COMMENT ⊗   VALID 00003 PAGES
C REC  PAGE   DESCRIPTION
C00001 00001
C00002 00002	toy[s87,jmc]		In defense of toy domains
C00004 00003		Some people ask why we who are trying to develo the use of logic in AI
C00006 ENDMK
C⊗;
toy[s87,jmc]		In defense of toy domains

	Some people ask why we who are developing the use of logic in AI
use toy examples, when the expert system writers are writing programs
that deal with large amounts of real world knowledge.  The reason
is that we are attacking problems with these ``toy'' examples taat
expert system writers have so far evaded but which must be faced
in expert systems sooner or later.  The complaint is like that of
bridge builder grumbling about metallurgists testing the strength
of new steels on samples in testing machines rather than building
bridges.

	The expert system builders might complain that they don't
expect ever to need to face the problems we are studying, although
no-one, to my knowledge, has ever complained very specifically.
Some of the problems I believe they will face are listed in my ``Some
Expert Systems Need Common Sense'' (McCarthy 1983).

	Some people ask why we who are trying to develo the use of logic in AI
use toy examples, when the expert system writers are writing programs
that deal with large amounts of real world knowledge.  The reply is 
that we are attacking problems with these ``toy'' examples that
expert system writers have so far evaded but which must be faced
in expert systems sooner or later.  The complaint is analogous to a 
bridge builder grumbling about metallurgists testing the strength
of new steels on samples.

	Of course the expert system builders might complain that they don't
expect ever to need to face the problems we are studying, although
no-one, to my knowledge, has ever complained very specifically.
Some of the problems I believe they will face are listed in my ``Some
Expert Systems Need Common Sense'' (McCarthy 1983).