Extending Universal Intelligence Models with Formal Notion of Representation
This work addresses the theoretical challenge of bridging universal and practical intelligence models, which is incremental as it builds on existing RMDL principles.
The paper tackles the gap between universal but incomputable Solomonoff induction and its practical approximations like MDL by extending the Representational MDL principle to universal intelligence agents, showing that introducing representations is essential for efficient general intelligence.
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.