To understand is to describe briefly. A learner that finds the shortest account of its data has found its regularities, and regularity is what understanding grasps. Intelligence and compression may be two names for one thing.
The intuition
The idea that compression equals intelligence holds that finding the shortest description of data is equivalent to understanding it. Patterns are precisely what allow brevity. To compress well is to have found the structure.
The structure
The mechanism is that regularity permits shortening. Random data cannot be compressed, but structured data can be captured by a rule far shorter than the data itself. The rule is the understanding.
The subtlety
The turn is that prediction and compression coincide. A model that predicts the next observation can encode it cheaply, so a good predictor is a good compressor and the reverse. Foreseeing and shortening are the same act.
The price
The equivalence has limits. Not all useful understanding minimizes description length, and the shortest code may be uncomputable or useless in practice. The principle inspires more than it operates.
The boundary
The implication is a definition of learning. To learn is to compress experience into rules that reach beyond it, trading many instances for few principles. Brevity is the fingerprint of comprehension.
The larger point
Compression-as-intelligence equates understanding with finding the shortest description, since regularity is what permits brevity. Prediction and compression coincide. To learn is to trade instances for principles, and brevity marks comprehension. What makes the idea durable is not that it settles a question but that it reframes many. It teaches where to look and what to discount, which is often more valuable than any particular answer it yields. Understood in this spirit, it becomes a habit of attention rather than a doctrine, and habits of attention are what distinguish deep comprehension from mere knowledge.
