Terrence Deacon

In the AI Story

Born in Boston in 1950, Deacon trained in neuroscience and biological anthropology at Harvard, conducting comparative neuroanatomical research on brain evolution and language. His empirical work focused on the disproportionate enlargements of specific brain regions in humans—the prefrontal cortex, perisylvian areas including Broca's and Wernicke's areas, and regions supporting vocal-motor control—none of which follow the expected scaling relationships of a simply enlarged ape brain. These targeted reorganizations correspond with remarkable precision to the computational demands of symbolic language: enhanced working memory to hold syntactic structures, refined prefrontal inhibition to suppress immediate indexical responses, precise vocal articulation to produce phonemic distinctions. The question this neuroanatomy posed—why would evolution produce these structures in advance of language?—drove Deacon to his central inversion.

The co-evolutionary thesis Deacon developed reverses the causal arrow: language did not follow the brain; language and the brain built each other through reciprocal selection pressure operating across hundreds of thousands of years. Proto-linguistic communication created a selection advantage for brains slightly better suited to symbolic processing; those brains enabled slightly more complex communication; the spiral fed itself across thousands of generations. The mechanism is the Baldwin Effect applied to symbolic cognition—learned behaviors becoming progressively more innate as the neural substrate reorganizes to support them more efficiently. This framework inverted not only the standard account of human origins but the entire relationship between tools and their users: the tool reached back into the toolmaker and restructured the toolmaker's architecture.

Deacon's 2012 Incomplete Nature broadened this co-evolutionary insight into a comprehensive theory of emergence, introducing the concepts of thermodynamic, morphodynamic, and teleodynamic processes as three levels of organization, each introducing constraints absent from the level below. Teleodynamic processes—self-maintaining, boundary-forming, purposively oriented systems—exhibit 'absential' properties: function, purpose, meaning defined by orientation toward what is not present. A cell is oriented toward its continuation; a conscious mind is oriented toward purposes it can examine and question. This hierarchy provided Deacon with the analytical apparatus for distinguishing genuine emergent properties from mere complexity, and for locating consciousness precisely within the dynamical architecture that produces it.

In recent work engaging directly with artificial intelligence, Deacon has argued that large language models operate at the symbolic surface without the indexical grounding that makes symbols genuinely meaningful. A 2025 paper co-authored with Parham Pourdavood and Michael Jacob reframed LLMs as cultural DNA—compressed repositories of statistical regularities requiring interaction with purposive human cognition to produce genuine meaning. His 2023 NYU talk 'In the Shadow of Descartes' diagnosed two oversimplifications obstructing AI understanding: collapsing information into signal properties and reducing symbols to conventional tokens. His 2025 conversations with theologian Ilia Delio and AI safety researcher Stuart Russell extended the framework into the contemporary moment, arguing that AI represents a natural extension of millennia-long cognitive offloading—but one operating at the unprecedented level of symbolic processing itself.

Origin

Deacon's intellectual formation combined rigorous neuroscience training with a philosophical sensibility rare in empirical brain science. At Harvard in the 1970s and 1980s, he studied under the generation of neuroscientists mapping brain evolution through comparative neuroanatomy, learning the techniques that would allow him to identify, with empirical precision, the specific reorganizations that distinguish the human brain from the brains of other primates. Simultaneously, he absorbed the semiotic philosophy of Charles Sanders Peirce—the American pragmatist whose three-part classification of signs (icon, index, symbol) would become the conceptual backbone of Deacon's account of human cognitive uniqueness.

The synthesis of these two traditions—rigorous neuroscience and rigorous semiotics—was unusual enough to make Deacon's work difficult to categorize and powerful enough to transform the study of language origins. The Symbolic Species appeared in 1997 to a reception that mixed enthusiasm and resistance: linguists appreciated the inversion but questioned the neurological details; neuroscientists appreciated the comparative anatomy but questioned the semiotic framework. The book's enduring influence came from its refusal to simplify either side—insisting that the neurological evidence and the semiotic analysis were equally indispensable, and that only their integration could explain the co-evolutionary process that produced the human mind.

Key Ideas

Co-evolution of language and the brain. Language did not emerge from a pre-existing brain capacity; language and the brain shaped each other across hundreds of thousands of years, each providing selection pressure for the other.

Semiotic hierarchy as cognitive architecture. Peirce's three levels—iconic (resemblance), indexical (correlation), symbolic (convention)—describe not merely sign types but the layered foundations of meaning, each dependent on the level below.

Absential properties as the signature of mind. The most important features of life and consciousness—function, purpose, meaning—are constituted by orientation toward what is not present, what is missing, what matters.

Thermodynamic-morphodynamic-teleodynamic hierarchy. Three levels of emergent dynamics, each introducing new constraints: thermodynamic dissipation produces morphodynamic regularity produces teleodynamic self-maintenance and purpose.

LLMs as cultural DNA. Large language models function as externalized informational substrates analogous to DNA—requiring interaction with grounded, purposive human cognition to produce genuine meaning, not intelligent agents themselves.