Extending Universal Intelligence Models with Formal Notion of Representation

TL;DR

This paper explores extending the Representational MDL principle within universal intelligence models, emphasizing the importance of representations as a key step toward achieving efficient general intelligence through hierarchical structures and information-theoretic optimization.

Contribution

It introduces an extension of the RMDL principle for universal intelligence agents, highlighting the role of representations as a crucial meta-heuristic for efficiency.

Findings

Representation integration enhances universal intelligence models.

Hierarchical representations improve model efficiency.

Information-theoretic optimization aids in model adaptation.

Abstract

Solomonoff induction is known to be universal, but incomputable. Its approximations, namely, the Minimum Description (or Message) Length (MDL) principles, are adopted in practice in the efficient, but non-universal form. Recent attempts to bridge this gap leaded to development of the Representational MDL principle that originates from formal decomposition of the task of induction. In this paper, possible extension of the RMDL principle in the context of universal intelligence agents is considered, for which introduction of representations is shown to be an unavoidable meta-heuristic and a step toward efficient general intelligence. Hierarchical representations and model optimization with the use of information-theoretic interpretation of the adaptive resonance are also discussed.