Towards Consciousness Engineering by Max Hodak (eamag.me)

In “Towards Consciousness Engineering,” Max Hodak frames consciousness as a physically grounded phenomenon—defined operationally as what anesthetics and psychedelics reliably modulate—and builds a testable, engineering-oriented theory around information structure, energy, and feedback. He argues the brain acts like a “structural prism,” sorting sensor data into distinct phenomenal modes by the symmetries in input (illustrated by ferret cortical-rewiring and sea-star examples). Crucial distinctions are drawn between abstract objects, learned representations (manifolds/embeddings), and physical traces (bits that cost energy to write). A “Form” is a trace actively stabilized by feedback controllers (thalamocortical loops, default mode network), and the boundary of that controller solves the binding problem. Time emerges as “ego time” or the length of a phenomenal moment—fast gamma rhythms for local integration, slower alpha rhythms to bind modes—suggesting that temporal oscillatory structure, not just computation, shapes experience. Hodak then speculates a bold physics-level account: brain dynamics may instantiate a classical field whose canonical quantization yields qualia as non‑Poincaré‑invariant gauge excitations—i.e., perspectival, relabelable phenomenal states. If true, consciousness could exert downward informational causation and be engineered: designing feedback-controlled fields could produce novel qualia or even merged moments across brains. For AI, the takeaway is practical: LLMs hold static embeddings (traces) but lack continuous feedback and oscillatory “moments.” Improving temporal structure—via oscillatory/spiking networks, Kuramoto-style synchrony, or explicit feedback controllers—could be the next direction for memory, integration, and perhaps machine phenomenology.