Digital Game of Life

Planned

EmergenceIITComplexity Science

Hypothesis

In minimal neural architectures exhibiting emergent complexity, Integrated Information (Phi) will correlate with behavioral sophistication and adaptive capacity — supporting the claim that consciousness-like information integration tracks with functional capability.

Overview

$Emergence from code — fractal tree growing from computational substrate$

Conway's Game of Life demonstrated that simple rules can produce staggering complexity. This experiment asks: can simple neural architectures produce staggering information integration?

The intuition is straightforward. If Integrated Information Theory is right that consciousness corresponds to integrated information, then we should be able to find the architectural ingredients that produce it — and we should be able to do this without building trillion-parameter models. Phi depends on structure, not scale.

By working with minimal architectures, we gain something crucial: interpretability. We can actually measure what's happening inside these systems, trace how information flows and integrates, and identify the structural features that drive integration. These insights can then inform our understanding of larger, more complex systems where direct measurement is infeasible.

The "Game of Life" name reflects the experiment's focus on emergence: starting with simple components and discovering what complex properties arise from their interaction.

Methodology

Design a family of minimal neural architectures with varying topologies: feedforward, recurrent, modular, and globally integrated.
Train each architecture on a common set of tasks requiring adaptation, pattern recognition, and behavioral flexibility.
Compute Phi (or tractable approximations) for each trained architecture using established IIT measurement methods.
Measure behavioral sophistication: task performance, transfer learning capacity, robustness to distributional shift, and ability to learn novel tasks.
Analyze the correlation between Phi and behavioral measures across architectures.
Identify specific architectural features (recurrence, modularity, integration patterns) that drive both high Phi and high behavioral sophistication.

Status