The integration of logic engines, deep neural nets, and evolutionary learning systems in cognitive processes should involve a common representation for all types of knowledge. This common representation should be practical yet interconvertible, so it can be used by both declarative and sensory knowledge to solve problems effectively.