Intuition & Insight
The leap across logic — knowledge that arrives whole.
What we mean.
Intuition is fast, parallel, embodied pattern-matching, often inaccessible to introspection. Insight is the sudden re-organization of a problem space — the gestalt 'aha!'.
Both rely on tacit knowledge: the elephant beneath the rider, Polanyi's 'we know more than we can tell'.
In the brain
Chess masters perceive board positions in chunks — not 30 pieces but 7 patterns. Firefighters smell a flashover before reading instruments. Mathematicians dream their proofs.
Neurally, insight correlates with right anterior temporal lobe activity and an alpha-band 'eureka' burst.
In silicon
Chain-of-thought, tree-of-thought, and o-style reasoning models simulate deliberate insight. They can backtrack, hypothesize, verify.
Yet intuition — the cheap fast guess that is usually right — remains the LLM's quiet superpower (single forward pass) but also a source of its hallucinations.
How we arrived here.
- 1908
Poincaré's omnibus insight
- 1958
Polanyi: tacit knowledge
- 1973
Kahneman: System 1 / System 2
- 2022
Chain-of-thought prompting
- 2024
o1 — deliberative AI
“Intuition is the supra-logic that cuts out the routine processes of thought.”
Where the edge moves next.
Will hybrid systems pair LLM intuition with formal verifiers (symbolic, theorem-proving) and become indispensable scientific collaborators?
Where it touches the world.
Medical diagnosis: pattern-recognition assistants.
Cybersecurity threat hunting.
Scientific hypothesis generation.
Why it matters.
If insight is the brain's gift back to itself, an AI that has it has — at minimum — a profoundly different cognitive geometry than yesterday's calculator.