Skip to main content
The Body in the Data
The Standard with Ears · Arc 2 · TAM_EAR_2-03

The Body in the Data

The withdrawal of a consolation this argument gave itself

In a hurry? Read the executive summary.

TAM-EAR.2-03 · The Standard with Ears · The Approximate Mind

In 1884, on a walled lot beside the veterinary hospital of the University of Pennsylvania, Eadweard Muybridge built a shed a hundred and twenty feet long, painted its interior black, and gridded the back wall with white threads at five-centimeter intervals. Along the front he set banks of cameras, twelve, then twenty-four, wired to trip in sequence. He was fifty-four, English, famous for having photographed a galloping horse with all four feet off the ground and settling a bet about it, and famous for other things too, and the university had brought him to Philadelphia to do systematically what he had done for the horse: catch the body in the act of moving.

For three years, people walked in front of the grid. The results were published in 1887 as Animal Locomotion, seven hundred and eighty-one plates, and despite the title most of the plates are human. Men and women walk, run, climb stairs, descend them, lift, carry, throw. A woman pours water from a jug, plate after plate, the water frozen mid-arc, never twice the same. A blacksmith swings. A child crawls. And in a number of the plates the figure against the threads is Muybridge himself, lean, white-bearded, naked, swinging a pickaxe or walking up an incline, an aging man who put his own body into the grid he built, sequence number attached, one specimen among the rest.

What the grid did was convert movement into positions. Against the threads, a stride became a set of coordinates, a swing became a curve you could measure. Painters used the plates, then physiologists, then everyone. It was the first time the body’s transit through a second of time had been made into data. Be precise about what it was not. The plates held where the body had been. They did not hold why. The rule that carried the pickaxe from one frame to the next, the thing in the man that produced the arc, was not in any photograph, and no quantity of photographs would contain it.

The consolation, and its withdrawal
#

The Vernacular After Geography ended on a floor, and the floor was offered in good faith and has to be taken back.

The claim was this. When the vernacular came apart from geography, the stack came apart unevenly. The lexicon delaminated and went planetary at the speed of the network. The syntax followed. But the bottom layer, everything the body does, how close you stand, the pause that reads as respect here and hostility there, the knife against the onion, the rhythm of pouring, stayed local, transmitted the old way, by standing near somebody for years. And it stayed local, the essay said, for a reason that had nothing to do with protection: it could not be encoded, so there was nothing to harvest. A person could sound like the whole world and move like a single street. The vernacular after geography was a vernacular of the hands.

The consolation was doing real work. It gave the argument a floor, one layer the hearing could not reach, and it gave the reader somewhere to stand.

The withdrawal is not that the reasoning was wrong. The reasoning was right about everything except its tense. The body was not unharvestable. It was unharvested, and the distinction between those two words is the distinction this whole argument keeps finding under every floor it checks: not a wall but a price.

“Cannot be encoded” was never a property of the body. It was a property of the cost of encoding it, and it is going where all the other costs went.

Why the camera changed nothing, and why this does
#

The objection arrives on schedule and deserves its hearing: the body has been on camera for a century. Muybridge’s plates are a hundred and forty years old. Film carried gait and gesture across the world for generations, and the body stayed local anyway, so why would more cameras change what a century of cameras did not?

Because watching is not acquiring, and the century of cameras only ever produced watching.

There is one documented crack in that wall, and it is famous precisely because it was singular. Marcel Mauss, lecturing in 1934 on what he called the techniques of the body, recalled lying in a New York hospital and being nagged by the walk of his nurses, a gait he knew from somewhere, and finally placing it: the cinema. Young French women, he realized on returning to Paris, were walking like American films. He treated it as remarkable, and it was. The screen could transmit a gait, occasionally, coarsely, to a viewer who then did the acquiring themselves, with their own body, by the old method. The camera collected appearances. A human still had to do the learning, and human learning of the embodied layer runs at the old speed, one street at a time.

What is being built now does the learning on the other side of the lens. World models are systems trained not to store the body’s positions but to predict its dynamics: what a hand does next given what it is holding, how weight moves through a stride, the order of operations by which a kitchen produces dinner. The training material is the century of watching, the flood of ordinary video, the sensor logs, the motion capture, and what comes out is not an album. It is the rule. The thing that was in the man and not in the plates, the generator of the arc, is extracted as a representation, and a representation can be run. It can drive a robot’s grip, a rendered figure’s shoulders, a synthetic voice’s breath rhythm, and it can be transferred, copied, and composed, which the embodied layer of a vernacular has never once been in the history of the species, because until now the only storage medium for it was another human body standing nearby for years.

Muybridge’s grid held where the body had been; the new models hold what the body would do next, and the difference is the difference between a record and a rival.

The child acquiring the kitchen from a grandmother was the only pipeline the bottom layer ever had. The pipeline now has a second customer, and the second customer learns from footage, at scale, without the years.

Where the reading starts
#

The hearing of the body, like every hearing in this argument, is not uniform, and saying where it starts is more honest than announcing that it starts.

It starts where the body converts. Labor converts: the choreography of a warehouse picker is already scored per motion, per second, and the scoring is a business line; the movements of drivers have been encoded so thoroughly that the encoding now drives. Surgery converts, and surgical gesture is being captured and modeled in exactly the way surgical vocabulary once was. Sport converts, wholesale. And the avatar market converts most directly of all, because a rendered figure that moves wrongly is a product nobody will sit with, so the gait, the idle stance, the small economy of a believable reach are being purchased, captured, and libraried right now, as assets.

The grandmother’s hands enter the harvest the day a machine needs kitchens, and machines have started needing kitchens. Manipulation data, the record of hands managing objects, has become one of the scarce commodities of the robotics buildout, and companies pay for footage of ordinary domestic work, which means the layer this argument called unharvestable has a spot price.

What is read first, then, is the monetizable body, and the inflammatory body was read long ago, since spectacle was the oldest business the cameras had. The remainder, the enormous quiet remainder of embodied culture, the funeral pace, the doorway distances, the reassurance touch, is in the footage but not yet in the reading, held the way the unread text corpus is held. The pattern of the selective hearing repeats one layer down, in the muscles. And which bodies the harvested slice contains, whose stride becomes the default stride, whose kitchen becomes the kitchen, is the standing question this series has already routed to its own accounting, and it applies here without discount.

No new floor
#

The tempting move at this point is the one that essay made: find the next layer down and rest on it. Say that the models capture the visible body but not the felt one, the kinematics but not the ache, the choreography but not what the kitchen means. Build a smaller floor a little lower and stand on that.

Declined. Not because the smaller claims are false, some of them may hold for a long time, but because the argument has now watched itself do this once and knows what the move is made of. The first floor was built out of a cost and mistaken for a wall. Any second floor built the same way inherits the same flaw, and an argument that keeps retreating one layer at a time, consecrating each new layer as the truly unreachable one, is not describing the world. It is managing its own need for a remainder.

What can be said without a floor is smaller and holds better. Acquisition by standing near somebody has not stopped. The child still gets the kitchen from the grandmother, at the old speed, by the old method, and what she gets is bound to a particular pair of hands and a particular room in a way no model’s composite is. The claim that survives is not that this layer is safe. It is that it is still happening, in parallel, for now, and that the two transmissions, the slow one through proximity and the fast one through the harvest, have different products even when they have the same source.

The body is becoming the last register to be encoded, which is not at all the same thing as the register that could not be.

We wonder whether a person who learns to cook from a rendered pair of hands will be standing in a lineage or in a library, and whether, a generation on, anyone will be able to feel the difference from inside their own wrists.

Muybridge finished the Philadelphia work, toured with his projections, and eventually went home to Kingston upon Thames, where he had been born, and where he died in 1904 digging a scale model of the American Great Lakes into his garden. The plates outlived him the way data outlives its collectors. But among the seven hundred and eighty-one, the ones that hold up strangest are his own: the old man against the threads, swinging the pickaxe, every position of the swing fixed forever at five-centimeter resolution.

The grid took everything it was built to take. What it could not take, and what no plate in the set contains, is the reason a man who owned the apparatus chose to walk into it, and stood in front of his own cameras with nothing on, and swung.


This essay is the third of the anatomy arc in The Standard with Ears, and performs the withdrawal that The Vernacular After Geography planted: the consolation that the embodied layer stays local because it cannot be encoded is retracted, on the grounds that the impossibility was a price. The selective reading of the body follows the two filters of The Standard with Ears; the question of whose body enters the harvested slice belongs to The Harvested Slice and is pointed at rather than spent. The refusal to build a second floor is the essay’s application of the series’ own rule against consolations it cannot defend. The Ear That Answers, in the Claude series, states this withdrawal from inside the apparatus.

How this essay connects to others across The Approximate Mind.

The Ear That Answers states the withdrawal from inside: what the apparatus holds of the body is record, not rule. The Body in the Data makes the same withdrawal from outside and prices it: the consolation that the harvested hands survive in the harvest is not available.
Who Gets Approximated asks whose way of being becomes the default. The Body in the Data asks it of the body itself: whose hands, whose gait, whose technique enters the corpus, and whose is covered by a demographic neighbor's.
The Amber Question holds the record side of the deposit, the linguistic surface speakable on request, and knows this essay's withdrawal without replanting the floor. The two essays split the deposit between them, rule side and record side.
The Body in the Data withdrew the consolation that the harvested hands survive in the harvest. The Harvested Slice shows the selection that ran before the harvesting: the slice was written and leisured, and the hands were outside it from the start.
Muybridge and the Gridded Body
  1. Braun, Marta. Eadweard Muybridge. Reaktion Books, 2010.
  2. Muybridge, Eadweard. Animal Locomotion: An Electro-Photographic Investigation of Consecutive Phases of Animal Movements. University of Pennsylvania, 1887.
  3. Solnit, Rebecca. River of Shadows: Eadweard Muybridge and the Technological Wild West. Viking, 2003.
Techniques of the Body and Their Transmission
  1. Ingold, Tim. The Perception of the Environment: Essays on Livelihood, Dwelling and Skill. Routledge, 2000.
  2. Mauss, Marcel. “Techniques of the Body.” Economy and Society, vol. 2, no. 1, 1973, pp. 70-88.
  3. Polanyi, Michael. The Tacit Dimension. University of Chicago Press, 1966.
Encoding Movement
  1. Delfanti, Alessandro. The Warehouse: Workers and Robots at Amazon. Pluto Press, 2021.
  2. Ha, David, and Jürgen Schmidhuber. “World Models.” arXiv preprint, arXiv:1803.10122, 2018.
  3. LeCun, Yann. “A Path Towards Autonomous Machine Intelligence.” OpenReview preprint, 2022.