/LMAR=0/XLINE=3/FONT#0=BASL30/FONT#1=BASI30/FONT#2=BASB30/FONT#10=BAXM30/FONT#11=ZERO30/FONT#3=STA200/FONT#4=NGB25
␈↓ ↓H␈↓␈↓βS␈↓∧ Department of Computer Science, STANFORD UNIVERSITY, Stanford, California 94305

␈↓ ↓H␈↓∧Telephone: 415 497-4430␈↓ 
≥July 16, 1984 

␈↓ ↓H␈↓∧Electronic mail: JMC␈↓@␈↓∧SU-AI␈↓
.␈↓∧ARPA




␈↓ ↓H␈↓Professor John Haugeland
␈↓ ↓H␈↓Philosophy Department
␈↓ ↓H␈↓Pittsburgh University
␈↓ ↓H␈↓Pittsburgh, PA

␈↓ ↓H␈↓Dear John:

␈↓ ↓H␈↓        Thanks for your "An Overview of the Frame Problem".  I have the following comments.

␈↓ ↓H␈↓        1.␈α
I␈α
doubt␈αthat␈α
many␈α
of␈α
the␈αlanguage␈α
translation␈α
people␈α
were␈αas␈α
dogmatic␈α
as␈α
Weaver.␈α By
␈↓ ↓H␈↓the␈α
way␈αit␈α
seems␈α
to␈αme␈α
that␈α
Weaver␈αwas␈α
more␈α
a␈αpublicizer␈α
than␈α
a␈αcreator␈α
of␈αstatistical␈α
information
␈↓ ↓H␈↓theory.

␈↓ ↓H␈↓        2.␈α
You␈αhave␈α
gotten␈α
into␈αtechnical␈α
ground␈α
in␈αyour␈α
discussion␈α
of␈αresolving␈α
the␈α
ambiguity␈αof
␈↓ ↓H␈↓"pen".␈α The␈α
point␈αis␈α
that␈αthere␈αis␈α
much␈αscope␈α
for␈αinvention␈αin␈α
≡nding␈αthe␈α
relevant␈αfacts␈α
in␈αlarge
␈↓ ↓H␈↓databases,␈α
and␈α
I␈αthink␈α
some␈α
of␈α
the␈αmethods␈α
the␈α
computation␈α
linguists␈αand␈α
AI␈α
people␈αhave␈α
devised
␈↓ ↓H␈↓will work quite well in the "pen" case.  Perhaps some of them have already told you that.

␈↓ ↓H␈↓        For␈α
example,␈α
pursuing␈αboth␈α
meanings␈α
at␈α
once␈αwill␈α
lead␈α
quickly␈α
to␈αa␈α
scenario␈α
in␈α
the␈αplay␈α
pen
␈↓ ↓H␈↓case␈α
and␈αto␈α
inconclusive␈αresults␈α
in␈αthe␈α
other.␈α
 The␈αuse␈α
of␈αhash␈α
tables␈αof␈α
associations␈α
can␈αquickly
␈↓ ↓H␈↓get␈αthe␈α
relevant␈αfacts␈αeven␈α
among␈αmillions␈α
of␈αirrelevant␈αfacts.␈α
 If␈αyou␈αare␈α
not␈αfamiliar␈α
with␈αhash
␈↓ ↓H␈↓coding, you will ≡nd it surprising.

␈↓ ↓H␈↓        The␈α⊂general␈α⊂point␈α⊂about␈α⊂translation␈α∂often␈α⊂requiring␈α⊂understanding␈α⊂has␈α⊂become␈α∂accepted
␈↓ ↓H␈↓except␈α
by␈α∞the␈α
most␈α∞recalcitrant␈α
Chomskyans.␈α∞ As␈α
to␈α
what␈α∞understanding␈α
is,␈α∞there␈α
is␈α∞no␈α
accepted
␈↓ ↓H␈↓dogma.

␈↓ ↓H␈↓        The␈αframe␈αproblem␈αhas␈αbeen␈αtackled␈αin␈αseveral␈αways␈αin␈αAI,␈αnone␈αof␈αwhich␈αdo␈αyou␈α
precisely
␈↓ ↓H␈↓describe.␈α
 It␈α
seems␈α
to␈α
me␈α
you␈α
invent␈α
your␈α
own␈α
approach.␈α
 My␈α
own␈α
approach␈α
is␈α
via␈α
circumscription,
␈↓ ↓H␈↓and␈αit␈αis␈αnow␈αgenerally␈αagreed␈αthat␈αsome␈αform␈αof␈αnon-monotonic␈αreasoning␈αshould␈αbe␈αinvoked.␈α I
␈↓ ↓H␈↓enclose␈α∩a␈α∩draft␈α∩of␈α∪a␈α∩paper␈α∩that␈α∩will␈α∩be␈α∪given␈α∩at␈α∩the␈α∩October␈α∩conference␈α∪on␈α∩non-monotonic
␈↓ ↓H␈↓reasoning.␈α∂ I␈α⊂believe␈α∂that␈α∂the␈α⊂level␈α∂of␈α⊂logical␈α∂technicality␈α∂of␈α⊂that␈α∂paper␈α∂is␈α⊂essential␈α∂for␈α⊂even␈α∂a
␈↓ ↓H␈↓philosophical discussion of the AI approach to the frame problem.

␈↓ ↓H␈↓        I␈α⊂hope␈α⊂you␈α⊂will␈α⊂continue␈α⊂formulating␈α⊂problems␈α⊃for␈α⊂AI␈α⊂to␈α⊂try␈α⊂to␈α⊂solve.␈α⊂ Try␈α⊂to␈α⊃≡nd␈α⊂the
␈↓ ↓H␈↓simplest␈α∞examples␈α∞that␈α∞seem␈α
di≠cult␈α∞to␈α∞you.␈α∞ The␈α∞lack␈α
of␈α∞sharp␈α∞challenges,␈α∞clearly␈α∞expressed,␈α
is
␈↓ ↓H␈↓one reason people keep making programs with the same weaknesses.
␈↓ ↓H␈↓Best regards,
␈↓ ↓H␈↓αProfessor John Haugeland␈↓ ¬tJuly 16, 1984␈↓ 
nPage 2␈↓




␈↓ ↓H␈↓John McCarthy
␈↓ ↓H␈↓Professor of Computer Science