22 c : 145 Artificial Intelligence Hantao Zhang http://www.cs ...

vinegarclothΤεχνίτη Νοημοσύνη και Ρομποτική

17 Ιουλ 2012 (πριν από 5 χρόνια και 29 μέρες)

402 εμφανίσεις

22c:145ArtificialIntelligence
HantaoZhang
http://www.cs.uiowa.edu/

hzhang/c145
TheUniversityofIowa
DepartmentofComputerScience
ArtificialIntelligence–p.1/??
Texbook
M.TimJones
ArtificialIntelligence:ASystemsApproach
JonesandBartlettPublishers,2009.
ClassWebSite
www.cs.uiowa.edu/˜hzhang/c145
Checktheclasswebsite
daily
!
ArtificialIntelligence–p.2/??
OtherAItexts
TitleAuthorsPublisherYear
AIApplicationProgrammingM.TimJonesCourseTechnology2005
AI,AModernApproachS.Ruessel,P.NorvigPrenticeHall2009
AI:ANewSynthesisNilssonMorganKaufmann1998
ComputationalIntelligencePoole,Mackworth,GoebelOxford1998
ArtificialIntelligenceWinstonAddison-Wesley1992
ArtificialIntelligenceRich,KnightMcGraw-Hill1991
AI:TheoryandPracticeDean,Allen,AloimonosBenjaminCummings1995
MathematicalMethodsinAIBenderIEEEComp.Press1996
LogicalFoundationsofAIGenesereth,NilssonMorganKaufmann1987
ArtificialIntelligence–p.3/??
CourseOverview
TopicChapters
Introduction1
UninformedSearch2
InformedSearch3
GamePlaying4
KnowledgeRepresentation5
MachineLearning6
EvolutionaryComputation7
NeuralNetworks8,9
BiologicallyInspiredMethods12
LanguagesofAI13
ArtificialIntelligence–p.4/??
Prerequisites
Thecoursewillbeself-contained,butanelementary
backgroundinCSandMathisrequired.
Expecttheclassmaterialtobecomealittletechnicalat
times.
Youwillimplementsomeofthetechniquesseenin
class.Programmingassignmentswillbeinyourfavorite
programminglanguagesandProlog.
ArtificialIntelligence–p.5/??
WhereisAIinComputerScience
ComputerScience:Itisaboutproblemsolvingusing
computers.
ComputerArchitectureandSystemSoftwarestudyhow
tobuildgoodcomputers.
ComputationTheoryandComplexityTheoreystudy
whatcanbecomputed,whatcannotbecomputed,i.e.,
thelimitsofdifferentcomputingdevices.
ProgrammingLanguagesstudyhowusecomputers
convenientlyandefficiently.
AlgorithmsandDataStructuresstudyhowtosolve
popularcomputationproblemsefficiently.
ArtificialIntelligence
,Databases,Netoworking,Security,
etc.,studyhowtoextendtheuseofcomputers.
ArtificialIntelligence–p.6/??
WhatisArtificialIntelligence?
Ascientificandengineeringdisciplinedevotedto:
understandingprinciplesthatmakeintelligentbehavior
possibleinnaturalorartificialsystems;
developingmethodsforthedesignandimplementation
ofuseful,intelligentartifacts.[Poole,Mackworth,
Goebel]
ArtificialIntelligence–p.7/??
Whatisintelligencethen?
Fastthinking?
Knowledge?
Abilitytopassasahuman?
Abilitytoreasonlogically?
Abilitytolearn?
Abilitytoperceiveandactuponone’senvironment?
Abilitytoplaychessatgrand-master’slevel?
ArtificialIntelligence–p.8/??
Actinghumanly:TheTuringtest
LoebnerPrizeGoldMedalhttp://www.loebner.net/Prizef/loebner-prize.html
In1990HughLoebneragreedwithTheCambridgeCenterfor
BehavioralStudiestounderwriteacontestdesignedto
implementtheTuringTest.Dr.LoebnerpledgedaGrand
Prizeof$100,000andaGoldMedalforthefirstcomputer
whoseresponseswereindistinguishablefromahuman’s.
Suchacomputercanbesaid"tothink."Eachyearanannual
prizeof$2000andabronzemedalisawardedtothemost
human-likecomputer.Thewinneroftheannualcontestisthe
bestentryrelativetootherentriesthatyear,irrespectiveof
howgooditisinanabsolutesense.
ArtificialIntelligence–p.13/??
ThinkingHumanly:CognitiveScience
1960s“
cognitiverevolution
”:information-processingpsychology
replacedprevailingorthodoxyof
behaviorism
Requirescientifictheoriesofinternalactivitiesofthebrain
Whatlevelofabstraction?“
Knowledge
”or“
circuits
”?
Howtovalidate?Itrequires
1.Predictingandtestingbehaviorofhumansubjects(top-down)
2.Directidentificationfromneurologicaldata(bottom-up)
Bothapproaches,
CognitiveScience
and
CognitiveNeuroscience
,
sharewithAIon:
theavailabletheoriesdonotexplain(orengender)
anythingresemblinghuman-levelgeneralintelligence
ArtificialIntelligence–p.14/??
ThinkingRationally:LawsofThought
SeveralGreekschoolsatthetimeofAristotledeveloped
variousformsof
logic
:
Notation
and
rulesofderivation
forthoughts;theymay
ormaynothaveproceededtotheideaofmechanization
Directlinethroughmathematicsandphilosophyto
modernAI
Problems:1.Notallintelligentbehaviorismediatedbylogical
deliberation
2.Whatisthepurposeofthinking?Whatthoughts
should
Ihave?
ArtificialIntelligence–p.15/??
ActingRationally
Rational
behavior:doingtherightthing,thatwhichis
expectedtomaximizegoalachievement,giventhe
availableinformation
Doesn’tnecessarilyinvolvethinking—e.g.,blinking
reflex—butthinkingshouldbeintheserviceofrational
action
Aristotle:
Everyartandeveryinquiry,andsimilarly
everyactionandpursuit,isthoughttoaimatsomegood
ArtificialIntelligence–p.16/??
OperationalDefinitionofAI
Systemsthatactlikehumans
Turingtest.
Systemsthatthinklikehumans
CognitiveScience
Systemsthatthinkrationally
Logic-basedAI
Systemsthatactrationally
RationalAgents
ArtificialIntelligence–p.17/??
AIPrehistory
Philosophy
logic,methodsofreasoning
mindasphysicalsystem
foundationsoflearning,language,rationality
Mathematics
formalrepresentationandproof
algorithms,computation,(un)decidability,(in)tractability
probability
Psychology
adaptation
phenomenaofperceptionandmotorcontrol
experimentaltechniques(psychophysics,etc.)
Economics
formaltheoryofrationaldecisions
Linguistics
knowledgerepresentation
grammar
Neuroscience
plasticphysicalsubstrateformentalactivity
Controltheory
homeostaticsystems,stability
simpleoptimalagentdesigns
ArtificialIntelligence–p.18/??
PottedHistoryofAI
1943
McCulloch&Pitts:Booleancircuitmodelofbrain
1950
Turing’s“ComputingMachineryandIntelligence”
1950s
EarlyAIprograms,includingSamuel’scheckersprogram,
Newell&Simon’sLogicTheorist,Gelernter’sGeometryEngine
1956
Dartmouthmeeting:“ArtificialIntelligence”adopted
1965
Robinson’scompletealgorithmforlogicalreasoning
1966–74
AIdiscoverscomputationalcomplexity
Neuralnetworkresearchalmostdisappears
1969–79
Earlydevelopmentofknowledge-basedsystems
1980–88
Expertsystemsindustrybooms
1988–93
Expertsystemsindustrybusts:“AIWinter”
1985–95
Neuralnetworksreturntopopularity
1988
–Resurgenceofprobability;generalincreaseintechnicaldepth
“NouvelleAI”:ALife,GAs,softcomputing
1995
–AIeverywhere...
ArtificialIntelligence–p.19/??
Stateoftheart
Whichofthefollowingcanbedoneatpresent?
1.Playadecentgameoftabletennis
2.Drivealongacurvingmountainroad
3.DriveinthecenterofCairo
4.Buyaweek’sworthofgroceriesatastore
5.Buyaweek’sworthofgroceriesontheweb
6.Playadecentgameofbridge
7.Discoverandproveanewmathematicaltheorem
8.Writeanintentionallyfunnystory
9.Givecompetentlegaladviceinaspecializedareaoflaw
10.TranslatespokenEnglishintospokenSwedishinrealtime
11.Performacomplexsurgicaloperation
ArtificialIntelligence–p.20/??
WhyStudyAI?
AIhelps
computerscientistsandengineersbuildmoreuseful
anduser-friendlycomputers,
psychologists,linguists,andphilosophersunderstand
theprinciplesthatconstitutewhatwecallintelligence.
AIisaninterdisciplinaryfieldofstudy.
ManyideasandtechniquesnowstandardinCS(symbolic
computation,timesharing,objects,declarativeprogram-
ming,...)werepioneeredbyAI-relatedresearch.
ArtificialIntelligence–p.21/??
AIisamongus!
RecentapplicationsusingAItechniques:
SonyAiboEntertainmentrobotwithpet-likebehaviour
(http://www.us.aibo.com)
DragonNaturallySpeaking(Dictationandvoicerecognitionsoftware)
(http://www.dragonsys.com/naturallyspeaking)
AnanovaVirtualnewscasterontheweb
(http://www.ananova.com/video)
HondaHumanoidRobotDemowalkingrobot
(http://www.honda.co.jp/robot)
ArtificialIntelligence–p.22/??
AIisamongus!
MoreapplicationsusingAItechniques:
DeepBlueChessprogramthatbeatchessgrand-grand-master
Kasparov
(http://researchweb.watson.ibm.com/deepblue)
MarsPathfinderAutonomouslandvehiclesenttoMars
(http://mars.jpl.nasa.gov/MPF)
Aaron
TheRobotasanArtist
http://www.scinetphotos.com/aaron.html
AstronomyandSpaceExploration
http://www.aaai.org/AITopics/html/astro.html
andmanymore!
ArtificialIntelligence–p.23/??
AIisprettyhardstuff
Iwenttothegrocerystore,IsawthemilkontheshelfandI
boughtit.
WhatdidIbuy?
Themilk?
Theshelf?
Thestore?
Anawfullotofknowledgeoftheworldisneededtoanswer
simplequestionslikethisone.
ArtificialIntelligence–p.24/??