Philosophically inspired concept acquisition for artificial general intelligence

Abstract

We describe a Bayesian network implementation of a theory of concepts that is motivated by observations from the philosophical debate between Lexical Concept Empiricism and Lexical Concept Nativism. According to our theory, Baptizing Meanings for Concepts (BMC), concepts are acquired by hypothesizing latent kinds/categories to explain observed co-occurrences of sets of properties in a group of objects. The hypothesized kind/category is given a name and inferential relationships are stored between the name and representations for the observable properties. We argue that this process appeases tensions in the philosophical debate by allowing for the acquisition of concepts via perception and inference, while yielding the concepts simple, in the sense of being contingently associated with other representations. The BMC is inspired by a well-known process in the philosophy of language for assigning meanings to linguistic terms [1, 2, 3, 4]. © 2011 Springer-Verlag Berlin Heidelberg.