东盟人工智能治理与伦理指南-东南亚国家联盟（英）.docx

《东盟人工智能治理与伦理指南-东南亚国家联盟（英）.docx》由会员分享，可在线阅读，更多相关《东盟人工智能治理与伦理指南-东南亚国家联盟（英）.docx（89页珍藏版）》请在课桌文档上搜索。

1、ASEANGuideonAlGovernanceandEthicsContentsExecutiveSummary03A. Introduction061. Objectives082. Assumptions083. TargetAudience094. Definitions09B. GuidingPrinciplesfortheFramework101. TransparencyandExplainability-112. FairnessandEquity123. SecurityandSafety134. Human-centricity145. PrivacyandDataGove

2、rnance-146. AccountabilityandIntegrity157. RobustnessandReliability16C. AlGovernanceFramework171. Internalgovernancestructuresandmeasures182. DeterminingthelevelofhumaninvolvementinAl-augmenteddecision-making-233. Operationsmanagement274. Stakeholderinteractionandcommunication41D. National-levelReco

3、mmendations46E. Regional-levelRecommendations53F. Conclusion58Annex A: AlRiskImpactAssessmentTemplate_60Annex B: UseCases731. AboitizGroup-742. EY763. Gojek784. CARE.AI805. SmartNationGroup(SNG)1Singapore836. MinistryofEducation,Singapore85ExecutiveSummaryWhattheGuideisaboutThisdocumentservesasaprac

4、ticalguidefororganisationsintheregionthatwishtodesign,develop,anddeploytraditionalAltechnologiesincommercialandnon-militaryordual-useapplications.ThisGuidefocusesonencouragingalignmentwithinASEANandfosteringtheinteroperabilityofAlframeworksacrossjurisdictions.Italsoincludesrecommendationsonnational-

5、levelandregional-levelinitiativesthatgovernmentsintheregioncanconsiderimplementingtodesign,develop,anddeployAlsystemsresponsibly.GuidingPrinciplesfortheFrameworkx/-/4KeyComponentsInternalgovernancestructuresandmeasures Multi-disciplinary,centralgoverningbody,suchasanAlEthicsAdvisoryBoard,tooverseeAl

6、governanceefforts Developstandards,guidelines,tools,andtemplatestohelporganisationsdesign,develop,anddeployAlresponsibly Clearlylayouttherolesandresponsibilitiesofpersonnelinvolvedintheresponsibledesign,developmentand/ordeploymentofAlDeterminingthelevelofhumaninvolvementinAl-augmenteddecision-making

7、/Conductrelevantriskimpactassessmentstodeterminelevelofrisk Threebroadcategoriesofhumaninvolvementbasedonlevelofrisk一human-in-the-loop,human-over-the-loop,human-out-of-the-loopMitigatingriskshelpsbuildtrusttowardstheacceptanceandgreateruseofAltechnologiesintheregionOperationsmanagement TheAlSystem1.

8、ifecycleconsistsofvariousstagesandisoftenaniterativeprocess Conductrisk-basedassessmentsbeforestartinganydatacollectionandprocessingormodelling Mitigaterisksofunjustbiasduetoinsufficientlyrepresentativetraining,testingandvalidationdatasetsStakeholderinteractionandcommunicationDeveloptrustwithstakeho

9、ldersthroughoutthedesign,development,anddeploymentofAl ProvidegeneraldisclosureofwhenAlisusedinproductsand/orserviceofferings PutinplacemeasurestohelpemployeesadapttoanAl-augmentedworkenvironmentNational-levelRecommendations.NurturingAltalentandupskillingworkforceWorkcloselywithpublicandprivatesecto

10、rstoensurethatacountrysworkforcecanadapttothenewwaysofworkingandpossessesenoughdigitalskillstointeracteffectivelywithAlsystems.SupportingAlinnovationecosystemandpromotinginvestmentinAlstart-upsWorkcloselywithpublicandprivatesectorstocreateasupportiveenvironmentforAldevelopment,wherecompaniesareablet

11、oaccessandleveragedata,digitaltechnologies,andinfrastructure.InvestinginAlresearchanddevelopmentKeepabreastofthelatestdevelopmentsinAlandencourageresearchrelatedtothecybersecurityofAl,Algovernance,andAlethicstoensurethatthesafetyandresiliencyofAlsystemsandtoolsalsoadvanceinparallelwithnewusecases.Pr

12、omotingadoptionofusefultoolsbybusinessestoimplementtheASEANGuideonAlGovernanceandEthicsDeploytoolstoenabletheimplementationofAlgovernanceinoperationsandensurethatdocumentationandvalidationprocessesaremoreefficient.RaisingawarenessamongcitizensontheeffectsofAlinsocietyRaiseawarenessofthepotentialrisk

13、sandbenefitsofAlsocitizenscanmakeinformeddecisionsabouttheappropriateuseofAlandtakeappropriateactionstoprotectthemselvesfromharmfulusesofAlsystems.Regional-levelrecommendations.SettingupanASEANWorkingGrouponAlGovernancetodriveandoverseeAlgovernanceinitiativesintheregionTheWorkingGroupcanconsistofrep

14、resentativesfromeachoftheASEANmemberstateswhocanworktogethertorollouttherecommendationslaidoutinthisGuide,aswellasprovideguidanceforASEANcountrieswhowishtoadoptcomponentsofthisGuide,andwhereappropriate,includeconsultationwithotherindustrypartnersfortheirviewsandinput.AdaptationofthisGuidetoaddressgo

15、vernanceofgenerativeAlRisksinclude: Mistakesandanthropomorphism Factuallyinaccurateresponsesanddisinformation Deepfakes,impersonation,fraudulentandmaliciousactivities Infringementofintellectualpropertyrights Privacyandconfidentiality PropagationofembeddedbiasesGovernanceshouldinclude: Adaptationofex

16、istingframeworksandtools Guidanceondevelopingasharedresponsibilityframework GuidanceonincreasingthecapacitytomanagerisksofgenerativeAl GuidanceonhowtodistinguishAl-generatedcontentversusauthenticallygeneratedonesCompilingacompendiumofusecasesdemonstratingpracticalimplementationoftheGuidebyorganisati

17、onsoperatinginASEANAcompendiumofusecasesshowcasesthecommitmentoftheseorganisationstoAlgovernanceandhelpsthempromotethemselvesasresponsibleAlpractitioners.UseCasesIllustrationofcomponentsoftheASEANGuideonAlGovernanceandEthicsthroughusecasesoforganisationsoperatinginASEANthathaveimplementedAlgovernanc

18、emeasuresinAldesign,development,anddeployment.GojekAboitizGroupSmartNationGroup(SNG)1SingaporeUCARE.AIEYMinistryofEducation,SingaporeA.IntroductionArtificialIntelligence(AI)isthedisciplineofmakinganalyticalmachinesintelligent,enablinganorganisationtofunctionappropriatelyandwithforesight.Unlikeothert

19、echnologies,someformsofAladaptonitsown,learningthroughuse,sothedecisionsitmakestodaymaybedifferentfromthoseitmakestomorrow.Alandautomationhavebeenhottopics,bothfortheirtransformativepotentialandfortheircapacitytointroducenewopportunitiesbydisruptingoldmodels.SoutheastAsiaisnoexception.Alsystemsshoul

20、dbetreateddifferentlyfromothersoftwaresystemsbecauseofitsuniquecharacteristicsandrisks.CapabilitiesofAlsystemsfuelledbytechniquesevolutionandbreakthroughsarequicklyoutpacingthemonitoringandvalidationtools.ThedevelopmentofAlisalsodecentralisedduetolowbarrierstoentryandtheproliferationofopen-sourcetec

21、hnologies.GiventheprofoundimpactthatAlpotentiallybringstoorganisationsandindividualsinASEAN1itisimportantthatthedecisionsmadebyAlarealignedwithnationalandcorporatevalues,aswellasbroaderethicalandsocialnorms.Also,theASEANDigitalMasterplan2025(ADM2025)developedbyASEANMemberStates(“AMS”)envisionsASEANa

22、saleadingdigitalcommunityandeconomicbloc,poweredbysecureandtransformativedigitalservices,technologies,andecosystem.Inthatcontext,theADM2025hasidentifiedEnablingAction(EA)2.7thatsuggeststhedevelopmentandadoptionofaregionalpolicytodeliverbestpracticeguidanceonAlgovernanceandethics.Inrecentyears,govern

23、mentsandinternationalorganisationshavebegunissuingprinciples,frameworksandrecommendationsonAlethicsandgovernance.ExamplesincludeSingaporesModelAlGovernanceFrameworkInfocommMediaDevelopmentAuthority,ModelArtificialIntelligenceGovernanceFrameworkSecondEdition(21January2020)Vhttps:WVvW.Ddpc.gv.sq-media

24、filesPdPCpdf-filesresouce-foroqanisatio11aiSqmOdelaiqoVframeWOr2.pdf2OrganisationforEnomicCo-operationandDevelopment,RecommendationoftheCouncilonArtificialIntelligence(22May2019)andOECD,sRecommendationoftheCouncilonAlVhttps:/IeqalinStrUmentSQeCdQrQeninstrumentsOECD-1.EGA1.-0449.However,therehasnotye

25、tbeenanintergovernmentalcommonstandardforAlthatdefinestheprinciplesofAlgovernanceandprovidesguidanceforpolicymakersintheregiontoutiliseAlsystemsinaresponsibleandethicalmanner.IntheprocessofdraftingthisGuide,existingAlgovernanceframeworksandguidelinessuchasUNESCOsRecommendationontheEthicsofArtificial

26、IntelligenceUnitedNationsEducational,ScientificandCulturalOrganisation,RecommendationontheEthicsofArtificialIntelligence(23November2021)Vhttos:UneSdoC.unesco.orqaric48223Df00003811374EropeanCommissionEthicsGuidelinesforTrustworthyAlandEU,sEthicsGuidelinesforTrustworthyAlwerereferenced(8April2019).Th

27、eASEANGuideonAlGovernanceandEthicsaimstoempowerorganisationsandgovernmentsintheregiontodesign,develop,anddeploytraditionalAlsystemsresponsiblyandincreaseusers,trustinAl.1. ObjectivesThisdocumentservesasapracticalguidefororganisationsintheregionthatwishtodesign,develop,anddeploytraditionalAltechnolog

28、iesincommercialandnon-militaryordual-useapplications.ThisGuidefocusesonencouragingalignmentwithinASEANandfosteringtheinteroperabilityofAlframeworksacrossjurisdictions.Italsoincludesrecommendationsonnational-levelandregional-levelinitiativesthatgovernmentsintheregioncanconsiderimplementingtodesign,de

29、velop,anddeployAlsystemsresponsibly.TheASEANGuideonAlGovernanceandEthicsencompassesabroadrangeofconsiderationsthatneedstobetailoredtotheorganisationsdesigning,developing,anddeployingAldependingonthenatureoftheindustry,complexityofthetechnology,andassociatedrisksoftheAlsystems.1.ocalgovernmentauthori

30、tiesareencouragedtorefertothisGuidewhenformulatingtheirpoliciesandapproaches.Whileadoptionoftheframeworklaidouthereinisvoluntary,thisGuidecanhelporganisationsbuildtrustamongstakeholdersandthepublicaswellasaligntheirAlpracticeswithinternationalstandardsandbestpractices,amongothers.Organisationsareenc

31、ouragedtorefertotheguidelinesinthisdocumenttounderstandhowtoassesstherisksassociatedwithAlandtakemeasurestodesign,develop,anddeployAlresponsiblyinthecontextoftheirorganisations.ThisGuideismeanttobealivingdocumentthatshouldbeperiodicallyreviewedandassessedbyrelevantASEANsectorialbodies,inconsultation

32、withindustrypartners,toensurethatitisuptodatewiththelatestregulationsandadvancementsintheAlspace.UpdatestothisGuidemaybepublishedsubsequentlytokeepupwithevolutionsandgrowthingovernanceandstandards.2. AssumptionsAlsystemsneedtobemanagedholistically,includingitsecosystemandallcomponents-humanoperator,

33、InternetofThings(IoT)1robotics,traditionaltechnology,vendors,etc.InadditiontotherecommendationssetoutinthisGuidepertainingtoAlgovernance,organisationsarealsoencouragedtofollowandrefertointernationalstandardsandbestpracticesinrelatedfieldslikeinformationsecuritymanagementsystems(ISMS),datamanagementa

34、ndgovernance,softwaredevelopmentandtesting,cybersecurity,IoT,etc.Developersanddeployersneedtoadheretoapplicablenationallawsandregulations,includingsectorspecificlawsandconstitutionswhendesigning,developing,anddeployingAltechnologies.TheASEANGuideonAlGovernanceandEthicsdoesnotreplaceorsupersedeanyexi

35、stingorupcominglawsandonlyservesasaguideforresponsibledesign,development,anddeploymentofAlintheregion.BeforedeployingAl,itisalsoimportantfordevelopersanddeployerstoconsidertherelevantlegalandregulatoryrequirementsoftherespectivecountrieswheretheAlsystemswillbedeployed,aswellastheusecontextforlegal,p

36、olicy,andregulatoryconcerns.Giventhefast-evolvingspaceofAl,developersanddeployersofAlsystemsshouldbemindfulofthelateststateoftheartvis-visgovernancetoolsandtechnologiesandconducttherelevantassessmentstoevaluatethefeasibilityandusefulnessofthesetoolsintheimplementationofAlgovernancepracticesinthedesi

37、gn,development,anddeploymentofAlsystems.3. TargetAudienceTheASEANGuideonAlGovernanceandEthicsisanASEAN-endorsedframeworkfororganisationsintheregiontorefertowhendesigning,developing,anddeployingtraditionalAltechnologies.Itprovidesguidelinesandrecommendationsforadiverserangeofindividualsandorganisatio

38、nsalongtheentirevaluechain.TheseincludeAldevelopersanddeployers,academicprofessionals,andeveryonethatisinterestedinutilisingorscalingupAlsystems.TheGuidealsoincludessectionsonnational-levelandregional-levelrecommendationsthataremoretargetedtowardspolicymakersinASEAN.4. DefinitionsThedefinitionsofsev

39、eralkeytermsusedinthisGuidearesetoutbelow.ASEAN:referstotheAssociationofSoutheastAsianNations.Itisapoliticalandeconomicunionof10memberstatesinSoutheastAsia,whichpromotesintergovernmentalcooperationandfacilitateseconomic,political,security,military,educational,andsocioculturalintegrationbetweenitsmem

40、bersandcountriesintheAsia-Pacific.ArtificialIntelligence(Al):isanengineeredormachine-basedsystemthatcan,foragivensetofobjectives,generateoutputssuchaspredictions,recommendations,ordecisionsinfluencingrealorvirtualenvironments.Alsystem:isamachine-basedsystemthatiscapableofinfluencingtheenvironmentbyp

41、roducinganoutput(predictions,recommendationsordecisions)foragivensetofobjectives.Itusesmachineand/orhuman-baseddataandinputsto(i)perceiverealand/orvirtualenvironments;(ii)abstracttheseperceptionsintomodelsthroughanalysisinanautomatedmanner(e.g.,withmachinelearning),ormanually;and(iii)usemodelinferen

42、cetoformulateoptionsforoutcomes.Alsystemsaredesignedtooperatewithvaryinglevelsofautonomy.Deep1.earning:isasubsetofmachinelearning.Itisverylooselybasedontheinformation-processingarchitectureofthebrain-albeitfarfrommatchingitsability-enablingsystemstofindwaysofrerepresentinginputdatathatfacilitatemaki

43、nghighlyaccuratepredictions.Deployer:isanentitythatusesorimplementsanAlsystem,whichcouldeitherbedevelopedbytheirinhouseteamorviaathird-partydeveloper.Developer:isanentitythatdesigns,codes,orproducesanAlsystem.Machine1.earning:isasubfieldinAlwherealgorithmslearnbyidentifyingpatternsandcorrelationswit

44、hindatausingstatisticaltechniquestoenhanceperformance,allwithoutbeingexplicitlyprogrammed.User:isanentityorperson(internalorexternal)thatinteractswithanAlsystemoranAl-enabledserviceandcanbeaffectedbyitsdecisions.GuidingPrinciplesfortheFrameworkB.GuidingPrinciplesfortheFrameworkThesevenguidingprincip

45、lesbelowhelptoensuretrustinAlandthedesign,development,anddeploymentofethicalAlsystems.TheyalsoprovideguidanceonhowAlsystemsshouldbedesigned,developed,anddeployedinwayswhichconsiderthebroadersocietalimpact.1. TransparencyandExplainabilityTransparencyreferstoprovidingdisclosureonwhenanAlsystemisbeingu

46、sedandtheinvolvementofanAlsystemindecision-making,whatkindofdataituses,anditspurpose.BydisclosingtoindividualsthatAlisusedinthesystem,individualswillbecomeawareandcanmakeaninformedchoiceofwhethertousetheAl-enabledsystem.ExplainabilityistheabilitytocommunicatethereasoningbehindanAlsystemsdecisioninaw

47、aythatisunderstandabletoarangeofpeople,asitisnotalwaysclearhowanAlsystemhasarrivedataconclusion.ThisallowsindividualstoknowthefactorscontributingtotheAlsystemsrecommendation.InordertobuildpublictrustinAl,itisimportanttoensurethatusersareawareoftheuseofAltechnologyandunderstandhowinformationfromtheir

48、interactionisusedandhowtheAlsystemmakesitsdecisionsusingtheinformationprovided.Inlinewiththeprincipleoftransparency,deployershavearesponsibilitytoclearlydisclosetheimplementationofanAlsystemtostakeholdersandfostergeneralawarenessoftheAlsystembeingused.WiththeincreasinguseofAlinmanybusinessesandindustries,thepublicisbecomingmo