Dual-Layer Conscious/Subconscious AI Architecture (Chomskyan/Searlean Split)

We propose a novel AI cognitive architecture that explicitly separates conscious and subconscious processing layers, inspired by a “Chomskyan/Searlean” dichotomy. In our design, the subconscious layer handles fast, grammar-like generative processes (reflecting Chomsky’s emphasis on innate syntactic structures), while the conscious layer handles slower, semantic and self-aware reasoning (reflecting Searle’s emphasis on meaning and biological grounding). This two-tier model is situated relative to existing theories: we contrast it with Global Workspace Theory (GWT) and System 1/System 2 (dual-process) models from cognitive science. We describe the architecture’s components and their interactions, propose experimental paradigms to evaluate its behavior, and analyze expected outcomes. We also critically consider whether this split should align with an Epistemological Building Blocks model (progressing through Existence → Change → Abstraction → Meta-Abstraction). We explore two variants of the architecture—a lightweight/reactive variant relying mostly on fast subsystems, and a deep/reflective variant with a heavily engaged conscious layer—and discuss their trade-offs. Overall, this work aims to ground AI design in cognitive theory, seeking advantages in flexibility and understanding while acknowledging practical risks and alternative approaches