/LMAR=0/XLINE=3/FONT#0=BAXL30/FONT#1=BAXM30/FONT#2=BASB30/FONT#3=SUB/FONT#4=SUP/FONT#5=BASL35/FONT#6=NGR25/FONT#7=MATH30/FONT#8=FIX25/FONT#9=GRKB30
␈↓ α∧␈↓␈↓ u1


␈↓ α∧␈↓␈↓ αTAs␈α∂you␈α⊂may␈α∂know,␈α∂we␈α⊂are␈α∂arranging␈α⊂a␈α∂group␈α∂project␈α⊂at␈α∂the␈α∂Center␈α⊂for␈α∂1979-80␈α⊂on␈α∂the
␈↓ α∧␈↓general␈α
topic␈α
of␈α
artificial␈α
intelligence␈α
and␈α
philosophy.␈α
 The␈α
interdisciplinary␈α
group␈α
begins␈αwith␈α
the
␈↓ α∧␈↓premise␈αthat␈αboth␈αartificial␈αintelligence␈αand␈αphilosophy␈αare␈αconcerned␈αwith␈αintelligent␈αbehavior␈αin
␈↓ α∧␈↓physical␈α
and␈α
biological␈α∞systems.␈α
 Both␈α
face␈α
conceptual␈α∞problems␈α
in␈α
characterizing␈α∞behavior.␈α
 On
␈↓ α∧␈↓the␈αphilosophical␈αside,␈αDaniel␈αDennett,␈αamong␈αothers,␈αhas␈αcharacterized␈αintentional␈αsystems␈αwhich
␈↓ α∧␈↓are␈αphysical␈αsystems␈αto␈αwhich␈αcan␈αbe␈αascribed␈αintentional␈αqualities␈αsuch␈αas␈αbeliefs␈αand␈αwants.␈α On
␈↓ α∧␈↓the␈α∞artificial␈α∞intelligence␈α∂side,␈α∞John␈α∞McCarthy␈α∞has␈α∂indentified␈α∞conditions␈α∞when␈α∂mental␈α∞qualities
␈↓ α∧␈↓can be ascribed to machines.

␈↓ α∧␈↓␈↓ αTArtificial␈αintelligence␈αhelps␈αthe␈αphilosopher,␈αbecause␈αintelligent␈αprograms␈αprovide␈αa␈αdomain
␈↓ α∧␈↓where␈α∞behavior␈α∞is␈α
precisely␈α∞defined␈α∞but␈α∞to␈α
which␈α∞one␈α∞must␈α
ascribe␈α∞some␈α∞intellectual␈α∞qualities␈α
if
␈↓ α∧␈↓one␈α∂is␈α∞to␈α∂describe␈α∞what␈α∂one␈α∞knows␈α∂about␈α∞its␈α∂behavior.␈α∞ For␈α∂example,␈α∞what␈α∂a␈α∂particular␈α∞person
␈↓ α∧␈↓knows␈αabout␈αthe␈αstate␈αof␈αa␈αparticular␈αcomputer␈αoperating␈αsystem␈αmay␈αbe␈αexpressed␈αas␈αascribing␈αto
␈↓ α∧␈↓the␈α⊃program␈α∩an␈α⊃incorrect␈α⊃belief␈α∩that␈α⊃a␈α⊃certain␈α∩user␈α⊃of␈α⊃the␈α∩system␈α⊃does␈α⊃not␈α∩want␈α⊃to␈α∩run␈α⊃his
␈↓ α∧␈↓program.␈α
 From␈α
the␈α∞artificial␈α
intelligence␈α
point␈α∞of␈α
view,␈α
a␈α
program␈α∞that␈α
plans␈α
travel␈α∞must␈α
know
␈↓ α∧␈↓that␈αtravel␈αagents␈α
know␈αairline␈αschedules␈α
and␈αmust␈αknow␈αthat␈α
the␈αgate␈αat␈α
which␈αa␈αflight␈αwill␈α
leave
␈↓ α∧␈↓an␈α∩intermediate␈α∩stop␈α∩is␈α∩not␈α∩knowable␈α∩initially,␈α∩but␈α∩can␈α∩easily␈α∩be␈α∩discovered␈α∩by␈α∩an␈α⊃English-
␈↓ α∧␈↓speaking traveller in the United States at the time the information will be required.

␈↓ α∧␈↓␈↓ αTWhile␈α
no␈α
one␈α
expects␈αto␈α
solve␈α
all␈α
the␈αphilosophical␈α
puzzles␈α
concerning␈α
knowledge␈αand␈α
wants
␈↓ α∧␈↓in␈α
the␈αnear␈α
future,␈αjoint␈α
work␈α
by␈αphilosophers␈α
and␈αartificial␈α
intelligence␈αpeople␈α
can␈α
identify␈αand
␈↓ α∧␈↓solve␈αsome␈αof␈αthe␈αeasier␈αproblems.␈α Sorting␈αout␈αthe␈αproblems␈αinto␈αeasy␈αand␈αhard␈αwill␈αbenefit␈αboth
␈↓ α∧␈↓philosophy␈αand␈α
artificial␈αintelligence␈αand␈α
facilitate␈αbridging␈α
the␈αgap␈αbetween␈α
the␈αabstract␈αworld␈α
of
␈↓ α∧␈↓philosophy␈αand␈αartificial␈αintelligence␈αand␈αpractical␈α
real␈αworld␈αproblems␈αof␈αeducation␈αand␈α
industry.
␈↓ α∧␈↓Cognitive␈αpsychology␈αhas␈αalready␈αbenefitted␈αfrom␈αthe␈αconcreteness␈αof␈αartificial␈αintelligence␈αsystems
␈↓ α∧␈↓and␈αwill␈αalso␈α
benefit␈αfrom␈αthe␈αidentification␈α
and␈αsolution␈αof␈α
the␈αmore␈αstraightforward␈αproblems␈α
of
␈↓ α∧␈↓knowledge, wanting and obligation.

␈↓ α∧␈↓␈↓ αTThe␈α∞scientists␈α∂who␈α∞are␈α∞currently␈α∂scheduled␈α∞to␈α∞participate␈α∂in␈α∞this␈α∞group␈α∂include:␈α∞Professor
␈↓ α∧␈↓John␈αMcCarthy␈α(Chairman),␈αComputer␈αScience,␈αStanford␈αUniversity;␈αDaniel␈αDennett,␈αPhilosophy,
␈↓ α∧␈↓Tufts␈α→University;␈α→John␈α→Haugeland,␈α_Philosophy,␈α→University␈α→of␈α→Pittsburgh;␈α→Patrick␈α_Hayes,
␈↓ α∧␈↓Computer␈α∨Sciences,␈α∨University␈α∨of␈α∨Essex;␈α∨Marvin␈α∨Minsky,␈α∨Computer␈α≡Science-Electrical
␈↓ α∧␈↓Engineering,␈αM.I.T.;␈αRobert␈αMoore,␈αComputer␈αScience-Engineering,␈αSRI␈αInternational;␈αand␈αZenon
␈↓ α∧␈↓W. Pylyshyn, Psychology, University of Western Ontario.

␈↓ α∧␈↓␈↓ αTSome of the issues to be addressed by the study are

␈↓ α∧␈↓␈↓ αT1. What challenges must a system meet, before its behavior can be considered intelligent?

␈↓ α∧␈↓␈↓ αT2.␈αWhat␈αif␈αany␈αis␈αthe␈αrelation␈αbetween␈αthe␈αunsolvability␈αand␈αincompleteness␈αresults␈α
of␈αlogic
␈↓ α∧␈↓and recursive function theory and the possibility of intelligent machines?

␈↓ α∧␈↓␈↓ αT3.␈α∞Is␈α∞intelligent␈α∞behavior␈α∞all␈α∞of␈α∞a␈α∞piece,␈α
or␈α∞can␈α∞a␈α∞system␈α∞posess␈α∞some␈α∞important␈α∞aspects␈α
of
␈↓ α∧␈↓intelligent behavior and not others?

␈↓ α∧␈↓␈↓ αT4. If the latter, what are these components of intelligent behavior?

␈↓ α∧␈↓␈↓ αT5.␈α∃What␈α∀constitutes␈α∃a␈α∀common␈α∃sense␈α∀knowledge␈α∃of␈α∀the␈α∃world␈α∀apart␈α∃from␈α∀specialized
␈↓ α∧␈↓knowledge but sufficient so that specialized knowledge can be added?
␈↓ α∧␈↓␈↓ u2


␈↓ α∧␈↓␈↓ αT6.␈αSpecifically,␈αwhat␈αare␈αthe␈αbasic␈α
facts␈αabout␈αevents␈αoccurring␈αin␈αtime,␈αabout␈α
causality␈αand
␈↓ α∧␈↓about the effects of actions?

␈↓ α∧␈↓␈↓ αT7.␈αIs␈αthere␈αa␈αsense␈αin␈αwhich␈αthe␈αdeep␈αknowledge␈αof␈αthese␈αtopics␈αthat␈αphilosophers␈αare␈αtrying
␈↓ α∧␈↓to obtain can be bypassed?

␈↓ α∧␈↓␈↓ αT8. How much knowledge of the world is required to understand ordinary language?

␈↓ α∧␈↓␈↓ αTThese␈α⊂issues␈α⊂are␈α⊂basic␈α⊂to␈α⊂cognitive␈α⊂science.␈α⊂ The␈α⊂generality␈α⊂of␈α⊂artificial␈α⊃intelligence␈α⊂(AI)
␈↓ α∧␈↓systems␈α
has␈α
been␈α
limited␈α
by␈α
overly␈α
special␈α
views␈α
of␈α
causality,␈α
knowledge,␈α
belief␈α
and␈α
action.␈α This␈α
is
␈↓ α∧␈↓especially␈αtrue␈α
of␈αsystems␈α
using␈αnatural␈α
language,␈αbecause␈α
it␈αis␈α
at␈αpresent␈α
impossible␈αto␈α
give␈αa␈α
clear
␈↓ α∧␈↓statement␈α⊂of␈α⊂the␈α⊂level␈α⊂of␈α⊂understanding␈α⊂such␈α⊂systems␈α⊂must␈α⊂have.␈α⊂ Therefore,␈α⊂the␈α⊂critic␈α⊃has␈α⊂to
␈↓ α∧␈↓invent␈α∞proposed␈α∞counterexamples,␈α∞i.e.␈α∞to␈α∞invent␈α∞some␈α∞request␈α∞that␈α∞a␈α∞human␈α∞with␈α∞common␈α
sense
␈↓ α∧␈↓can do, but the system cannot.

␈↓ α∧␈↓␈↓ αTThe␈αgroup␈αis␈αthe␈αfirst␈αprolonged␈αjoint␈αstudy␈αinvolving␈αboth␈αartficial␈αintelligence␈αresearchers
␈↓ α∧␈↓and␈α∞philosophers.␈α∞ The␈α∞participants␈α∞are␈α∞motivated␈α∞by␈α∞finding␈α∞that␈α∞they␈α∞have␈α∞been␈α∞thinking␈α
on
␈↓ α∧␈↓parallel␈αlines␈αabout␈αa␈αnumber␈αof␈αproblems␈αand␈αare␈αlooking␈αforward␈αto␈αthe␈αopportunity␈αto␈αinteract
␈↓ α∧␈↓more␈α
closely.␈α
 However,␈α
each␈α
of␈α
the␈α
researchers␈αhas␈α
many␈α
problems␈α
about␈α
which␈α
he␈α
has␈αalready
␈↓ α∧␈↓been␈αthinking␈αand␈α
writing␈αand␈αon␈αwhich␈α
he␈αknows␈αhow␈αhe␈α
can␈αmake␈αprogress.␈α
 Therefore,␈αeven
␈↓ α∧␈↓more␈α
than␈α
is␈α
the␈α∞case␈α
with␈α
most␈α
interdisciplinary␈α
studies,␈α∞the␈α
participants␈α
are␈α
uncertain␈α∞of␈α
what
␈↓ α∧␈↓form␈α∞the␈α∞interaction␈α∞can␈α∞best␈α∞take.␈α∞ For␈α∞this␈α∞reason,␈α∞although␈α∞we␈α∞plan␈α∞to␈α∞begin␈α∞with␈α∞a␈α
seminar
␈↓ α∧␈↓meeting␈α
several␈α
times␈α
per␈α∞week␈α
that␈α
will␈α
allow␈α∞the␈α
participants␈α
to␈α
familiarize␈α∞themselves␈α
quickly
␈↓ α∧␈↓with the ideas of the others, no specific output is demanded.

␈↓ α∧␈↓␈↓ αTBecause␈α∞the␈α∞interest␈α∞in␈α∞AI␈α∞and␈α∞philosophy␈α∞extends␈α∞far␈α∞beyond␈α∞the␈α∞seven␈α∞participants,␈α
we
␈↓ α∧␈↓have also planned an October meeting for about 40 invitees.