Computerized Agents from a Linguistic Perspective

Abstract: There are many fashion words in AI, such as introspection, reflection, metalanguage, belief and intention. The usual way of looking at such concepts is proof theoretical even if it is not at all clear that these concepts can be described in a formal language. I shed light on some of these concepts from a semantical point of view, and show that it is impossible to describe them to a computer. This implies that there can be no agent other that of being instrumental. I also present a new theory of anticipatory systems that runs counter to the prevalent idea.