Memory and Learning are the most difficult AI techniques
in the programmer's arsenal. Memory, by far the easier of the two, is not
all that complex. Memory merely requires that a program store date
for reference in the future. Learning is the process of knowing when,
where, and how to effectively access that information in the future.
For instance, it is one thing for a baseball program to simply
remember that a certain pitcher struck it out with a curve-ball last
time at bat, and quite another for that program to figure out how to
hit that same pitch in the next situation. By creating programs that
can learn on their own, researchers hope to construct the first fully
functional AI engines. Computers that, but for their physical
appearance, could honestly be mistaken for human beings.