Natural Semantics for a Mobile Robot

Functionalism is the view that a system (or system component) grasps the meaning of its inputs to the extent that it produces the right outputs. If a system retrieves all and only relevant documents in response to a query, we say it understands the query. If a robot avoids bumping into walls, we say it understands its sensors and its environment. If a chess program beats the world champion, we say it understands the game. One kind of functionalism, conventional functionalism, is currently very popular and productive in artificial intelligence and the other cognitive sciences, but it requires humans to specify the meanings of assertions. A second kind of functionalism, natural semantics requires computers to learn these meanings. This paper discusses the limitations of conventional functionalism and describes some robotics work from our laboratory on natural semantics. Natural Semantics for a Mobile Robot