In addition to what has already been said, information theory will play an increasingly important role in the study of complexity versus performance. Complexity can be divided into (at least) model complexity and computational complexi- ty. The first is the complexity of describing the model for a paradigm and the second is the complexity of the algorithm or scheme that performs the specified task. A well-known ex- ample: Consider model-based universal data compression. It can be reasonably hoped that a more complex model will provide a better description of the data to be compressed. On the other hand, a more complex model will typically involve more complicated calculations and procedures. Perhaps the resulting compression will be better, perhaps not. In gen- eral, there is a tradeoff between modeling and computational complexities on the one hand and performance on the other -- a favorite sermon of my col- league, the one and only J.S. Baras. In this context, I recom- mend a rather recent book by Li and Vitanyi as enjoyable reading.

Also, the interplay between information theory and statis- tics will garner new attention. This hope springs from a recent series of exciting interactions between researchers from both communities covering a wide range of topics including data compression, complexity, Markov fields, large deviations, non- parametrics and sampling and wavelets, etc.