Planning to Be Surprised: Optimal Bayesian Exploration in Dynamic Environments

TL;DR

This paper derives an optimal Bayesian exploration strategy for AGI in dynamic environments, enabling more effective discovery of unknown worlds.

Contribution

It introduces a theoretical framework for optimal exploration in dynamic settings, filling a gap in existing exploration strategies.

Findings

Proves the existence of an optimal exploration policy for certain environments

Provides a mathematical derivation of the exploration strategy

Enhances understanding of exploration in AGI development

Abstract

To maximize its success, an AGI typically needs to explore its initially unknown world. Is there an optimal way of doing so? Here we derive an affirmative answer for a broad class of environments.