Intelligence as Dimensional Collapse: A Mechanistic Model Bridging Digital and Analog Computation

We propose a physical mechanism of intelligence: neurons fire discrete action potentials (digital spikes) that generate electric fields, which interact to form continuous, high-dimensional vector spaces (analog fields). These fields encode information through spatial, temporal, and intensity relationships among vectors—creating a representational space of far higher dimensionality than the physical 3D substrate. Through stochastic resonance, noise selectively amplifies meaningful patterns within these fields until they exceed thresholds, collapsing this hyperdimensional analog information back into discrete neural spikes. We term this cycle of expansion and contraction “dimensional collapse.” This framework provides a parsimonious explanation for neural phenomena including ephaptic coupling and coding efficiency, while suggesting principles for designing robust, energy-efficient artificial intelligence systems that transcend traditional computational paradigms.