␈↓␈↓↓␈↓α␈↓ β
␈↓αARTIFICIAL INTELLIGENCE AND THE STANFORD AI LABORATORY␈↓

␈↓ ↓H

␈↓ ↓H
␈↓ ↓H␈α?␈α⊃This␈α∪is␈α∪a␈α∪general␈α∪article␈α∪for␈α∩answering␈α∩inquiries␈α∩about␈α∩artificial␈α∩intelligence␈α∩and␈α∩the␈α∩Stanford
␈↓ ↓HArtificial␈α
Intelligence␈α
Laboratory.␈α
We␈α
hope␈α
it␈α
answers␈α
your␈α
questions.

␈↓ ↓H␈α?␈α⊃␈↓↓Artificial␈α
intelligence␈↓␈α
is␈α
the␈α
name␈α
given␈α
to␈α
the␈α
study␈α
of␈α
intellectual␈α
processes␈α
and␈α
how␈αcomputers
␈↓ ↓Hcan␈α⊂be␈α⊂made␈α⊂to␈α⊂carry␈α⊂them␈α⊂out.␈α∂Most␈α∂workers␈α∂in␈α∂the␈α∂field␈α∂believe␈α∂that␈α∂it␈α∂will␈α∂be␈α∂possible␈α∂to␈α∂program
␈↓ ↓Hcomputers␈αto␈αcarry␈αout␈αany␈αintellectual␈αprocess␈αnow␈αdone␈αby␈αhumans.␈αHowever,␈αalmost␈α
all␈α
agree␈α
that␈α
we␈α
are
␈↓ ↓Hnot␈αvery␈αclose␈αto␈αthis␈α
goal␈α
and␈α
that␈α
some␈α
fundamental␈α
discoveries␈α
must␈α
be␈α
made␈α
first.␈α
Therefore,␈α
work␈α
in␈α
AI
␈↓ ↓Hincludes␈αtrying␈αto␈αanalyze␈αintelligent␈αbehavior␈αinto␈αmore␈αbasic␈αdata␈αstructures␈α
and␈α
processes,␈α
experiments␈α
to
␈↓ ↓Hdetermine␈αif␈αprocesses␈αproposed␈αto␈αsolve␈αsome␈αclass␈αof␈αproblems␈αreally␈αwork,␈αand␈αattempts␈αto␈αapply␈αwhat␈αwe
␈↓ ↓Hhave␈α
found␈α
so␈α
far␈α
to␈α
practical␈α
problems.

␈↓ ↓H␈α?␈α⊃The␈α
idea␈α
of␈α
intelligent␈αmachines␈αis␈αvery␈αold␈αin␈αfiction,␈αbut␈αpresent␈αwork␈αdates␈αfrom␈αthe␈αtime␈αstored
␈↓ ↓Hprogram␈αelectronic␈αcomputers␈αbecame␈αavailable␈αstarting␈αin␈α1949.␈αAny␈αbehavior␈αthat␈αcan␈αcarried␈αout␈αby␈αany
␈↓ ↓Hmechanical␈αdevice␈αcan␈αbe␈αdone␈αby␈αa␈αcomputer,␈αand␈αgetting␈αa␈αparticular␈αbehavior␈αis␈αjust␈αa␈αmatter␈αof␈αwriting
␈↓ ↓Ha␈αprogram␈αunless␈αthe␈αbehavior␈αrequires␈αspecial␈αinput␈αand␈α
output␈α
equipment.␈α
It␈α
is␈α
perhaps␈α
reasonable␈α
to␈α
date
␈↓ ↓HAI␈α⊃from␈α⊃A.M.␈α⊃Turing's␈α⊃1950␈α⊃paper,␈α⊂␈↓↓Computing␈α⊂Machinery␈α⊂and␈α⊂Intelligence␈↓.␈α⊂Newell,␈α⊂Shaw␈α⊂and␈α⊂Simon
␈↓ ↓Hstarted␈αtheir␈αgroup␈αin␈α1954␈αand␈αthe␈αM.I.T.␈αArtificial␈αIntelligence␈αLaboratory␈αwas␈αstarted␈αby␈αMcCarthy␈αand
␈↓ ↓HMinsky␈α
in␈α
1958.

␈↓ ↓H␈α?␈α⊃Early␈α∞work␈α∞in␈α∞AI␈α∞included␈α∞programs␈α∞to␈α∞play␈α
games␈α
like␈α
chess␈α
and␈α
checkers␈α
and␈α
kalah␈α
and␈α
go.␈α
The
␈↓ ↓Hsuccess␈α∩of␈α∩these␈α∩programs␈α∩was␈α∩related␈α∩to␈α∩the␈α∩extent␈α∩that␈α⊃human␈α⊃play␈α⊃of␈α⊃these␈α⊃games␈α⊃makes␈α⊃use␈α⊃of
␈↓ ↓Hmechanisms␈α⊃we␈α⊃didn't␈α⊃understand␈α⊃well␈α⊃enough␈α⊃to␈α⊃program.␈α⊃If␈α⊃the␈α⊃game␈α⊃requires␈α⊃only␈α⊂well␈α⊂understood
␈↓ ↓Hmechanisms,␈α∩computers␈α∩play␈α∩better␈α∩than␈α∩humans.␈α∩Kalah␈α⊃is␈α⊃such␈α⊃a␈α⊃game.␈α⊃The␈α⊃best␈α⊃rating␈α⊃obtained␈α⊃in
␈↓ ↓Htournament␈α⊂play␈α⊂by␈α⊂a␈α⊂chess␈α⊂program␈α⊂so␈α⊂far␈α⊂is␈α⊂around␈α⊂1700␈α⊂which␈α⊂is␈α⊂a␈α⊂good␈α⊂amateur␈α⊂level.␈α⊂The␈α∂chess
␈↓ ↓Hprogrammers␈α
hope␈α
to␈α
do␈α
better.

␈↓ ↓H␈α?␈α⊃Another␈αearly␈αproblem␈αdomain␈αwas␈αtheorem␈αproving␈αin␈αlogic.␈αThis␈αis␈α
important␈α
for␈α
two␈α
reasons.␈α
First,
␈↓ ↓Hit␈αprovides␈αanother␈αarea␈αin␈αwhich␈αour␈α
accomplishments␈α
in␈α
artificial␈α
intelligence␈α
can␈α
be␈α
compared␈α
with␈α
human
␈↓ ↓Hintelligence.␈α∪Again␈α∪the␈α∪results␈α∪obtained␈α∪depend␈α∪on␈α∩what␈α∩intellectual␈α∩mechanisms␈α∩the␈α∩theorem␈α∩proving
␈↓ ↓Hrequires,␈αbut␈αin␈αgeneral␈αthe␈αresults␈αhave␈αnot␈αbeen␈αas␈αgood␈αas␈αwith␈αgame␈αplaying.␈αThis␈αis␈αpartly␈αbecause␈αthe
␈↓ ↓Hmathematical␈α⊃logical␈α⊃systems␈α⊃available␈α⊃were␈α⊃designed␈α⊃for␈α⊃proving␈α⊃metatheorems␈α⊃about␈α⊃rather␈α⊃than␈α⊂for
␈↓ ↓Hproving␈αtheorems␈αin.␈αThe␈αsecond␈αreason␈αwhy␈αtheorem␈αproving␈αis␈αimportant␈αis␈αthat␈αlogical␈αlanguages␈αcan␈αbe
␈↓ ↓Hused␈αto␈αexpress␈αwhat␈αwe␈αwish␈αto␈αtell␈αthe␈αcomputer␈αabout␈αthe␈αworld,␈αand␈αwe␈αcan␈αtry␈αto␈αmake␈αit␈αreason␈αfrom
␈↓ ↓Hthis␈αwhat␈αit␈αshould␈αdo␈αto␈αsolve␈αthe␈αproblems␈αwe␈αgive␈αit.␈αIt␈αturns␈αout␈αto␈αbe␈αquite␈αdifficult␈αto␈αexpress␈αwhat
␈↓ ↓Hhumans␈αknow␈αabout␈αthe␈αworld␈αin␈αthe␈αpresent␈αlogical␈αlanguages␈αor␈αin␈αany␈αother␈αway.␈αSome␈αof␈αwhat␈αwe␈αknow
␈↓ ↓His␈αreadily␈αexpressed␈αin␈αnatural␈αlanguage,␈αbut␈αmuch␈αbasic␈αinformation␈αabout␈αcausality␈αand␈αwhat␈αmay␈αhappen
␈↓ ↓Hwhen␈αan␈αaction␈αis␈αtaken␈αis␈αnot␈αever␈αexplicitly␈αstated␈αin␈αhuman␈αspeech.␈αThis␈αgives␈αrise␈αto␈αthe␈αrepresentation
␈↓ ↓Hproblem␈αof␈αdetermining␈αwhat␈αis␈αknown␈αin␈αgeneral␈αabout␈αthe␈αworld␈αand␈αhow␈αto␈αexpress␈αit␈αin␈αa␈αform␈αthat␈αcan
␈↓ ↓Hbe␈αused␈αby␈αthe␈αcomputer␈αto␈αsolve␈αproblems.␈αAs␈αthis␈αis␈αwritten␈α-␈αsummer␈α1973␈α-␈αthe␈αrepresentation␈αproblem
␈↓ ↓Hseems␈α
to␈α
be␈α
the␈α
key␈α
problem␈α
in␈α
artificial␈α
intelligence.

␈↓ ↓H␈α?␈α⊃The␈αStanford␈αArtificial␈αIntelligence␈αLaboratory␈αwas␈αstarted␈αin␈α1963␈αand␈αmoved␈αto␈αits␈αpresent␈αlocation
␈↓ ↓Hat␈α
1600␈α
Arastradero␈α
Road,␈α
Palo␈α
Alto␈α
in␈α1966.␈αSince␈αthe␈αbeginning,␈αour␈αwork␈αhas␈αbeen␈αmostly␈αsupported␈αby
␈↓ ↓Hthe␈αAdvanced␈α
Research␈α
Projects␈α
Agency␈α
of␈α
the␈α
Defense␈α
Department,␈α
but␈α
we␈α
have␈α
also␈α
been␈α
supported␈α
by␈α
the
␈↓ ↓HNational␈α
Science␈α
Foundation,␈α
the␈α
National␈αAeronautics␈αand␈αSpace␈αAgency,␈αthe␈αNational␈αInstitute␈αof␈αMental
␈↓ ↓HHealth,␈α
and␈α
private␈α
foundations.

␈↓ ↓H␈α?␈α⊃Our␈αresearch␈αgoals␈α
include␈α
all␈α
areas␈α
of␈α
artificial␈α
intelligence,␈α
mathematical␈α
theory␈α
of␈α
computation␈α
(the
␈↓ ↓Hproblem␈α∃of␈α∃specifying␈α∃properties␈α∀of␈α∀computer␈α∀programs␈α∀and␈α∀proving␈α∀that␈α∀the␈α∀programs␈α∀meet␈α∀their
␈↓ ↓Hspecifications),␈α⊃studies␈α⊃in␈α⊃natural␈α⊃language,␈α⊂and␈α⊂studies␈α⊂in␈α⊂time-sharing␈α⊂and␈α⊂other␈α⊂aspects␈α⊂of␈α⊂computer
␈↓ ↓Hsystems.␈α
However,␈α
the␈α
main␈α
work␈α
of␈α
the␈α
laboratory␈α
since␈α
its␈α
inception␈α
has␈α
included␈α
the␈α
following:

␈↓ ↓H␈α?␈α⊃1.␈α
Computer␈α
vision.␈α
Images␈α
are␈α
obtained␈α
from␈α
a␈α
television␈α
camera␈αand␈αare␈αprocessed␈αto␈αdescribe␈αthe
␈↓ ↓Hscene␈α∞in␈α∞ways␈α∞appropriate␈α∞to␈α∞the␈α∞purpose␈α∞of␈α∞the␈α∞program.␈α∞These␈α∞purpose␈α∞include␈α∞manipulation,␈α∞driving␈α∞a
␈↓ ↓Hvehicle,␈α
and␈α
simply␈α
checking␈α
our␈α
understanding␈α
of␈α
perceptual␈α
mechanisms.

␈↓ ↓H␈α?␈α⊃2.␈α∂Manipulation.␈α∂Programs␈α∂have␈α∂been␈α∂and␈α∞are␈α∞being␈α∞developed␈α∞to␈α∞assemble␈α∞objects␈α∞out␈α∞of␈α∞parts.␈α∞A
␈↓ ↓Hrecent␈α
program␈α
assembles␈α
the␈α
water␈α
pump␈α
from␈α
a␈α
Model␈α
T␈α
Ford.

␈↓ ↓H␈α?␈α⊃3.␈α
Driving␈α
a␈α
vehicle.

␈↓ ↓H␈α?␈α⊃4.␈α∂Theorem␈α∂proving.␈α∂Programs␈α∂using␈α∂J.␈α∞Alan␈α∞Robinson's␈α∞resolution␈α∞method␈α∞of␈α∞proving␈α∞theorems␈α∞in
␈↓ ↓Hfirst␈αorder␈αlogic␈αare␈αused␈αto␈αprove␈αtheorems␈αin␈αmathematics,␈αto␈αprove␈αproperties␈αof␈αcomputer␈αprograms,␈α
and
␈↓ ↓Hto␈α
generate␈α
computer␈α
programs␈α
having␈α
prescribed␈α
properties.

␈↓ ↓H␈α?␈α⊃5.␈α∞Mathematical␈α∞theory␈α∞of␈α∞computation.␈α∞Methods␈α∞for␈α∞proving␈α
properties␈α
of␈α
programs␈α
are␈α
developed.
␈↓ ↓HPrograms␈αfor␈αchecking␈αproofs␈αin␈αfirst␈αorder␈αlogic␈αand␈αin␈αa␈αspecial␈αlogic␈αof␈αcomputable␈αfunctions␈αhave␈αbeen
␈↓ ↓Hdeveloped.

␈↓ ↓H␈α?␈α⊃6.␈α
Game␈α
playing.␈α
Some␈α
work␈α
in␈α
checkers,␈α
chess␈α
and␈α
go␈α
has␈α
been␈α
done,␈αbut␈αwe␈αare␈αnot␈αactive␈αin␈αthis
␈↓ ↓Hfield␈α
at␈α
present.

␈↓ ↓H␈α?␈α⊃7.␈α
Speech␈α
recognition.

␈↓ ↓H␈α?␈α⊃8.␈α∩Computer␈α∩graphics.␈α∩Programs␈α∩have␈α∩been␈α∩written␈α∩to␈α∩allow␈α∩a␈α∩computer␈α∩designer␈α∩to␈α∩put␈α∩logic
␈↓ ↓Hdiagrams␈αin␈αthe␈αcomputer␈αand␈αdiagrams␈αfor␈αprinted␈αcircuit␈αboards.␈αAn␈αintegrated␈αsystem␈αchecks␈αconsistency
␈↓ ↓Hof␈α∩the␈α∩various␈α∩diagrams,␈α∩permits␈α∩changes␈α∩to␈α∩be␈α∩made␈α∩easily,␈α∩and␈α∩produces␈α⊃output␈α⊃for␈α⊃the␈α⊃automatic
␈↓ ↓Hmanufacture␈α
of␈α
printed␈α
circuit␈α
boards␈α
and␈α
for␈α
automatic␈α
wire␈α
wrapping␈α
machines.

␈↓ ↓H␈α?␈α⊃A␈α
separate␈α
project␈α
allows␈αthe␈αdesign␈αof␈αthree␈αdimensional␈αobjects␈αand␈αtheir␈αdisplay␈αin␈αvarious␈αviews,
␈↓ ↓Hand␈α
the␈α
simulation␈α
of␈α
their␈α
motion␈α
and␈α
operation␈α
of␈α
joints.

␈↓ ↓H␈α?␈α⊃9.␈α
There␈α
is␈α
also␈α
development␈α
of␈α
time-sharing␈α
techniques␈αespecially␈αfor␈αdisplay␈αoriented␈αtime-sharing
␈↓ ↓Hsystems.


␈↓ ↓H␈α?␈α⊃The␈α∂computer␈α∂facilities␈α∂of␈α∂the␈α∂laboratory␈α∞comprise␈α∞PDP-10␈α∞and␈α∞PDP-6␈α∞processors,␈α∞256K␈α∞words␈α∞of
␈↓ ↓Hcore,␈αa␈αswapping␈αdisk,␈αan␈αIBM␈α3330␈αfor␈αfile␈αstorage,␈αa␈αsix␈αterminal␈αvector␈αtype␈αkeyboard␈αand␈αdisplay␈αsystem
␈↓ ↓Hand␈αa␈α60␈αterminal␈αraster␈αtype␈αdisplay␈αsystem,␈αA-D␈αand␈αD-A␈αconverters,␈αa␈αconnection␈αto␈αthe␈αARPA␈αnetwork,
␈↓ ↓Hand␈α∞a␈α∞few␈α∞external␈α∞teletype␈α∞speed␈α∞lines.␈α
Connected␈α
to␈α
the␈α
system␈α
are␈α
television␈α
cameras,␈α
mechanical␈α
arms,
␈↓ ↓Hand␈α
a␈α
computer␈α
controllable␈α
cart␈α
with␈α
a␈α
TV␈α
camera.

␈↓ ↓H␈α?␈α⊃The␈α∞personnel␈α∞of␈α∞the␈α∞laboratory␈α∞include␈α
faculty␈α
and␈α
students␈α
of␈α
the␈α
Computer␈α
Science␈α
Department
␈↓ ↓Hand␈α∞a␈α∞few␈α∞other␈α∞departments␈α∞at␈α∞Stanford␈α∞University,␈α∞the␈α∞professional␈α∞staff,␈α
and␈α
some␈α
research␈α
associates.
␈↓ ↓HPotential␈αgraduate␈αstudents␈αwishing␈αto␈αspecialize␈αin␈αartificial␈αintelligence␈αand␈αdo␈αresearch␈αin␈αthe␈αlaboratory
␈↓ ↓Hshould␈α
usually␈α
apply␈α
for␈α
admission␈α
to␈α
the␈α
Computer␈α
Science␈α
Department␈α
at␈α
Stanford␈α
University.␈α
Potential
␈↓ ↓Hresearch␈α
associates␈α
should␈α
apply␈α
directly␈α
to␈α
the␈α
laboratory.

␈↓ ↓H␈α?␈α⊃The␈α⊃director␈α⊃of␈α⊃the␈α⊃Laboratory␈α⊃is␈α⊃John␈α⊃McCarthy,␈α⊃professor␈α⊂of␈α⊂computer␈α⊂science.␈α⊂The␈α⊂associate
␈↓ ↓Hdirectors␈α⊃are␈α⊃Jerome␈α⊃Feldman,␈α⊃associate␈α⊃professor␈α⊃of␈α⊃computer␈α⊃science␈α⊃and␈α⊃Lester␈α⊂Earnest␈α⊂who␈α⊂is␈α⊂also
␈↓ ↓Hexecutive␈α
officer␈α
of␈α
the␈α
laboratory.