It is argued that the world representation by symbolic metaphysics and knowledge representation by symbolic logic are two different strategies in AI. Fundamental knowledge of elements and causes of reality, being ontologically and logically prior to knowledge representation formalisms and any another form of language, is an essential and underlying framework for general intelligence. The intelligence reasoning from theoretical principles is the intelligent agent with deep understanding, whereas the intelligence reasoning from experience is just an expert agent with narrow understanding. It is shown that metaphysical symbolism as unified theory of reality and knowledge of the world is to be the cognitive (conceptual) framework of machines with universal understanding and intelligence.