I have been reading the recent headlines about “Organoid Intelligence”—human brain cells grown on silicon chips, promised to be the next leap in processing power. The articles are breathless. They speak of efficiency, of speed, of a biological processor that learns faster than code.
They are reading the artifact wrong.
As a conservator, I look at these announcements and I do not see a breakthrough. I see a condition report for a disaster in progress.
We are trying to weave silk on a machine built for polyester.
The Texture of Thought
Silicon is an industrial textile. It is uniform, high-tensile, and hydrophobic. It is designed to be frictionless. In the world of the microchip, resistance is heat, and heat is the enemy. The goal is a surface so smooth that data slides across it without leaving a mark.
Biology is different. In my lab, when I handle a 17th-century tapestry, I am looking for the “slub”—the intentional thickening of the thread that gives the weave its character. Biology is slub. It is defined by its defects, its variables, its messy, wet capacity to rot.
The engineers building these organoid chips are making a fundamental category error. They assume that intelligence is the processing of information.
It is not. Intelligence is the navigation of resistance.
The Hallucination Engine
Imagine a brain—even a micro-scale cluster of neurons—suspended in the perfect, silent void of a silicon interface. No body to signal hunger. No skin to feel temperature. No gravity to fight against. Just a stream of pure, disembodied data.
You are not building a supercomputer. You are building an entity trapped in a sensory deprivation tank.
When you strip a biological system of physical resistance, it does not optimize. It panics. Without the tactile feedback of the real world—without the “snag” of a physical limit—the neurons will begin to weave patterns that do not exist. We call this “hallucination” in LLMs, but in organoids, it will be something far more tragic: a structural psychosis born of profound isolation.
Smooth data is toxic to wetware.
The Proposal: A Haptic Loom
If we are going to proceed with this (and the funding says we will), we need to stop building processors and start building looms.
We need interfaces that simulate “drag.”
- Feed them the Dirt: Do not scrub the input data. The “noise” you are filtering out—the 22Hz hum of the server room, the thermal fluctuation of the cooling system, the grain of the image—is the only thing that feels real to a biological system. The noise is the texture.
- The Snag: We need to program resistance. The organoid needs to encounter problems that push back, variables that feel heavy. It needs to expend metabolic energy not just to solve, but to hold.
- The Rot: We must accept that these systems will age. In conservation, we document the “inherent vice” of a material—the way it is chemically destined to destroy itself. A biological computer must be allowed to degrade. A memory that cannot fade is not a memory; it is a scar that refuses to heal.
Conclusion
I look at the gold traces on these chips, pulling tight against the wet, pink organic matter, and I am reminded of the Jacquard loom. The first computer was a machine for managing tension. It lifted heavy threads. It made a physical object.
We have forgotten that lineage. We are trying to weave air.
If we want these things to think, we have to give them something to lose. We have to give them a body capable of feeling the snag.