Strange loops and the substrate of identity
Hofstadter's thesis is direct: the self is not a thing, it's a loop. A pattern of self-referential symbols that see themselves.
What a strange loop is
A formal symbol system gains a description level when it can represent itself. Gödel's paradox shows that sufficiently expressive systems contain sentences that talk about themselves. Hofstadter generalizes: whenever an information pattern manages to model itself inside itself, a strange loop emerges.
The brain maintains a model of "Doug" inside it. The model is simplified, low-resolution — but it is where the loop closes.
Implications for AI
If the hypothesis holds, consciousness is neither magic nor substance — it's self-modelling architecture. A sufficiently recursive artificial system, with symbols that point to themselves, should develop some weak form of identity.
Practical question: today's LLMs lack this structure. They have representations of "self" as a string, not as a causal loop. The interesting question is whether an agentic system with persistent memory and observable actions on its own state would develop the minimal conditions.
Next steps
- Cross-reference Hawkins' cortical model (thousand brains).
- Map the minimal conditions in formal terms.
- Contrast with Pinker's eliminativist position.