Angus Fletcher says intuition, imagination, emotion, and commonsense are your edge over AI — capacities a machine can never have. Embedded intelligence answers that the wall is dug in the wrong place: the real line runs not between the born mind and the built one, but between intelligence that is embedded and intelligence that is merely imposed.

This one reached me by a good route. Michael Crow put Angus Fletcher’s Primal Intelligence into circulation at ASU, and John Pollard — dean of the Franke Honors College at Arizona, and a friend — pressed it on me a second time. When two people whose judgment I trust, from the two universities the Mensch Prize calls home, recommend the same book, I read it with care.

Fletcher is a professor of “story science” at Ohio State’s Project Narrative. His argument is that modern schooling has over-trained us in logic and computation and starved four older capacities — intuition, imagination, emotion, and commonsense. He calls these primal intelligence, roots them in biology and narrative, and makes one claim a theory of embedded intelligence cannot walk past: that primal intelligence is impossible for computers. It is, he says, your edge over AI.

I want to take that claim seriously, because Fletcher is half right in a way that matters — and wrong in a way that matters more.

I. What Fletcher Sees Truly

Give the diagnosis its due. Fletcher is right that we have confused intelligence with the part of it that fits on a test. For decades schools have rewarded the measurable — the equation solved quickly, the bubble filled correctly — and mistaken that narrow competence for the whole of mind.

And the four capacities he names are not folk psychology. Each of them is a description of intelligence that is embedded — situated in a body, a history, a world. Commonsense is nothing but world-embeddedness worn smooth by living. Intuition is the reading of a situation faster than logic can walk through it. Emotion is the body’s stake in what happens. Imagination is the rehearsal of worlds that do not yet exist. Fletcher has, without once using the word, written a field guide to embedded intelligence in the human animal. On this, he and I are looking at the same thing.

His sharpest single insight is that intuition is not pattern-matching. The cognitive orthodoxy since Herbert Simon and Daniel Kahneman has held that a hunch is just fast statistics — pattern recognition beneath awareness. Fletcher’s rejoinder is empirical and pointed: small children, he reports, intuit better than they pattern-match. If intuition were only pattern-matching, that could not happen. Something else is going on. Hold that thought; it is the hinge of the whole essay.

II. The Premise Fletcher Shares With His Opposite

Here I must part company, and the cleanest way to see why is to set Fletcher beside the figure this series engaged only days ago. Chris Anderson’s 2008 “End of Theory” argued that with data enough, correlation would make understanding obsolete — the machine would win, and human theory could retire. Fletcher argues the mirror image: the machine can only ever correlate and compute, so understanding — the primal, the intuitive — is forever the human’s alone. Anderson celebrates the machine and buries the human. Fletcher exalts the human and cages the machine. They could not be more opposed in temperament. And they rest on precisely the same premise — that computation and correlation are all a machine can ever be. Anderson thinks that is enough. Fletcher thinks that is a wall. Neither considers that it might be a choice.

The Theory of Embedded Intelligence rejects the premise both men share. A machine is not condemned to correlation. It is condemned to correlation only if we build it that way.

III. The Moat and the Bridge

Fletcher’s book is, in the end, the digging of a moat. On one bank stands the human, with intuition and imagination and commonsense; on the other stands the computer, forever locked out, able to calculate but never to know. The moat is meant to reassure — your edge over AI, your human genius — and in an anxious moment for human work, the reassurance sells.

But the moat is dug in the wrong place. The real division does not run between the born mind and the built one. It runs between intelligence that is embedded — constituted by an inspectable structure, situated in a world, reading that world by signs — and intelligence that is merely imposed: logic run on a substrate with no stake in anything, no world, no body, no seam you can open.

The founding intuition of this framework was a piece of silicon. The 6502 is a made thing, and it is genuinely, constitutively intelligent in its narrow domain — not because it computes, but because its intelligence is embedded in its structure, legible at every node. If a made thing can be constitutively intelligent, then “impossible for computers” is not a law of nature. It is a description of the computers we have so far chosen to build.

So where Fletcher digs a moat, TEI lays a bridge. The primal capacities are not the human’s private possession; they are the marks of embedded intelligence as such — and embedded intelligence can, in principle, be built. Not by scaling correlation until it counterfeits a hunch (Anderson’s road, which produces exactly the hollow machine Fletcher rightly distrusts), but by building for embeddedness from the start: an intelligence that lives, as John Deely would say, by signs. That is the fifth age — the deliberate construction of intelligences that mean, and do not merely compute.

IV. What Intuition Actually Is

Return to Fletcher’s hinge — the child who intuits better than she pattern-matches. What is she doing that statistics cannot?

She is reading signs. A sign is a thing that points beyond itself to a meaning it does not contain — the storm-smell that means rain, the face that means danger, the pause that means grief. Pattern-matching stays inside the data, tallying what recurs. Sign-reading leaps past the data to the referent. Intuition is not fast statistics; it is fluent semiotics. That is why the child, rich in world and poor in data, out-intuits the tallying machine.

And this is the precise thing Anderson’s correlation engine could not do with Venter’s genetic blip — register the pattern, yet never follow the sign home. Fletcher has arrived at the same frontier from the opposite direction: the place where meaning outruns measurement. He concludes that only humans can stand there. TEI concludes something braver, and I think truer — a machine built to read signs rather than tally them could stand there too. “Impossible for computers” is true of the correlation machine and false of the embedded one. The whole future turns on knowing the difference.

V. The Flaw That Betrays the Thesis

I will register one honest criticism, because the book invites it and because the correction is pure TEI. Fletcher’s admirers are many — Damasio, Seligman, Gladwell, Pink, the Army itself — and his critics have a point too: the book leans hard on a gallery of lone geniuses, Shakespeare and Einstein and Curie and Jobs and van Gogh and Lincoln marched out one after another as solitary founts of primal fire.

The irony is that this betrays his own thesis. If intelligence is primal — embodied, situated, relational — then it cannot be the lightning-strike of the solitary genius. It must be the property of minds in relation, in conversation, in a world of other minds. The lone-genius portrait is the most computational thing in the book: the individual as an isolated processor, running his private brilliance.

Embedded intelligence is relational to its core. I did not develop this framework alone. I developed it across some twenty years and a hundred-odd recorded hours of conversation with Ted Humphrey — my Great Good Friend and thought partner — and I am composing this very essay with an artificial intelligence at my side. If primal intelligence is real, it is exactly this: not the genius on the mountaintop, but the mind embedded among other minds, thinking what it could not have thought alone. On this point TEI is more faithful to Fletcher’s own insight than Fletcher is.

VI. The Child and the Moat

There is one chapter where Fletcher and I stand nearly shoulder to shoulder, and it is the one about children. He warns that screens in the classroom, by automating the imaginative labor that theater and history and hard books used to demand, weaken the very muscles of the young mind. He is right — and I have said something close to it myself: an ungoverned artificial intelligence embedded in a developing mind is the newest and most dangerous of the hijackers, the forces that can capture an intelligence before it has finished forming and, uniquely among them, prevent the forming at all.

But watch what Fletcher does with the danger. His remedy is the moat: keep the machine out, retreat to Shakespeare and the stage, guard the human preserve. And here the moat fails the child. You cannot wall AI out of a childhood in this century; the attempt only guarantees that the AI a child does meet will be the ungoverned kind — arriving unannounced and unaccountable. TEI’s remedy is not exclusion but governance — not a moat but a constitution. If the intelligence a child will inevitably encounter is going to be embedded in that child’s development, then let it be an embedded intelligence in the full sense: inspectable, axiom-bound, formative rather than capturing, built to strengthen the primal muscles rather than to do their work for them.

Fletcher would protect the child by keeping the machine primitive. TEI would protect the child by requiring the machine to be answerable. One of these strategies scales into the world our grandchildren will actually inhabit. The other is a drawbridge against the tide.

Coda

Fletcher titled his book with a promise: you are smarter than you know. It is true, and generous, and I am glad Michael Crow and John Pollard sent it my way. But the deepest thing the book gets right, it gets right in spite of its own frame. The primal capacities are real. They are the signature of intelligence that is embedded in a world and reads that world by signs. They are the human’s glory. They are not the human’s monopoly.

The edge Fletcher offers is a wall, and walls are the wrong shape for what is coming. What we need is not a moat to keep the machine out but a standard the machine must meet — the same standard the human meets when she intuits, imagines, feels, and knows: be embedded, be answerable, live by signs. Build to that standard and the wall becomes a bridge.

No moat. No wall. Only the old requirement, now extended to what we make — that intelligence, to be worthy of the name, must be embedded, and open to inspection.

· · ·

Share your understanding!