The separation principle: where beliefs and desires come from?

LessWrong reports that a new essay argues the classic belief‑desire model of agents can be derived from optimal control and reinforcement‑learning theorems. By invoking the separation principle—a result that lets a complex optimization be solved by tackling its sub‑tasks independently—the author shows that an agent’s internal model, its beliefs, and its goals, its desires, emerge naturally from optimal policies. The piece draws on cybernetic ideas like the Good Regulator Theorem and on information‑theoretic concepts such as Shannon’s source‑channel separation theorem to build this bridge. In short, the essay proposes that rather than assuming beliefs and desires, we can see them as formal properties of optimal control solutions.

Listen to this story

Hear this and more stories in a personalized audio briefing.

Open The Chonkerton