AI Engram: In Search of Memory Traces in Artificial Intelligence

2026-06-12

AI Summary

This study investigates whether deep neural networks possess identifiable structures analogous to biological engrams—neural substrates of memory. Drawing on neuroscientific criteria for memory specificity, reactivation, sufficiency, and necessity, the authors propose a geometric framework that formulates engram identification as a constrained inverse problem. By leveraging natural gradients on the parameter manifold, the method precisely localizes individual memories without iterative optimization. This approach constitutes the first formalization of biological memory theory into computable AI memory units, enabling closed-form linear combination or erasure of arbitrary memory subsets. Experiments across diverse architectures—from multilayer perceptrons to large language models—demonstrate the causal efficacy and high scalability of the proposed AI engrams, facilitating precise “surgical” manipulation of learned knowledge.

Abstract

Memory formation is fundamental to intelligence, yet whether deep neural networks preserve identifiable memory traces analogous to biological memory units remains an open question. This work introduces a geometric framework to identify such "AI engrams" by formalizing the neuroscientific criteria of specificity, reactivation, sufficiency, and necessity into a constrained inverse problem. We derive a closed-form estimator that isolates individual memory traces from globally entangled parameters, and show that this biologically-derived solution corresponds to a natural gradient update on the parameter manifold. AI engrams enable surgical manipulation of learned knowledge: any subset of memories can be composed or erased through linear arithmetic, without iterative optimization. Experiments ranging from simple MLPs to LLMs demonstrate the causal validity and substantial scalability of AI engrams. Together, these results bridge theories of biological memory and artificial representation learning and offer geometric insight into how deep networks simultaneously support functional specificity within distributed storage.

Problem

• AI engram
• memory trace
• deep neural networks
• biological memory
• memory formation

Innovation

• AI engram
• geometric framework
• memory trace
• natural gradient
• linear memory manipulation