Challenges to Machine Learning: Relations Between Reality and Appearance

Abstract

Machine learning research, e.g. as described in [4], has as its goal the discovery of relations among observations, i.e. appearances. This is inadequate for science, because there is a reality behind appearance, e.g. material objects are built up from atoms. Atoms are just as real as dogs, only harder to observe, and the atomic theory arose long before there was any idea of how big atoms were. This article discusses how atoms were discovered, as an example of discovering the reality behind appearance. We also present an example of the three-dimensional reality behind a two-dimensional appearance, and how that reality is inferred by people and might be inferred by computer programs. Unfortunately, it is necessary to discuss the philosophy of appearance and reality, because the mistaken philosophy of taking the world (or particular phenomena) as a structure of sense data has been harmful in artificial intelligence and machine learning research, just as behaviorism and logical positivism harmed psychology.