Imagine a world where software will be able to automatically schedule doctor's appointments based on mutual availability, put together travel itineraries that match very specific criteria, and exchange data of various forms seamlessly--be they biological, commercial, or statistical in nature. This is exactly what the Semantic Web--basically today's Web enhanced with machine-readable information--is trying to foster. In supplementing the human-readable content of the current Web with agent-processable metadata, the Semantic Web community will be attempting to leverage half a century of AI research on knowledge representation.
Therein lies the challenge: how do we expose the power of descriptive logic to users unfamiliar with terms like "directed graph" and "ontology"? How will the Semantic Web analog of the Web browser hide these notions from the user? Beyond natural language interfaces, there has been little work done on the problem of creating user interfaces for managing what are in essence distributed knowledge representations. We need a "sandbox" that will allow both users and developers to incrementally construct decentralized portions of the Semantic Web, as Netscape Navigator did for the Web. The nature of this sandbox forms the basis for this talk.