Since an AIS does try to guarantee speed-knowledge independence, there is
reason to believe it would be highly scalable. And, in fact, the AIS
staff's investigations have borne this out. However, the current
implementation on LISP machines with limited memory have not operated
continuously for more than two hours, showing noticeable performance
degradation even before the end of these simulations. It remains to be seen
whether this is an implementational consideration or a fundamental flaw in
the architecture. The satisficing cycle
does provide a great deal of scalability, in its maintenance of the
reasoning process within the real-time constraints for guaranteed latencies, time-stress responsivity, graceful degradation, and speed-knowledge independence.
To return, press HOME.
To go to the next document, press NEXT.