In order to deal with the inherent combinatorial nature of many tasks in ar
tificial intelligence, domain-specific knowledge has been used to control s
earch and reasoning or to eliminate the need for general inference altogeth
er. However, the process of acquiring domain knowledge is an important bott
leneck in the use of such "knowledge-intensive" methods. Compute-intensive
methods, on the other hand, use extensive search and reasoning strategies t
o limit the need for detailed domain-specific knowledge. The idea is to der
ive much of the needed information from a relatively compact formalization
of the domain under consideration. Up until recently, such general reasonin
g strategies were much too expensive for use in applications of interesting
size but recent advances in reasoning and search methods have shown that c
ompute-intensive methods provide a promising alternative to knowledge-inten
sive methods.