As a field, knowledge representation has often been accused of being o
ff in a theoretical noman's land, removed from, and largely unrelated
to, the central issues in AI. This article argues that recent trends i
n KR instead demonstrate the benefits of the interplay between science
and engineering, a lesson from which all Al could benefit. This artic
le grew out of a survey talk on the Third International Conference on
Knowledge Representation and Reasoning (KR'92) (Nebel, Rich, and Swart
out 1992) that I presented at the Thirteenth International Joint Confe
rence on Artificial Intelligence (IJCAI'93).