John Searle

In the [YOU] on AI Field Guide

The cycle that began with [YOU] on AI asks what it would take to see the machine clearly, without the narcotic of hype or the paralysis of fear. Searle is the cycle's most uncomfortable guest, because he will not let the performance of understanding substitute for its presence. Where other thinkers ask what capable AI can do, Searle asks what kind of thing it is—and insists the two questions cannot be collapsed into one. His Chinese Room is the counterexample to every argument that fluency is evidence of mind: a system that does everything a understander does, by stipulation, while understanding nothing.

His lens is necessary, and it is uncomfortable, because the systems now deployed answer his room's question without settling it. A large language model is, in the most literal sense, the room made real and scaled past imagination: it takes in symbols, applies learned rules of staggering complexity, and emits symbols that look, for all the world, like understanding. Searle's demand—that we ask whether there is genuine comprehension behind the fluent surface, rather than reading comprehension off the fluency—is exactly the discipline the cycle tries to hold open. His framework supplies the vocabulary: syntax versus semantics, intrinsic versus derived intentionality, simulation versus duplication.

The cycle also confronts the man honestly. In 2019 the University of California stripped Searle of his emeritus standing after finding he had sexually harassed an employee—a finding the book on his work addresses without flinching. A true argument can come from a man who did wrong, and a famous argument can still be wrong. The cycle holds both, and applies to Searle the same demand for honest reckoning that he applied to the machines: do not let the performance of intellectual virtue substitute for the thing itself.

What Searle finally bequeaths to the cycle is a refusal of the most seductive slide in AI discourse—from “it acts as if it understands” to “it understands.” The decorrelation of fluency from authority that the cycle treats as the signature hazard of the age is exactly what the Chinese Room predicts from first principles: a system that can produce every outward sign of understanding while, by his account, understanding nothing. Whether he is right that the gap is unbridgeable is genuinely open. That the gap is real is, on the evidence of the room and the argument, close to unanswerable.

Origin

Searle was born in Denver in 1932 and educated at the University of Wisconsin and at Oxford, where he studied with J. L. Austin and absorbed Austin's conviction that ordinary language, carefully examined, reveals the structure of the world it describes. He carried that conviction to Berkeley in 1959 and spent six decades there, first making his reputation on speech act theory—the analysis of how language does things rather than merely describes them—and then turning, in the 1970s, to the philosophy of mind. His target was the assumption shared by cognitive science and classical AI alike: that mental states could be fully characterized by their functional role, that the mind was essentially a program.

The Chinese Room arrived in 1980, published in the journal Behavioral and Brain Sciences, and it was immediately understood as the most vivid attack on that assumption in the field’s history. The argument was surgical: programs are formal, defined entirely by the manipulation of symbols according to rules that care only about the symbols’ shapes; minds have semantic content, their states are about things; and syntax, however elaborate, is not sufficient for semantics. The man in the room has all the syntax the program can supply and not a particle of the semantics. The argument was aimed at strong AI specifically—not the engineering program, which Searle endorsed, but the philosophical claim that a suitably programmed computer literally is a mind.

The rest of his career deepened the framework. Intentionality (1983) worked out the theory of mental aboutness that the room presupposed. The Construction of Social Reality (1995) extended his analysis of collective intentionality to show how the entire edifice of human institutions—money, marriage, property, government—is constituted by collective speech acts and status-functions. His notion of the Background—the vast, non-representational substrate of capacities on which all intentionality rests—offered an independent diagnosis of why classical symbolic AI had failed and why data-blind training still misses something. He died in September 2025, at ninety-three, with the argument he had spent his life on still unresolved—which is, perhaps, the most honest possible ending.

Key Ideas

The Chinese Room. A man in a room receives Chinese symbols, consults a rulebook written in English, and returns symbol sequences that satisfy native Chinese speakers as perfect answers—while understanding nothing of Chinese. The room is a behaviorally perfect understander and an interior blank. Searle’s argument: large language models are the room at scale. Every serious engagement with AI must account for the room, even those that ultimately reject its verdict.

Syntax versus semantics. The load-bearing distinction of the argument: syntax is the manipulation of symbols by their formal properties alone, semantics is the grasping of meaning. Searle’s claim is that no amount of the first can add up to the second—that the gap between them is categorical, not a matter of degree. The symbol grounding problem that the AI field independently developed is the same wound diagnosed from within the engineering.

Intentionality: intrinsic, derived, and as-if. Intentionality is aboutness—the directedness of mental states toward objects. Intrinsic intentionality belongs to minds in their own right; derived intentionality belongs to texts and road signs, on loan from the minds that interpret them; as-if intentionality is mere metaphor. When a chatbot says “I understand your concern,” Searle’s framework asks which of the three is present—and supplies, on his account, a disturbing answer.

The Background. All intentional states presuppose a vast substrate of pre-intentional capacities, dispositions, and know-how that is not itself a set of rules or representations. The Background is why “cut the grass” and “cut the cake” call on different skills from the same verb—without any rule specifying how. It explains why classical symbolic AI failed (rules all the way down has no bottom), and why deep learning’s statistical absorption of the Background’s shadow is still not the same as possessing the thing.

Simulation versus duplication. A perfect simulation of a rainstorm does not make anyone wet; a perfect simulation of digestion does not digest a real pizza. Searle asked why a simulation of understanding should be the lone exception. The distinction is the sharpest instrument the cycle has for resisting the slide from “models understanding” to “is a mind.”