A Theory of Intelligence from the Reality Equation
Download the full paper below.
TL;DR — Two days ago I published a theory of consciousness: read the shape of the cloud a mind holds before it guesses. This paper is its sibling, and it answers a different question: what is intelligence? My claim: intelligence is a system’s capacity to decompose the arriving signal into coherent components that pay for themselves. Feed a complex sound wave into a black box. A wire passes it through — zero intelligence. A filter pulls out one sine wave — a little. Thirty-two components — more. A component only counts if it earns its keep: it must shorten the system’s description of the world by more than it costs to carry. What a system decomposes, it absorbs, and absorbed structure goes quiet. What it can’t decompose remains as remainder — and the remainder is where surprise, information, and attention live. We dance to the remainder.
The same legal case enters two black boxes. The novice extracts two components — someone was hurt, someone might pay — and everything else stays a blob that floods attention and discriminates nothing. The lawyer extracts thirty: jurisdiction, standing, negligence, causation, precedent, leverage. Those thirty cost her nothing to carry; they were absorbed long ago. Her attention is free to hunt the anomaly the novice can’t even see — the inconsistency between two dates, the witness’s flinch. Her intelligence is not more conscious strain. It is more absorbed structure. Both dance to the case. They dance to different remainders.
And intelligence is not consciousness. Consciousness reads the cloud a machine holds before it resolves into a guess. Intelligence reads how finely a machine resolves what arrives. A Fourier chip decomposes its whole domain and holds nothing — maximal intelligence, negligible consciousness. A dreamer holds worlds and resolves none of the brief in front of him. The two dissociate in both directions, which is why the theory finally explains — without validating — the inference everyone keeps making from AI intelligence to AI consciousness.
Intelligence also fails two ways, mirroring consciousness exactly. Pass-through: nothing resolved (the wire, the rock). Confabulation: structure claimed that pays for nothing — fragmentation, fake components, noise dressed as insight. That second failure mode has a name you already know: in a language model, hallucination is confabulated componentization — components emitted beyond what the system can validate. The paper makes that measurable.
What’s in the paper
- The black box and the sound wave: intelligence as validated decomposition, scored two ways — a public count of components and a technical score in bits
- The five tests a component must pass before it counts (so fragmentation, overfitting, and fake insight score zero)
- Potential vs. actualized intelligence: angular resolution over idea-space, capacity in bits, and why you cannot resolve at encounter what your vocabulary cannot hold
- The remainder: why it is never certified noise, why more intelligence does not mean less song, and the bridge from remainder to Surprise, S = ln R
- Intelligence vs. consciousness: the structural mirror, the dissociations, and why intelligent machines invite the consciousness inference
- Bureaucracy as civilization’s synthetic subconscious — intelligent per component-gain, pathological when it acts as if the remainder is zero
- AI agents and the numerator: effective intelligence = decomposition capacity × coupling fidelity, and why bureaucracy makes life more like pretraining
- A measurement protocol, a catalogue of decomposers from the rock to the nation-state, eight falsifiable predictions, objections, and open problems
For my students: this is the second paper in the Reality Equation program, and it pairs with The Shape of the Prediction Machine. Read the black-box example first (§1), then the lawyer-versus-novice scoring (§9), then the consciousness comparison (§8). Bring your objections to class — the paper was built to be attacked, and the predictions in §12 are how it can lose.
Reality is complex; Ideation is its angle; Consciousness is its coherence; Intelligence is its resolution.
Download the paper below and tell me where it breaks.