Artificial Intelligence is Artificial Thinking — Krezl

Image: Adobe Stock

Artificial Intelligence is a misnomer. The honest name is Artificial Thinking.

Intelligence is preverbal. Thinking is representational. The seeing comes first as a whole, not in pieces. Language is thinking made portable: the words arrive after, to carry what was already seen.

Thinking lifts. We meet the jaguar in the grass, and from that encounter lift a predator, a recurring danger, something that may return. The organism can act on the pattern before the pattern comes back.

Once something is lifted, it is no longer bound to the encounter that produced it. We can divide it, join it, compare it, rearrange it, carry it through time, or imagine what has never existed. We split the lifted day into hours that exist nowhere in the day. We split the self into watcher and watched. Past from future, inside from outside, mind from body. None of these seams are in the thing. They are in the copy, and we live as if they were in the thing.

The power of this is immense. By lifting what repeats in the sky we found the season, and planted against it, and lived. We split objects beyond what the eye can see and found atoms. Then we split the atoms and released the enormous energy bound within them. Prediction begins here: in finding what repeats, what relates, and what follows from what.

The root of it is survival. The predator came for our young as it comes for the young of the giraffe and the elephant, for every animal facing its first winter. The animal that could lift the recurring danger from the single event was the animal whose child saw a second winter. Thinking emerges from survival because survival favors the ability to carry what happened once into what may happen again.

We taught the machine to work on these representations: ideas, categories, relations, rules, everything the lift leaves behind. It finds patterns among things already lifted from the world, at a scale no person can approach. Its material is not the world itself, but the record we have made of it.

A representation can outlive the intelligence that made it. A person sees something, lifts it into thought, and leaves the thought behind in language. Another person can inherit the thought without inheriting the seeing. The record accumulates these inherited representations, and the machine works across that accumulation.

So the name is wrong. We called it artificial intelligence. It is artificial thinking, the seeing left out: an operation until now performed by biological organisms. This does not diminish it. If this immense thing is only thinking, then thinking is no small thing.

Its governing move is center-finding, but not finding an average. Given the right question, it can find relations across the record that no one explicitly stated, and surface information that remained latent within it. The center is constrained by what it has been given, but it can be a figure no person in the record ever expressed. It does exactly what the material permits.

That material is the record. It is gargantuan, and mostly restatement: the kernel a thin seam in a vast fill. A centering instrument inherits the distribution it is given. If repetition dominates the record, greater fidelity to the record does not recover the seeing. It recovers the repetition more faithfully. More data may widen the field, but volume alone does not pull the center toward the kernel. Where the common copy overwhelms the rare seeing, scale can deepen the mistake.

There is an old word for the middle height. Mediocre: medius, the middle, and ocris, the mountain, the one who climbs halfway up. It is not an insult, and it does not describe the machine. It describes a pairing. The machine climbs whatever it is given, all the way to the top. Given the bloated record, the structure available to be recovered is the structure of restatement, and the top of that climb is the middle height. Given a well-put question and material that was seen, the same climb reaches genuinely clever ground, arrangements the record never contained. The mountain sets the height. The machine only climbs.

Then it generates, without fatigue and without end. It can compress the stack, restate it shorter, surface what was implied in it and never said. What it cannot do is discard the stack and begin again from the thing. The human record at least had the occasional accident of someone seeing afresh and collapsing the whole accumulation back to one line. That collapse was not compression. It was replacement, and replacement requires contact. The machine has no contact, so it cannot replace. It removes even that accident.

Grant the machine a flawless operation. The errors live elsewhere. Feed it restatement and it returns the center of restatement. Find the center among seeings that point different ways and it may return a place none of them was going. And some things have no center at all. Found flawlessly, the center is still false....