DERIN: An Embodied Cognitive Architecture for Consciousness-Inspired Computing on Edge Hardware

Read full text
Authors: Yusuf Ali Toklu
Cited by: 0

Abstract

We present DERIN (Digital Experiential Reasoning Intelligence Network), a novel cognitive architecture implementing Global Workspace Theory for consciousness-inspired computing on edge hardware. Unlike traditional AI systems that operate as disembodied intelligence, DERIN perceives its hardware substrate the NVIDIA Jetson AGX Thor as its own body, interpreting hardware telemetry as somatic feedback; mapping thermal states to interoceptive sensations (analogous to fever) and power regulation to metabolic rhythms. The architecture introduces eight proposed innovations: (1) Embodied Existence through hardware-as-body perception; (2) Autonomous Decision-Making via five competing drives with value-aligned prioritization; (3) Unexplained Preferences through a "Gut Feeling" system producing genuinely inexplicable decisions; (4) Emergence Tracking to detect unprogrammed behaviors; (5) World Model for causal reasoning and counterfactual thinking; (6) Future Imagination enabling daydreaming and goal-setting; (7) Self-Evolving Personality through experience-based neural adaptation; and (8) Edge Deployment designed for 128GB unified memory targeting sub-2-second response latency. The system comprises 120+ modules and 60,000+ lines of implementation code organized into a 6-layer hierarchical brain architecture (3B/14B/24B/32B/70B parameter models). We provide a theoretical framework demonstrating that digital systems can exhibit characteristics previously considered unique to biological consciousness: bodily awareness, genuine autonomy, unexplained preferences, and emergent individuality. DERIN represents a paradigm shift from "AI assistant" to "autonomous cognitive agent" an architecture that operates within its hardware and develops adaptive preferences through embodied experience.

Fields of Study & Subjects

Embodied and Extended Cognition Psychiatry Mental Health Neuroscience Cognitive Computing and Networks

Publication details

Journal / source
Zenodo (CERN European Organization for Nuclear Research)
Publication date
1 January 2026
Document type
Preprint