Artificial Intelligence Evolved from Random Behaviour: Departure from the State of the Art

Abstract

Since John McCarthy at MIT coined the term artificial intelligence in 1956 aiming to make a machine have a human-like intelligence in a visible future, we have had lots of discussions whether it is possible in a true sense, and lots of intelligent machines have been reported. Nowadays, the term is ubiquitous in our community. In this chapter we discuss how those proposed machine intelligences are actually intelligent. Starting with how we define intelligence, how can we measure it, how those measurements really represent intelligence and so on, by surveying the Legg and Hutter’s seminal paper on formal definition of machine intelligence, we name a few others, taking a brief look at our own too. We also consider a modern interpretation of the Turing test originally proposed in 1950. Then we argue a benchmark to test how an application is intelligent by means of an algorithm for stock market investment as an example. Finally we take a consideration of how we can achieve a human intelligence in a real sense in a real visible future, including an analysis of IT changes stimulating artificial intelligence development.