What-is-AI

10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.1WhatisAI?Chapter1•Quotes&Concepts•Goals&Approaches•ApplicationAreas•FrameworkforAISystems•FundamentalIssuesforAIProblems10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.2WhatisAI?“AIisthedesign,studyandconstructionofcomputerprogramsthatbehaveintelligently.”TomDean10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.3WhatisAI?Whatisartificialintelligence(AI)?isafieldconcernedintelligentbehaviorin-machineshumansanimalsotherattemptsatnamingthisfieldincludedmachineintelligencecomplexinformationprocessingheuristicprogrammingcognology10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.4WhatisAI?“AIisthestudyofcomplexinformationprocessingproblemsthatoftenhavetheirrootsinsomeaspectofbiologicalinformationprocessing.Thegoalofthesubjectistoidentifysolvableandinterestinginformationprocessingproblems,andsolvethem.”DavidMarr(1945-1980)10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.5WhatisAI?“Itisthescienceandengineeringofmakingintelligentmachines,especiallyintelligentcomputerprograms.Itisrelatedtothesimilartaskofusingcomputerstounderstandhumanintelligence,butAIdoesnothavetoconfineitselftomethodsthatarebiologicallyobservable.”“Intelligenceisthecomputationalpartoftheabilitytoachievegoalsintheworld.”seemore:©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.6WhatisAI?Whatisintelligentbehavior?perceivingone’senvironmentactingincomplexenvironmentslearningandunderstandingfromexperienceusingreasoningtosolveproblemsandtodiscover“hidden”knowledgeapplyingknowledgesuccessfullyinnewsituationsthinkingabstractly,usinganalogiescommunicatingwithotherswhatabout-creativity?ingenuity?expressiveness?curiosity?10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.7WhatisAI?“Ibelievethatunderstandingintelligenceinvolvesunderstandinghowknowledgeisacquired,represented,andstored;howintelligentbehaviorisgeneratedandlearned;howmotives,andemotions,andprioritiesaredevelopedandused;howsensorysignalsaretransformedintosymbols;howsymbolsaremanipulatedtoperformlogic,toreasonaboutthepast,andplanforthefuture;andhowthemechanismsofintelligenceproducethephenomenaofillusion,belief,hope,fear,anddreams-andyesevenkindnessandlove.”seemore:©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.8GoalsofandApproachestoAIWhataregoalsofAI?toreplicatehumanintelligence(stilldistantgoal)tosolveknowledge-intensivetaskstomakeanintelligentconnectionbetweenperceptionandactiontoenhancehuman-human,human-computerandcomputer-computerinteraction/communication10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.9GoalsofandApproachestoAIEngineeringGoaldevelopconcepts,theoryandpracticeofbuildingintelligentmachinesemphasisonsystembuildingScienceGoaldevelopconcepts,mechanismsandvocabularytounderstandbiologicalintelligentbehavioremphasisonunderstandingintelligentbehavior10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.10GoalsofandApproachestoAIChoosethegoalsofthecomputationalmodelthebasisforevaluatingperformanceofthesystemhuman-like:rational:think:1)thinklikehumanscognitivescience2)thinkrationallylawsofthoughtact:3)actlikehumansbehavioristapproach4)actrationallysatisficingmethods10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.11GoalsofandApproachestoAIBox1:CognitiveScienceApproachfocusonbehaviorandI/OmodelreasoningprocessescomputationalmodelshouldreflecthowresultswereobtainedGoalnotjusttoproducehumanlikebehavior(box3)buttoproduceareasoningprocesssimilartothestepsusedbyhumans10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.12GoalsofandApproachestoAIBox2:LawsofThoughtfocusoninferencemechanismsthatareprovablycorrectandguaranteeanoptimalsolutionGoalformalizethereasoningprocessasasystemoflogicalrulesandproceduresforinferencingdevelopsystemsofrepresentationtoallowinferencestobeperformed10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.13GoalsofandApproachestoAIBox3:BehavioristApproachfocusonactionintheworld,intelligentbehaviornotcenteredaroundrepresentationoftheworldGoaldevelopsystemsthatarehuman-likeTuringTestAlanTuring1950canamachinefoolaninterrogatorintobelievingthatitisahuman?10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.14GoalsofandApproachestoAIBox4:SatisficingMethodsfocusonsystemsthatactsufficientlyifnotoptimallyinallsituationsit'sOKtohaveimperfectreasoningifthejobgetsdoneGoaldevelopsystemsthatarerationalandsufficient10/10/2020©2001-2004JamesD.SkrentnyfromnotesbyC.Dyer,et.al.15GoalsofandApproachestoAITherearetwokindsofpeopleintheAIworld:classicalAI,thesymbol-processingapproachtopdown,knowledgetosymboltoimplementationuseslogicaloperationsanddeclarativeknowledgebasesNewellandSimon’sphysicalsymbolsystemsub-symbolicAI,theanimatapproachbottomup,fromsignaltosymboldesignproceedsinstepsoftheevolutionaryladderintelligenceisanemergentprope