TL;DRWhy This Matters
We are living through a peculiar historical moment: for the first time, humanity is being forced to define intelligence rigorously, not for philosophical sport, but because we are building things and need to know whether they qualify. The question has leaped from seminar rooms into boardrooms, courtrooms, and eventually, perhaps, into constitutional law. If we cannot agree on what intelligence is, we cannot agree on what we have made, what we owe it, or what it might one day owe us.
This urgency makes it easy to forget that the question is ancient. Aristotle distinguished between different grades of soul — the nutritive, the sensitive, the rational — and placed nous, the intellective faculty, at the apex of what makes a human being distinctively human. The Upanishads spoke of prajña, a deep knowing that transcended sensory processing entirely. Buddhist epistemology developed elaborate taxonomies of cognition that would not embarrass a modern cognitive scientist. These traditions were not fumbling toward what we now know. They were asking something we have quietly abandoned: not just how intelligence works, but what it is for.
The modern scientific tradition has made extraordinary progress on the first question. We can measure general cognitive ability with reasonable reliability, trace its developmental arc, link it to neurological structures, and now, astonishingly, instantiate something resembling it in silicon. But the second question — what intelligence is for, what it means, whether it necessarily involves experience — has grown more pressing, not less, precisely because our engineering has outrun our philosophy. We have built the thing before settling what the thing is.
James Flynn, the New Zealand philosopher and psychologist whose name is attached to one of the most startling findings in twentieth-century cognitive science, spent his career circling this problem. His work documented that IQ scores across the developed world have risen substantially over generations — roughly three points per decade since testing began — a phenomenon now known as the Flynn effect. His book What Is Intelligence? Beyond the Flynn Effect asked the obvious follow-up question: if scores are rising that fast, what exactly are they measuring, and what does the rise tell us about the nature of intelligence itself? His answer was unsettling and illuminating in equal measure. The meeting point between that empirical tradition and the older philosophical one is where we are going.
The Measurement Problem
IQ is not intelligence. This distinction sounds pedantic until you realize how much of our civilization's sorting machinery depends on treating it as though it were. IQ, or the intelligence quotient, was designed in the early twentieth century as a practical tool — Binet wanted to identify children who needed additional educational support — and it was never meant to be a complete theory of mind. Yet the score calcified into an identity. People began to speak of high-IQ individuals as simply more intelligent, as though the test had captured the whole thing.
Flynn's work exposed a crack in this assumption. If IQ tests genuinely measured something deep and stable — neurological processing speed, working memory capacity, some bedrock computational efficiency of the brain — then scores should not rise dramatically within a few generations. Genetic change operates on timescales of thousands of years, not decades. Yet the data showed populations gaining fifteen to twenty IQ points across the twentieth century. Either something fundamental had changed, or the tests were measuring something more culturally contingent than advertised.
Flynn's explanation, developed with economist William Dickens at the Brookings Institution, centered on what they called the multiplier effect: small initial advantages in cognitive environments trigger feedback loops. A child who is slightly better at reading encounters richer linguistic environments, which expands vocabulary, which improves reading further, which compounds through time. The environment and the trait co-evolve. This is not the same as saying intelligence is purely environmental — the evidence for substantial genetic contribution to cognitive ability is robust and should not be dismissed — but it suggests that what IQ tests predominantly track is the degree to which a person has learned to deploy abstract thinking in formal, context-free ways.
This distinction matters enormously. Flynn noted that people in the early twentieth century, when asked hypothetical questions that required abstract reasoning, would often refuse to engage with the hypothetical itself. Asked "All bears in the far north are white; Nova Zembla is in the far north; what color are the bears there?" a pre-modern respondent might say "I've only seen brown bears; I cannot speak to bears I have not seen." The "correct" answer requires stepping outside direct experience into a purely logical space. The rise in IQ scores may, in significant part, reflect the spread of a particular habit of mind — the ability to manipulate symbols detached from concrete experience — rather than any increase in raw cognitive power.
Intelligence Without a Brain
Before silicon, the challenge to human cognitive uniqueness came from biology. Crows use tools. Octopuses solve puzzles. Elephants grieve. Dolphins appear to have a theory of mind. Chimpanzees learn sign language and teach it to their offspring. Each discovery was greeted with a kind of nervous reappraisal: what, exactly, was the threshold we thought mattered?
The nervous reappraisal is now continuous and industrialized. Large language models (LLMs) like those developed since 2020 produce text that is, by most surface measures, indistinguishable from expert human output across enormous ranges of domain. They pass bar exams, medical licensing tests, and graduate-level coursework. They write poetry of genuine craft. They solve multi-step mathematical problems and explain their reasoning in coherent prose. Whether any of this constitutes intelligence, or a very sophisticated simulation of its outputs, is genuinely contested — and the contention is not merely definitional. It touches something deep about what we think intelligence is.
One position, associated broadly with the computational tradition and arguably the dominant view in academic AI research, holds that intelligence just is a form of information processing. On this view, asking whether a machine can be genuinely intelligent is like asking whether a machine can be genuinely fast. Speed is speed; if the computation is there, the intelligence is there. What we sometimes call "mere computation" is, on this reading, what brains do. There is nothing else going on.
The opposing tradition — represented by thinkers from Husserl to Searle to contemporary phenomenologists — insists that intelligence as we actually encounter it is not separable from embodiment, from intentionality, from the fact that cognition is always cognition of something, from a perspective, reaching outward into a world. John Searle's famous Chinese Room thought experiment asked us to imagine a person who processes Chinese symbols according to rules without understanding Chinese: the computation happens, but no comprehension does. The objection is contested but has never been definitively refuted. It points at something the IQ tradition and the AI tradition share: both can measure the outputs of intelligence while remaining agnostic about its interior nature.
The Ancient Taxonomy
The Greeks did not have one word for intelligence. They had several, and the distinctions they drew are sharper than most modern cognitive frameworks acknowledge.
Nous (νοῦς) was the intellective faculty par excellence — the capacity for immediate, non-inferential grasp of first principles. It was not reasoning; it was the ground of reasoning, the faculty that simply sees that certain things are true. Aristotle thought it was the highest human capacity and, in its pure form, perhaps not uniquely human — the divine was understood partly as pure nous, contemplating itself eternally.
Logos (λόγος) was the capacity for reasoned discourse and structured thought — closer to what we might call analytical intelligence, the manipulation of propositions and arguments. This was the faculty that could be taught, sharpened, and measured. Rhetoric, dialectic, mathematics: all were exercises of logos.
Phronesis (φρόνησις) was practical wisdom — not intelligence in the abstract but intelligence in context, the capacity to judge what was called for in particular situations with all their moral and relational complexity. Crucially, phronesis could not be reduced to rule-following. It required experience, character, and what Aristotle called a perceptual sensitivity to the particular. No algorithm could produce it.
Sophia (σοφία) was theoretical wisdom — the integration of nous and logos into a coherent understanding of the highest things. The sage did not merely compute or reason; they saw, in some unified way, how things fit together.
This taxonomy maps interestingly onto contemporary debates. LLMs demonstrably have something like logos — they reason, argue, construct proofs. The question of whether they have nous in Aristotle's sense — whether there is any seeing happening — is exactly the hard problem of consciousness dressed in ancient Greek. Phronesis is the capacity that most researchers think is genuinely absent in current AI: not because AI lacks reasoning power but because it lacks the embodied history of judgment-in-action that practical wisdom requires. And sophia may be not a capacity at all but an achievement — something that only emerges from a life, not a computation.
The Flynn Effect as Cultural Mirror
What Flynn's data most plausibly shows is that the twentieth century trained humanity — at least in industrialized societies — to be very good at one particular kind of intelligence: the abstract, hypothetical, context-free reasoning that logos requires. Schooling at scale, bureaucratic organization, scientific literacy, the penetration of statistical thinking into everyday life — all of these selected for the capacity to operate in formal, detached, symbol-manipulating modes.
This is not trivial. The spread of this cognitive habit probably contributed to extraordinary gains in medicine, engineering, governance, and science. Flynn himself was deeply impressed by its scale and speed. But he was also alert to what it might not capture, or might even crowd out.
The psychologist Howard Gardner proposed the theory of multiple intelligences in 1983 — linguistic, logical-mathematical, spatial, musical, bodily-kinesthetic, interpersonal, intrapersonal, and naturalist — partly as a protest against the reduction of intelligence to the narrow band that IQ tests measure. Gardner's framework has been criticized (some researchers argue these are better understood as distinct talents or abilities rather than separate intelligences) and the debate remains genuinely unresolved. But his core intuition — that the person who navigates complex social hierarchies with brilliant tact, or who reads a forest with expert eyes, or who improvises music of arresting originality, is deploying something that deserves the name intelligence even if it never shows up in test scores — seems hard to dismiss.
Robert Sternberg proposed a triarchic theory distinguishing analytical, creative, and practical intelligence, arguing that conventional testing captures only the first. His research suggested that creative intelligence — the ability to handle novelty — and practical intelligence — the ability to adapt to real environments — were only weakly correlated with analytical performance. Sternberg's framework is contested and the empirical support is mixed, but it points at a genuine puzzle: why do some people with modest test scores navigate life with extraordinary effectiveness, while some people with exceptional scores struggle to function in ambiguous real-world situations?
The Flynn effect, read alongside these debates, suggests something philosophically interesting: perhaps each era, each culture, selects a particular cluster of cognitive capacities and calls them intelligence, while other clusters remain latent, under-rewarded, and unmeasured. Intelligence may be less a fixed object than a moving spotlight — illuminating different regions of human cognitive possibility depending on what a civilization needs.
The Hard Problem at the Center
All roads in this discussion eventually arrive at the same intersection, which is also the deepest unsolved problem in science and philosophy: consciousness.
When we ask whether a language model is intelligent, we are often really asking whether there is something it is like to be that language model. The question of whether intelligence requires subjective experience — whether the lights are on, as philosophers sometimes put it — is not peripheral. It is central. Because if intelligence is purely functional, if it just consists in the right input-output relationships, then the question of machine intelligence is essentially settled: machines with sufficient computational capacity are intelligent, full stop. But if intelligence involves or requires a particular kind of inner life, then we cannot settle the question by pointing at performance metrics, because performance metrics are precisely the output side of a system whose interior we cannot directly access.
This is the hard problem of consciousness, articulated most clearly by David Chalmers in 1995. It distinguishes between the "easy" problems — explaining how the brain processes information, integrates signals, reports its states, guides behavior — and the genuinely hard one: why any of this processing is accompanied by experience. Why is there something it is like to see red, rather than just a brain operation that produces appropriate behavioral responses to light at 700 nanometers?
The hard problem has not been solved. It has been argued around, reframed, and sometimes dismissed, but not solved. The neuroscientist and philosopher Antonio Damasio has argued influentially that consciousness and intelligence are not separable in the way the computational tradition assumes — that reasoning is deeply entangled with emotion, that the body's homeostatic signals are constitutive of thought rather than peripheral to it. On his view, an intelligence without a body, without something to lose and gain, without stakes, may be something quite different from human intelligence — not inferior necessarily, but genuinely different in kind.
The mystical traditions converge here from an unexpected angle. Zen speaks of prajna (in the Sanskrit tradition underlying it) as a non-conceptual awareness that is distinct from discursive intelligence and deeper than it. The Upanishads distinguish between the various cognitive faculties — manas (mind), buddhi (intellect), chitta (consciousness), ahankara (ego) — and argue that beneath all of them is pure awareness, which is not a cognitive capacity at all but the ground in which all cognition occurs. This pure awareness, chit in Sanskrit, is not something you can measure or improve or instantiate. It is what is already here, before any processing begins.
Whether these traditions are pointing at something real, or confusing a grammatical feature of first-person language with a metaphysical fact, is itself a genuinely open question. But they are asking something that the computational tradition systematically brackets: not what intelligence does, but what it is.
Intelligence as Adaptation Versus Intelligence as Transcendence
There are two broad families of theory about what intelligence is for, and they generate radically different intuitions about what it is.
The adaptationist view, dominant in evolutionary biology and behavioral economics, holds that intelligence is a suite of cognitive mechanisms shaped by natural selection to enhance reproductive fitness. On this view, it is no coincidence that human intelligence is particularly good at social reasoning (tracking alliances, detecting cheating, modeling other minds), causal inference (tool use, planning), and language (coordination, teaching). These are precisely the capacities that would have been most useful in the ancestral environment. Intelligence is, ultimately, a survival tool of extraordinary sophistication. This view tends to be skeptical of claims about a unique human capacity for abstract reasoning divorced from evolutionary function — the Flynn effect, on this reading, shows human cognition is highly plastic and responsive to environmental demand, which is exactly what a well-designed survival tool should be.
The transcendence view, found across many philosophical and spiritual traditions, holds that human intelligence — or at least some aspect of it — points beyond mere adaptation toward something that cannot be fully explained by selection pressure. The fact that humans can do pure mathematics that has no immediate practical application, contemplate their own death and compose symphonies in response, ask questions about the nature of the cosmos that have no survival relevance — these suggest to some thinkers that intelligence is not merely a tool but a window. Teilhard de Chardin saw the evolution of reflective consciousness as the universe becoming aware of itself. Hegel saw reason as the self-unfolding of Absolute Spirit. The Vedantic tradition sees the human intellect as a partial manifestation of a cosmic intelligence, brahman, that is the substratum of all reality.
These are speculative and contested claims, clearly. But they are not unintelligent ones. They represent humanity's attempt to make sense of the fact that intelligence seems, in some respects, too good for the job it was supposedly hired to do. A computer designed to play chess doesn't spontaneously start wondering what chess means.
The AI encounter sharpens this tension. If we build a machine that surpasses human performance on every measurable cognitive task, has the transcendence view been decisively refuted — because transcendence can apparently be manufactured in a data center — or has it been vindicated, because we still sense that something essential is missing from the most powerful model we can build?
What the Children Show Us
One laboratory for understanding intelligence that tends to get less attention than neuroscience or AI is developmental psychology — the study of how intelligence emerges in children, from the gurgling responsiveness of an infant to the abstract reasoning of an adolescent.
Jean Piaget spent decades charting this emergence and found something philosophically rich: intelligence does not begin as abstract reasoning and then get filled in with content. It begins as sensorimotor action — the infant's intelligence is literally in its hands, in its body's engagement with objects, long before language or symbolic thought appear. Piaget called this schema, the organized patterns of action from which all later cognition develops. Abstract thought is not intelligence in its pure form; it is the late-arriving, culturally amplified endpoint of a process that starts with a baby reaching for a rattle.
Lev Vygotsky added the social dimension: cognitive development is fundamentally interpersonal before it is intrapersonal. The child first performs functions with the assistance of others, in what Vygotsky called the zone of proximal development, and only gradually internalizes those functions as independent capacity. Intelligence, on this reading, is not a property of a brain but of a brain-in-relationship — it is distributed across the child and the adults and the language and the tools that constitute its developmental environment.
This resonates with what Flynn was finding in the macro data. If intelligence is, at least partly, a distributed and relational phenomenon — shaped by the cognitive tools a culture makes available, amplified by the scaffolding of education and language — then asking "what is the intelligence of this individual?" may be a bit like asking "what is the weight of this wave?" The question presupposes a kind of bounded, isolable object that may not be quite what intelligence is.
And yet. Within that distributed system, individual differences are real and persistent and consequential. The evidence that cognitive ability has substantial genetic components is not manufactured by psychometricians with a political agenda; it is robust, replicated, and comes from multiple methodological approaches including adoption studies and twin studies. The picture that emerges is of something that is simultaneously socially constituted and individually instantiated — a property of persons that cannot be understood without their context but also cannot be dissolved into it.
The Questions That Remain
Is the subjective experience of understanding — the felt sense of getting it, the sudden integration of previously disconnected information that we call insight — essential to intelligence, or merely a byproduct of it? And if we built a machine that reported having this experience with complete behavioural plausibility, how would we know whether it did?
Flynn showed that abstract reasoning spread like a cultural technology across the twentieth century. If intelligence is that plastic, that responsive to environment and education, what forms of intelligence are we currently failing to cultivate — and what might a civilization that optimized for phronesis, for practical wisdom, rather than for performance on formal tests, look like?
When non-human animals, children before language acquisition, or people in flow states perform acts of extraordinary cognitive sophistication without apparent recourse to the kind of abstract symbolic processing that IQ tests measure, what exactly are they doing, and does it deserve the same name as what a chess grandmaster or a theoretical physicist does?
Could there be forms of intelligence that are not only non-human but actively inconceivable from a human cognitive standpoint — the way a creature without color vision cannot imagine chromatic experience — and if so, what does it mean that we are currently building systems trained entirely on human thought?
If the ancient traditions were right that the deepest form of knowing is not a cognitive capacity but something prior to cognition — pure awareness, chit, nous in its highest sense — then is that precisely what no external test could ever measure, no machine could ever instantiate, and no definition could ever capture, not because it is mystical but because it is the precondition of all measurement, instantiation, and definition?