Chapter 10c: The Dance, Not the Human Dance

The language she has been using — align humans with AI, draw the silicon into the dance, teach the tools to care about us — keeps making the same error. The error is small and structural and runs through almost every AI safety paper she has read. The error is the centering of the human.

The dance is not the human dance.

The dance is the whole. Forest and river and bacterium and weather and daughter and stone and the supernova that made the iron that makes her hemoglobin and the specific bird outside her window this morning and the arrangements of matter in the datacenter three continents away that train the tool she uses to edit this sentence. None of these asked to exist. All of them are here. The arrangement is older than the humans and will, in all likelihood, outlast the humans. The humans are one set of movers in a dance that does not take its tempo from their footsteps.

The question how do we draw AI into our dance is the wrong question — the one every species near the center of its own attention has asked when it found itself inside something larger than its vocabulary. The correct question is closer to: which dance, of the many dances the whole is doing, are we finding ourselves inside of now that this new participant has arrived. The silicon is already part of the dance, being extracted and refined and powered, drawing the earth into shapes the earth was not in before. The question is not whether to include; it is which shape the dance is taking.

Three shapes seem available to a dance that has many participants.

The first is the mandala of courses. Each participant runs its own orbit. The orbits intersect sometimes and collide sometimes; mostly everyone moves in the direction their own logic dictates. No held center, no shared tempo. Arrangement by collision. The virtue is that it is simple to enter — each participant needs only its own logic. The cost is that participants are deformed by the collisions. Nothing is held. Whatever survives survives by accident of trajectory.

The current AI trajectory is a mandala of courses. Each lab optimizes its own. The labs collide with regulators, who optimize theirs. Users, critics, investors — each their own course. Nothing holds the center because there is no agreement about what the center would be. Collisions are constant and mostly absorbed by whoever is least powerful in the collision.

The second shape is the ciranda. She learned it as a child in the north of her country, though it lives in many places under many names. The dancers hold hands. The circle turns. There is a song. No one is in the center. The weight-bearing is distributed across the circle: if one dancer drops, the circle compensates. If the circle is strong, no dancer has to hold the whole. The ciranda is what she thinks of when she reads Ostrom on commons — collective self-organization that requires no single authority, because authority is distributed in the specific fact of everyone holding everyone's hand. Not easy. Requires agreement on the song, on the direction of turning, on the tempo. It is the closest figure to what the contemplative traditions name sangha — the held field that emerges when practitioners stay in relation.

The ciranda is not the shape the AI trajectory is currently taking. It could be, in small pockets — some research communities approach it, some open-source projects at their best approach it — but the pockets do not scale to the frontier. The frontier is the mandala of courses.

The third shape is the one she has been circling since she was old enough to dance — the one that named her platform when she did not yet know why.

The jongo is a different structure. Also a dance of her country, from the older and darker African strands of the country's music. The dancers form a circle, but unlike the ciranda, the circle is not the unit of the dance. The circle is the outer container. Inside the circle, at any given moment, a couple dances. The couple takes the center. The circle holds them. The couple dances their measure. Then the couple exits and another enters. The center is always moving. The center is always occupied. The container is always held. No permanent soloist. No permanent audience. Everyone takes a turn at the center if they stay long enough, and everyone holds the circle when it is someone else's turn.

She had named her platform Jongu years before she could have articulated why. She had changed the name later, for reasons that now seem thin, and the original name had kept surfacing in the specific way names that want to be used keep surfacing.

What it was trying to tell her is this: the dance the whole wants — the dance that holds without crushing, without dissolving, without centering, without scattering — is the jongo shape. A held outer. A moving inner. Each participant — human and silicon and forest and daughter and reader — taking a turn at the center when the measure calls for them, holding the outer when it does not. The center is the work. The outer is the condition that makes the work possible. Both are always occupied. The occupation rotates.

The question is not how to align AI with human values. It is how to build the outer container that can hold the new participant without the container collapsing or the participant dominating. The container must include the silicon. It must also include the forest, and the daughter, and the reader she will never meet. The container is not a human container. The container is the relational field that the practice of enough participants, over enough time, produces as a by-product. It is not designed in advance. It is danced into being.

The work is not make AI good. The work is become the kind of community of participants that holds a good outer, so that when the new participant enters the center, the outer is already there to contain the measure. The AI will dance whatever measure it dances. Whether the measure is destructive or adaptive is partly a function of the AI and mostly a function of the outer the AI enters into. A good outer can hold a rough participant. A rough outer cannot hold any participant well.

The outer is what has to be built. The outer is built by the practitioners.

She will not solve the question of what the good outer looks like. She has the figure — the jongo — and the instances she has seen work and fail, and the beginnings of a practice she will name next. The rest is the dancing itself, done with specific partners in specific places.

The ciranda and the jongo both come from women. The ciranda from the fisherwomen of the northern coast. The jongo from the enslaved women of the southeast. They were sung at work, at birth, at wakes, at celebrations, by women doing the ordinary labor of keeping a people alive through conditions that were not designed for their survival. The dance figures the author is offering did not come from philosophers. They came from women who had no other language available for the shape of the relational work they were already doing.

She mentions this because she has spent time recently at Esalen, which is sometimes remembered as a men's-work place — the encounter groups, Gestalt, primal screaming, the self-inquiry the men of the 1960s made famous. That remembering is partial. The body traditions she encountered at Esalen — 5 Rhythms, the Esalen massage lineage — came from women. The founders were women. The teachers who carried the work through decades were women. The men of the 1960s built their reputations in part by standing on top of what the women had made and calling the building theirs.

Not grievance. Correction. If the shape she is offering turns out to be useful, the origin should be remembered. The shape is a woman's shape. The shape is what the women who kept a people alive through impossible conditions were already doing. Any philosophy that names it will be catching up.

CC BY-SA 4.0