Can brains generate random numbers?

TL;DR

This paper demonstrates that large neural networks inspired by brain activity can generate all possible neuron firing patterns over time without external input.

Contribution

It introduces a method to construct large McCulloch-Pitts neural networks capable of producing all neuron firing subsets autonomously.

Findings

Networks can generate all neuron firing patterns

No external input needed for pattern generation

Scalable to arbitrarily large networks

Abstract

Motivated by EEG recordings of normal brain activity, we construct arbitrarily large McCulloch-Pitts neural networks that, without any external input, make every subset of their neurons fire in some iteration (and therefore in infinitely many iterations).