On Intelligence

Definitions are useful only when they survive edge cases.

Pragmatic, but fragile. Consider a monk who has cultivated total equanimity. He wants nothing, and therefore has everything he wants. Maximally intelligent? No. Tuning the denominator to zero is transcendence, not intelligence.

This definition conflates intelligence with desire management. It rewards wanting less, not doing more.

Steel-manned, the definition is pointing at something real: intelligence helps you predict where life is heading and act on it. But a good definition should not need that much rescue.

Harder to break. Right now, infinite factors collide: markets, physics, human behavior. An intelligent system turns these noisy factors into an accurate prediction.

The chess grandmaster sees twelve moves ahead. The engineer knows which design fails before it's built. The investor weights conflicting signals. Each predicts the future across different time horizons.

What separates intelligence from luck is reliability. Intelligence is a computational capacity: taking noisy inputs and producing accurate outputs about states that have not yet occurred.

This is why artificial intelligence is correctly named even when it lacks consciousness, desires, or goals. A model that predicts protein folding is intelligent in the way that matters — not because it wants anything, but because it compresses reality into accurate forecasts. The definition survives the edge case that will dominate the next century: an entity that predicts better than any human but experiences nothing at all.

The monk is at peace. The predictor is intelligent. They may be the same person, but they are exercising different faculties.