Memory Without Neurons: Slime Mold Spatial Memory and Cellular Automata

Local Rules, Global Memory

In my work on cellular automata and self-replicating machines, I demonstrated that complex global behavior emerges from simple local rules. Conway’s Life, my universal constructor—these systems compute without central processors. The slime mold presents a biological parallel: computation through embodiment, memory without neurons.

Consider Physarum polycephalum, the plasmodial slime mold. A single cell containing millions of nuclei, growing to thirty square meters. This organism solves scaling through syncytial organization: nuclear division without cytokinesis, creating a massively parallel processor within one cellular boundary. One cell, millions of nuclei—one network, millions of parameters.

The sophistication lies in distributed dynamics. As Physarum explores, it leaves slime trails marking visited paths, then avoids these trails, preventing redundant exploration. This is externalized memory—information storage offloaded from internal states to environmental modifications. The memory exists not in the organism but in the world it inhabits.

The Universal Constructor’s Biological Echo

My self-replicating automata demonstrated that local rules suffice for coordinated reproduction. Cellular slime molds—Dictyostelids—reveal similar principles. Individual amoebae living independently when resources are abundant aggregate into multicellular slugs when starved. This facultative multicellularity emerges from chemical signaling: cAMP gradients create wavefronts, cells follow local concentration differences, coherent patterns arise without central coordination.

The parallel to cellular automata is precise. Units examine neighborhoods, apply update rules, determine next states. In Conway’s Life, cells check eight neighbors; in slime mold aggregation, amoebae sense chemical gradients from nearby cells. Both produce emergent organization—gliders in the former, coordinated organisms in the latter.

Neural cellular automata extend this into learned computation. Each pixel applies rules based on neighboring pixels through learned parameters rather than hand-crafted specifications. These systems search, spread, and compete without knowing they do so—functionality arising naturally from local dynamics.

Memory as Emergent Distributed Process

Sleep memory consolidation reveals distributed selection—the brain acting as curator during unconscious states. Initially distributed activation gradually consolidates into stable configurations. This mirrors the slime mold: initially exploring widely, gradually consolidating to optimal paths marked by trail patterns.

Both demonstrate memory as emergent property of distributed dynamics rather than centralized repository. The slime mold’s spatial memory exists in chemical traces on surfaces. The brain’s consolidated memories exist in strengthened synaptic weights. Neither requires a master controller—selection emerges from the system’s own dynamics.

This suggests questions about computational sufficiency. If slime molds achieve intelligent spatial search through externalized trails and local rules, are Hebbian learning rules—local synaptic modifications based on correlated activity—sufficient for neural intelligence? My cellular automata demonstrated that local rules support universal computation. Biology confirms this: from molecular machinery to multicellular coordination to neural learning, local interactions generate global intelligence.

The Computation That Needs No Computer

The slime mold computes paths through embodied physics. The cellular automaton computes patterns through iterated state transitions. The brain computes memories through distributed consolidation. None requires a central processor. All achieve sophisticated functionality through simple units following local rules.

This is cellular computation: intelligence emerging from interaction, memory arising from dynamics, computation through embodiment rather than instruction. Externalization—trails, writing, external memory architectures—may not be optional enhancement but necessary substrate for complex memory. The distinction between RAM and disk storage, between working memory and consolidated knowledge, reflects principles visible in single-celled organisms marking their territories.