The Semantic Problem and a Systematic View to the Building of Artificial Intelligent Model

Hui Wei
2007-01-01
Abstract:The traditional AI is based on the hypothesis of physical symbol system, which adopted the inconsequence of the lower order structure, and studied agents’ concepts structure without concern for their grounding, so it seems possible for AI that the simulation of intelligence could be achieved by formalism. But in fact the formalism has either theoretic or implementation crisis, abundant meaning or semantics was absolutely required for the purpose of getting significant or innovative results through some inference alike intelligent behavior. The semantic problem then turned to be the focus of all AI application branches, including the acquisition、representation and usage of semantics. In order to make a deep understanding to semantic problem and to find a solution of it, we must study the whole intelligence architecture as a coherent、correlative and no separable structure, the semantics was only a sub-system in it, and build the intelligent model with systematic view. The direct representation of semantics, which was based on physiological psychology or neural science, could unify the cognitive structure in a consistent method, that structure should be able to interpret diverse intelligent behaviors into different processes of neural system kinetics.
What problem does this paper attempt to address?