Documents

Ready to Scale Ai Idc 88025788USEN

Description
Description:
Categories
Published
of 17
All materials on our website are shared by users. If you have any questions about copyright issues, please report us to resolve them. We are always happy to assist you.
Related Documents
Share
Transcript
    May2019,IDC#US45083519 White Paper Ready to Scale AI? Don't Suffer from Core Starvation Sponsoredby:IBMPeterRuttenMay2019 IDC OPINION ThebusinessopportunitiesthatcanbeachievedwithAIareexceptionallypromising.BusinessesandotherorganizationsknowthatnotactingonAIcouldpotentiallybeabusinessdisasterascompetitorsgainawealthofpreviouslyunavailabledataandcapabilitiestogrowanddelighttheircustomerbase.Few,ifany,businessestodaybelievethat AIisnotforus orthat AIismostlyhype. Rather,seriousAIinitiativesarebeingundertakenworldwide,acrossindustries,andacrosscompanysizes.Manyorganizations'linesofbusiness(LOBs),ITstaff,datascientists,anddevelopershavebeenworkingtolearnaboutAI,understandtheusecases,defineanAIstrategyfortheirbusiness,launchinitialAIinitiatives,anddevelopandtesttheresultingAIapplicationsthatdelivernewinsightsandcapabilitiesusingmachinelearning(ML)algorithms,especiallydeeplearning(DL).Organizationsarenowreadytoscaletheseinitiativesandnewquestionsthatemerge.Theyknow  —  indeed,theymayhavefirst-handexperience  —  thattheycannotusestandard,multipurposeinfrastructure.Also,theyhaveestablishedthatAItraining(thetrainingoftheAImodel)andAIinferencing(theuseofthetrainedmodeltounderstandorpredictanevent)requiredifferenttypesofcompute.Butwhatisthatdifferentcompute?Also,shouldtheydeployon-premise,inthecloud,orahybridcloudmodel?AIapplicationsandespeciallydeeplearningsystems,whichparseexponentiallygreateramountsofdata,areextremelydemandingandrequirepowerfulparallelprocessingcapabilitiesbasedonlargenumbersofcores,andstandardCPUscannotsufficientlyexecutetheseAItasks.IDCresearchshowsthat,intermsofcorecapacity,alargegapbetweenactualandrequiredCPUcapabilitywilldevelopinthenextseveralyears.Toovercomethisgap,AIusersthathaveexperimentedwithexistinginfrastructureandthatarereadytoscaleneedtooverhaultheirinfrastructuretoobtaintherequiredparallelprocessingperformance,whichisachievedwithmultithreadedCPUscombinedwithGPUs,fastinterconnects,largeamountsofmemory,andadvancedI/Ocapabilities. SITUATION OVERVIEW BusinessesaroundtheworldarerespondingvigorouslytothenewopportunitiesofferedbyAIworkloads.IDCdefinesAIasasetoftechnologiesthatusenaturallanguageprocessing(NLP),image/videoanalytics,machinelearning,knowledgegraphs,andothertechnologiestoanswerquestions,discoverinsights,andproviderecommendations.Thesesystemshypothesizeandformulatepossibleanswersbasedonavailableevidence,canbetrainedthroughtheingestionofvastamountsofcontent,andadaptandlearnfromtheirmistakesandfailuresthroughretrainingorhumansupervision.  ©2019IDC#US450835192 MachinelearningisasubsetofAItechniquesthatenablecomputersystemstolearnandimprovetheirbehaviorforagiventaskwithouthavingtobeprogrammedbyahuman.Machinelearningmodelsarealgorithmsthatcanimproveovertimebytestingthemselvesoverandoveragainusinglargeamountsofstructuredand/orunstructureddatauntiltheyaredeemedtohave learned atask(e.g.,recognizingahumanface).Figure1illustrateshowdeeplearningisasubsetofML.TypicalDLarchitecturesaredeepneuralnetworks(DNNs),convolutionalneuralnetworks(CNNs),recurrentneuralnetworks(RNNs),generativeadversarialnetworks(GAN),andmanymore. FIGURE 1 Machine Learning and Deep Learning Source:IDC,2019 AIsoftwareplatformsinclude: ▪   ConversationalAIsoftware(e.g.,digitalassistants) ▪   Predictiveanalyticstodiscoverhiddenrelationshipsindataandmakepredictions ▪   Textanalyticsandnaturallanguageforrecognizing,understanding,andextractingvaluefromtext ▪   Voice/speechanalyticsforrecognizing,identifying,andextractinginformationfromaudio,voice,andspeech ▪   Imageandvideoanalyticsforrecognizing,identifying,andextractinginformationfromimagesandvideo,includingpatternrecognition,objects,colors,andotherattributessuchaspeople,faces,emotion,cars,andsceneryManybusinessesarewellontheirwaywithAIinitiativesandhavereachedastagewheretheyarereadytostartdeployingAIatproductionscale.OthersarestillexperimentingwithAI,whileathirdgroupiscurrentlyatthestageofevaluatingwhatAIapplicationscanmeanforitsorganization.    Machine Learning (ML) DNNEtc.CNNRNNGAN Deep Learning (DL) K-MeansNaïve BayesRandom ForestEtc. Structured DataUnstructured Data  ©2019IDC#US450835193 Withregardtothefirstgroup(readytodeploy),IDCisseeingarangeofAIusecasesthatbusinesses,governments,andotherorganizationshavebeguntoimplement.Thefivemostcommonusecasestodayare(rankedbytheamountthatbusinessesspendonthemintermsofhardware,software,andservices): ▪   Automated customer service agents. Inthebankingindustry,forexample,theseAIapplicationsprovidecustomerserviceviaalearningprogramthatunderstandscustomerneedsandproblemsandhelpsabankreducethetimeandresourcesneededforresolvingcustomerissues.Theseagentsarebecomingwidelyusedacrossindustries. ▪   Sales process recommendation and automation. Usedininvariousindustries,theseareAIapplicationsthatworkwithcustomerrelationshipmanagement(CRM)systemstounderstandcustomercontextinrealtimeandrecommendrelevantactionstosalesagents ▪   Automated threat intelligence and prevention systems. Becomingacriticalpartofthreatpreventionacrossgovernmentsandindustries,theseAIapplicationsprocessintelligencereports,extractinformationfromthem,establishrelationshipsbetweendiversepiecesofinformation,andthenidentifythreatstodatabases,systems,websites,andsoforth. ▪   Fraud analysis and investigation. Intheinsuranceindustry,butusedwidelyelsewhereaswell,theseAIapplicationsuserule-basedlearningtoidentifyfraudulenttransactions,andtheyautomaticallylearntoidentifyvariousinsurance-relatedfraudschemes. ▪   Automated preventative maintenance. Inthemanufacturingindustries,theseAIapplicationsarebasedonmachinelearningalgorithmsthatbuildanaccuratepredictivemodelofpotentialplantandmachineryfailures,reducingdowntimeandmaintenancecost.AdditionalAIusecasesthathavegainedtractioninenterprisesare(rankedinorderofspendingonhardware,software,andservices): ▪   Programadvisorsandrecommendationsystems ▪   Diagnosisandtreatmentsystems ▪   Intelligentprocessingautomation ▪   Qualitymanagementinvestigationandrecommendationsystems ▪   ITautomation ▪   Digitalassistantsforenterpriseknowledgeworkers ▪   Expertshoppingadvisorsandproductrecommendations ▪   Supplyandlogistics ▪   Regulatoryintelligence ▪   Asset/fleetmanagement ▪   Automatedclaimsprocessing ▪   Digitaltwin/advanceddigitalsimulation ▪   Publicsafetyandemergencyresponse ▪   Adaptivelearning ▪   Smartnetworking ▪   Freightmanagement ▪   Pharmaceuticalresearchanddiscovery  ©2019IDC#US450835194 Cloud Versus On-Premise Theapplicationsthataddresstheseusecasesmaybecustomdevelopedbyanorganization,maybebasedoncommercialAIsoftware,ormaybedeliveredasAISaaS.Deploymentconsiderationsforthecustomdevelopedandcommercialsoftwareareon-premise,inthecloudonIaaS,orasahybridcloud,whereintheon-premiseenvironmentinteractswithapubliccloudenvironment.Forthevariousdeploymentscenarios,solutionsmustbeconsideredfor: ▪   SecurelyprocessingthevolumeofdatathatisrequiredfortrainingAImodelswithextremelyhighperformance.TheperformancerequirementsfordeeplearningtraininginvolvetheabilitytoexecutemassivelyparallelprocessingusingGPUscombinedwithhigh-bandwidthdataingestion. ▪   SecurelyprocessingthevolumeofdatathattheAImodelwillperforminferencingonwithextremelyhighperformance.PerformancewithregardtoinferencingmeanstheabilitytoprocessincomingdatathroughthetrainedAImodelanddelivernearreal-timeAIinsightsordecisions.Fordatascientistsanddevelopers,itcansometimesbeeasiertostartanAIinitiativeinthecloud,savingthemfromhavingtoarrangeon-premisecomputethat,fordeeplearning,typicallyneedstobeaccelerated.AcceleratedAIcloudinstancesareavailableonmostpublicclouds,usuallywithopensourceAIstacks.Ofcourse,withacceleratedcloudinstancesforAItraining,thecloudSPdictateswhat'savailabletotheenduserintermsofprocessors,coprocessors,interconnects,memorysizes,I/Obandwidth,andsoforth.NotallcloudSPsofferthemostoptimizedcombinationsofthesecomponents,whichultimatelydeterminethespeedandqualitywithwhichdatascientistscandeveloptrainingmodels.Asaresult,manyorganizationsoptforon-premisedeployment.DuringtheirAIexperimentsinthepastfewyears,manyorganizationsfoundthemselves hittingthewall withtheirstandardinfrastructureorwiththebasiccloudinstances.Trainingmodelstooktoolong,andinferencingwastooslow.IDCresearchshowsthat77.1%ofrespondentssaytheyranintooneormorelimitationswiththeirAIinfrastructureon-premiseand90.3%ranintocomputelimitationsinthecloud.Figure2showsthekindsofhardwarelimitationsthatarebeingencounteredmostoften,on-premiseandinthecloud.Mostorganizationsexperienceacombinationofthesehurdles.Theresponseshavebeenrankedbymostoftencitedon-premisehurdles.

Folding Gate

Sep 10, 2019
Search
Tags
Related Search
We Need Your Support
Thank you for visiting our website and your interest in our free products and services. We are nonprofit website to share and download documents. To the running of this website, we need your help to support us.

Thanks to everyone for your continued support.

No, Thanks
SAVE OUR EARTH

We need your sign to support Project to invent "SMART AND CONTROLLABLE REFLECTIVE BALLOONS" to cover the Sun and Save Our Earth.

More details...

Sign Now!

We are very appreciated for your Prompt Action!

x