| Many of the problems that machine intelligence (MI) is expected to solve require extensive knowledge about the world. In order to make informed decisions regarding MI, computing and policy professionals need to understand how to measure intelligent machines. Multiple perspectives (TOP), machine intelligence measurement (MIM), and fuzzy set theories were used to determine the commonalities and differences among the current diverse machine intelligence quotient (MIQ) theories and the means to synthesize them. Three perspectives of MIM were synthesized: the technical perspective (T) focused on features that are of no qualitative meaning to humans; the organizational perspective (O) focused on whether a machine violated any regulation during the course of the intelligent actions; and the personal perspective (P) focused on subtle features parallel to human intelligence. An analysis of 185 scholarly articles on MIQ, suggested that autonomy-theoretic, information-theoretic, performance-theoretic, and semiotic-theoretic measures were in accord with the T-perspective; and the Turing, Searle, and phenomenology tests were in accord with the P perspective. No measure was discovered with the O perspective. An MIQ calculus based on the theoretic framework and three scientific contexts in which a machine could be tested was created. Linguistic Choquet fuzzy integral and complex fuzzy set were used to analyze a case data to demonstrate the calculus. Recommendation was made to use TOP to measure the intelligence of machines. The MIQ tools suggested by this study can be used by computing professionals to measure and make informed decisions on MI. |
