Common sense, as we shall use the term, consists of a certain
collection of knowledge and reasoning abilities.  

***list of common sense abilities

	The human common sense abilities may be divided into
what we shall call physical-observational abilities and logical abilities.
We will emphasize the logical abilities in this paper, and it may
be better to try to make AI systems accomplish certain functions
using the logical abilities that humans do with the physical-observational
abilities.  We will explain the distinction with the aid of examples.

	Quite primitive animals, including fish, can predict
where a moving object will reappear when it goes behind a barrier to vision.
For this reason it is implausible to suppose that humans do it by logic,
since the ability to reason logically is quite a late evolutionary development.
Evidently a large number of specialized abilities to reach conclusions
from observations have evolved.  Some of them are redundant in humans
and perhaps even interfere with our more advanced capabilities.  Others
are still more efficient than other ways of accomplishing certain tasks.
The balance between what is best done with specialized mechanisms and
what with the general intellectual mechanism is likely to be different
for artificial intelligence than what humans have evolved.  However, even
for AI, it is likely that some tasks are best done by specialized
mechanisms.  In these cases, suitable forms of interaction between the
specialized and the general will be needed.

	Here are some more examples.

	1. A crudely logical visual system would operate by generating
logical sentences for each pixel in the visual field at each resolvable
time point.  Each such sentence would say the brightness and hue of that
point at that time, and the presence and location of complex objects would
be obtained by logical inference from these sentences.