The Sexer's Hand

In 1933 the Japanese professors Masui and Hashimoto published a short paper on how to read the sex of a day-old chick. You lift the chick in the left hand, press until the cloaca, the single rear opening a bird uses for everything, turns out, and look for the genital eminence, a pale bead at the lip of the fold. Males usually carry it. Females usually do not. Some males show almost none. The differences run to fractions of a millimeter and they never close into a rule.

What grew around the paper was a school. At the Zen-Nippon Chick Sexing School the course ran two years and had almost no content to teach. A student picked up a chick, looked, said male or female, and heard one word back, yes or no. Then the next chick, and the next, ten thousand times and then a hundred thousand, the eye and the thumb tuning themselves under a verdict whose reasons were never handed over. A finished sexer worked at a thousand birds an hour, one every three seconds, and was right ninety-eight times in a hundred. Ask him how, and he gave the honest answer, that he did not know.

Kevin Kelly has a great essay describing a world where everything is mapped, everything is legible.

Kelly invites us to see the inside of an LLM as a new place, a “latent space,” and to work it the way an artist works a medium. His account is largely right and worth keeping. A frontier model is a feat of compression, the readable internet pressed onto a card you could close in your palm, holding no whole play and no whole face, only the abstracted relations among all plays and all faces at once. It does not store Shakespeare so much as the shape of Shakespeare, not faces but the relations by which facehood can be generated. The model throws away the library and keeps the directions.

Everything that can be named becomes a direction: catness, redness, Africa, gunpowder, truth, completion, resemblance.

A prompt ןs the start of a path. The answer is grown as the model travels, word by word, through the space. This is why Kelly thinks latent space will become a medium. It can prototype by pushing one thing toward another, search white space between known forms, carry analogies across fields no single human has mastered, measure likeness between complex objects, find unnamed patterns, build parallel worlds, and eventually produce personal latent spaces tilted by one life. Everything the model has read lies on a single map, the real and the merely possible on one continuous sheet, and Kelly’s proposal is that invention now means to “prospect in the gaps.”

Three centuries ago Leibniz imagined a characteristica universalis, an alphabet of thought so exact that men who disagreed could say, let us calculate, and settle the matter like a sum. Ferdinand de Saussure, the Swiss founder of modern linguistics, taught that in language “there are only differences without positive terms,” each word fixed by its distance from the others. A generation on, the English linguist J. R. Firth gave the maxim a machine could finally obey: “You shall know a word by the company it keeps.”

Latent space is that sentence run across a hundred thousand chips. Kelly stands at the end of the oldest ambition in Western thought, to make all knowing into the geometry of the already-said.

Michael Polanyi, a chemist who turned to philosophy in the middle of the same century, raises a challenge, though: “we know more than we can tell.” It is the challenge of the sexer who cannot explain his craft.

Gilbert Ryle, in his plainer English way, had already separated knowing that from knowing how. A cyclist does not balance by consulting propositions about balance. A pianist does not find the note by reading a rule inside the wrist. The sexer’s knowing was tacit, and the reason it cannot enter latent space is exact. The model is made of what was written down. The tacit is the part that was never written down, that could not be, that crosses from a master’s silence to a student’s thumb across two years of yes and no. How do you scrape tacit knowledge from the internet?

A map drawn from all the sentences in the world is still a map of sentences, and the deepest human competences might withdraw from sentences.

Martin Heidegger spent his life in this gap.

In Being and Time he held up a hammer: you know it by hammering, when it vanishes into the hand and the nail goes home, a mode he called readiness-to-hand and set before the detached inspection that arrives only later, when the hammer breaks. In titled What Is Called Thinking, he argues that thinking was never calculation over what is already present. Truth itself he traced behind the Latin notion of correctness to the Greek aletheia, unconcealment, the prior event of the world coming into the open at all, so that a true sentence would have ground to stand on.

Kelly makes truth a direction you steer toward. Aletheia is what opens the field in which any direction could be found.

Henri Bergson split the intellect, which lays things side by side on a map, from intuition, which enters the moving thing and knows it from within; the map belongs to the intellect and the walk to intuition. The rabbinic tradition has its own harder language for the same distinction. In Berakhot, Rabbi Yohanan says that

the service of Torah is greater than its study

and proves it from Elisha, who is praised not as the man who learned from Elijah but as the man who poured water over his hands. The point is that some knowledge passes only through nearness, repetition, and apprenticeship.

To “prospect in the gaps” is to find what already waits between the mapped points, and Plato named that motion first. In the Meno, Socrates draws a geometry proof out of an unschooled slave boy by question alone and calls it recollection, anamnesis, the soul remembering what it always held. Latent space is Meno’s boy at planetary scale, everything already inside, discovery collapsed into retrieval.

When the map surprises us, and it does, the surprise is a fresh arrangement of directions it already carried, a corner of the sheet no one had walked to. In Plato’s world and in the concept of latent space, there is no true novelty.

There is a name for meaning kept permanently on call. When Kelly offers to “add more Africa to anything,” Africa becomes what Heidegger calls standing-reserve.

Perhaps in time, AI will take away the sexer’s job, showing 99% accuracy on the basis of its map of latent space.

Yet even then, the mystery of touch will remain. The knowing hand that reads its own map without knowing it.