates text that represents the products of sense-making without performing any sense-making itself: there is no organism, no need, no stake in the outcome.

The hard problem dissolved. Thompson does not answer Chalmers’s hard problem of consciousness. He dissolves it by refusing the separation that generates it. The neuron firing and the experience of red are not two things requiring a bridge; they are one process—the organism’s enacted engagement with the world—described from two perspectives. The dissolution removes the philosophical foundation on which functionalist arguments for machine consciousness rest.

The processing-enacting gap. The gap between a system that processes and a system that enacts is not a matter of architectural sophistication. It is the difference between a mechanism that produces outputs without any stake in them and an organism whose entire existence is at stake in every act of sense-making. This gap is what Thompson’s Nature letter calls “never”—and the never is structural, not a forecast about engineering timelines.