Artificial Intelligence at Work (Part 1): Evolution, Behaviour and Learning of Machines

Abstract

The idea of simulating the intelligent behaviour from 1960s and 1970s being enough to speak about smart machines is still alive. It is the so called non-symbolic approach based on the training of artificial neural networks capable of learning by identifying typical features, classifying them and correcting their own errors, that is in the foreground now. They gave rise to sets of extremely powerful AI models which are gradually dissolving borders between the natural and artificial intelligence. The paper analyses the concept of intelligence in terms of perception, memory and learning, then briefly outlines the evolution of AI stating some crucial historical and well known nowdays events, and finally within the symbolic and non-symbolic approach to the simulation of smart behaviour introduces machine learning, more specifically, under what conditions the training of a given artificial system takes place.