Artificial-IntelligenceP教學講解課件

ArtificialIntelligenceCS165AFall2004Lecture1Prof.TerrySmithCourseintroductionandoverviewCourseadministriviaOverviewofAIToday1ArtificialIntelligenceCS165AGoalsofthiscourseToteachyouthemainideasofAITointroduceyoutoasetofkeytechniquesandalgorithmsfromAITohelpyouunderstandwhat’shardinAIandwhyToseehowAIrelatestotherestofcomputerscienceTogetyouthinkingabouthowAIcanbeappliedtoavarietyofrealproblemsTohavefun2GoalsofthiscourseToteachyCourseadministriviaWebsites/~cs165a

/group/UCSB-CS165A

SyllabusDiscussionsessionsScheduleAssignmentsAssignment#0dueonTuesday!ExpectationsCometoclass,andcomepreparedParticipate:Askquestions,offerinsight,tellmeI’mwrong...Think!3CourseadministriviaWebsites3WhatisArtificialIntelligence?AIinthemediaPopularmovies2001:ASpaceOdysseyStarTrekTheTerminatorAI:TheMoviePopularpress,novelsOftenportrayedasApropertyofevilcomputersComputersdoingimpossiblethingsPublicviewBooksandmovieshaveinspiredmanyAIresearchersBooksandmovieshaveraisedthepublic’sexpectations4WhatisArtificialIntelligencWhatisArtificialIntelligence?(cont.)“Thescienceandengineeringofmakingintelligentmachines,especiallyintelligentcomputerprograms.”“Thebusinessofgettingcomputerstodothingstheycannotalreadydo,orthingstheycanonlydoinmoviesandsciencefictionstories.”“Thestudyofhowtomakecomputersdothingsatwhich,atthemoment,peoplearebetter.”“Thedesignofflexibleprogramsthatrespondproductivelyinsituationsthatwerenotspecificallyanticipatedbythedesigner.”“Theconstructionofcomputationsthatperceive,reason,andacteffectivelyinuncertainenvironments.”“ThebranchofCSconcernedwithenablingcomputerstosimulatesuchaspectsofhumanintelligenceasspeechrecognition,deduction,inference,creativeresponse,theabilitytolearnfromexperience,andtheabilitytomakeinferencesgivenincompleteinformation.”“Modelingaspectsofhumancognitiononcomputers”“WhatAIpeopledo”5WhatisArtificialIntelligencGoalsofAIScientificTounderstandtheprinciplesandmechanismsthataccountforintelligentactionEngineeringTodesignintelligentsystemsthatcansurviveandoperateintherealworldandsolveproblemsofconsiderablescientificdifficultyathighlevelsofcompetenceTounderstandandbuildintelligentsystemsTocreatemodelsandmechanismsofintelligentaction6GoalsofAIScientificToundersIntelligentsystemsAnintelligentsystemischaracterizedasonethatcan:Exhibitadaptive,goal-orientedbehaviorLearnfromexperienceUsevastamountsofknowledgeExhibitself-awarenessInteractwithhumansusinglanguageandspeechTolerateerrorandambiguityincommunicationRespondinreal-time7IntelligentsystemsAnintelligWhatAIpeoplestudyLogicKnowledgerepresentationSearchReasoning/inferenceNon-monotonicreasoningPlanningProbabilisticreasoningNa?vephysicsMachinelearningSpeechrecognitionNaturallanguageprocessingComputervisionPatternrecognitionIntelligentagentsRoboticsNeuralnetworksDataminingExpertsystems…andmore…8WhatAIpeoplestudyLogicSpeecWhatAIpeople(andprograms)doProvetheoremsEmulate/modelhumancognitiveabilities(Attemptto)solveexponentiallyhardproblemsBuildexpertsystemsfordiagnostictasks(e.g,medicaldiagnosis,erroranalysis)BuildrobotsBuildmachinevisionsystemsforindustrialtasks,surveillance,consumerapps,etc.CreatespeechrecognitionandunderstandingsystemsforvariousdomainsProcesstextto{understand,summarize,correct,respond,etc.}Createdataminingsystemstoprocessverylargeamountsofinformation(e.g.,bioinformatics)BuildintelligentagentstolookandactinsociallyusefulwaysDevelopcomputergames…andmore…9WhatAIpeople(andprograms)SomenotableAIsystemsIBM’sDeepBlueBeatworldchesschampionGaryKasparovin1997Kasparovvs.(Israeli-built)DeepJunior,January2003(endedinadraw)Kasparovvs.X3DFritz,November2003ExpertsystemsMedicaldiagnosisAcomputerizedLeukemiadiagnosissystemdidabetterjobcheckingforblooddisordersthanhumanexpertsSpeechrecognitionCommercialsystemsbyDragon,IBM,andothersPhone-basedsystems(e.g.,airlinereservations)AutomaticschedulingformanufacturingoperationsUserinterfaceGrammarandspellingcheckers,automatedhelp10SomenotableAIsystemsIBM’sDSomenotableAIsystems(cont.)DataminingFrauddetection,creditscoring,customerprofilesandpreferences,genomeanalysisCycDougLenat’s18-yearoldprojecttogivecomputercommonsenseComputervisionE.g.,“HandsAcrossAmerica”1995FacerecognitionsystemsforbiometricsRoboticsMarsRover,robotsforhazardenvironments,factoryautomationSony,Honda,others:robotpetsCMUNavlabdroveacrosscountry(2797/2849miles)1980s–DARPAALVProgramDARPAGrandChallenge2004Failedin2004…repeatinOctober200511SomenotableAIsystems(cont.DARPAGrandChallenge“DARPAintendstoconductachallengeofautonomousgroundvehiclesbetweenLosAngelesandLasVegasinMarchof2004.Acashawardof$1millionwillbegrantedtotheteamthatfieldsthefirstvehicletocompletethedesignatedroutewithinaspecifiedtimelimit.”/grandchallenge12DARPAGrandChallenge“DARPAinTerrainbetweenLAandLasVegas13TerrainbetweenLAandLasVegPerspectivesonAI/DisciplinesinvolvedAIfunctionsasachannelofideasbetweencomputingandotherfields,ideasthathaveprofoundlychangedthosefieldsLogicMathematicsStatisticsPhilosophyPsychologyLinguisticsNeuroscienceComputerscienceCognitivescienceAI14PerspectivesonAI/DisciplinFoundationsofAIPhilosophyFramedtheideasofAIDualism/materialism,logical/rational/empirical,causality,consciousness,mind/body…MathematicsFormalizedcomputation,logic,probabilityPossibilitiesandlimitationsofcomputationPsychologyExperimental:thebrainasaninformationprocessingdevice(CognitiveScience)ComputerEngineeringBuiltrealmachines,Moore’sLawprogress15FoundationsofAIPhilosophy15AIandComputerScienceAIismostlyaboutsoftware(usuallylargeandcomplex)Important:Algorithms,tools,complexity,etc.EarlyadvancedinCSduetoAIresearchersinclude:SearchalgorithmsListstructures,pointersVirtualmemoryDynamicmemoryallocationGarbagecollectionLogicalprogrammingCS165AwillbetaughtprimarilyfromaCSperspectiveNottheonlyperspective,though16AIandComputerScienceAIismUCSBCSAISequence:165Aand165B165A.ArtificialIntelligence(Fall)(4)TURKPrerequisites:ComputerScience130A;opentocomputersciencemajorsonlyAnintroductiontothefieldofArtificialIntelligence,whichattemptstounderstandandbuildintelligentsystems.TopicsincludeAIprogramminglanguages,search,logic,knowledgerepresentationandreasoning,gameplaying,planning,uncertainty,perception,andintelligentagents.

165B.MachineLearning(Winter)(4)SMITH/SUPrerequisites:ComputerScience165AThecoursecoversthemostimportanttechniquesofmachinelearning(ML)andincludesdiscussionsof:well-posedlearningproblems;artificialneuralnetworks;conceptlearningandgeneraltospecificordering;decisiontreelearning;geneticalgorithms;learningsetsofrules;Bayesianlearning;analyticallearning;andcombininginductiveandanalyticallearning.Thecourseintegratestheseapproachestolearningwithfundamentalaspectsofmachineintelligence(MI),includingsearch,knowledgerepresentationandreasoning,andapplications.

17UCSBCSAISequence:165Aand“Proper”backgroundBlindsearch(depth-first,breadth-first)CS130ATrees(programming)CS20,50,130ABooleanlogic,Propositionallogic,First-orderlogicCS40Probability,BayesrulePSTAT120AParsingCS20,160(some)C++/Javaseveral18“Proper”backgroundBlindsearcAI=“A”+“I”ArtificialAsin“artificialflowers”or“artificiallight”?IntelligenceWhatisintelligence?ThecapacitytoacquireandapplyknowledgeThefacultyofthoughtandreasonSecretinformation,especiallyaboutanactualorpotentialenemySymbolmanipulation,groundedinperceptionoftheworldThecomputationalpartoftheabilitytoachievegoalsintheworldWhatmakessomeonemore/lessintelligentthananother?Are{monkeys,ants,trees,babies,chessprograms}intelligent?Howcanweknowifamachineisintelligent?TuringTest(AlanTuring,1950),a.k.a.TheImitationGame19AI=“A”+“I”ArtificialTuringReplicatinghumanintelligence?AIdoesn’tnecessarilyseektoreplicatehumanintelligenceSometimesmore,sometimesless…“EssenceofX”vs.“X”ExamplesPhysicalvs.electronicnewspaperPhysicalvs.virtualshoppingBirdsvs.planes“SayingDeepBluedoesn’treallythinkaboutchessislikesayinganairplanedoesn’treallyflybecauseisdoesn’tflapitswings.”DrewMcDermott20Replicatinghumanintelligence“StrongAI”vs.“WeakAI”“StrongAI”Makestheboldclaimthatcomputerscanbemadetothinkonalevel(atleast)equaltohumansOneversion:ThePhysicalSymbolSystemHypothesisAphysicalsymbolsystemhasthenecessaryandsufficientmeansforgeneralintelligentactionIntelligence=symbolmanipulation(perhapsgroundedinperceptionandaction)“WeakAI”Some“thinking-like”featurescanbeaddedtocomputerstomakethemmoreusefultoolsExamples:expertsystems,speechrecognition,naturallanguageunderstanding….21“StrongAI”vs.“WeakAI”“Stro“StrongAI”vs.“WeakAI”(cont.)Principlesof“StrongAI”Intelligentbehaviorisexplicableinscientificterms;arigorousunderstandingofintelligenceispossibleIntelligencecantakeplaceoutsidethehumanskullThecomputeristhebestlaboratoryinstrumentforexploringthesepropositionsMaybe…StrongAIisscience?WeakAIisengineering?22“StrongAI”vs.“WeakAI”(conPhilosophicalandethicalimplicationsIs“StrongAI”possible?Ifso(orevenifnot)…Shouldwebeworried?Isthistechnologyathreat?(BillJoy)Isitokaytokillanintelligentmachine?Whenwillithappen?(Willweknow?)Willtheykeepusaround?(Kurzweil,Moravec)Mightwebecometoodependentontechnology?Terrorism,privacyMaincategoriesofobjectionstoAINonsensical(Searle)Impossible(Penrose)Unethical,immoral,dangerous(Weizenbaum)Failed(WallStreet)23PhilosophicalandethicalimplAnotherwayoflookingatAIThoughtprocessesandreasoningBehaviorHumanIdealSystemsthatthinklikehumansSystemsthatthinkrationallySystemsthatactlikehumansSystemsthatactrationally24AnotherwayoflookingatAIThHuman/BiologicalIntelligenceThinkinghumanly(Cognitivemodeling)Cognitivescience1960s–InformationprocessingreplacedbehaviorismasthedominantviewinpsychologyCognitiveneuroscienceNeurophysiologicalbasisofintelligenceandbehavior?Actinghumanly(Operationalintelligence)TheTuringTest–operationaltestforintelligentbehaviorWhatdoesitrequire?Required:knowledge,reasoning,languageunderstanding,learning…Problem:Itisnotreproducibleoramenabletomathematicalanalysis;rathersubjective25Human/BiologicalIntelligenceTIdeal/AbstractIntelligenceThinkingrationally(LawsofThought)Rationalthought:governedby“LawsofThought”Logicapproach–mathematicsandphilosophyActingrationally(Rationalagents)Rationalbehavior:doingtherightthingMaximizegoalachievement,giventheavailableinformation(knowledge+perception)Can/shouldincludereflexivebehavior,notjustthinkingGeneralrationalityvs.limitedrationalityBasicdefinitionofagent–somethingthatperceivesandacts26Ideal/AbstractIntelligenceThiHowcanyoutellit’sAI?Itdoessomethingthatisclearly“human-like” …or…Separationofdata/knowledgeoperations/rulescontrolHasaknowledgerepresentationframeworkproblem-solvingandinferencemethods27Howcanyoutellit’sAI?ItdoWhystudyAI?It’sfascinatingDeepquestionsaboutintelligence,consciousness,thenatureofbeinghumanGrandchallenges–creatingintelligentmachinesMultidisciplinaryendeavorItleadstoadifferentperspectiveoncomputerscienceissuesLevelsofexplanationSearch,problemsolving,etc.–higherlevelapproachExponential,NP-hardproblemsIt’sgoodbackgroundforrelatedareasComputervision,speechrecognition,naturallanguageunderstanding,probabilisticreasoningsystems,machinelearning,etc.28WhystudyAI?It’sfascinating2AquoteThehardestapplicationsandmostchallengingproblems,throughoutmanyyearsofcomputerhistory,havebeeninartificialintelligence–AIhasbeenthemostfruitfulsourceoftechniquesincomputerscience.Itledtomanyimportantadvances,likedatastructuresandlistprocessing...artificialintelligencehasbeenagreatstimulation.Manyofthebestparadigmsfordebuggingandforgettingsoftwaregoing,allofthesymbolicalgebrasystemsthatwerebuilt,earlystudiesofcomputergraphicsandcomputervision,etc.,allhadverystrongrootsinartificialintelligence.—DonaldKnuth29AquoteThehardestapplicationWillitgetmeajob?Well….NotasmanyAIjobsasJavaprogrammingjobs….But…Seewebsite(Announcements)forrelevantarticlesAIisacomponentofmanyadvancedtechnologiesAthoroughunderstandingoftheconceptscoveredinthecoursewillmakeyouabettercomputerscientistYouwillhaveabroaderarrayoftoolswithwhichtoapproachproblemsYouwillbetterbeabletoevaluatetechnologieswithAIcomponentsAIrelatedresearchusuallyrequiresagraduatedegree30Willitgetmeajob?Well….3AnoteonAIprogrammingLispListprocessingInterpreter–greatforfastprototypingFeatures:garbagecollection,dynamictyping,….PrologLogicprogrammingProgram=setoflogicalstatements+generaltheoremproverOtherhigh-levellanguages(Java,C++,etc.)31AnoteonAIprogrammingLisp31經(jīng)常不斷地學習,你就什么都知道。你知道得越多,你就越有力量StudyConstantly,AndYouWillKnowEverything.TheMoreYouKnow,TheMorePowerfulYouWillBe寫在最后32經(jīng)常不斷地學習,你就什么都知道。你知道得越多,你就越有力量寫謝謝你的到來學習并沒有結束，希望大家繼續(xù)努力LearningIsNotOver.IHopeYouWillContinueToWorkHard演講人：XXXXXX時間：XX年XX月XX日

33謝謝你的到來演講人：XXXXXX33ArtificialIntelligenceCS165AFall2004Lecture1Prof.TerrySmithCourseintroductionandoverviewCourseadministriviaOverviewofAIToday34ArtificialIntelligenceCS165AGoalsofthiscourseToteachyouthemainideasofAITointroduceyoutoasetofkeytechniquesandalgorithmsfromAITohelpyouunderstandwhat’shardinAIandwhyToseehowAIrelatestotherestofcomputerscienceTogetyouthinkingabouthowAIcanbeappliedtoavarietyofrealproblemsTohavefun35GoalsofthiscourseToteachyCourseadministriviaWebsites/~cs165a

/group/UCSB-CS165A

SyllabusDiscussionsessionsScheduleAssignmentsAssignment#0dueonTuesday!ExpectationsCometoclass,andcomepreparedParticipate:Askquestions,offerinsight,tellmeI’mwrong...Think!36CourseadministriviaWebsites3WhatisArtificialIntelligence?AIinthemediaPopularmovies2001:ASpaceOdysseyStarTrekTheTerminatorAI:TheMoviePopularpress,novelsOftenportrayedasApropertyofevilcomputersComputersdoingimpossiblethingsPublicviewBooksandmovieshaveinspiredmanyAIresearchersBooksandmovieshaveraisedthepublic’sexpectations37WhatisArtificialIntelligencWhatisArtificialIntelligence?(cont.)“Thescienceandengineeringofmakingintelligentmachines,especiallyintelligentcomputerprograms.”“Thebusinessofgettingcomputerstodothingstheycannotalreadydo,orthingstheycanonlydoinmoviesandsciencefictionstories.”“Thestudyofhowtomakecomputersdothingsatwhich,atthemoment,peoplearebetter.”“Thedesignofflexibleprogramsthatrespondproductivelyinsituationsthatwerenotspecificallyanticipatedbythedesigner.”“Theconstructionofcomputationsthatperceive,reason,andacteffectivelyinuncertainenvironments.”“ThebranchofCSconcernedwithenablingcomputerstosimulatesuchaspectsofhumanintelligenceasspeechrecognition,deduction,inference,creativeresponse,theabilitytolearnfromexperience,andtheabilitytomakeinferencesgivenincompleteinformation.”“Modelingaspectsofhumancognitiononcomputers”“WhatAIpeopledo”38WhatisArtificialIntelligencGoalsofAIScientificTounderstandtheprinciplesandmechanismsthataccountforintelligentactionEngineeringTodesignintelligentsystemsthatcansurviveandoperateintherealworldandsolveproblemsofconsiderablescientificdifficultyathighlevelsofcompetenceTounderstandandbuildintelligentsystemsTocreatemodelsandmechanismsofintelligentaction39GoalsofAIScientificToundersIntelligentsystemsAnintelligentsystemischaracterizedasonethatcan:Exhibitadaptive,goal-orientedbehaviorLearnfromexperienceUsevastamountsofknowledgeExhibitself-awarenessInteractwithhumansusinglanguageandspeechTolerateerrorandambiguityincommunicationRespondinreal-time40IntelligentsystemsAnintelligWhatAIpeoplestudyLogicKnowledgerepresentationSearchReasoning/inferenceNon-monotonicreasoningPlanningProbabilisticreasoningNa?vephysicsMachinelearningSpeechrecognitionNaturallanguageprocessingComputervisionPatternrecognitionIntelligentagentsRoboticsNeuralnetworksDataminingExpertsystems…andmore…41WhatAIpeoplestudyLogicSpeecWhatAIpeople(andprograms)doProvetheoremsEmulate/modelhumancognitiveabilities(Attemptto)solveexponentiallyhardproblemsBuildexpertsystemsfordiagnostictasks(e.g,medicaldiagnosis,erroranalysis)BuildrobotsBuildmachinevisionsystemsforindustrialtasks,surveillance,consumerapps,etc.CreatespeechrecognitionandunderstandingsystemsforvariousdomainsProcesstextto{understand,summarize,correct,respond,etc.}Createdataminingsystemstoprocessverylargeamountsofinformation(e.g.,bioinformatics)BuildintelligentagentstolookandactinsociallyusefulwaysDevelopcomputergames…andmore…42WhatAIpeople(andprograms)SomenotableAIsystemsIBM’sDeepBlueBeatworldchesschampionGaryKasparovin1997Kasparovvs.(Israeli-built)DeepJunior,January2003(endedinadraw)Kasparovvs.X3DFritz,November2003ExpertsystemsMedicaldiagnosisAcomputerizedLeukemiadiagnosissystemdidabetterjobcheckingforblooddisordersthanhumanexpertsSpeechrecognitionCommercialsystemsbyDragon,IBM,andothersPhone-basedsystems(e.g.,airlinereservations)AutomaticschedulingformanufacturingoperationsUserinterfaceGrammarandspellingcheckers,automatedhelp43SomenotableAIsystemsIBM’sDSomenotableAIsystems(cont.)DataminingFrauddetection,creditscoring,customerprofilesandpreferences,genomeanalysisCycDougLenat’s18-yearoldprojecttogivecomputercommonsenseComputervisionE.g.,“HandsAcrossAmerica”1995FacerecognitionsystemsforbiometricsRoboticsMarsRover,robotsforhazardenvironments,factoryautomationSony,Honda,others:robotpetsCMUNavlabdroveacrosscountry(2797/2849miles)1980s–DARPAALVProgramDARPAGrandChallenge2004Failedin2004…repeatinOctober200544SomenotableAIsystems(cont.DARPAGrandChallenge“DARPAintendstoconductachallengeofautonomousgroundvehiclesbetweenLosAngelesandLasVegasinMarchof2004.Acashawardof$1millionwillbegrantedtotheteamthatfieldsthefirstvehicletocompletethedesignatedroutewithinaspecifiedtimelimit.”/grandchallenge45DARPAGrandChallenge“DARPAinTerrainbetweenLAandLasVegas46TerrainbetweenLAandLasVegPerspectivesonAI/DisciplinesinvolvedAIfunctionsasachannelofideasbetweencomputingandotherfields,ideasthathaveprofoundlychangedthosefieldsLogicMathematicsStatisticsPhilosophyPsychologyLinguisticsNeuroscienceComputerscienceCognitivescienceAI47PerspectivesonAI/DisciplinFoundationsofAIPhilosophyFramedtheideasofAIDualism/materialism,logical/rational/empirical,causality,consciousness,mind/body…MathematicsFormalizedcomputation,logic,probabilityPossibilitiesandlimitationsofcomputationPsychologyExperimental:thebrainasaninformationprocessingdevice(CognitiveScience)ComputerEngineeringBuiltrealmachines,Moore’sLawprogress48FoundationsofAIPhilosophy15AIandComputerScienceAIismostlyaboutsoftware(usuallylargeandcomplex)Important:Algorithms,tools,complexity,etc.EarlyadvancedinCSduetoAIresearchersinclude:SearchalgorithmsListstructures,pointersVirtualmemoryDynamicmemoryallocationGarbagecollectionLogicalprogrammingCS165AwillbetaughtprimarilyfromaCSperspectiveNottheonlyperspective,though49AIandComputerScienceAIismUCSBCSAISequence:165Aand165B165A.ArtificialIntelligence(Fall)(4)TURKPrerequisites:ComputerScience130A;opentocomputersciencemajorsonlyAnintroductiontothefieldofArtificialIntelligence,whichattemptstounderstandandbuildintelligentsystems.TopicsincludeAIprogramminglanguages,search,logic,knowledgerepresentationandreasoning,gameplaying,planning,uncertainty,perception,andintelligentagents.

165B.MachineLearning(Winter)(4)SMITH/SUPrerequisites:ComputerScience165AThecoursecoversthemostimportanttechniquesofmachinelearning(ML)andincludesdiscussionsof:well-posedlearningproblems;artificialneuralnetworks;conceptlearningandgeneraltospecificordering;decisiontreelearning;geneticalgorithms;learningsetsofrules;Bayesianlearning;analyticallearning;andcombininginductiveandanalyticallearning.Thecourseintegratestheseapproachestolearningwithfundamentalaspectsofmachineintelligence(MI),includingsearch,knowledgerepresentationandreasoning,andapplications.

50UCSBCSAISequence:165Aand“Proper”backgroundBlindsearch(depth-first,breadth-first)CS130ATrees(programming)CS20,50,130ABooleanlogic,Propositionallogic,First-orderlogicCS40Probability,BayesrulePSTAT120AParsingCS20,160(some)C++/Javaseveral51“Proper”backgroundBlindsearcAI=“A”+“I”ArtificialAsin“artificialflowers”or“artificiallight”?IntelligenceWhatisintelligence?ThecapacitytoacquireandapplyknowledgeThefacultyofthoughtandreasonSecretinformation,especiallyaboutanactualorpotentialenemySymbolmanipulation,groundedinperceptionoftheworldThecomputationalpartoftheabilitytoachievegoalsintheworldWhatmakessomeonemore/lessintelligentthananother?Are{monkeys,ants,trees,babies,chessprograms}intelligent?Howcanweknowifamachineisintelligent?TuringTest(AlanTuring,1950),a.k.a.TheImitationGame52AI=“A”+“I”ArtificialTuringReplicatinghumanintelligence?AIdoesn’tnecessarilyseektoreplicatehumanintelligenceSometimesmore,sometimesless…“EssenceofX”vs.“X”ExamplesPhysicalvs.electronicnewspaperPhysicalvs.virtualshoppingBirdsvs.planes“SayingDeepBluedoesn’treallythinkaboutchessislikesayinganairplanedoesn’treallyflybecauseisdoesn’tflapitswings.”DrewMcDermott53Replicatinghumanintelligence“StrongAI”vs.“WeakAI”“StrongAI”Makestheboldclaimthatcomputerscanbemadetothinkonalevel(atleast)equaltohumansOneversion:ThePhysicalSymbolSystemHypothesisAphysicalsymbolsystemhasthenecessaryandsufficientmeansforgeneralintelligentactionIntelligence=symbolmanipulation(perhapsgroundedinperceptionandaction)“WeakAI”Some“thinking-like”featurescanbeaddedtocomputerstomakethemmoreusefultoolsExamples:expertsystems,speechrecognition,naturallanguageunderstanding….54“StrongAI”vs.“WeakAI”“Stro“StrongAI”vs.“WeakAI”(cont.)Principlesof“StrongAI”Intelligentbehaviorisexplicableinscientificterms;arigorousunderstandingofintelligenceispossibleIntelligencecantakeplaceoutsidethehumanskullThecomputeristhebestlaboratoryinstrumentforexploringthesepropositionsMaybe…StrongAIisscience?WeakAIisengineering?55“StrongAI”vs.“WeakAI”(conPhilosophicalandethicalimplicationsIs“StrongAI”possible?Ifso(orevenifnot)…Shouldwebeworried?Isthistechnologyathreat?(BillJoy)Isitokaytokillanintelligentmachine?Whenwillithappen?(Willweknow?)Willtheykeepusaround?(Kurzweil,Moravec)Mightwebecometoodependentontechnology?Terrorism,privacyMaincategoriesofobjectionstoAINonsensical(Searle)Impossible(Penrose)Unethical,immoral,dangerous(Weizenbaum)Failed(WallStreet)56PhilosophicalandethicalimplAnotherwayoflookingatAIThoughtprocessesandreasoningBehaviorHumanIdealSystemsthatthinklikehumansSystemsthatthinkrationallySystemsthatactlikehumansSystemsthatactrationally57AnotherwayoflookingatAIThHuman/BiologicalIntelligenceThinkinghumanly(Cognitivemodeling)Cognitivescience1960s–InformationprocessingreplacedbehaviorismasthedominantviewinpsychologyCognitiveneuroscienceNeurophysiologicalbasisofintelligenceandbehavior?Actinghumanly(Operationalintelligence)TheTuringTest–operationaltestforintelligentbehaviorWhatdoesitrequire?Required:knowledge,reasoning,languageunderstanding,learning…Problem:Itisnotreproducibleoramenabletomathematicalanalysis;rathersubjective58Human/BiologicalIntelligenceTIdeal/AbstractIntelligenceThinkingrationally(LawsofThought)Rationalthought:governedby“LawsofThought”Logicapproach–mathematicsandphilosophyActingrationally(Rational