版權說明:本文檔由用戶提供并上傳,收益歸屬內容提供方,若內容存在侵權,請進行舉報或認領
文檔簡介
2
Contents
Acknowledgement 3
Abstract 4
Preface 4
KT'sAITransformationutilizingAgentandData 4
NTTDOCOMO'sStrategicJourneytowardsDigitalTransformationandEnhanced
CustomerExperience 5
ChinaMobile'sTransitiontoAI+toAmplifyScaleEmpowerment 5
1LLMAdoptionStrategiesinIndustry 6
2EmergingChallengesandTechnicalForesights 7
2.1AIApplicationPerspective 7
2.2DataFuelingPerspective 9
3ApplicationToolingPlatforms 11
3.1ChinaMobileJiutianLargeLanguageModelApplicationPlatform 11
3.2DOCOMOLLMValue-AddedPlatform 12
3.3KTSLM/LLMPlatform 13
4GenerativeAIApplicationCases 14
4.1GenerativeAIforNetworkO&M 14
4.2GenerativeAIforCustomerService 17
5FutureOutlookandIndustrySuggestions 21
6Abbreviations 22
3
Acknowledgement
SCFAwasestablishedin2011byChinaMobile,Korea'sKT,andJapan'sNTTDOCOMO,aimingtopromoteatripartitecooperationframeworkforglobaltechnologystandardsandindustryecosystems.
In2022,theAIWorkgroupwasestablished,focusingonthedevelopmentandapplicationofAItechnology,promotingtechnicalexchangesamongmembercompanies,andguidingandfacilitatingtheapplicationandcooperationofAItechnologywithintheindustry.
ThisWhitePaperhasbeenproducedasacollectiveeffortwithintheSCFAAIWG,andonitsbehalfthefollowingeditingteam(listedinalphabeticalorder):
ChinaMobile:
LingliDeng,BoYuan,XuefengZhao,XiangyangYuan,DiJin
KT:
JiyoungKim,JaehoOh
NTTDOCOMO:
IsseiNakamura,KuanyinLiu,AoguYamada,SatomiKura,TakeshiKato
SCFAAIWG
ChinaMobileContact:
liukaixi@
KTContact:
zeeyoung.kim@
NTTDOCOMOContact:
issei.nakamura.zs@
4
Abstract
ThisdocumentanalyzesthechallengesofscaleadoptionofLargeLanguageModels(LLMs)intoindustrialapplications,highlightingtheproblemofreinventingthewheelofcommoncapabilities,theperformancebottleneckofnetworkcommunication,theimprovementofproductivitybyutilizingwork-orientedSLM/LLMbasedAIagents,andproposestechnologicaldevelopmenttrendssuchasinnovationinfundamentalalgorithms,standardizationofapplicationtoolplatforms,andCloud-Edgecollaboration.ItshowcasescontributingCSPs’strategiclayoutinAItechnology,dataintegration,applicationtoolingplatforms,aswellasavarietyofgenerativeAIapplications,andlooksforwardtothefuturedevelopmentofAItechnology,dataintegrationandindustrycollaborationrecommendations.
Preface
KT'sAITransformationutilizingAgentandData
WiththerapidadvancementofAIHWandSWtechnologies,generativeAImodelsareevolvingintovariousversions.Alongsidethis,generativeAIAgentsareswiftlypermeatingourdailylives.TheparadigmshiftstoapracticalAIAgentcompetition,reflectingusers'GenAIdemands,iscloselyrelatedtothehandlingandaccommodationofextensivecustomerdata.AsAIadvances,theimportanceofdataincorporateactivitieshasbecomeevengreater,andData-drivenAIAgentsbasedoncustomersandcompaniesareatthecenterof"CorporateTransformationUsingAI".TosucceedinAX,itisessentialtocollectandutilizedatafromcorporateactivitieseffectively,andtheprimaryinnovationofAIcompaniesmustbedrivenbyData-drivenAX.
Inthe"EraofAIAgents",whereAIisbecomingcentraltocorporateandpersonaldailyservices,KTispursuingtheenhancementofAIcompetitivenessusingAIAgentsasoneofitssuccessfultransformationdirectionsintoanAICTcompany.Underthemulti-modelline-upstrategy,whichcombinesitsself-developedAIlanguagemodelMi:dmwithmodelsbasedonopen-source,KTaimstoprovideavarietyofcustomer/industry-specificmodelsandAIAgentstothemarket,basedonhigh-qualitydatalearningandutilization.KTismovingforwardwiththegoalofenhancingproductivitybyutilizingworkAIAgentsforitsemployees,anditalsoplanstospreadnewAIexperiencestocustomersbyapplyingthemtoitsGenieTV.BydevelopingtheseAIAgentsandlaunchingservices,KTexpectstosecurecustomerAIdataandconceivespecificAIbusinessmodelsutilizingthedata.StrengtheningAIMSPcompetitivenessbyprovidingModelasaServicecomprehensivelyandthroughglobalAIAgenttechnology/businesscooperation,KTwillleadtheAImarketandecosystemconstruction.
5
NTTDOCOMO'sStrategicJourneytowardsDigitalTransformationandEnhancedCustomerExperience
NTTDOCOMO(DOCOMO)setthegoalofimprovingcustomerexperienceandreformingbusinessstructurewithdigitalizationofbusinessmanagement,andpromotionandexecutionofdatautilizationasourmedium-termstrategytoward2025.InitiativesindigitaltransformationatDOCOMOincludenetworkoptimizationthroughdatautilization,AIandhumanresourcetraining,andthepromotionofdigitalmarketing.AIplatformsforimagerecognition,voicerecognition,andcustomeranalysisarebeingofferedtoenhanceDOCOMO'scompetitivenessbyapplyingthesetechnologiestoitsservices.
Since2014,DOCOMOhasbeenbuildingabigdatainfrastructurethatcollectsdatasuchasuserinformation,usagehistory,networktrafficandpaymenthistoryfromalmost100millionusersandmorethan270,000basestationsasanefforttopromotedigitalizationofbusinessmanagementanddatautilization.TheplatformincorporatesexternaldatafrombusinesspartnersandAItechnologiestocreatevalueacrossvariousbusinessfields,suchasMobilityasaService,retail,banking,andthemetaverse.
LeveragingnewtechnologieslikegenerativeAItofindnewrevenuestreamsandgrowthebusinessisnotaneasytask.Itrequiresstrategicplanning,includingtrainingpersonnel,andalotoftrialanderror.DOCOMOisnotonlyfocusingondevelopingthefoundationaltechnologiesforgenerativeAIbutisalsoactivelyworkingonvariousinitiativestocreateusecasesandtrainpersonnelthroughcontinuousexperimentationandrefinement.
ChinaMobile'sTransitiontoAI+toAmplifyScaleEmpowerment
Inthefaceofthewaveofchange,ChinaMobile,asthelargestmobilecommunicationoperatorintheworld,hasalwaysanchoreditsstrategicpositioningof"world-classinformationservicetechnologyinnovationcompany".
Intermsofnetworkcomputinginfrastructure,acommunicationnetworkwiththewidestcoverageandthelargestuserscaleintheworldhasbeenbuilt,withmorethan1.9million5Gbasestationsaccountingfor30%oftheworld'stotal,over90landandseacablesystemsconnecting78countries,andthelargestsingleintelligentcomputingcenterofglobaloperatorswith18000GPUcards.
Jiutian,aseriesoflargefoundationmodelsoflanguage,vision,voice,structureddataandmulti-modalityhavebeenconstructed,ontopofwhichmorethan40largeindustrymodelsarelaunched,formingacomprehensiveAIportfolioincludingplatforms,capabilities,andlarge-scaleapplications.Over10,000"AI+"projectshavebeenlaunchedtopromotetheintelligentandgreendevelopmentofvariousindustries,suchasenergy,manufacturing,medicalcaring,transportationandothers.
Alongtheway,itisnoticedthatthetransitionto"AI+"signifiestheshiftofAItechnologyfromameretechnicalapplicationtoacomprehensiveempowermentdeeplyintegratedintoindustrialdevelopment.Thechallengesfacedinthisprocessincludethe
6
limitationsofLLMsincriticaltaskexecution,thewasteofresourcescausedbytherepetitivedevelopmentofcommoncapabilities,andthebottleneckeffectofnetworkcommunication.
Toaddressthesechallenges,ChinaMobilecallsonallpartiesintheindustrytoworktogetherinbuildingacomprehensive"AI+"industryecosystemtopromoteinnovationsatthefundamentalalgorithmlevel,standardizationofapplicationtoolingplatforms,andnewmodelsofCloud-Edgecollaboration
1LLMAdoptionStrategiesinIndustry
Artificialintelligence,representingthenewgenerationofinformationtechnology,israpidlyemergingasasignificantdrivingforcefornewqualityproductivity.Amongthese,generativeAItechnologybasedonLLMsissignificantlyempoweringvariousindustries,leadingtoanexplosivegrowthintheapplicationofAImodelsacrossindustries,heraldingthearrivalofatechnologicalandindustrialrevolution,wheretheinformationservicesystemandtheeconomicandsocialoperationsystemsaredeeplyintegrated,profoundlychangingpeople'slifestylesandmodesofproduction.
LLMshavedemonstratedextensiveandprofoundimpactsoncurrentindustrialapplications,emergingaspivotaltoolsinthedigitaltransformationofenterprises.Fromknowledgemanagementtohandlingcomplextasks,LLMsareprogressivelyintegratingintocorebusinessprocesses.Onenotableapplicationisretrieval-augmentedgeneration(RAG),whichcombinesexternalknowledgebaseswithgenerativecapabilitiestoeffectivelyaddresscomplexqueries.Thisapproachisparticularlyeffectiveincustomerservice,whereLLMsassistcompaniesinextractingpreciseanswersfrommassiveinternaldocuments,therebyenhancingserviceefficiency.Moreover,LLMsplayasignificantroleinbuildingandmanagingenterpriseknowledgebases,facilitatingintelligentqueryingandupdatingthroughnaturallanguageunderstandingandknowledgeextraction.Inhandlingcomplextasks,LLMsexhibitpowerfulcapabilitiessuchasautomatedreportwriting,marketingcopygeneration,andcodegeneration,significantlyboostingproductivityandautomatingbusinessprocesses.LLMshavealsofoundwidespreaduseinautomatedcustomerservicesystems,wheretheirdeepunderstandingofnaturallanguageallowsthemtohandlecomplexcustomerintentionsandcontextualinteractionsbeyondthereachoftraditionalchatbots.Additionally,LLMscontributetopersonalizedrecommendationsbygeneratingcustomizedcontent,offeringprecisesuggestionsthathelpbusinessesachievehighercustomersatisfaction.Torealizetheseapplications,LLMsleveragevarioustechniquestooptimizetheirperformanceinspecificscenarios.TheadoptionofLLMsinindustrycanproceedindifferentways,dependingonthetechnologicalrequirementsandapplicationcontext.Forapplicationswithlowertechnicalbarriers,enterprisescanquicklydeployL0andL1modelsbyintegratingdomain-specificknowledgebases,makingthisapproachsuitableforscenariosthatrequirerapidimplementationwithoutintensivemodeloptimization.Inscenariosrequiringdomain-specificcustomization,L0modelscanbefine-tunedbyuploadingcustomizeddatasetsandapplyinglow-codeconfigurationtoproduceL1modelsadaptedtospecifictasks.Thismethodsuitssituationswheredata
7
accumulationandmodeladaptabilityareneeded,allowingformorepreciseresponsestoparticularbusinessrequirements.Forapplicationswithhighertechnicaldemandsandmorecomplexcontexts,enterprisescanadoptacomprehensivemodeldevelopmentprocess,encompassingdatacollection,processing,pre-training,andfine-tuning,ensuringmodelperformanceandstabilityinintricateapplicationsandmeetingtheneedsofhigh-precision,high-reliabilityoperations.Furthermore,LLMdeploymentcanberealizedthroughmulti-modelconvergenceplatforms,enablingbroadercollaborativeapplications.Enterprisescanutilizemodularpluginsandcentralizedagentstobuildcomplexbusinesssystemsthatintegratemultiplemodels,therebyfacilitatingcross-industryapplicationexpansionandfulfillingtherequirementsofsophisticatedapplicationecosystems.
Inconclusion,theindustrialdeploymentofLLMsspansfrombasicknowledgebaseintegrationtofull-scalemodelcustomizationandmulti-modelmanagement,creatingamulti-layeredapplicationsystemthatrangesfromlowtechnicalbarrierstohighlycustomizedimplementations.Throughthesediverseapproaches,LLMsaredrivingthedevelopmentofintelligentindustries,providingflexibleandpersonalizedsolutionsacrosssectors,andempoweringenterpriseswithefficientoperationsandintelligentdecision-makingcapabilities.
2EmergingChallengesandTechnicalForesights
Withthein-depthdevelopmentofthefourthindustrialrevolutioncharacterizedbydigitalintelligence,thereisaforeseeabletrendofthemutualembracebetweentraditionalindustriesandAItechnologytoaddressemergingchallengesforLLMscaleadoption:ontheonehand,thedeepeningintegrationofindustryinformationresourcesanddatagovernanceempowerstheinnovationofLLMapplicationsbyprovidingdesiredrawdatamaterials;ontheotherhand,continuousinnovationinLLMalgorithmsandengineeringtoolsaddressestheapplicabilityandeconomicissuesoflarge-scaleproductionenvironmentapplications.
2.1AIApplicationPerspective
Challenge:Largelanguagemodelscurrentlydonotpossessthecapabilitytobedirectly
appliedinkeydecision-makingprocessesinproductionenvironments.
Foresight:Innovationinbasictheoriesforreasoningacceleration,full-processautonomouscontrolatthefundamentalalgorithmlevel,torealizeautonomouscognition,autonomousevolution,andautonomousbreakthroughofAIagents.
Currently,LLMsserveaspowerfulinformationprocessingtoolscapableofexecutingtaskssuchasnaturallanguageprocessing,imagerecognition,languagetranslation,textgeneration,andimagerecognition.However,largelanguagemodelsthemselveslackenvironmentalperceptioncapabilitiesanddonotpossessautonomyandproactivedecision-makingabilities,usuallyrequiringhumaninputortriggeringtoprocess
8
informationinapresetmanner.Therefore,theyfacedifficultiesinexecutingdynamicandcomplextasks,asthesetaskstypicallyrequireperceptionandunderstandingoftherealworld,theabilitytoadapttoenvironmentalchanges,andmakingdecisionsthatalignwiththegoals.Hencefutureinnovationatthebasicalgorithmlevelwillfocusonthefollowingareas:
lAutonomouscognitionFuturealgorithmswillplacegreateremphasisontheautonomouscognitivecapabilitiesofintelligentagents,enablingthemtobetterunderstandandpredicttheirenvironment,withenhancedperception,reasoning,anddecision-makingcapabilitiesoftheenvironment,aswellasadaptabilityincomplexenvironments.
lAutonomousevolutionAlgorithmswillbedesignedtoevolveontheirown,continuouslyoptimizingtheirperformancethroughmachinelearning.Intelligentagentswillbeabletolearnfromexperience,automaticallyadjusttheirbehaviortoadapttonewtasksandenvironments,therebyimprovingtheirgeneralizationcapabilities.
lAutonomousbreakthroughToachieveahigherlevelofintelligence,algorithmsneedtobeabletoachievebreakthroughsontheirownwithouthumanintervention.Thisinvolvesinnovativealgorithmdesign,enablingAIagentstodiscovernewsolutionsandevensurpasstheperformanceofhumanexpertsinsomecases.
Moreover,tosupportthedevelopmentoftheabovecapabilities,algorithmsandAIagentoperationoptimizationandcontroltechnologyalsoneediterativeinnovation,includingreasoningaccelerationtechnologytoimprovetheresponsivenessandefficiencyofAIagentsforcomplextasks,andfull-processautonomouscontrollablealgorithmstoensuretheirstabilityandreliability.
Challenge:Theverticalrepetitivedevelopmentofalargenumberofcommon
capabilitiesleadstoresourcewasteandslowsupdatesandupgrades.
Foresight:TheriseofapplicationtoolingplatformsservingasLLMsplusdomainspecificknowledgebases,withplugins,tools,enhancingprofessionalcapabilitieswhilenotlosingbasiccapabilitiesforAIagentcustomizationdevelopment.
Inthecurrentfieldofartificialintelligence,wefaceasignificantchallenge,thatis,theverticalrepetitivedevelopmentofalargenumberofcommoncapabilities,whichnotonlyleadstoresourcewastebutalsomakestheprocessofupdatesandupgradesslow.ThisphenomenonisparticularlyprominentintherapidlydevelopingAItechnologybecauseitinvolvesalargeamountofresearchandapplicationdevelopment.
Toaddressthischallenge,itisforeseenthatanimportantdirectionforfuturetechnologicaldevelopmentistheinnovationofapplicationtoolplatforms.Inparticular,AIagentcustomizationanddevelopmentplatformswillbekey,whichcanprovidelow-codesolutionstoenablenon-technicaluserstocreateofficeagents,financialagents,andotherprofessionaltoolseasily.SuchplatformsprovidebasicLLMscombinedwithprofessionalknowledgebases,aswellaspluginsandtools,whichcanenhanceprofessionalcapabilitieswhilekeepingbasiccapabilities.
Throughsuchplatforms,onemaynotonlyreduceresourcewastebutalsoacceleratetheadvancementofAItechnology,therebypromotingthehealthydevelopmentofthe
9
entireindustry.
Challenge:The"bottleneckeffect"ofnetworkinconnectingdataandcloudcomputing
infrastructureishighlightedasthe"lastmile"ofLLMdeploymentanduserempowerment.
Foresight:Cloud-Edgecollaborationisleveragedtoenablepremise(networkedge,hometerminal)personalizedAIagentservices.
Intoday'sdigitalera,thebottleneckeffectofnetworkcommunicationhasbecometherestricting"lastmile"forLLMstoreachandempowerusers.Tosolvethisproblem,itisforeseeablethatthenewmodelofCloud-Edgecollaborationwillbecomemainstream,especiallyontheend-sideofthenetworkedgeandhometerminal,byprovidingpersonalizedintelligentagentservicesasasolution.
Thenetworkedgeandhometerminalontheend-sidearekeylinksintheCloud-Edgecollaboration,andAIagentservicescanbedeployedattheseendpointstoreducethedependenceoncentralizedcloudcomputingresources.Inthisway,datapre-processing,analysis,andresponsecanbeexecutedclosertotheuser,reducingdatatransmissionlatencyandbandwidthrequirements.e.g.,bydeployingintelligentgatewaysathometerminals,functionslikehomeautomationcontrolandsecuritymonitoringcanberealizedwithimprovedresponsivenessandreducednetworkload.
Inaddition,basedontheAIagentcustomizationanddevelopmentplatform,personalizedAIagentservicescanbecustomizedaccordingtothespecificneedsandusagehabitsofusers,providingmoreaccurateandefficientservices.Thisnotonlyincludesapplicationsinprofessionalfieldssuchasofficeagentsandfinancialagentsbutcanalsobeextendedtovariousaspectsoflifesuchaspersonalhealthmanagement,education,andentertainment.BycallingontheLLMsandprofessionalknowledgebasesdistributedintheend-to-endnetworkondemand,integratingpluginsandtools,etc.,personalizedAIagentscanenhancetheirprofessionalcapabilitieswhilenotlosingresponsivenessorcustomerexperience.
Insummary,throughthedevelopmentofCloud-EdgecollaborationandpersonalizedAIagentservices,thebottleneckproblemofnetworkcommunicationcanbeeffectivelysolved,promotingthewidespreadapplicationofLLMsinvariousfieldsandachievingatrueintelligenttransformation.
2.2DataFuelingPerspective
Challenge:Thelackofstandardizationofscattereddatahindersthestartingpointfor
data-drivenAX.
Foresight:DataGovernancefordataclassification,datastandardizationandsystematization,andgrademanagementofdata.
DatagovernanceisaseriesofprocessesrelatedtodatastandardizationforAI,toensureconsistencyindatanames,datadescriptions,anddataformats.
Thefollowingthreestagesarenecessarytoimplementdatagovernancesuccessfully.Meaningfulclassificationofcompany-widedataItiscrucialtosystematically
10
classifyvarioustypesofcompany-widedata,suchasenterprisedata,customerdata,managementdata,andinfrastructuredata,accordingtotheirtypesandpurposes.Systematicclassificationofdataisthestartingpointforefficientmanagement,utilization,andexecutionofAXinthenearfuture.
StandardizationandsystematizationofclassifieddataItisnecessarytomanageandunifystandardssothatcustomerscanunderstandfromthesameperspectiveatanycontactpointwiththepossibilityofconnectionsbetweencompany-widedata.Additionally,toimprovethereadabilityofbusinessdatabyapplyingdatastandardizationandsecureAIutilizationisneeded.
Managingdatagradesandconstructinggrade-basedcloudsconnectedwiththeappropriatesecuritysystemsItisessentialtoestablishagradingsystembycreatingmanagementindicators(quality,utilization,andcost)fordataandaccordinglyconfiguringgrade-basedclouds.Fromthesecurityenhancementperspective,itshouldbeavailabletochooseaccesscontrol,monitoring,andlogmanagementaccordingtothedatagrade.
Challenge:Dataintegrationisrequiredtomanagedatathatmakesunfragmentedinoneplace.
Foresight:Cloud-basedintegratedplatformfordatacentralization,analysis,andmodeling.
Itisrequiredtobuildacloud-basedMLdataplatformthatcancentralizecompany-widedatatoresolveexistingdataissues.
Buildinganintegrateddataplatformhelpscentralizethedataandgraduallyresolvetheissuescausedbydatasilos.
Tocontinuouslymanagethedataintegrationeffectively,itisnecessarytoconsistentlyalignamodernizationofAI,Data,andITinfrastructuresothattheprocessofdataaccumulationbythealignmentbetweenAIandDataandavailabilityofassetsbythealignmentbetweenDataandITcontinuestocirculate.
Throughthedirectionofdatacollectionandavailabilityofassets,itisexpectedtoachievetheeffectssuchasimprovingdecision-making,andpredictingissuesbyutilizingcustomerdata,managementdata,andinfrastructuredata.
Challenge:DataServingshouldbepreparedtointegrateanddistributethedataappropriately.
Foresight:Company-widecollaboration,secureandaccumulationofcapabilities,datamonetization.
Eveniftheprocessofintegrateddatagovernanceandmanagementiscarriedoutproperly,itcannotbesaidthatdata-drivenAXhasbeenfullyrealized.
Toeffectivelyintegratetheaccumulateddataanddistributeitasneeded,adedicatedorganizationthatleadsdataplanningandexecutionmustbeestablishedaswellasacollaborativesystembasedondomain-specificMLOps.
Anexpertiseindatagovernanceanddomain-specificdatacanbesecuredthroughsuchacollaborativesystem.
Additionally,itisnecessarytoexpanddatautilizationbusinessesbasedontheacquired
11
dataoperationandmanagementcapabilitiesandtoconvertthisexperienceintoexternalbusinesscapabilities.
3ApplicationToolingPlatforms
Inresponsetonumerouschallengesthatgreatlylimittheefficiencyofusersinbuildingintelligentagentsduringthedevelopmentprocess,suchashightechnicalbarriers,longdevelopmentcycles,difficultiesinimprovingmodelperformance,complexdeploymentandmaintenance,insufficientcustomizationandflexibility,difficultiesinteamcollaboration,andensuringsecuritycompliance,bothChinaMobile'sJiutianLargeLanguageModelApplicationPlatformandDOCOMO'sLLMValue-AddedPlatformenableone-stopintelligentagentapplicationdevelopment.
3.1ChinaMobileJiutianLargeLanguageModelApplicationPlatform
ChinaMobile'sJiutianLargeLanguageModelApplicationPlatformhascapabilitiessuchasapplicationconstruction,pluginintegration,modelplayground,andinferenceservices,offeringafull-process,one-stopproductiontoolforLLMapplications.Itprovidesacombinationofautonomousplanningandschedulingwithcontrollablemanualschedulingtoimproveschedulingaccuracyandreducemodelhallucinations,achievesenhancedmanagementofprivatedomainknowledgebasestoimprovetheaccuracyandprofessionalismofanswers,integratesarichsetofofficialpluginstofacilitatetheconstructionofabroaderrangeofapplicationcapabilities,integratesvariousmemorycapabilitiestopersonalizemodelresponsesandintegrateswiththird-partyapplicationstoprovideaccesstoAPIsandotherinferenceservices,whichhelpsindividualandenterprisecustomerstodeveloptheirownAIapplicationsatalowcostandinatimelyfashion,promotingtheapplicationandimplementationofLLMsinvariousindustries.
Figure1IllustrativeWorkflowofJiutianLargeLanguageModelApplicationPlatform
12
AsshowninFigure1,theJiutianLargeLanguageModelApplicationPlatformprovidesone-stopintelligentagentservicesforindividualandenterprisecustomers,insupportingmorethan100,000userstoquicklybuildmorethan1,500customizedintelligentagentapplications,coveringmultiplescenariossuchasoffice,social,entertainment,anddailylife,helpingAItoempowervariousindustries.
Lookingtothefuture,consumers'needsarebecomingincreasinglycomplex,andhigherrequirementswillbeproposedforthequality,stability,andrefinementofservices.Toempoweruserstobuilddiverseandcomplexapplications,theplatformwillfocusonstandardizingprocesses,supportingmultimodaldata,low-codeworkflows,andoptimizingthecorecapabilitiesofintelligentagents.Bycomprehensivelyupgradingintelligentagentservices,itensuresexcellentquality,stability,andreliability,enrichesthepluginecosystem,andprovidesanefficient,intelligent,andcomprehensiveconstructionexperience,inordertohelpitscustomersseizetheinitiativeindigitaltransformation,acceleratethepaceofinnovation,andachievealeapinbusinessvalue.
3.2DOCOMOLLMValue-AddedPlatform
SinceAugust2023,DOCOMOhavebeendevelopingtheLLMValue-AddedPlatformtopromotedigitaltransformationwithinourinternaloperationsandprovidenewservicesusingLLMs.ThisplatformisutilizedwithintheDOCOMOGroup,boastingapproximately7,000monthlyactiveusersandaround1,000,000callspermonth.
Themajorfeaturesavailableontheplatforminclude:
lLLMTherearevariousLLMsavailableasopen-sourcesoftware(OSS)orsoftwareasaservice(SaaS).TheseLLMsdifferintermsofcost,inp
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,僅對用戶上傳內容的表現(xiàn)方式做保護處理,對用戶上傳分享的文檔內容本身不做任何修改或編輯,并不能對任何下載內容負責。
- 6. 下載文件中如有侵權或不適當內容,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。
最新文檔
- 2025版?zhèn)€性化定制門窗安裝與綠色建材供應合同2篇
- 二零二五版木地板工程進度與成本管理合同4篇
- 二零二五年度游戲角色形象授權合同4篇
- 二零二五年度嬰幼兒奶粉安全風險評估與管理體系建設合同4篇
- 二零二五年度城市綠化景觀提升項目種植合同3篇
- 二零二五年度影視MV拍攝與藝人肖像權授權合同
- 二零二五年度木材貿易代理與倉儲管理合同3篇
- 二零二五年度人防工程防雷接地檢測合同2篇
- 二零二四年度信用證項下跨境貿易融資合同模板3篇
- 二零二四年度液化氣供應與綜合能源服務合同范本3篇
- 2024-2025學年山東省濰坊市高一上冊1月期末考試數(shù)學檢測試題(附解析)
- 江蘇省揚州市蔣王小學2023~2024年五年級上學期英語期末試卷(含答案無聽力原文無音頻)
- 數(shù)學-湖南省新高考教學教研聯(lián)盟(長郡二十校聯(lián)盟)2024-2025學年2025屆高三上學期第一次預熱演練試題和答案
- 決勝中層:中層管理者的九項修煉-記錄
- 幼兒園人民幣啟蒙教育方案
- 臨床藥師進修匯報課件
- 軍事理論(2024年版)學習通超星期末考試答案章節(jié)答案2024年
- 《無人機法律法規(guī)知識》課件-第1章 民用航空法概述
- 政治丨廣東省2025屆高中畢業(yè)班8月第一次調研考試廣東一調政治試卷及答案
- 網(wǎng)絡設備安裝與調試(華為eNSP模擬器)整套教學課件
- 銀行卡凍結怎么寫申請書
評論
0/150
提交評論