Ronald J. Brachman
Knowledge representation (KR) has traditionally been thought of as the heart of artificial intelligence. Anyone who has ever built an expert system, a natural language system-almost any AI system at all-has had to tackle the problem of representing its knowledge of the world. Despite it ubiquity, for most of AI’s history KR has been a backstage activity. But in the 1980’s it emerged as a field unto itself, with its own burgeoning literature. Along with this growth, the last decade has seen major changes in KR methodology, important technical contributions, and challenges to the basic assumptions of the field. I survey some of these developments, and then speculate about some of the equally interesting changes that appear on the horizon. I also look at some of the critical problems facing KR research in the near future, both technical and sociological.