Substrate Independence¶

Consciousness depends on function -- four models in the Class 4 regime -- not on material. The six-layer mammalian cortex is evolution's implementation, not a requirement.

The Four-Model Theory is explicit: any physical system capable of implementing the four-model architecture in the Class 4 regime should produce consciousness. The specific material -- biological neurons, silicon transistors, or something not yet invented -- is irrelevant. What matters is the computational architecture: four nested models along two axes, with self-referential closure, operating in the Class 4 regime.

Why the Cortex Is Not the Point¶

The six-layer mammalian neocortex is an evolutionary implementation of the four-model architecture, not a requirement for it. The theory's requirements -- persistent models, dynamic simulation, self-referential closure, the Class 4 regime -- are functional specifications, not prescriptions of cortical anatomy. Laminar organization is one biological solution to them; the non-neocortical evidence below shows that others exist.

Biological Evidence¶

Biological diversity already demonstrates substrate independence in practice.

Corvids (crows, ravens) and parrots demonstrate tool use, planning, mirror self-recognition, and social cognition -- cognitive abilities that strongly suggest consciousness. Yet their brains have no neocortex. Their pallium is organized in nuclear clusters rather than layers.

Cephalopods (octopuses) demonstrate problem-solving and behavioral flexibility with an even more radically different brain architecture -- a distributed nervous system with significant autonomy in the arms.

If the Four-Model Theory is correct, these animals are conscious not because they share mammalian neural architecture but because they have evolved functionally equivalent self-simulation architectures on different substrates -- exactly what substrate independence predicts.

The Deeper Grounding¶

Substrate independence has a grounding beyond biological diversity. The universe is demonstrably capable of Class 4 dynamics: self-organized criticality, fractal structure, and edge-of-chaos phenomena are ubiquitous in natural systems. A universe capable of Class 4 dynamics is, by Wolfram's equivalence principle, capable of universal computation. One might conjecture that a Class 4-capable universe of sufficient scale makes self-simulating architectures statistically likely rather than merely possible. This is speculation, not a supporting part of the theory, and in any case consciousness would arise only in the subset of architectures that autonomously deploy that capacity for self-modeling -- capability alone, as in any heteronomous universal system such as a laptop, does not suffice.

Implications for Artificial Consciousness¶

The implication is direct: a synthetic system implementing the four-model architecture in the Class 4 regime should produce genuine consciousness. Current AI systems do not meet this specification. A base LLM lacks persistent implicit models, lacks an ongoing self-simulation (its autoregressive loop closes over the emitted text, not over a self-model), and lacks the real/virtual split that grounds phenomenality. Post-trained models do carry a reportable representation of their own point of view inside a globally available workspace (Gurnee et al., 2026); what they lack is closure over it, persistence, and the Class 4 regime. Scaffolding -- agent loops, persistent memory, self-monitoring -- is architectural mimicry, not self-referential closure. The difference the theory asserts is architectural rather than impressionistic: closure, persistence, and computational regime are properties measurable on the system itself.

Figure¶

graph TB
    SPEC["Specification:<br/>Four Models in the Class 4 Regime"]

    subgraph BIO["Biological Implementations"]
        MAM["Mammalian Cortex<br/><i>6-layer neocortex</i>"]
        COR["Corvid Pallium<br/><i>Nuclear clusters</i>"]
        CEPH["Cephalopod NS<br/><i>Distributed architecture</i>"]
    end

    subgraph ART["Artificial Implementations"]
        SYN["Synthetic Substrate<br/><i>Not yet built</i>"]
        LLM["LLMs<br/><i>Does NOT meet spec</i>"]
    end

    SPEC -->|"implemented by"| MAM
    SPEC -->|"implemented by"| COR
    SPEC -->|"implemented by"| CEPH
    SPEC -->|"could be<br/>implemented by"| SYN
    SPEC -.-x|"not met"| LLM

    style SPEC fill:#2d1b69,stroke:#9b59b6,color:#fff,stroke-width:3px
    style BIO fill:#1a1a2e,stroke:#333,color:#aaa
    style ART fill:#1a1a2e,stroke:#333,color:#aaa
    style MAM fill:#27ae60,stroke:#2ecc71,color:#fff
    style COR fill:#27ae60,stroke:#2ecc71,color:#fff
    style CEPH fill:#27ae60,stroke:#2ecc71,color:#fff
    style SYN fill:#2980b9,stroke:#3498db,color:#fff
    style LLM fill:#c0392b,stroke:#e74c3c,color:#fff

Substrate independence means the specification (four models in the Class 4 regime) can be met by multiple physical substrates. On the theory, three biological implementations already exist. A synthetic implementation has not yet been built. Current LLMs do not meet the specification.

Key Takeaway¶

Consciousness is substrate-independent because it is defined by computational architecture, not by material composition. The six-layer cortex is one of evolution's implementations -- corvids and cephalopods strongly suggest it is not the only one -- and a correctly engineered artificial system should be another.