賦能數(shù)字經(jīng)濟:人工智能在金融行業(yè)的機遇和風險_第1頁
賦能數(shù)字經(jīng)濟:人工智能在金融行業(yè)的機遇和風險_第2頁
賦能數(shù)字經(jīng)濟:人工智能在金融行業(yè)的機遇和風險_第3頁
賦能數(shù)字經(jīng)濟:人工智能在金融行業(yè)的機遇和風險_第4頁
賦能數(shù)字經(jīng)濟:人工智能在金融行業(yè)的機遇和風險_第5頁
已閱讀5頁,還剩51頁未讀 繼續(xù)免費閱讀

下載本文檔

版權(quán)說明:本文檔由用戶提供并上傳,收益歸屬內(nèi)容提供方,若內(nèi)容存在侵權(quán),請進行舉報或認領(lǐng)

文檔簡介

MONEYANDCAPITALMARKETSAND

INFORMATIONTECHNOLOGYDEPARTMENTS

PoweringtheDigital

Economy

OpportunitiesandRisksofArtificial

IntelligenceinFinance

PreparedbyElBachirBoukherouaaandGhiathShabsigh

incollaborationwith

KhaledAlAjmi,JoseDeodoro,AquilesFarias,EbruS.Iskender,

AlinT.Mirestean,andRangacharyRavikumar

DP/2021/024

2021

OCTOBER

MONEYANDCAPITALMARKETSANDINFORMATIONTECHNOLOGYDEPARTMENTS

DEPARTMENTALPAPERS

PoweringtheDigitalEconomy

OpportunitiesandRisksofArtificialIntelligencein

Finance

PreparedbyElBachirBoukherouaaandGhiathShabsigh

incollaborationwith

KhaledAlAjmi,JoseDeodoro,AquilesFarias,EbruS.Iskender,AlinT.Mirestean,andRangacharyRavikumar

Copyright?2021InternationalMonetaryFund

PoweringtheDigitalEconomy:OpportunitiesandRisksofArtificialIntelligenceinFinance

DP/2021/024

Authors:ElBachirBoukherouaaandGhiathShabsigh

incollaborationwith

KhaledAlAjmi,JoseDeodoro,AquilesFarias,EbruS.Iskender,AlinT.Mirestean,andRangacharyRavikumar

1

Cataloging-in-PublicationData

IMFLibrary

Names:Boukherouaa,ElBachir.|Shabsigh,Ghiath.|AlAjmi,Khaled.|Deodoro,Jose.|Farias,Aquiles.|Iskender,EbruS.|Mirestean,Alin.|Ravikumar,Rangachary.|InternationalMonetaryFund,publisher.

Title:Poweringthedigitaleconomy:opportunitiesandrisksofartificialintelligenceinfinance/preparedbyElBachirBoukherouaaandGhiathShabsighincollaborationwithKhaledAlAjmi,JoseDeodoro,AquilesFarias,EbruS.Iskender,AlinT.Mirestean,andRangacharyRavikumar.

Description:Washington,DC:InternationalMonetaryFund,2021.|2021September.|Departmentalpaperseries.|Includesbibliographicalreferences.

Identifiers:ISBN9781589063952(paper)

Subjects:LCSH:Artificialintelligence—Economicaspects.|Machinelearning—Economicaspects.|Financialservicesindustry—Technologicalinnovations.

Classification:LCCHC79.I55B682021

ISBN

978-1-59806-395-2(Paper)

JELClassificationNumbers:

C40,C510,C550,E17,G21,G23,G280,O310,O330

Keywords:

ArtificialIntelligence,MachineLearning,FinancialStability,EmbeddedBias,FinancialRegulation,Cybersecurity,RiskManagement,DataPrivacy

Author’sE-MailAddress:

GShabsigh@;EBoukherouaa@;KAlAjmi@;JDeodoro@;AFarias@;ESonbulIskender@;AMirestean@;RRavikumar@

TheDepartmentalPaperSeriespresentsresearchbyIMFstaffonissuesofbroadregionalorcross-countryinterest.Theviewsexpressedinthispaperarethoseoftheauthor(s)anddonotnecessarilyrepresenttheviewsoftheIMF,itsExecutiveBoard,orIMFmanagement.

Publicationordersmaybeplacedonlineorthroughthemail:

InternationalMonetaryFund,PublicationServices

P.O.Box92780,Washington,DC20090,USA

T.+(1)202.623.7430

publications@

IMF

elibrary.IMF.org

1WearegratefultoAdityaNarainandotherIMFcolleaguesforvaluablecomments,andtoJavierChangforproductionsupport.

1IMFDEPARTMENTALPAPERSPoweringtheDigitalEconomy

ExecutiveSummary

Thispaperdiscussestheimpactoftherapidadoptionofartificialintelligence(AI)andmachinelearning(ML)inthefinancialsector.Ithighlightsthebenefitsthesetechnologiesbringintermsoffinancialdeepeningandefficiency,whileraisingconcernsaboutitspotentialinwideningthedigitaldividebetweenadvancedanddevelopingeconomies.Thepaperadvancesthediscussionontheimpactofthistechnologybydistillingandcategorizingtheuniquerisksthatitcouldposetotheintegrityandstabilityofthefinancialsystem,policychallenges,andpotentialregulatoryapproaches.Theevolvingnatureofthistechnologyanditsapplicationinfinancemeansthatthefullextentofitsstrengthsandweaknessesisyettobefullyunderstood.Giventheriskofunexpectedpitfalls,countrieswillneedtostrengthenprudentialoversight.

AIandMLaretechnologieswiththepotentialforenormoussocietalandeconomicimpact,bringingnewopportunitiesandbenefits.Recenttechnologicaladvancesincomputinganddatastoragepower,bigdata,andthedigitaleconomyarefacilitatingrapidAI/MLdeploymentinawiderangeofsectors,includingfinance.TheCOVID-19crisishasacceleratedtheadoptionofthesesystemsduetotheincreaseduseofdigitalchannels.

AI/MLsystemsarechangingthefinancialsectorlandscape.CompetitivepressuresarefuelingrapidadoptionofAI/MLinthefinancialsectorbyfacilitatinggainsinefficiencyandcostsavings,reshapingclientinterfaces,enhancingforecastingaccuracy,andimprovingriskmanagementandcompliance.AI/MLsystemsalsoofferthepotentialtostrengthenprudentialoversightandtoequipcentralbankswithnewtoolstopursuetheirmonetaryandmacroprudentialmandates.

Theseadvances,however,arecreatingnewconcernsarisingfromrisksinherentinthetechnologyanditsapplicationinthefinancialsector.Concernsincludeanumberofissues,suchasembeddedbiasinAI/MLsystems,theopaquenessoftheiroutcomes,andtheirrobustness(particularlywithrespecttocyberthreatsandprivacy).Furthermore,thetechnologyisbringingnewsourcesandtransmissionchannelsofsystemicrisks,includinggreaterhomogeneityinriskassessmentsandcreditdecisionsandrisinginterconnectednessthatcouldquicklyamplifyshocks.

AI/MLinfinanceshouldbebroadlywelcome,togetherwithpreparationstocapturetheirbenefitsandmitigatepotentialriskstothefinancialsystem’sintegrityandsafety.Preparationsincludestrengtheningthecapacityandmonitoringframeworksofoversightauthorities,engagingstakeholderstoidentifypossiblerisksandremedialregulatoryactions,updatingrelevantlegalandregulatory,andexpandingconsumereducation.ItisimportantthattheseactionsaretakeninthecontextofnationalAIstrategiesandinvolveallrelevantpublicandprivatebodies.

Cooperationandknowledgesharingattheregionalandinternationallevelisbecomingincreasinglyimportant.ThiswouldallowforthecoordinationofactionstosupportthesafedeploymentofAI/MLsystemsandthesharingofexperiencesandknowledge.Cooperationwillbeparticularlyimportanttoensurethatless-developedeconomiessharethebenefits.

2IMFDEPARTMENTALPAPERSPoweringtheDigitalEconomy

Contents

ExecutiveSummary1

AcronymsandAbbreviations

4

1.Introduction

5

2.ArtificialintelligenceintheFinancialSector

7

A.Forecasting

7

B.InvestmentandBankingServices

7

C.RiskandComplianceManagement

9

D.PrudentialSupervision

9

E.CentralBanking

12

3.RisksandPolicyConsiderations

14

A.EmbeddedBias

14

B.Unboxingthe“BlackBox”:ExplainabilityandComplexity

15

C.Cybersecurity

16

D.DataPrivacy

17

E.Robustness

17

F.ImpactonFinancialStability

18

4.Conclusion

20

Annexes

Annex1.HowMachineLearningAlgorithmsWork

21

Annex2.ArtificialIntelligenceinFinance—RiskProfile

24

Annex3.NationalArtificialIntelligenceStrategies

25

References

28

BOXES

Box1.ArtificialIntelligenceandMachineLearningCapabilities

6

Box2.ArtificialIntelligenceinInvestmentManagement—SampleUseCases

8

Box3.ArtificialIntelligenceinCreditUnderwriting

8

Box4.ArtificialIntelligenceinRegulatoryCompliance—SampleUseCases

10

Box5.ArtificialIntelligenceinSupervision—SampleApplications

11

Box6.ArtificialIntelligenceinCentralBanking—SampleApplications

13

Box7.Explainingthe"BlackBox"

16

FIGURES

Figure1.TopFiveTechnologiesEmployedinRegulatoryTechnologyOfferings

9

Figure2.TechnologiesUsedinSuprvisoryTechnologyTools

10

AnnexFigure1.1.MachineLearningParadigms

22

3IMFDEPARTMENTALPAPERSPoweringtheDigitalEconomy

AnnexFigure1.2.ExampleofanInputAttack

23

AnnexFigure3.1.NationalArtificialIntelligenceStrategyLandscape

25

AnnexFigure3.2.KeyFeaturesofNationalArtificialIntelligenceStrategies

26

4IMFDEPARTMENTALPAPERSPoweringtheDigitalEconomy

AcronymsandAbbreviations

AI

ArtificialIntelligence

AML/CFT

Anti-MoneyLaundering/CombatingtheFinancingofTerrorism

Fintech

FinancialTechnology

ML

MachineLearning

NLO

NaturalLanguageProcessing

OECD

OrganisationforEconomicCo-operationandDevelopment

Regtech

RegulatoryTechnology

Suptech

SupervisoryTechnology

5IMFDEPARTMENTALPAPERSPoweringtheDigitalEconomy

1.Introduction

Thispaperexplorestheuseofartificialintelligence(AI)andmachinelearning(ML)inthefinancialsectorandtheresultantpolicyimplications.

1

ItprovidesanontechnicalbackgroundontheevolutionandcapabilitiesofAI/MLsystems,theirdeploymentandusecasesinthefinancialsector,andthenewchallengestheypresenttofinancialsectorpolicymakers.

AI/MLsystemshavemademajoradvancesoverthepastdecade.Althoughthedevelopmentofamachinewiththecapacitytounderstandorlearnanyintellectualtaskthatahumanbeingperformsisnotwithinimmediategrasp,today’sAIsystemscanperformquitewelltasksthatarewelldefinedandnormallyrequirehumanintelligence.Thelearningprocess,acriticalcomponentofmostAIsystems,takestheformofML,whichreliesonmathematics,statistics,anddecisiontheory.AdvancesinMLandespeciallyindeeplearningalgorithmsareresponsibleformostoftherecentachievements,suchasself-drivingcars,digitalassistants,andfacialrecognition.

2

Thefinancialsector,ledbyfinancialtechnology(fintech)companies,hasbeenrapidlyincreasingtheuseofAI/MLsystems(Box1).Recentadoptionbythefinancialsectoroftechnologicaladvances,suchasbigdataandcloudcomputing,coupledwiththeexpansionofthedigitaleconomy,madetheeffectivedeploymentofAI/MLsystemspossible.Arecentsurveyoffinancialinstitutions(WEF2020)showsthat77percentofallrespondentsanticipatethatAIwillbeofhighorveryhighoverallimportancetotheirbusinesseswithintwoyears.McKinsey(2020a)estimatesthepotentialvalueofAIinthebankingsectortoreach$1trillion.

AI/MLcapabilitiesaretransformingthefinancialsector.

3

AI/MLsystemsarereshapingclientexperiences,includingcommunicationwithfinancialserviceproviders(forexample,chatbots),investing(forexample,robo-advisor),borrowing(forexample,automatedmortgageunderwriting),andidentityverification(forexample,imagerecognition).Theyarealsotransformingtheoperationsoffinancialinstitutions,providingsignificantcostsavingsbyautomatingprocesses,usingpredictiveanalyticsforbetterproductofferings,andprovidingmoreeffectiveriskandfraudmanagementprocessesandregulatorycompliance.Finally,AI/MLsystemsprovidecentralbanksandprudentialoversightauthoritieswithnewtoolstoimprovesystemicrisksurveillanceandstrengthenprudentialoversight.

TheCOVID-19pandemichasfurtherincreasedtheappetiteforAI/MLadoptioninthefinancialsector.BoE(2020)andMcKinsey(2020b)findthataconsiderablenumberoffinancialinstitutionsexpectAI/MLtoplayabiggerroleafterthepandemic.Keygrowthareasincludecustomerrelationshipandriskmanagement.BanksareexploringwaystoleveragetheirexperienceofusingAI/MLtohandlethehighvolumeofloanapplicationsduringthepandemictoimprovetheirunderwritingprocessandfrauddetection.Similarly,supervisorsrelyingonoff-siteintensivesupervisionactivitiesduringthepandemiccouldfurtherexploreAI/ML-supportedtoolsandprocessesinthepost-pandemicera.

TherapidprogressinAI/MLdevelopmentcoulddeepenthedigitaldividebetweenadvancedanddevelopingeconomies.AI/MLdeployment,andtheresultingbenefits,havebeenconcentratedlargelyinadvancedeconomiesandafewemergingmarkets.Thesetechnologiescouldalsobringsignificantbenefitstodevelopingeconomies,includingenhancedaccesstocreditbyreducingthecostofcreditriskassessments,particularlyincountriesthatdonothaveanestablishedcreditregistry(Syandothers2019).However,theseeconomiesarefallingbehind,lacking

1FollowingtheOxfordDictionary,AIisdefinedasthetheoryanddevelopmentofsystemsabletoperformintellectualtasksthatusuallyrequirehumanintelligence.MListhelearningcomponentofanAIsystem,andisdefinedastheprocessthatusesexperience,algorithms,andsomeperformancecriteriontogetbetteratperformingaspecifiedtask.GiventhatAIandMLheavilyoverlapandthatmoststatementsinthispaperholdtrueforbothconcepts,thetermsareoftenusedasapair(AI/ML).

2SeeAnnex1formoredetails.

3Thisincludesrevenuegainsandcostsavings.

6IMFDEPARTMENTALPAPERSPoweringtheDigitalEconomy

thenecessaryinvestment,accesstoresearch,andhumancapital.

4

Bridgingthisgapwillrequiredevelopingadigital-friendlypolicyframeworkanchoredaroundfourbroadpolicypillars:investingininfrastructure;investinginpoliciesforasupportivebusinessenvironment;investinginskills;andinvestinginriskmanagementframeworks(IMF2020).

Cooperationamongcountriesandbetweentheprivateandpublicsectorscouldhelpmitigatetheriskofawideningdigitaldivide.Sofar,globalinitiatives—includingthedevelopmentofprinciplestomitigateethicalrisksassociatedwithAI(UNESCO2021;OECD2019),callsforcooperationoninvestingindigitalinfrastructure(see,forexample,GoogleandInternationalFinanceCorporation(2020)),andtheprovisionofaccesstoresearchinlow-incomecountries(see,forexample,AI4G)—havebeenlimited.Multilateralorganizationscouldplayanimportantroleintransferringknowledge,raisinginvestments,buildingcapacity,andfacilitatingapeer-learningapproachtoguidedigitalpolicyeffortsindevelopingeconomies.Similarly,themembershipinseveralintergovernmentalworkinggroupsonAI(suchastheGlobalPartnershiponArtificialIntelligenceandtheOECDNetworkofExpertsonAI,amongothers)couldbeexpandedtoincludeless-developedeconomies.

AI/MLadoptioninthefinancialsectorisbringingnewuniquerisksandchallengesthatneedtobeaddressedtoensurefinancialstability.AI/ML-baseddecisionsmadebyfinancialinstitutionsmaynotbeeasilyexplainableandcouldpotentiallybebiased.AI/MLadoptionbringsinnewuniquecyberrisksandprivacyconcerns.FinancialstabilityissuescouldalsoarisewithrespecttotherobustnessoftheAI/MLalgorithmsinthefaceofstructuralshiftsandincreasedinterconnectednessthroughwidespreadrelianceonfewAI/MLserviceproviders.Chapter2explorestheadoptionofAI/MLinthefinancialsectorandpossibleassociatedrisks,Chapter3discussesrelatedpolicyconcerns,andChapter4providessomeconclusions.

Box1.ArtificialIntelligenceandMachineLearningCapabilities

?Forecasting.Machinelearningalgorithmsareusedforforecastingandbenefitfromusinglargedatasets.Theyusuallyperformbetterthantraditionalstatisticaloreconometricmodels.1Inthefinancialsector,thisisusedinsuchareasascreditriskscoring,economicandfinancialvariablesforecasting,riskmanagement,andsoon.

?Naturallanguageprocessing.Artificialintelligencesystemscancommunicatebyunderstandingandgeneratinghumanlanguage.Boostedbydeeplearningandstatisticalmodels,naturallanguageprocessinghasbeenusedinthefinancialsectorinsuchapplicationsaschatbots,contractreviewing,andreportgeneration.

?Imagerecognition.Facialandsignaturerecognitionisbeingusedbysomefinancialinstitutionsandfinancialtechnologycompaniestoassistwithcarryingoutcertainanti-moneylaundering/combatingthefinancingofterrorism(AML/CFT)requirements(forexample,theidentificationandverificationofcustomersforcustomerduediligenceprocess),andforstrengtheningsystemssecurity.

?Anomalydetection.Classificationalgorithmscanbeappliedtodetectrareitems,outliers,oranomalousdata.Inthefinancialsector,insidertrading,creditcardandinsurancefrauddetection,andAML/CFTaresomeoftheapplicationsthatleveragethiscapability(Chandola,Banerjee,andKumar2009).

4SeeAlonsoandothers(2020)forabroaderdiscussionaboutpossibleimplicationsofAIondevelopingeconomies.Inparticular,thepaperfindsthatthenewtechnologyriskswideningthegapbetweenrichandpoorcountriesbyshiftingmoreinvestmenttoadvancedeconomieswhereautomationisalreadyestablished,withnegativeconsequencesforjobsindevelopingeconomies.

7IMFDEPARTMENTALPAPERSPoweringtheDigitalEconomy

2.ArtificialIntelligenceintheFinancialSector

Thecapabilityofacquiringlargesetsofdatafromtheenvironmentandprocessingitwithartificialintelligence(AI)andmachinelearning(ML)ischangingthefinancialsectorlandscape.AI/MLfacilitatesenhancedcapacitytopredicteconomic,financial,andriskevents;reshapefinancialmarkets;improveriskmanagementandcompliance;strengthenprudentialoversight;andequipcentralbankswithnewtoolstopursuetheirmonetaryandmacroprudentialmandates.

A.Forecasting

AI/MLsystemsareusedinthefinancialsectortoforecastmacro-economicandfinancialvariables,meetcustomerdemands,providepaymentcapacity,andmonitorbusinessconditions.AI/MLmodelsofferflexibilitycomparedtotraditionalstatisticalandeconometricmodels,canhelpexploreotherwisehard-to-detectrelationshipsbetweenvariables,andamplifythetoolkitsusedbyinstitutions.EvidencesuggeststhatMLmethodsoftenoutperformlinearregression-basedmethodsinforecastaccuracyandrobustness(BolhuisandRayner2020).

WhiletheuseofAI/MLinforecastingoffersbenefits,italsoposeschallenges.Useofnontraditionaldata(forexample,socialmediadata,browsinghistory,andlocationdata)inAI/MLcouldbebeneficialinfindingnewrelationshipsbetweenvariables.Similarly,byusingAInaturallanguageprocessing(NLP),unstructureddata(forexample,theinformationinemailtexts)canbebroughtintotheforecastingprocess.However,theuseofnontraditionaldatainfinancialforecastingraisesseveralconcerns,includingthegoverninglegalandregulatoryframework;ethicalandprivacyimplications;anddataqualityintermsofcleanliness,accuracy,relevancy,andpotentialbiases.

B.InvestmentandBankingServices

Inthefinancialsector,advancesinAI/MLinrecentyearshavehadtheirgreatestimpactontheinvestmentmanagementindustry.Theindustryhasusedtechnologyfordecadesintrading,clientservices,andback-officeoperations,mostlytomanagelargestreamsoftradingdataandinformationandtoexecutehigh-frequencytrading.However,AI/MLandrelatedtechnologiesarereshapingtheindustrybyintroducingnewmarketparticipants(forexample,productcustomization),improvedclientinterfaces(forexample,chatbots),betteranalyticsanddecision-makingmethods,andcost-reductionthroughautomatedprocesses(Box2).

Comparedtotheinvestmentmanagementindustry,thepenetrationofAI/MLinbankinghasbeenslower.Thebankingindustryhastraditionallybeenattheforefrontoftechnologicaladvancements(forexample,throughtheintroductionofATMs,electroniccardpayments,andonlinebanking).However,confidentialityandtheproprietarynatureofbankingdatahaveslowedAI/MLadoption.Nonetheless,AI/MLpenetrationinthebankingindustryhasacceleratedinrecentyears,inpartonaccountofrisingcompetitionfromfinancialtechnology(fintech)companies(includingfintechlenders),butalsofueledbyAI/ML’scapacitytoimproveclientrelations(forexample,throughchatbotsandAI/ML-poweredmobilebanking),productplacement(forexample,throughbehavioralandpersonalizedinsightsanalytics),back-officesupport,riskmanagement,creditunderwriting(Box3),and,importantly,costsavings.

5

5TheaggregatepotentialcostsavingsforbanksfromAI/MLsystemsisestimatedat$447billionby2023(Digalaki2021).

8IMFDEPARTMENTALPAPERSPoweringtheDigitalEconomy

Box2.ArtificialIntelligenceinInvestmentManagement—SampleUseCases1

?Increasedmarketliquidityprovisionthroughawideruseofhigh-frequencyalgorithmictradingandmoreefficientmarketpriceformation.

?Expandedwealthadvisoryservicesbyprovidingpersonalandtargetedinvestmentadvicetomass-marketcustomersinacost-effectivemanner,includingforlow-incomepopulations.

?Enhancedefficiencywithartificialintelligenceandmachinelearning(AI/ML)takingonagrowingportionofinvestmentmanagementresponsibilities.

?MorecustomizedinvestmentportfoliosbasedonAI/MLtargetedcustomerexperiences.

?DevelopmentofnewreturnprofilesthroughtheuseofAI/MLinsteadofestablishedstrategies.

1SeeWEF(2018)foramoredetaileddiscussion.

Box3.ArtificialIntelligenceinCreditUnderwriting

?Artificialintelligence/machinelearning(AI/ML)predictivemodelscanhelpprocesscreditscoring,enhancinglenders’abilitytocalculatedefaultandprepaymentrisks.ResearchfindsthatMLreducesbanks’lossesondelinquentcustomersbyupto25percent(Khandani,Adlar,andLo2010).Thereisalsoevidencethat,givetheirgreateraccuracyinpredictingdefaults,automatedfinancialunderwritingsystemsbenefitunderservedapplicants,whichresultsinhigherborrowerapprovalrates(Gates,Perry,andZorn2002),asdoesthefacilitationoflow-costautomatedevaluationofsmallborrowers(Bazarbash2019).

?AI/ML-assistedunderwritingprocessesenabletheharnessingofsocial,business,location,andinternetdata,inadditiontotraditionaldatausedincreditdecisions.AI/MLreducesturnaroundtimeandincreasestheefficiencyoflendingdecisions.Evenifaclientdoesnothaveacredithistory,AI/MLcangenerateacreditscorebyanalyzingtheclient’sdigitalfootprint(socialmediaactivity,billspaymenthistory,andsearchengineactivity).AI/MLalsohasthepotentialtobeusedincommerciallendingdecisionsforriskquantificationofcommercialborrowers.1However,financialinstitutionsandsupervisorsshouldbecautiousinusingandassessingAI/MLincreditunderwritingandbuildrobustvalidationandmonitoringprocesses.

1SeeBazarbash(2019)foradiscussionofthepotentialstrengthsandweaknessesofAI/ML-basedcreditassessment.

AI/MLintroducesnewchallengesandpotentialrisks.TheuseofAI/MLininvestmentandbankingdependsontheavailabilityoflargevolumesofgood-quality,timelydata.Withthestorageanduseoflargequantitiesofsensitivedata,dataprivacyandcybersecurityareofparamountimportance.DifficultiesinexplainingtherationaleofAI/ML-basedfinancialdecisionsisincreasinglyanimportantissueasAI/MLalgorithmsmayuncoverunknowncorrelationsindatasetsthatstakeholdersmaystruggletounderstandbecausetheunderlyingcausalityisunknown.Inaddition,thesemodelsmayperformpoorlyintheeventofmajorandsuddenmovementsininputdataresultinginthebreakdownofestablishedcorrelations(forexample,inresponsetoacrisis),potentiallyprovidinginaccuratedecisions,withadverseoutcomesforfinancialinstitutionsortheirclients.

9IMFDEPARTMENTALPAPERSPoweringtheDigitalEconomy

C.RiskandComplianceManagement

AI/MLadvancesinrecentyearsarechangingthescopeandroleoftechnologyinregulatorycompliance.Regulatorytechnology(regtech)

6

hasassumedgreaterimportanceinresponsetotheregulatorytighteningandrisingcompliancecostsfollowingthe2008globalfinancialcrisis.Forthemostpart,technologyhasbeenusedtodigitizecomplianceandreportingprocesses(Arner,Barberis,andBuckley2017).However,advancesinAI/MLoverthepastfewyearsarereshapingriskandcompliancemanagementbyleveragingbroadsetsofdata,ofteninrealtime,andautomatingcompliancedecisions.Thishasimprovedcompliancequalityandreducedcosts.

MaturingAI/MLtechnologyhasthepotentialto

propelfurtheradoptionofregtechinthefinancial

sector.Accordingtoarecentglobalsurvey,AI/MLis

thetoptechnologyunderconsiderationamong

regtechfirms(Schizasandothers2019;Figure1).

IncreasedadoptionofAI/MLinregtechhas

significantlyexpandeditsusecases,cuttingacross

banking,securities,insurance,andotherfinancial

services,andcoveringawidevarietyofactivities.

Theseincludeidentityverification,anti-money

laundering/combatingthefinancingofterrorism,

frauddetection,riskmanagement,stresstesting,

microprudentialandmacroprudentialreporting,as

wellasco

溫馨提示

  • 1. 本站所有資源如無特殊說明,都需要本地電腦安裝OFFICE2007和PDF閱讀器。圖紙軟件為CAD,CAXA,PROE,UG,SolidWorks等.壓縮文件請下載最新的WinRAR軟件解壓。
  • 2. 本站的文檔不包含任何第三方提供的附件圖紙等,如果需要附件,請聯(lián)系上傳者。文件的所有權(quán)益歸上傳用戶所有。
  • 3. 本站RAR壓縮包中若帶圖紙,網(wǎng)頁內(nèi)容里面會有圖紙預(yù)覽,若沒有圖紙預(yù)覽就沒有圖紙。
  • 4. 未經(jīng)權(quán)益所有人同意不得將文件中的內(nèi)容挪作商業(yè)或盈利用途。
  • 5. 人人文庫網(wǎng)僅提供信息存儲空間,僅對用戶上傳內(nèi)容的表現(xiàn)方式做保護處理,對用戶上傳分享的文檔內(nèi)容本身不做任何修改或編輯,并不能對任何下載內(nèi)容負責。
  • 6. 下載文件中如有侵權(quán)或不適當內(nèi)容,請與我們聯(lián)系,我們立即糾正。
  • 7. 本站不保證下載資源的準確性、安全性和完整性, 同時也不承擔用戶因使用這些下載資源對自己和他人造成任何形式的傷害或損失。

評論

0/150

提交評論