Drew McDermott works on artificial intelligence, but disavows any theory of ``intelligence.'' Instead, he believes that the field should focus on basic problems of getting agents to react appropriately to events in the world around them, regardless of whether we normally call such behavior intelligent. Getting an agent to cope with a real environment requires an understanding of available sensory input and the ability to tailor behavior to changing circumstances. McDermott's current work includes the study of fast algorithms for checking for the presence of a known object in a visual scene; map-learning algorithms for robotic acquisition of robust map representations; and improved heuristic estimators for classical planning.
McDermott has been Area Editor for Artificial Intelligence for the Journal of the ACM since 1984. He is a Fellow of the American Association for Artificial Intelligence, and was General Chairman of the First International Conference on AI Planning Systems in 1992.