Todd, P.M. (1992). The animat approach to intelligent behavior. Computer, 25(11), 78-81.
Abstract
The traditional approach to building artificially intelligent systems can be characterized as a top-down methodology: creating large complex systems with massive amounts of knowledge dedicated to solving one particular aspect of supposedly intelligent behavior, such as theorem-proving, or playing chess, or understanding natural language. At some point, it is hoped, all of these separate modules can be connected together to create a wholly intelligent system, but until then, we have a collection of idiots savants, who may, for instance, be able to beat all but a handful of the humans on this planet at chess, but lack the ability possessed by even the lowliest housefly to navigate through a crowded room to the chessboard. The failure of three decades of AI research to even match the behavioral repertoire of insects, let alone humans, and a heightened awareness that what it takes to act intelligently in a challenging environment has little to do with game-playing and theorem-proving, has led to the emergence of a new approach to creating intelligent systems, one that starts from the bottom: the animat path to AI.
Back to main page.