This paper discusses the development of a new search algorithm for art
ificial intelligence systems. The search technique is based on the mat
hematical theory of the Cantor set. The search algorithm will be effic
ient for specialized search domains where the distribution of the data
elements to be searched is approximately normal. The approach uses th
e iterative procedure of deleted middle thirds. This facilitates quick
pruning of a search space. The new search technique has potential app
lications in computational systems with large input-output data handli
ng. Extensive experiments comparing the Cantor set search (CSS) to bin
ary search indicate that the new search technique holds good promise.