The generality of an AIS would appear to be limited only by the generality
of its global database. If we can
describe the domain completely in such a conceptual graph representation,
then it an AIS can probably operate successfully. However, we do not always
have all the control knowledge available a priori, even after asking
the supervisors. And since the system's learning mechanisms are not clear, it's not clear that
the system can acquire such knowledge independently.
In addition, although the AIS treatment of sensor-effector processes is quite general, there
do seem to be some missing cognitive capabilities. Most notable among these is
natural language. Also, the knowledge source method is a bit limited,
prohibiting implementation of more sophisticated memory mechanisms.
To return, press HOME.
To go to the next document, press NEXT.