Wittgenstein and the Concept of Learning in Artificial Intelligence

Abstract

The object of this investigation is to analyze the application of the concept of learning to machines and software as displayed in Artificial Intelligence (AI). This field has been approached from different philosophical perspectives. AI, however, has not yet received enough attention from a Wittgensteinian angle, a gap this thesis aims to help bridge. First we describe the use of the concept of learning in natural language by means of a familiar and of a less familiar case of human learning. This is done to give us a general idea about the meaning of this concept. By building two basic machine learning algorithms, we introduce one of the technical meanings of learning in computer science, i.e. the use of this concept in machine learning. Based on a study and comparison between both uses, the one in ordinary language and the one in machine learning, we conclude that both usages exemplify one and the same family resemblance concept of learning. We apply this insight further in a critical discussion of two specific philosophical positions about the applicability of psychological or mental concepts to software and hardware, especially in AI. One of the contributions of this investigation is that the use of mental concepts concerning machines does not imply the ascription of a mind.