CRANIS
CRANIS Abstract
CRANIS (Concept-Relation in Artificial and Natural Intelligent Systems) is:
- a theory of networked processing attempting to describe natural intelligence and to define artificial intelligence
- a practical model reflecting this theory
- an object-oriented software implementing this model
Regarding finality, CRANIS is an engineering project. The possible philosophical perspectives present in the theory have no other utility than the selection of appropriate parts in a thinking engine. This engine in turn will aim various applications. These applications all tend to provide support or substitution to humans in solving some problems. These problems include among others:
- associative calculus
- gaming
- pattern recognition
- disambiguation(considered as a kind of pattern recognition)
- translation
The two main features of CRANIS are (a) the strict separation of words from concepts and (b) the critical role of consciousness.
The kernel of CRANIS is a set of concepts (nodes) linked through relations. The relations defined in this jelly may be qualified or fuzzy.
Words are considered as weak correspondents to concepts. The mapping of words (percepts) towards concepts is called evocation. The mapping of concepts towards words (percepts) is called manifestation. The more general expression percept is used for representations of concepts using words, images, sounds,...
The consciousness is organized as a tree of focus, and each focus is a dynamic excitation state given to all concepts. Excitation states are determined from external signals, and form the propagation of signals across the concepts nodes. The nodes interfere as a resonating network.
More generally, the CRANIS study relies one one assumption, and tests one hypothesis. The key assumption is that intelligence - at least a basic form of intelligence - is nothing more than resonance effects appearing in a jelly of excited concept nodes linked through numerous relations. The hypothesis to test is the emergence of this form of basic intelligence in an artificial software or hardware construction.