People Are High-Dimensional Data

I once declined an invitation to hang out with friends by saying it was “too high-dimensional.” I mostly wanted to sound smart. But the metaphor has stayed with me, because I genuinely think of people as data — a lot of data. We like to regard consciousness or the soul as something higher, something ineffable. I regard it as a very large vector.

And vectors have pathologies. In machine learning there is a well-known failure mode called the curse of dimensionality: as the number of dimensions grows, distance stops meaning anything, every point becomes equidistant from every other point, and your models stop generalizing. Signal drowns in noise. You need exponentially more data to learn anything at all, and what you learn is mostly variance, not structure.

Relationships work the same way. Every person you meet adds dimensions: history, moods, obligations, opinions about pop culture, wounds you didn’t cause and can’t heal. Past a certain number of meaningful connections — my rule of thumb is around five — the whole thing degrades into garbage output. Not because people are bad, but because the representation is too complex, too high-variance, for anything true to survive. What we actually want from another person is their main form: the essential shape, not the trivial and superficial details. High dimensionality makes that ultimate form nearly impossible to reach, because everything becomes a compromise. And compromise, compounded across enough dimensions, outputs garbage.

This is why large social gatherings exhaust me. A dinner with many people is a high-dimensional object being sampled at low resolution: lots of laughing, almost no signal. I understand people who thrive there — I honestly don’t know how they do it without misery. For me it means wearing a mask, performing interest, and paying for the performance afterward. If life were a game with skill points, the social tree is the one I would refuse to level. Most people insist social fluency is mandatory for success. To hell with it; I am not pursuing what society files under success. The positive spin is that these events can bond me with friends and train a skill I lack. But training a skill you despise so you can perform better at events you dread is its own kind of curse of dimensionality: more dimensions, less truth.

There’s a darker corollary. Why does everyone seem content swimming in the trivial? My suspicion: we are increasingly fed by algorithms tuned toward the absurd and the stupid, precisely so we never look for what we should be looking for. This era’s censorship doesn’t hide information — it bloats you with it. The old censor removed the book from the shelf. The new censor shelves ten thousand books around it and calls the noise “the feed.” Attention, not access, is the scarce resource now, and attention is exactly what dimensionality devours.

So the practice, for me, is subtraction. Fewer connections, held deeper. Less input, chosen harder. Find the low-dimensional form of a person and love that — or admit the dimensionality is too high and step back before the output turns to garbage. It sounds cold to call a soul a dataset. But there is something respectful in it: it means the way to know someone is the way you know anything worth knowing — patiently, at low resolution first, refusing the noise until the structure appears.

Drafted with AI assistance from my personal journal (2024-12-18), then edited by me. See On AI Assistance.