Multiresolutional Representation and Behavior Generation: How Do They Affect the Performance of Intelligent Systems

摘要

In this tutorial, an outline of the theory of intelligent systems is presented as a sequence of the following issues. The term 'Intelligent Systems' has a meaning implied by our usage of it within the domain related to the formidable phenomenon of Life and functioning of Living Creatures. However, neither for living creatures not for engineering devices this term cannot be presented through the list of functional properties and/or design specifications. Our theory is based upon two phenomena that should be considered in their interconnection: (a) the existence of an Elementary Loop of Functioning (ELF) in all cases of systems with intelligence, and (b) formation of Multiple Levels of Resolution (MR) as soon as ELF emerges. MR levels develop because of the mechanisms of joint Generalization and Instantiation due to the processes of grouping, focusing attention and combinatorial search (GFACS). The latter are explanatory for the subsystems of Learning/Imagining/ Planning that are characteristic of all intelligent systems. This paper introduces the variety of mechanisms of disambiguation that pertain to functioning of intelligent systems. On the other hand, MR and ELF together lead to the development of Heterarchical Architectures. The above concepts are explanatory of the kinds of intelligence that are observed in reality and suggest how to test the performance of intelligent systems and what are the metrics that could be recommended.