Towards generally intelligent machines

Abstract

In spite of the impressive successes of many Artificial Intelligence techniques in narrow domains, one of the unrealized goals of AI remains the creation of a generally intelligent machine. It is hypothesized that the capacity for being creative in a wide variety of environments is a necessary and sufficient condition for general intelligence in machines. In defense of our proposition, we bring forth three arguments: Law of Requisite Variety Argument, Turing Test Argument, and Complex Systems Argument. Grounded in our hypothesis, we introduce a new test of machine intelligence.