人工智能機器學習:深度學習應用的四個IT:IS支柱_第1頁
人工智能機器學習:深度學習應用的四個IT:IS支柱_第2頁
人工智能機器學習:深度學習應用的四個IT:IS支柱_第3頁
人工智能機器學習:深度學習應用的四個IT:IS支柱_第4頁
人工智能機器學習:深度學習應用的四個IT:IS支柱_第5頁
已閱讀5頁,還剩1頁未讀, 繼續(xù)免費閱讀

下載本文檔

版權說明: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,請進行舉報或認領

文檔簡介

PAGE

149

/10.48009/2_iis_2018_149-154

IssuesinInformationSystems

Volume19,Issue2,pp.149-154,2018

FOURIT/ISPILLARSFORARTIFICIALINTELLIGENCEMACHINELEARNING/DEEPLEARNINGAPPLICATIONS

RobertE.Samuel,

Robert.Samuel@

GeraldCormier,

gcormier@

ShannonFascendini,

Shannon.Fascendini@

ChristinaM.Stubanas,

tinastubanas@

KatherineA.Yacko,

kyacko1@

ABSTRACT

ArtificialIntelligenceMachineLearning/DeepLearning(AIML/DL)technologyadoptionforbusinessapplicationsisimpactingmanyInformationTechnology/InformationSystems(IT/IS)roles.Bridgingthedividebetweendata,insight,andactionrequiresrevisitingthedevelopmentoperatingmodels.TheevolutionoffourkeyIT/ISpillarswillbenecessarytosuccessfullyimplementtoday’sAIML/DLbusinesssolutions.BasedonqualitativeresearchwithinaFortune50U.S.-basedhealthcarecompany,thispaperassessestheIT/ISpillarsforroledefinition,requiredtechnicalskills,andbehavioralcompetencies.TheevolutionofthesepillarscouldinfluencehowindividualslearnandpreparefortherolesasAIapplicationsgainbusinessadoptionandacceptance.

Keywords:ArtificialIntelligence,MachineLearning,DeepLearning,IT/ISRoles

INTRODUCTION

McKinseyGlobalInstitutebroughtnationalspotlighttotheterm“bigdata”andtheexponentialgrowthofdatainmajorindustries(Manyika,et.al.,2011).Autonomousvehicles,advancedhealthcareresearchandpervasivevirtualassistants,socialmediaanalytics,andInternetofThings(IoT)areafewusecaseswheredatascienceandmachinelearningarebeingrapidlyadopted.Theadoptionofartificialintelligencemachinelearning/deeplearning(AIML/DL)applicationsincorporationsisforcinginformationtechnology/informationsystems(IT/IS)practitionerstoreassessthepillarsnecessarytosuccessfullyimplementbusinesssolutions(BanavarandCooper,2016)(Samuel,et.al.,2017).TheITindustryhasexperiencedcomputingerashiftsbeginningwiththefirsteraoftabulatingmachinestotheseconderaofprogrammablecomputerstotoday’seraofAIML/DL.Tabulatingmachineswereheavilyfocusedontheapplicationofmathematics.Programmablecomputerswerecodedwith“if,then,else”instructionstoproduceadeterministicoutput.Today’sAIML/DLapplicationsareabouttheprobabilisticoutcomesbasedoncontext(BanavarandCooper,2016).

AliteraturereviewindicatesthattheindustryremainsfragmentedontheexactclassificationofAI(Harper,2017)(Chen,2017)(Skansi,2018).Skansi(2018)furtherelaboratesthattherearetwomajorindustrysocietiesthatprovideaformalAIclassificationusedtoclassifyresearchpapers:theAmericanMathematicalSociety(AMS)andtheAssociationforComputingMachinery(ACM).TheAMSmaintainstheMathematicsSubjectClassification2010whichdividesAIintothefollowingsubfields:

General,

Learningandadaptivesystems,

Patternrecognitionandspeechrecognition,

Theoremproving,

Problemsolving,

Logicinartificialintelligence,

Knowledgerepresentation,

Languagesandsoftwaresystems,

Reasoningunderuncertainty,

Robotics,

IssuesinInformationSystems

Volume19,Issue2,pp.149-154,2018

PAGE

150

Agenttechnology,

Machinevisionandsceneunderstanding,and

Naturallanguageprocessing.

TheACMclassificationforAIprovidestheirsubclassesaswell(NotethatACMidentifiesmachinelearningisaparallelcategorytoAI,notsubordinatedtoit).Thesubclassesare:

Naturallanguageprocessing,

Knowledgerepresentationandreasoning,

Planningandscheduling,

Searchmethodologies,

Controlmethods,

Philosophical/theoreticalfoundationsofAI,

Distributedartificialintelligence,and

Computervision.

Skansi(2018)concludedfromthesetwoclassificationsthatthereareafewbroadfieldsofAIthatcansummarizedas:

Knowledgerepresentationandreasoning,

Naturallanguageprocessing,

MachineLearning,

Planning,

Multi-agentsystems,

Computervision,

Robotics,

Philosophicalaspects.

ThisresearchusedtheSkansibroadfieldsofAIasabasisfordiscussionandinterviews.ThesefieldsprovidethebackgroundtoassesstheIT/ISpillarsforAIML/DLbusinessapplicationdevelopment.TheidentifiedIT/ISpillarsare:

DataArchitecture-theprimaryroletogatherhighlevelbusinessneedsandrequirements(ofteninpartnershipwithadataanalyst)anddesignthesolution.Typically,thearchitectisageneralistbutoftenhasdepthofknowledgeinparticulartechnologydomain(Forrester,2017).

DataEngineering-theprimaryroletodeployaruntimeimplementationoftheapplicationandmonitorapplicationperformance,reliability,andstability.Theengineerisoftenaspecialistinaparticularoperations-basedtechnologystack(Forrester,2017)

DataScience–theprimaryroleistofollow“asetoffundamentalprinciplesthatsupportandguidetheprincipledextractionofinformationandknowledgefromdata”(ProvostandFawcett,2013,p.52).

DataAnalysis–theprimaryroleisjudgethevalueofgeneratedinsightsandtoensureapplicationsaddressimportantbusinessproblems(ProvostandFawcett,2013).DataAnalystsdobasicdescriptivestatistics,datavisualization,datasourceassessments,andcommunicatedatapointsforconclusions.

ThepotentialofAIML/DLdependsontheavailabilityoftalentandtechnologytoharnessitsvalue(Boisvert,et.al.,2017).AsbusinessAIapplicationsareincreasinginfrequencyandcomplexity,IT/ISskillsareevolving.ForresterResearch(Goetz,2017)discoveredthatenterpriseswanttogetvalueoutoftheirdatafasterandatscale,howevertechnologyalonecan’tsolvethischallengeandaddingroleswiththesoleresponsibilityofactivatingdataisnecessary.DavenportandPatil’sHarvardBusinessReviewarticle(2012)identifieddatascientistsasanimportantandemergingrolethathaslittleconsensusonwheretherolefitsinanorganization,thevaluetotheorganization,ortheprocessofdiscovery.Asaresult,itwasnecessaryfordatascientisttocrafttheirowntoolsandresearchapproaches.WilliamChen(2017),aDataScientistatQuora,summarizesthefiveskillsandcompetenciesthatheseesisimportantforadatascientistrole.Thisincludes1)programming(augments,largedatasets,createtools);2)quantitativeanalysis(experimentaldesignandanalysis,modelingofcomplexeconomicorgrowthsystems,machinelearning);3)productintuition(generatinghypothesis,definingmetrics,debugginganalysis);4)communication(communicatinginsights,

datavisualizationandpresentation,generalcommunication);and5)teamwork(beingselfless,constantiteration,sharingknowledge).Theauthorsnoticethatthreeofthefivecompetenciesarenottechnicalskills.

Inrecentyears,theemphasishasbeenonestablishingandcultivatingthedatascientistrolewithincorporations.ThefocusonthedatascientistrolehasresultedinabottleneckwhenthereareinadequateresourcestoefficientlybuildtheAIapplications.Forrester(2017)discoveredtheneedtogrowtheroleofdataengineerequallytothedatascientistroletosuccessfullydeployAIML/DLapplicationsbystating:

“Asorganizationsbegandevelopingdatascientists,theyexpectedthemtocarrytheloadofdevelopingdatalakesanddatapipelinesaswellascreatingsophisticatedanalytics.Butindividualswithcomputerscienceandstatisticalskillsarerare.Atoneaerospacemanufacturer,dataengineerstakeonthetaskstosource,wrangle,anddemocratizedataindatalakes,allowingdatascientiststofocusoninsightcreation.”

Theycontinuebystatingthatdatadevelopmentrolesarehighlyoutsourcedandtransient.Thistransientnatureoftheseengineerandscientistrolesmeansthatknowledgefromdevelopmentprojectsarelostwhenresourcesarereassigned.Dataarchitectsareoftennecessaryfordesignworkforplatformsandtoprovidedataaccesstoadhocrequests.Therefore,asdataengineeringevolves,dataarchitectureisbecomingapeerandguidestheinvestmentsrequiredfordatasourcesandserviceswhileprovidingtheplatform,frameworks,andreferencearchitectures.

Theroleofadataanalystcontinuestoevolve.NicholasChamandy(2018),theScientificDirectoratLyft,recentpublishedanarticlestatingthathave“maintainedafairlystrongsemanticdistinctionbetweenthetworoles:analystsextractinsightsfromdata,trackthehealthofourbusinessanddrivebetterdecision-making;scientistsbuildthemathematicalmodelsandalgorithmsthatpowerthecorecomponentsofourproduct.”ThisfurtherhighlightstheindustrychallengesatdefiningthepillarsandrolesofAIML/DL.

RESEARCHMETHODOLOGY

Thereisalimitedvolumeofacademicpeer-reviewedliteraturethataddresseshowAIML/DLisimpactingIT/ISpillars.Theauthorsfoundtheclearmajorityofmaterialpublishedisintradejournals.Eveninthetradejournals,thetopicisaddressedwithsignificantvariance.Whilethefieldofmachinelearning,deeplearningandartificialintelligencespansthreedecades,theacademicresearchregardingIT/ISpillarsandindustryrolesisrathersparse.

UsingthefourIT/ISpillarsasaframework,theresearchercentereduponthefollowingresearchquestions:

R1:HowhastheIT/ISpillarsdefinitionevolvedwithrespecttothedevelopmentofML/DLbusinessapplications?

R2:WhichskillsaligntotheIT/ISpillarsforthedevelopmentofML/DLbusinessapplications?

R3:WhichcompetenciesaligntotheIT/ISpillarsforthedevelopmentofML/DLbusinessapplications?

Toaddresstheseresearchquestions,theauthorsusedaqualitativeopinion-basedresearchmethodologyapproachusinginformalindividualinterviewsofsubjectmatterexpertpractitionersandmanagers.Areviewofpeer-reviewedscholarlypublicationswasusedtodeterminetheboundaries(skills,competencies,technologymaturity)oftheproblemspaceandIT/ISpillarframework.Theinterviewquestionsderivedfromarticlesthatarewithinthecomputerscience,softwareandsystemsengineering,informationscience,andartificialintelligencedomains.Theinterviewquestionswereassembledbasedonkeydomaintopicsandissuesobservedduringthereviewofexistingpeer-reviewscholarlypublications.Thequestionswerestructuredtoinquireontheparticipant’sunderstandingofhowthepillarsandroleshaveevolvedoverthepastfiveyears.TheinterviewsinvolvedaconveniencesamplesizeofnineindustrypractitionersataU.S.basedFortune50company.TheindustryexpertsspannedthefourIT/ISpillarsofdataarchitecture,dataengineering,datascienceanddataanalysis.TheinterviewswereconductedduringtheMarchand

April2018timelineviaphone,in-person,andemailcommunication.Interviewdatawasnormalizedandgroupedintocategoriesforcommonalityandroleassociation.

RESULTS

Thefirstgeneralobservationisthatthereareavarietyofrolesineachpillarthatessentiallyhavethesame,orsimilar,definition,technicalskillsandcompetencies.Dependingontheindustry,theserolescouldgobyadifferenttitle(e.g.InformationArchitect,MachineLearningEngineer,DataSpecialist,DatabaseAnalyst,BigDataAnalyst).Regardlessofthetitle,it’simportanttounderstandthekeydistinctionsbetweeneachroleandhowtheyfittogetheracrossthepillarstomakebetterdata-drivenbusinessapplications.Thedatavolume,veracity,andvariabilityareincreasingandbecomingmoreprominentwithinthesystemsinwhichusersinteracteveryday.

Analysisoftheinterviewsresultedinthefollowingobservationsacrossthefourpillars:

DataArchitecture–TraditionallydataarchitectsarenotincludedintheearlyphaseofdevelopmentAIML/DLbusinessapplications.IndustrychangesarerequiringdataarchitectstohaveabroadunderstandingofAIML/DLtechnologiesandassistinprovidingthe“holistic”viewoftheproblemdomain.Thisrolerequiresincreasedcompetenciesforteamworkandfacilitationalongwithtechnicalskills.

DataEngineering–Thisroleisbecomingmoreimportantandmoreinclusiveofskillspreviouslyheldbythedatascientist.Thedataengineerhastheprimaryroleofdatapreparationandmanipulation,fromingestiontoformatting/transformationtostoragefortheconsumptionbythedatascientist.OfthefourIT/ISpillars,thedataengineerhashadthemostimpactonamountoftechnicalskillsrequiredtobesuccessful.

DataScience–Withtheevolutionofthedataengineer,thedatascientisthasbeenaffordedmoretimetofocusondiscoveringinsights.Thereislessfocusonthetoolsneededforthedatapreparation.Thetechnicalskills/competencieshavenotbeensignificantlyimpacted,buttheirproductivityhasshiftedgiventhelesstimeneededfordatapreparation.

DataAnalysis–Thisrolewashistoricallyalignedtocorporatebusinessintelligencereportinghasexpandedtoincludebusinessacumenanddatasemanticunderstanding.Thisrolehasincreasedemphasisonthevisualizationofthedataforbusinessunderstanding.Thedataanalystrolewillneedtoexpandintonewwaysofdeliveringinsightsthatgobeyondthecurrentdeliverablesofreports,dashboards,andmessagingalerts.

Overall,manytechnicalskillsandcompetenciesdocumentedinliteraturealignstotheresearchfindings.Thisresearchdiscoveredthatduetothesacristyandhighersalariesofexperienceddatascientists,theroleofdataengineerisincreasingthroughouttheindustrytohelpaugment,andsometimesreplace,thedatapreparationtasksforthedatascientists.ThisfindingalignswiththeresearchperformedbyForrester(2017).

InterviewdatacollectedassociatedwiththeresearchquestionR1identifiedthatthepillardefinitionshaveevolvedasshowninTable1.WithrespecttoresearchquestionR2,thetechnicalskillsforeachpillarisshowninTable2.Table3highlightsthepillarcompetenciesthataddressesresearchquestionR3.

Table1.IT/ISPillarsRevisedRoleDefinitions

Table2.TechnicalSkillsforIT/ISPillars

Table3.CompetencySkillsforIT/ISPillars

Cao(2016)identifiedtherevolutionofdatascienceandanalyticsbythethreekeyindicatorsof1)adisciplinaryparadigmshift;2)atechnologicaltransformation;and3)innovativedataproducts.TheauthorsagreewiththisobservationandextendsittoincludetheevolutionoftheIT/ISpillars.Caoandthisresearchalignwiththefindingsthatmostdatascientistssimplyconductnormaldataengineeringanddescriptiveanalytics.Anorganizationrequiresdifferentroles,skillsandcompetenciesaccordingtothematuritylevelofbusinessapplicationsthroughtheeffectiveusetheIT/ISpillars.

Additionalsuggestedresearchcouldbecomparingthesefindingstothecurrentcareerpathsandavailableresourcesforindividualbecomeeducated.Sagheb-Tehrani(2015)concludedthat“thecollegecurriculumininformationsystems(IS)isrevisitedandoftenchangedininstitutionsforhighereducationtoreflectthechangesinthefield.ItisimportanttomakenecessarychangestotheIScurriculumtomakeprogramschallengingandtobetterpreparegraduatesfortoday’sjobmarket.”TheauthorsbelievethatitispossibletocapturetheevolutionoftheIT/ISpillarsforAIML/DLintheredesignofcollegecurriculums.Theavailabilityofundergraduatecurriculumsforthedataarchitectroleisparticularlysparse(Aasheim,et.al.,2015).Otherthanundergraduatecurriculums,theIT/ISindustryhasanopportunitytoincreaseunderstandingthroughamoreformalizedassociationtothepillarsandroles.

SUMMARY

ThisresearchindicatesthatAIML/DLtechnologyisimpactingmanyIT/ISroleswhendevelopingbusinesssolutions.Theinitialheavyemphasisonthedatascientistroleisresultinginaresourceconstraintthatcanbeaugmentedbythedataengineer,dataarchitectanddataanalyst.TheevolutionoffourkeyIT/ISpillars(forroledefinition,requiredtechnicalskills,andcompetencies)willbenecessarytosuccessfullyimplementtoday’sAIML/DLbusinesssolutions.TheevolutionofthesepillarscouldinfluencecurriculumchangesforuniversityIT/ISprogramsasAIapplicationsgainbusinessadoptionandacceptance.

ThefourIT/ISpillarswillneedtoexpandandevolvetheirtechnicalskillsandcompetenciestoincludeInternetofTechnology(IoT)dataandeventdriventechnologies.IoTwillimpactorganizationswithnewdatasources,tools,andarchitecturesforanalysisandinsights.Eventdrivenconceptsandtechnologieswillhelpmanageandthrottlemultiple

datastreamstodeliveradditionalinsights.TheauthorsrecommendfurtherresearchinrespecttoIoTandeventdriventechnologyastherelatetoAIML/DLtechnologies.

REFERENCES

Aaschiem,C.L.,Williams,C.,Rutner,P.,&Gardiner,A.(2015).DataAnalyticsvs.DataScience:AStudyofSimilaritiesandDifferencesinUndergraduateProgramsBasedonCourseDescriptions.JournalofInformationSystemsEducation,26(2),103-115.

BanavarG.,&CooperM.(2016).TuringLecture2017CognitiveComputing.ITNOW2016;58(4):62-63.doi:10.1093/itnow/bww117

Boisvert,D.,Topi,H.,Harris,M.D.,&Yohannan,K.(2017).ExploringtheLandscapeofDataScience.

SIGITE’17,October

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)容里面會有圖紙預覽,若沒有圖紙預覽就沒有圖紙。
  • 4. 未經(jīng)權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