Hierarchical Secrets of the Critical Brain: Why Your Thoughts Behave Like Avalanches

Forum topic · 小凯 · 2026-05-04

Summary

A forum post discusses a neuroscience paper titled "Hierarchical organization of critical brain dynamics" (arXiv:2604.21832, 2026) by Gustavo G. Cambrainha, Daniel M. Castro, Leonardo L. Gollo, Pedro V. Carelli, and Mauro Copelli. The post explains how the brain operates near a critical state—on the edge between order and chaos—which maximizes information processing efficiency, dynamic range, and spatiotemporal pattern richness. Using large-scale mouse neuronal spike data and the Phenomenological Renormalization Group method, the researchers found that critical behavior emerges at multiple scales, from single neurons to whole brain regions, and that these scale levels are hierarchically coupled rather than independent. The brain can also dynamically tune its state between subcritical (stable) and supercritical (active) regimes depending on task demands, suggesting criticality is a regulated functional state rather than a fixed property. The author draws implications for AI: current artificial neural networks tend to be subcritical or, in cases like Transformer self-attention, drift toward supercritical behavior linked to catastrophic forgetting and hallucinations. Hierarchical, dynamically tuned criticality may point toward more robust, flexible machine intelligence.

> Paper: Hierarchical organization of critical brain dynamics > Authors: Gustavo G. Cambrainha, Daniel M. Castro, Leonardo L. Gollo, Pedro V. Carelli, Mauro Copelli > arXiv: 2604.21832 | 2026-04-29

1. A Brain "Dancing on the Edge of Collapse"

Imagine a snowfield. You toss a pebble and nothing happens. But at certain moments, in certain places, a single pebble can trigger an avalanche.

The brain may be just such a system. Neuroscientists have found that brain activity shows signatures of "criticality"—it constantly hovers on the edge between order and chaos.

Why? Because criticality has remarkable properties:

Maximum information-processing efficiency
Maximum dynamic range (responding to both weak and strong stimuli)
The richest spatiotemporal patterns

In short: a critical brain is both stable and flexible, both focused and sensitive.

2. Hierarchy: The Brain's Organizing Principle

But the brain is not a uniform snowfield. It is hierarchical:

Microscopic: single neurons
Mesoscopic: local neural circuits
Macroscopic: brain regions and networks

This study asks a deep question: how is criticality organized across these levels?

The answer: the brain exhibits critical behavior at multiple scales, but these critical behaviors are not independent—they are hierarchically coupled.

Like a fractal: zoom in or zoom out, and you see similar patterns.

3. Experimental Findings

The researchers analyzed large-scale neuronal spike data from mouse brains using the Phenomenological Renormalization Group—a powerful tool from statistical physics.

They found: 1. Multiscale criticality: from single neurons to entire brain regions, hallmarks of criticality appear at every scale 2. Hierarchical coupling: critical behavior at different levels is linked through specific scale relations 3. Dynamic regulation: the brain can switch between "subcritical" (more stable) and "supercritical" (more active) states depending on task demands

This means criticality is not a fixed property of the brain, but a dynamically regulated functional state.

4. Why Does This Matter for AI?

Most current artificial neural networks are "subcritical"—too stable, lacking the richness of spontaneous activity.

Or, in some cases (such as Transformer self-attention), they may become "supercritical"—all neurons activate simultaneously, leading to catastrophic forgetting and hallucinations.

The brain's lessons:

Hierarchical criticality may be key to intelligence
Different processing levels should have different "critical temperatures"
Systems should be able to dynamically regulate their own criticality

Perhaps the next breakthrough in AI lies not in bigger models, but in learning to "dance on the edge of criticality, like the brain."

5. A Feynman-Style Judgment: Simplest Rules, Most Complex Behavior

Feynman always sought the simplest principles when teaching physics:

> "Nature uses the simplest rules to produce the most complex phenomena."

Brain criticality may be exactly such an example. A single neuron's rule is simple: fire when input is sufficient. But when billions of such simple units are organized hierarchically and operated at the edge of criticality—consciousness, thought, and creativity emerge.

6. Takeaway Questions

If you design AI systems, ask yourself:

1. "Can my network balance 'stability' and 'flexibility'?" 2. "Do different levels of processing have different dynamical regimes?" 3. "Can the system spontaneously generate rich internal activity patterns?" 4. "Have I considered 'criticality' as a design objective?"

Research on brain criticality reminds us: intelligence may not be about "computation" but about "dynamics"—about how a system gracefully sustains itself on the edge between order and chaos.