Project SOMNUS: Technical Overview and Process Guide

1.0 Project Title and Mandate

Full Title: Project SOMNUS (Subconscious and Unconscious Modeling via Neuromorphic Unsupervised Simulation).

Symbolic Meaning: The name is derived from the Roman god of sleep and dreams, chosen to evoke the project's focus on exploring the full spectrum of consciousness, particularly non-waking and altered states.

Initiative Type: A dedicated splinter initiative from the A.L.C.H.E.M.I. framework.

Primary Objective: To create a high-fidelity computational simulation of the brain's filtering mechanisms and the dynamics of transitioning between different states of consciousness.

Explicit Limitation: The project's goal is not to create a sentient or "truly conscious" AI. Its focus is on the tractable and scientifically valuable aim of building a model that can accurately replicate the measurable neural correlates of consciousness (fMRI, EEG) observed in human studies.

2.0 Architectural Design: A Hybrid GWT-IIT Model

The core of Project SOMNUS will be a computational model built upon the G.E.N.O.M.E. (General Embodied Neuromorphic Operating Machine Environment) engine. The architecture is a novel hybrid of Global Workspace Theory (GWT) and Integrated Information Theory (IIT).

Macro-architecture (GWT-inspired):

• Specialist Processors: The model will feature a multitude of parallel, unconscious modules implemented as specialized Spiking Neural Networks (SNNs) within the G.E.N.O.M.E. environment.
• Global Workspace: The central G.E.N.O.M.E. neuromorphic core will serve as the global workspace where information is integrated and broadcast back to the entire system.
• Communication: The A.L.C.H.E.M.I. NeuroMesh infrastructure will provide the necessary low-latency, high-bandwidth communication.

Micro-principles (IIT-inspired):

• Competition via Integrated Information: The competition between specialist processors for access to the workspace will be resolved based on principles of integrated information.
• IIT-based Ignition: The "ignition" event (selection and broadcasting of information) will be triggered when a coalition of information forms a maximally irreducible cause-effect structure, optimizing for a computationally tractable proxy for Phi (Φ).

3.0 Core Simulation Mechanics: Modeling the Filter

A key objective is to explicitly model the brain's dual-filter system by implementing two specific, dynamically tunable modules.

DMN Simulation (The Content Filter):

The Default Mode Network will be modeled as a high-level specialist SNN trained on internally generated data representing autobiographical memory and future planning. Its output will act as a modulatory input to the global workspace's selection mechanism. In its default state, the DMN module's activity will be high, biasing the workspace to select self-relevant information, and its activity will be programmed to be inversely correlated with the activity of primary sensory modules.

Thalamic Gate Simulation (The Volume Gate):

This will be a set of parameters that function as a dynamic gain-control mechanism, scaling the amplitude or frequency of spike trains from sensory modules before they compete for workspace access. In the default "Normal Waking Consciousness" state, this gate will be in a high-filtering mode (low gain), allowing only strong or highly salient signals to pass.

4.0 Experimental Protocols and Validation

The primary test of the model will be its ability to simulate transitions into various Altered States of Consciousness (ASCs) and replicate corresponding human neurobiological data.

5.0 Subproject Implementation Plan

The project is broken down into five distinct subprojects with clear objectives.

• SOMNUS-01, Foundational Architecture: Implement the hybrid GWT-IIT architecture on the G.Ε.Ν.Ο.Μ.Ε. engine.
• SOMNUS-02, NWC Baseline Modeling: Calibrate the model to replicate resting-state fMRI and EEG data of Normal Waking Consciousness.
• SOMNUS-03, Sensory Deprivation Simulation: Simulate Flotation-REST by removing sensory input and validate against human EEG signatures.
• SOMNUS-04, Psychedelic State Emulation: Simulate psychedelic effects by programmatically suppressing the DMN and opening the thalamic gate, validating against human fMRI data.
• SOMNUS-05, Ethical Oversight & Reporting: Conduct continuous ethical review and documentation of the model's capabilities and limitations.

6.0 Ethical and Validation Framework

The project will adhere strictly to the principles and tools established in the A.L.C.H.E.M.I. Master Project Plan. The model's performance will be continuously benchmarked against a large, curated dataset of real human neuroimaging data (fMRI, EEG). The A.L.C.H.E.M.I. CLARITY module will be essential for this process, allowing for a transparent comparison between the simulation's output and the empirical ground truth. The project will operate entirely within the ethical framework outlined in A.L.C.H.E.M.I. Subproject 6, which addresses AI bias, fairness, and safety. The scope is explicitly limited to the simulation of neural correlates, not the creation of a phenomenally conscious entity.