Replay and compositional computation

TL;DR

This paper proposes that brain replay mechanisms enable compositional computation by assembling entities into relational structures, facilitating the derivation of new knowledge, with implications for AI generalization.

Contribution

It introduces a novel hypothesis that replay supports compositional reasoning through role-binding and compound statement formation, advancing understanding of neural and AI generalization.

Findings

Replay can implement compositional computation.

Hippocampus binds objects to roles for flexible reasoning.

Implications for AI systems to enhance generalization.

Abstract

Replay in the brain has been viewed as rehearsal, or, more recently, as sampling from a transition model. Here, we propose a new hypothesis: that replay is able to implement a form of compositional computation where entities are assembled into relationally-bound structures to derive qualitatively new knowledge. This idea builds on recent advances in neuroscience which indicate that the hippocampus flexibly binds objects to generalizable roles and that replay strings these role-bound objects into compound statements. We suggest experiments to test our hypothesis, and we end by noting the implications for AI systems which lack the human ability to radically generalize past experience to solve new problems.