One model has one set of blind spots; many, averaged, cover for each other. Combine diverse imperfect predictors and the group often beats any member. There is wisdom in the crowd of models.
The principle
An ensemble aggregates the predictions of several models into one. Where individuals err differently, their combination cancels idiosyncratic mistakes. The whole is steadier than its parts.
The mechanism
The mechanism is the cancellation of uncorrelated error. If members fail in different ways, averaging suppresses the noise while preserving the shared signal. Diversity is the source of the gain.
An unexpected turn
The turn is that diversity matters more than individual excellence. A group of varied mediocre models can outperform a set of identical strong ones. Difference, not just quality, drives improvement.
The hidden cost
Ensembles cost more and reveal less. They multiply computation and obscure a single line of reasoning, trading efficiency and transparency for accuracy. The gain is real but not free.
The limit
The implication echoes beyond machines. Aggregating independent judgments improves estimates wherever errors are uncorrelated, from forecasts to committees. Independence is the crucial ingredient.
The larger point
Ensembles combine diverse models so that uncorrelated errors cancel and shared signal survives. Diversity matters more than individual strength, at the cost of efficiency and clarity. The wisdom of crowds is real when the members are independent. 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.
