Problems Computers Have Too
Wiring isn't the only problem a brain shares with a computer. Two more are strikingly similar in shape.
The first is what to do with the flood of incoming information. Any computer with limited storage and processing power has to decide what to keep and what to throw away. The brain faces this in an extreme form. Our senses gather information at something like a billion bits a second. In 2024, two Caltech neuroscientists, Jieyu Zheng and Markus Meister, gathered evidence from typing, speaking, reading, games, and memory contests, and estimated that the rate at which a person actually uses information, to think, decide, and act, is about ten bits a second. That's a hundred million times slower. Their figure is new and debated, but the general picture isn't: nearly all the information arriving at our senses is discarded or boiled down long before we're aware of anything. Neuroscientists have recognized the principle since the 1950s, under the name "efficient coding": sensory nerve cells avoid wasting capacity on repetition, and pass on mainly what's new. A brain doesn't record the world. It edits, ruthlessly, before we know anything was left out.
The second shared problem is retrieval: not just keeping information, but finding the right piece again, quickly, in an enormous store. Computers solve this with indexes, small separate structures that record where things are, so a search can go straight to the right place instead of checking everything. The brain seems to use a physical version of the same idea. According to the leading account, the hippocampal indexing theory, a memory isn't stored in one place as a complete file. Its sights, sounds, and other details are scattered across whichever brain regions first processed them. The hippocampus, a curled structure deep in the brain, holds a compact index linking the pieces together. A partial cue, a smell or a few notes of a song, reactivates the index, and the index reactivates the rest, rebuilding the memory from a fragment. The comparison now runs both ways: in 2024, computer scientists published an AI memory system, called HippoRAG, modeled directly on this theory.
Retrieval assumes there's a whole, coherent thing to retrieve, and even that is less settled than it sounds. Think of remembering a flower, its color and its shape held together as one simple memory. Color and shape were never handled in the same place; separate groups of neurons, in separate parts of the brain, process them. Nobody has found where, or how, those separately handled pieces are put back together into the single thing we experience. It's a well-known open question in neuroscience, called the binding problem.
It has a cousin from the operating room. The neurosurgeon Wilder Penfield, working in the mid-twentieth century with patients whose brains he could stimulate directly during surgery, found that he could trigger movements, sensations, and sometimes fragments of memory, but never a decision, a belief, or an act of will. A brain does something entirely ordinary and effortless from the outside, and what happens underneath remains only partly mapped.
Last revised 2026-11-18