Documente Academic
Documente Profesional
Documente Cultură
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact support@jstor.org.
American Philosophical Association is collaborating with JSTOR to digitize, preserve and extend access to
Proceedings and Addresses of the American Philosophical Association.
http://www.jstor.org
MACHINESAND THE
MENTAL1.
FredDretske
University
ofWisconsin/Madison
Computersare machinesand thereare a lot of thingsmachinescan't do. But
thereare a lot of thingsI can't do: speakTurkish,
understand
JamesJoyce,or recognize a nasturtium
when I see one. Yet, numerousas are my disabilities,
theydo
not materially
affectmystatusas a thinking
I
lack
being.
specializedskills,knowledge
and understanding,
but nothingthat is essentialto membership
in the societyof
rationalagents. Withmachines,though,and this includesthe mostsophisticated
it is different.
moderncomputers,
thatis essential.
Theydo lacksomething
Or so somesay. And so sayI. In sayingit,though,one should,as a philosopher,
be preparedto say whatis essential,
whatare theconditionsformembership
in this
exclusiveclub. If an abilityto understand
JamesJoyceisn't required,
what,then,
mustone be able to understand?If one doesn'thaveto knowwhatnasturtiums
look
else one mustbe able to identify?Whatmightthisbe? If
like,is theresomething
one is toldthatthereis no specificthingone hasto understand,
orknowbut,
identify
or
other
towards
which
one
have
a
must
of
nonetheless,
something
degree competence,
it is hardto see how to denycomputers
admissionto the club. For eventhesimple
robotsdesignedforhome amusement
and learn. Or so I keep
talk,see, remember
readingin thepromotional
catalogs.Isn'tthisenough?Whynot?
I happen to be one of those philosophers
who, thoughhappyto admitthat
mindscomputer,and in thissenseare computers,
have greatdifficulty
seeinghow
computerscould be minded. I'm not (not now at least) goingto complainabout
the impoverished
innerlifeof the computer--how
theydon'tfeelpain,fear,love or
innerlightof consciousness.
anger. Nor am I goingto talk about the mysterious
For I'm not at all sureone needs feelingsor selfconsciousness
to solveproblems,
and
understand
can't
the
stories.
games,
recognize
play
patterns
Why
purethought,
sort of thingcomputerspurportedly
have, standto ordinarythought,the sortof
strollstandsto a hecticwalkdowna crowdedstreet?
thingwe have,thewaya solitary
The same thing--walking--is
goingon in both cases. It just seemsdifferent
because,
in the lattercase, so muchelse is goingon at the same time. A mathematician's
calculationsare no lessbrilliant,
no lessdeserving
ofclassification
as mental,
certainly
because he or she is blind,deaf and emotionally
in otherwords,
stunted--because,
Addressdeliveredbeforethe Eighty-third
AnnualMeetingof the
*Presidential
WesternDivisionof the AmericanPhilosophicalAssociation,Chicago,Illinois,April
26, 1985.
23
24
APAPROCEEDINGS
thecalculations
occurwithina comparatively
anemicsensoryand emotionalenvironment. Whycan'twe thinkof our machinesas occupyinga positionon the farright
of thismentalcontinuum?Justa bit to therightof StarTrek'sDoctorSpock? We
don't,afterall, denysomeonethecapacityforlovebecausetheycan'tdo differential
or understand
stories
calculus. Whydenythe computertheabilityto solveproblems
becauseit doesn'tfeellove,experience
nausea,or suffer
indigestion?
areat doingwhatmostchildren
Nor am I goingto talkabouthowbad computers
theirnativelanguage,
makeup a storyor appreciate
can do--e.g.,speakandunderstand
a joke. For such comparisons
make it soundlike a competition,
a competition
in
whichhumans,withtheirenormoushead start,and barring
dramaticbreakthroughs
in AI, willremainunchallenged
fortheforeseeable
future.I don'tthinkthecomparisonshouldbe putin thesetermsbecauseI don'tthinkthereis a genuinecompetition
in thisarea at all. It isn't thatthe best machinesare stillat thelevelof two-yearto growup.
olds, requiringonly greaterstoragecapacityand fancierprogramming
Nor shouldwe thinkof themas idiot savants,exhibiting
a spectacularabilityin a
fewisolatedareas,but havingan overallIQ too low forfraternal
association. For
machines,eventhe best of them,don't have an IQ. They don't do whatwe do-at least none of the thingsthat,whenwe do them,exhibitintelligence.And it's
notjust thattheydon't do themthe way we do themor as well as we do them.
They don'tdo themat all. Theydon't solveproblems,
playgames,provetheorems,
recognizepatterns,let alone think,see and remember.They don't even add and
subtract.
to variousinstruTo convinceyou of this,it is usefulto look at our relationship
will not take us far,but it willset
examination
mentsand tools. The preliminary
of what I take to be the fundamental
difference
the stagefor a clearerstatement
betweenmindsandmachines.
of instruments
and toolswe tendto assignthemthecapaciIn our descriptions
ties and powersof the agentswho use them. We oftenthink,or at least talk,of
instruments
and machines--as
artifacts--tools,
tellingus things,recognizing,
sensing,
moments,
and,in general,doingthingsthat,in ourmoreserious,literal,
remembering
to be the provinceof rationalagents. In mostcases thisfigurative
we acknowledge
use of languagedoes no harm. No one is reallyconfused.Thoughwe open doors,
and keys open (locked) doors, no one seemsto worryabout whetherkeysopen
doorsbetterthanwe do, whetherwe are stillaheadin thiscompetition.No one is
to builda fifthgeneration
keythatwillsurpassus in thisenterprise.
Whynot?
trying
Since both keysand peopleopen doors,whydoesn'tit makesenseto ask who does
it better? Because,of course,we all understand
thatdoorsare openedwithkeys.
Thatwe sometimes
We are the agents. The keyis the instrument.
speakof theinto theagent,speakof thekeyas doingwhattheagent
in termsappropriate
strument
thereare
does with the key, shouldnot temptus into supposingthat,therefore,
some thingswe do thatkeyscan also do. We catchfishwithworms;we, not the
worms,catchthefish.
Beforeconcluding,
are,likekeys,merelyfancyinstruhowever,thatcomputers
unableto do what
mentsin our cognitivetool box--and,thus,takenby themselves,
anothercase. Whoreallypicksup thedust,themaid
we can do withthem--consider
thatthemaid
or the vacuumcleaner? Is the vacuumcleanermerelyan instrument
uses to pickup dust? Wellyes,but notquite thewayone usesa keyto opena door
or a hammerto pound a nail. One pushesthevacuumcleaneraroundbut it picks
25
up the dust. In thiscase (unlikethe key case) the question:"Who picksup dust
better:peopleorvacuumcleaners?"does makegoodsense,and theanswer,obviously,
is the vacuumcleaner. We mayneverhavehad any realcompetition
fromkeysfor
but
we
to
seem
have
for
lost
the
race
dust
to vacuum
doors,
opening
pickingup
cleaners.
Whatsuch examplesrevealis thattheagent-instrument
is no certain
distinction
Wedo things.Machines
guideto who or whatis to be givencreditfora performance.
do things.Sometimeswe do thingswithmachines.Whogetsthe creditdependson
what is done and how it is done. To ask whethera simplepocketcalculatorcan
orwhether
it is we whomultiply
is to ask,whether,
withthecalculator
reallymultiply
relativeto thistask,the agent-instrument
relationis morelike our use of a keyin
openinga dooror morelikeouruse of a vacuumcleanerin picking
up dust.
ourcomputational
Well,then,arecomputers
keys? Or aretheymorelikevacuum
do thecomputational
cleaners?Do theyliterally
do without
tasksthatwe sometimes
thembut do it better,faster,and morereliably?Thismaysoundlikea rathersimplemindedway to approachtheissueof mindsand machines,
butunlessone getsclear
about the relatively
simplequestionof who does thejob, thepersonor the pocket
one is unlikelyto makesuchprogress
calculator,in addingup a columnof figures,
in penetrating
the morebaffling
machines
questionof whethermoresophisticated
exhibit(or willsomeday) someof thegenuinequalitiesofintelligence.
For I assume
thatif machinescan reallyplay chess,provetheorems,
a text,diagnose
understand
an illness,and recognizean object--allachievements
thatare routinely
creditedto
these
modernmachinesby sobermembers
of theartificial
intelligence
community--if
are literallytrue,thento thatdegreetheyparticipate
in theintellectual
descriptions
enterprise.To thatdegreetheyare minded.To thatextenttheybelongin the club
howevermuchwe, withour prejudicein favorofbiologicallook-alikes,
maycontinue
to denythemfulladmission.
So let me beginwitha naivequestion:Can computersadd? We maynot feel
if thisis all theycan do. Nevertheless,
if theydo eventhismuch,
verythreatened
thenthe barriersseparating
mindand machinehave been breachedand thereis no
reasonto thinktheywon'teventually
be removed.
is an attemptto showthatwhatever
it is thatcomputers
The following
argument
are doingwhenwe use themto answerour arithmetical
questions,it isn'taddition.
Additionis an operationon numbers.We add 7 and 5 to get12,and 7, 5 and 12 are
on numnumbers.The operationscomputersperform,
are not operations
however,
bers. At best,theyare operationson certainphysicaltokensthatstandfor,or are
don'tadd.
as standing
for,thenumbers.Therefore,
computers
interpreted
In thinking
about this argument(longerthanI care to admit)I decidedthat
therewas something
rightabout it. And something
wrong. Whatis rightaboutit
is the perfectly
valid(and relevant)distinction
it invokesbetweena representation
and what it represents,
betweena signand what it signifies,
betweena symboland
its meaningor reference.We havevariouswaysof representing
or designating
the
numbers.The writtennumeral"2" standsforthenumber2. So does "two." Unless equippedwithspecialpatternrecognition
machinesare notprepared
capabilities,
to handlethese particularsymbols(the symbolsappear on the keyboardforour
convenience).But theyhave theirown systemof representation:
open and closed
theorientation
of magneticfields,the distribution
of holeson a card. But
switches,
26
APAPROCEEDINGS
theformof representation,
to operations
whatever
themachineis obviouslyrestricted
on the symbolsor representations
themselves.It has no access,so to speak,to the
to thethingstherepresentations
to thenumbers.
meaningof thesesymbols,
represent,
Wheninstructedto add two numbersstoredin memory,the machinemanipulates
in some electromechanical
wayuntilit arrivesat anotherrepresentarepresentations
that (if thingsgo right)standsforthe sum of what the firsttwo
tion--something
in contrast
stood for. At no pointin the proceedings
do numbers,
representations
to numerals,
in
to
two
to perform
add
one
has
involved.
And
order
numbers,
if,
get
thenwhatthe computeris doingis not
some operationon the numbersthemselves,
additionat all.
This argument,
as I am sureeveryoneis aware,showstoo much. It showsthat
in or by ourcentral
we don't add either. For whatever
operationsmaybe performed
it quiteclearlyisn'tan operationon the
nervoussystemwhenwe add two numbers,
numbersthemselves.Brainshavetheirown codingsystems,
theirownway of representingthe objects(includingthe numbers)about whichits (or our) thoughtsand
thana computer.
calculationsare directed. In thisrespecta personis no different
have
different
of
the
Biologicalsystemsmay
ways representing objectsof thought,
but they,likethecomputer,
are necessarily
limitedto manipulating
theserepresentations. This is merelyto acknowledgethe natureof thoughtitself.It is a vicarious
about two
business,a symbolicactivity.Addingtwo numbersis a way of thinking
and thinking
about X and Y is not a way of pushingX andY around.It is
numbers,
a wayof pushing
aroundtheirsymbolic
representatives.
Whatis wrongwiththe argument,
thatin orderto add
then,is theassumption
two numbers,
a systemmustliterally
someoperationon thenumbersthemperform
selves. Whattheargument
is thatin orderto carryout
shows,if it showsanything,
arithmetical
thenumbersand
operations,a systemmusthave a way of representing
the capacityfor manipulating
these representations
in accordancewitharithmetic
whatcomputers
have?
principles.Butisn'tthisprecisely
I have discussedthisargument
at some lengthonly to make the pointthatall
or naturalbiologicalsystems)will necescognitiveoperations(whetherby artifacts
sarilybe realizedin some electrical,chemicalor mechanical
operationoverphysical
structures.(Or, ifmaterialism
isn'ttrue,theywillbe realizedin or bytransformations
of mind-stuff.)
nature
This factalone doesn'ttell us anything
about thecognitive
of the operationsbeingperformed--whether,
forinstance,it is an inference,
a thought
or the takingof a squareroot. For whatmakestheseoperationsintothoughts,
incalculationsis, amongotherthings,
themeaningor,ifyou
ferences,or arithmetical
overwhichtheyare performed.To think
prefer,the semanticsof thosestructures
about thenumber7 or yourcousinGeorge,youneedn'tdo anything
withthenumber
7 or your cousin George,but you do need the internalresourcesforrepresenting
7 and Georgeand the capacityformanipulating
theserepresentations
in waysthat
standforactivities
andconditions
of thethings
beingrepresented.
This shouldbe obviousenough. Openingand closingrelaysdoesn'tcountas
addition,or as movesin a chessgame,unlesstherelays,or theirvariousstates,stand
fornumbersand chessmoves. But whatmaynot be so obviousis thatthesephysical
activitiescannotacquirethe relevantkindof meaningmerelyby assigning
theman
to or for us. Unlessthe symbols
by lettingthemmean something
interpretation,
meansomething
to thesystemmanipulating
them,theirmeaning,
beingmanipulated
27
it is, is irrelevant
to evaluating
whatever
whatthesystemis doingwhenit manipulates
them.2 I cannotmakeyou,someone'sparrot,or a machinethinkaboutmycousin
in accordancewithwhichthis
George,or the number7, just by assigning
meanings
is what your (the parrot's,the machine's)activitiesstandfor. If thingswere this
thinkaboutmycousinGeorge.Everything
easy,I couldmakea taperecorder
depends
on whetherthisis themeaningtheseeventshaveto you, theparrot,or themachine.
of symbolsis
Despite some people's tendencyto thinkthatthe manipulation
as "addingnumbers,"
itselfa wondrousfeat,worthyof such inflateddescriptions
out its nextmove" the processis, in fact,ab"drawingconclusions,"or "figuring
a symI once watcheda gerbilmanipulate
solutelydevoidof cognitive
significance.3
thatI, butnotthe
standards--standards
bol, a symbolthat,accordingto conventional
stoodformybankbalance. I didn'thavetheslightest
gerbil--understood,
temptation
to see in thissymbolmanipulation
ofspecial
(actuallyconsumption)
processanything
if
Even
I
trained
a
fleet
of
to
in
some
significance.
gerbils arrangesymbols
computationallysatisfying
way (e.g., to balancemy checkbook),I don't thinktheyshould
be creditedwithbalancingmy checkbook. I would merelybe usingthegerbilsto
balancemycheckbookin thewayI use wormsto catchfish.
whata systemis doingwhenit manipulates
To understand
itis necessary
symbols,
to know,not just what these symbolsmean,whatinterpretation
theyhave been,
or can be, assigned,but whattheymean to the systemperforming
theoperations.
JohnSearle and Ned Block have dramatizedthispoint.4 Searle,forinstance,asks
one to imaginesomeonewho understands
no Chinesemanipulating
Chinesesymbols
in accordancewithrulesexpressedin a languagehe does understand.Imaginethe
rulescleverlyenoughdesignedso thatthispersoncan carryon a correspondence
in
to (written)Chinesequestionswith (written)Chineseanswers
Chinese--responding
in a way thatis indistinguishable
fromtheperformance
of a nativespeakerofChinese.
Clearly,thougha correspondent
mightnot be able to discoverthisfact,thesymbol
himselfdoesn'tunderstand
Chinese. Nor does thesystemof whichhe
manipulator
is a part. Understanding
Chineseis not just a matterof manipulating
meaningful
to thesystem
symbolsin someappropriate
way. Thesesymbolsmustmeansomething
theoperations.
performing
This shouldnot be takento implythatmachinescannotserveas usefulmodels
forcognitive
use in cognitive
processes.On thecontrary.Theirprevalent
psychology
indicatesotherwise. Whatit does implyis thatthe machinesdo not literallydo
whatwe do whenwe engagein thoseactivitiesforwhichtheyprovidean effective
model. Computersimulations
of a hurricane
do not blow treesdown. Whyshould
of problemsolvingmustthemselves
solve
anyonesupposethatcomputersimulations
problems?
But how does one builda systemthatis capablenot onlyof performing
operations on (or with) symbols,but one to whichthesesymbolsmean something,a
machinethat,in thissense,understands
themeaningof thesymbolsit manipulates?
Onlywhenwe can do thiswillwe havemachinesthatnot onlyproducemeaningful
in producing
thatoutputbearthemarkof the
output,but machineswhoseactivities
mental. Only thenwill we have machinesthatwe can not onlyuse to balanceour
checkbook,but machinesthatwill do it forus, machinesthatwill not only print
out answersto ourquestions,butmachinesthatwillanswerourquestions.
One thingseems reasonablyclear: if the meaningof the symbolson whicha
machineperforms
its operationsis a meaningwhollyderivedfromus, its users--if
28
APAPROCEEDINGS
a
it is a meaningthat we assignthe variousstatesof the machineand, therefore,
thewaythesesymbolsare promeaningthatwe can changeat willwithoutaltering
thereis no way the machinecan acquireundercessedby the machineitself--then
no way thesesymbolscan have a meaningto the machineitself. Unless
standing,
a meaningtheypossess
thesesymbolshave whatwe mightcall an intrinsic
meaning,
and purposes,thenthismeanintentions
of our communicative
whichis independent
to assessingwhatthe machineis doingwhenit manipulates
ing mustbe irrelevant
them. The machineis processing
(to us) symbols,to be sure,but the
meaningful
of whattheymean--hence,
way it processesthemis quite independent
nothingthe
machinedoes is explicablein termsof the meaningof the symbolsit manipulates
and data
or, indeed,of theirevenhavinga meaning. Giventherightprogramming
base, we can contriveto make the sentencesa machineproducesanswersto our
ourquestionsthanis an autoquestions.But themachineitselfis no moreanswering
in the bankingindustry)
maticteller(now so prevalent
embezzlingmoneywhenit
ouraccount.
keepsourdepositwithoutcrediting
to approximate
of genuinecognitive
In order,therefore,
significance,
something
thatbearsa mark,if notall themarks,of the
in orderto givea machinesomething
mustbe givena meaningof theirown,
mental,the symbolsa machinemanipulates
of theiruser'spurposesandintentions.
a meaningthatis independent
Onlybydoing
thiswill it becomepossibleto makethemeaningof thesesymbolsrelevantto what
themachinedoes withthem,possible,in otherwords,to makethemachinedo someto makethesesymbols
thingbecauseof whatits symbolsmean,possible,therefore,
to themachineitself.
meansomething
And how mightthisbe done? In thesameway,I submit,thatnaturearranged
it in our case. We mustput the computerinto the head of a robot,into a larger
thatenable
theperceptual
resources,
systemthathas thekindof sensorycapabilities,
in Paul Grice'snaturalsense
whatgoes on insidethe computerto mean something,
of meaning5,about whatgoes on outsidethe computer.The elementsoverwhich
the computerperforms
its operationswill thenhave a meaningthatis independent
in thesamewaythe
of itsusers. Theywillthenmeansomething
of theconventions
in
the
of
means
electricalactivity
a
needle
something
regarding
swing
galvanometer
meanssomething
the circuitto whichit is connected,the way expandingmercury
the way a voltagespike in our visualcortex
about the surrounding
temperature,
on theretina.Thiskind
of lightimpinging
about the distribution
meanssomething
It is thekindof meaningwe associate
calledinformation.6
of meaningis sometimes
thekindof meaning
withreliablesignsand trustworthy
indicators,
possessedbydark
and theelectrical
smoke,acousticvibrations,
clouds,shadows,prints,leaf patterns,
a
robot
and
thedisembodied
between
in
the
The
difference
sensory
activity
pathways.
is thattheformer,
unlikethe
and laboratories
computerfoundin our officebuildings
latter,have symbolsystemsthatare also signsystems:signsbeingsymbolshavinga
of whatwe mightsay or thinktheymean. The only
meaningquite independent
derivedfromthearrayof pressure
in
is themeaning
most
intrinsic
meaning
computers
on its keyboard. The activitiesin the computermay meana
sensitivetransducers
move to KB-3 to us, but all theymeanto thecomputeris thatkey37 has been depressed.
assothekindof meaning
real information,
Thisis onlyto say thatinformation,
to theoperationofhighspeeddigitalcomputers
ciatedwithnaturalsigns,is irrelevant
to theoperationof livingsystems.If a sea snaildoesn't
in a wayit is notirrelevant
29
30
APAPROCEEDINGS
31
APAPROCEEDINGS
32
forinterpreting
signs,understanding
meanings,
possessedof theconceptualresources
his or her behaviorin the lightof experiencedcorrelations.Thisis
and modifying
at almost
can be illustrated
true,but irrelevant.For the verysame phenomenon
an
level
at
It
in
occurs.
which
is,
fact,
level,
learning
merely
biological
every
every
thewaya
theorists
describeas thecontingencies
instanceof whatlearning
modifying
systemprocesses,and hence respondsto, the internalsignsforstimulusconditions.
earlierchangesthewayitprocessessignsbyexposing
Eventhelowlysnailmentioned
the meaningof thesesigns. And it is, surely,the
it to the correlations
constituting
factthatour internalstatesare correlatedwithcertainkindsof externalconditions
the ultimateoutcomeof the motoractivitiesproducedby
thathelpsto determine
what
thathelp to determine
theseinternalstates. It is the correlations,
therefore,
kindof feedbackwe receivedfromsuch activities
and,hence,thelikelihoodof our
not
the correlations,
repeatingthemin the same circumstances.It is, therefore,
behavior.
internal
that
the
correlates,
Learning,
explain--learned
merely
shape--hence,
with
in fact,is a processin whichthemeaningof internalsigns(i.e., theircorrelation
not (merely)thesignsthemselves,
howthese
externalconditions),
helpsto determine
signsare exploitedfor purposesof motorcontrol. For such systemsthe internal
thismeaningaffectstheway thesystemmanagesthese
signsnot onlyhavemeaning,
to thesystemin which
signs;and it is in thissensethatthe signsmean something
theyoccur.
betweenthe signprocessing
difference
This, it seemsto me, is a fundamental
that
various
It
is
a
difference
of
helpsexplainwhyit seemsso
systems.
capabilities
naturalto say of some of them(humanbeingsand some animals)but not others
meansomething
thatthesymbolstheymanipulate
(machinesand simpleorganisms)
ourconviction
thatunderlies
I submit,
to the symbolmanipulator.It is a difference,
aregenuinethinkers
of simpleorganisms,
thatwe, but not themachinesand a variety
is a long and
thisdifference
of thoughts. Whatgivesus the capacitiesunderlying
ourmultiple
I think,issuesin learning
sensory
theory,
complicatedstory. It involves,
accessto thethingswe requireto satisfyour needs,and thekindof feedbackmechainternalsignsby the
nismswe possessthatallow us to modifyhow we manipulate
have produced. But this,clearly,is a
kind of resultsour previousmanipulations
All I have
not fromphilosophers.
storythatwe expectto hearfromneurobiologists,
for
been tryingto tell is a simplerstory,a storyabout the entrancerequirements
admissionto the club. I leave it to othersto worryabout how different
systems
theserequirements.
manage,eachin theirownway,to satisfy
Footnotes
and manyusefulsuggestions.
1. My thanksto DennyStampeforcarefulcriticism
I also want to acknowledgethe help givenme by Fred Adamsand the other
scepticsin theaudienceat AugustanaCollegewhereI readan earlydraftof this
paper. They convincedme thatthe draftI read themwas earlierthanI ever
suspected.
that,according
2. This is whatHaugelandcalls"originalintentionality,"
something
themselves
to Haugeland,computersdon't have: "'To put it bluntly:computers
books
more
than
their
tokens
don't meananything
onlymean
do)--they
(any
by
on theotherhand,is intentional
whatwe say theydo. Genuineunderstanding,
3.
4.
5.
6.
33
fromsomething
in its own right"and not derivatively
else. MindDesign,John
Books;Montgomery,
Vt., 1981, pp. 32-33. A number
Haugeland(ed.), Bradford
of authorshave made essentially
thispointin theirown way;e.g.,Jerry
Fodor,
"Tom Swiftand His ProceduralGrandmother,"
Cognition,6 (1978), reprinted
in Representations,
1981; HilaryPutnam,"Brainsin a Vat,"
MIT/Bradford,
Reason, Truthand History,Cambridge
Press,1981, pp. 10-11; Rob
University
Cummins,TheNatureof Psychological
1983, p. 94;
Explanation,MIT/Bradford,
Tyler Burge,"BeliefDe Re," The Journalof Philosophy,LXXILV, 6 (1977);
JohnSearle,"Minds,Brainsand Programs,"
The Behavioraland BrainSciences
3:3 (1980).
In explainingwhyhe thinkscomputers
can (or will someday),MarvinMinsky
Can't,"AI Magazine,Fall 1982), seemsmost
(in "WhyPeopleThinkComputers
forinstance,withthefactthat"computers
can manipulate
impressed,
symbols."
JohnSearle,"Minds,Brainsand Programs,"
TheBehavioraland BrainSciences,
3:3 (1980); Ned Block,"TroubleswithFunctionalism,"
in Wade Savage(ed.),
Minnesota
Perceptionand Cognition:Issues in the Foundationsof Psychology,
Studiesin thePhilosophy
ofScience,Vol. 9, Minneapolis,
Minn.;1978.
Paul Grice, "Meaning,"PhilosophicalReview, vol. 66 (1957), pp. 377-388.
See my Knowledgeand the Flow of Information,
MIT/Bradford:
Cambridge,
Mass.,1981.