We review some of the scientific opportunities and technical challenges posed by the exploration of the large digital sky surveys, in the context of a Virtual Observatory (VO). The VO paradigm will profoundly change the way observational astronomy is done. Clustering analysis techniques can be used to discover samples of rare, unusual, or even previously unknown types of astronomical objects and phenomena. Exploration of the previously poorly probed portions of the observable parameter space are especially promising. We illustrate some of the possible types of studies with examples drawn from DPOSS; much more complex and interesting applications are forthcoming. Development of the new tools needed for an efficient exploration of these vast data sets requires a synergy between astronomy and information sciences, with great potential returns for both fields.