




版权说明:本文档由用户提供并上传,收益归属内容提供方,若内容存在侵权,请进行举报或认领
文档简介
ShapingtheFutureofGenerativeAI
TheImpactofOpenSourceInnovation
AdriennLawson,TheLinuxFoundation
StephenHendrick,TheLinuxFoundationNancyRausch,TheLinuxFoundation
JeffreySica,TheLinuxFoundation
MarcoGerosa,Ph.D.,NorthernArizonaUniversity
Forewordby
HilaryCarter,TheLinuxFoundation
November2024
ShapingtheFutureofGenerativeAI
84%oforganizationshavemoderate,high,orveryhighadoptionofGenAI.
41%ofGenAI
infrastructurecodeisopensource.
For92%ofsurveyedcompanies,GenAIisimportant,and51%
consideritextremelyimportant.
For71%oforganizations,theopensourcenatureofa model/toolhasapositiveinfluenceonitsadoption,duetotransparencyandcostefficiency.
82%ofrespondentsagreethatopensourceAIiscriticalforapositive
AIfuture.
78%oforganizations
believeitisimportanttouseopensourcetools
hostedbyaneutralparty,primarilyduetostandards®ulationscomplianceandtrust.
Amongthosewhoserveorself-hostGenAImodels,50%useKubernetesfortheirinference
workloads.
30%oforganizationsuseproprietarydatafortheirproprietarymodels,and22%useitforopen
sourcemodels.
Mostorganizationsadopt multiplestrategiesfor hostingGenAIinference, includingself-hosting inthecloud(49%)andmanagedAPIservices(47%).
65%ofsurveyed
ForthefutureofGenAI,83%ofrespondentsagreethatAIneedstobeincreasinglyopen.
organizationsbuildandtrainGenAImodelsoncloud-based
infrastructure.
GenAIhasimproved
productivityfor79%ofrespondentsandhasallowedthemtolearnnewskillsandimprovecreativityandinnovation.
Copyright©2024
TheLinuxFoundation
|November2024.Thisreportislicensedunderthe
CreativeCommonsAttribution-NoDerivatives4.0InternationalPublicLicense
Contents
Foreword
0
4
Executivesummary
0
5
Introduction
0
7
GenAIadoptionanduseinorganizations
0
8
GenAIadoption
0
8
GenAIactivitybreakdown:Consumptiondominatesas
custommodelbuildinggainstraction
0
9
PrimaryGenAIusecases
12
HowopensourceisexpandingtheroleofGenAI
16
HighadoptersofGenAIaremorelikelytouseopen
sourcetoolsthanlowadopters
16
Thecriticalroleofopensourcetoolsandframeworks
inmodelbuildingandinference
18
Theopensourcenatureofatoolhasapositiveinfluence
onitsadoption
20
GenAIandthecloudnativeapproach
23
Cloudnativeandhybridcloudstrategiesarefoundational
tohoworganizationsdeployandhosttheirGenAImodels
23
Kubernetesasakeyenablerforhostingscalable
GenAIworkloads
25
Cloud-basedinfrastructureleadsthewayin
GenAImodelbuilding,withhybridandon-premises
solutionsremainingkey
27
ChallengesinGenAIadoption
29
PrimaryconcernsofGenAIadoption
30
InvestmentinGenAI
31
Impactonemployment
33
ThefutureofGenAIisopen
34
Topprioritiesforopensourceprojects
34
TheroleofopensourceAIinthefutureofGenAI
36
Conclusionsandrecommendations
38
Methodology
40
Aboutthesurvey
40
Data.Worldaccess
42
Respondentdemographics
42
Abouttheauthors
43
Acknowledgments
44
SHAPINGTHEFUTUREOFGENERATIVEAI3
SHAPINGTHEFUTUREOFGENERATIVEAI4
Foreword
Afewdaysbeforethisreportwaspublished,myson,whoispursuinghisBachelorofMusicdegree,calledmetoaskwhatIthoughttheimpactwouldbeof“opensourceGenerativeAI”onthemusicindustry.“DidIthinkthatopensourceGenerativeAIwouldhelp
creators,orhurtthem?”heasked.Inearlydroppedthephone.OfcourseItookhimthroughmyreasoningforwhyopennesswasthewaytobuildatrustedfuturefordigitalcreationsofallkinds,betheydigitalmusic,ordigitalapplicationsusedinanindustrycontext.
Ievenshowedhimsomedata!
WeknowthatGenAIistransformingindustriesatanunprecedentedpace.Asthistechnologymovesintothemainstream,
organizationsarerallyingaroundtheideathatAI’sfuturemustbeopen.Infact,82%oforganizationsbelieveopensourceAIiscriticalforensuringapositiveAIfuture,and83%agreethatAIneedstobeincreasinglyopentofostertrust,collaboration,andinnovation.Tome,thisistheheadlinetakeawayfromthisreport.
TheLinuxFoundationisproudtochampionthisvisionbynurturinganecosystemwhereopennessdrivesprogress.ProjectslikePyTorchandinitiativessuchastheGenerativeAICommonsexemplifyhowopensourcefuelsinnovation.Meanwhile,theLFAI&DataFoundation’sModelOpennessFrameworkanditscompaniontoolsareempoweringmodelcreatorsanduserswithpractical,transparentguidanceforbuildingandadoptingopenAIsystems.
Cloudnativetechnologiesarealsocentraltothisevolution.NotonlycancloudnativeprovideascalableandreliableplatformforrunningAIworkloadsoncloudinfrastructure,butAIisenhancingcloudnativeofferingsthemselves.Throughsharedstandards,robustframeworks,andsecureinfrastructure,theCloudNativeComputingFoundation(CNCF)isenablingenterprisestoreducecostsandacceleratetheperformanceofAIapplications.Thissymbiosisunderscoresthetransformativepowerofopensourcetomeettoday’sbusinessandtechnicalchallenges.
GenerativeAI’spotentialislimitless,butitssuccessreliesontrust,accessibility,andglobalcollaboration.IamgratefultoLFAI&DataandCNCFforsponsoringhisresearch,andindoingso,creatingdatathatcanhelpdecisionmakingtoscaleandsustainopensourceAIprojects.
Thisreportisatestamenttowhatwecanachievewhentheworldworkstogether,openlyandtransparently.Fornextgenerationcreatorslikemyson,andbusinessdecisionmakers,itprovidesareasontobeoptimistic.
HILARYCARTER
SeniorVicePresident,ResearchTheLinuxFoundation
SHAPINGTHEFUTUREOFGENERATIVEAI5
Executivesummary
Thereport,ShapingtheFutureofGenerativeAI,writtenby
theLinuxFoundation,supportstheimportantroleofopen
sourceintheevolutionandintegrationofgenerativeAI(GenAI)technologieswithinorganizations.Basedonasurveyof316
professionalsacrossdiverseindustries,thereportshowshowopensourceplatformsandtoolsarenotonlyacceleratingGenAIadoptionbutarealsosettingafoundationalframeworkfor
futureAIadvancements.Currently,94%oforganizationsareusingGenAI.Leadingusecasesincludeprocessautomation,contentgeneration,andcodegeneration.
OpensourcesoftwareisalreadyaforceshapingGenAI.On
average,41%ofanorganization’scodeinfrastructurethat
supportsGenAIisopensource.HigheradoptersofGenAIare
morereliantonopensourcecode(47%)comparedtolower
adopters(35%).OrganizationsthatarehigheradoptersofGenAIarenotjustheavyusersofopensourcetechnology;63%arealsosignificantcontributorstoopensource.Consequently,71%of
respondentsreportthatopensourcepositivelyinfluencestheirdecision-making,and73%oforganizationsexpecttoincreasetheiruseofopensourceGenAItoolsoverthenexttwoyears.
CentraltothesuccessoftheGenAIspaceareopensource
frameworkssuchasTensorFlowandPyTorchforbuildingand
trainingGenAImodelsandapplicationframeworksincluding
LangChainandLlamaIndexforinferencing.Theseopensource
frameworksenableorganizationstobuild,train,anddeploy
modelsatafractionofthecostassociatedwithproprietary
tools.Opensourcemodelsempowerorganizationstodevelop
customizedsolutionswhilepreservingtransparencyand
reducingdependencyonclosed-sourceplatforms.Thisflexibilityhasprovedessentialinindustrieswheretrust,transparency,andregulatorycompliancearecritical.
LookingtothefutureofGenAItechnology,opensource’s
influenceintheAIdomainisexpectedtoexpandfurther.This
surveyrevealsthat83%oforganizationsstronglyagreeoragreethatAIneedstobeincreasinglyopen.Additionally,82%reportthatopensourceAIisacriticalcomponentforasustainable
AIfuture,with61%expressingconfidencethatthebenefitsofopensourceoutweightheassociatedrisks.ThegrowthofopensourceGenAItechnologyislikelytobesignificant,with73%oforganizationsexpectingtoincreasetheiruseofopensource
generativetoolsoverthenexttwoyearsand26%anticipatingasubstantialriseinuse.
OrganizationsthatintegrateopensourceGenAItoolsnot
onlybenefitfromreducedcostsbutalsooftencontributetoacollaborativeecosystemthatdrivestechnologicalprogress.Thereportalsodiscussescloudnative’scriticalroleinsupporting
scalableGenAIsolutions.Cloud-basedinfrastructure,
combinedwithopensourceframeworksandtools,allows
organizationstomanageanddeploycomplexAImodelsmoreefficiently.Kubernetes,forinstance,hasemergedasakey
enablerfororchestratingscalableGenAIworkloads,with50%oforganizationsusingKubernetestohostsomeoralloftheirGenAIinferencingworkloads.
“Organizationswithhigherlevelsof
GenAIadoptionarehelpingtoshape
next-generationframeworksand
models,aligningthemmorecloselywithadvanced,real-worldusecases.”
SHAPINGTHEFUTUREOFGENERATIVEAI6
Thisreportrecommendsthatorganizationscontinueto
prioritizeopensourceintheirGenAIstrategiestoremain
competitiveandalignedwithindustrytrends.Italsohighlightstheimportanceofneutralorganizations,suchastheLinux
Foundation,CloudNativeComputingFoundation(CNCF),
andLFAI&DataFoundation,inprovidingopengovernance
structuresthatimprovetrustandcollaboration.AsAIcontinues
toreshapeindustries,opensourcewillremainindispensable,
offeringabalanced,transparent,andcommunity-ledpathway
toinnovationthatwilldefinethefutureofAItechnologies.By
offeringaccessible,adaptable,andcommunity-drivenresources,opensourcehasdemocratizedaccesstoGenAI,allowing
organizationsofallsizestoleveragecutting-edgeAIcapabilitiessecurelyandeffectively.
SHAPINGTHEFUTUREOFGENERATIVEAI7
Introduction
Thisreportexploresthedeployment,use,andchallengesofGenAItechnologiesinorganizationsandtheroleandfutureofopensourceinthisdomain.
LinuxFoundationResearchanditspartnersconductedawebsurveyfromAugustthroughSeptember2024,whichprovidedtheempiricalbasisforthisstudy.Surveyrespondentscreeningensuredthatrespondents:
•Werefamiliar,veryfamiliar,orextremelyfamiliarwiththeadoptionofGenAIintheirorganization
•Workedforanorganization
•Hadprofessionalexperience
Atotalof316respondentscompletedthesurvey.
Therearealsoavarietyofcalloutsthroughoutthisreport.
Thesecalloutsincludeselectedverbatimcommentsin(italicizedbluetextwithnobackgroundcolour)responsetoanopentextquestioninthesurvey,whichasked,“Doyouhaveanyfinal
commentsorthoughtsaboutGenAI?”
SHAPINGTHEFUTUREOFGENERATIVEAI8
GenAIadoptionanduseinorganizations
OrganizationsareadoptingGenAIbecauseof
itsabilitytoaddressabroadarrayofstrategic
andtacticalneeds,includingcontentcreation,
personalizedcustomerexperiences,decision
support,processautomation,employeetraining,andresearchandplanning.Tounderstandthe
developmentanduseofGenAIandhowopen
sourceisimpactingtheevolutionofGenAI,we
needtofirstevaluatehoworganizationsare
involvedwithGenAI,itsleadingusecases,andthematurityofGenAIdeployments.
GenAIadoption
Figure1showstheextentoftheadoptionofGenAI.ThetopchartinFigure1indicatesthat94%of
organizationsareinvolvedwithGenAIandcanbesegmentedintotwocategories:organizationswhohaveveryhighorhighGenAIadoption(42%,higherGenAIadopters)andorganizationswhohaveslightormoderateGenAIadoption(52%,lowerGenAI
adopters).WealsoseeinFigure1that84%of
organizationshaveamoderate,high,orveryhighadoptionofGenAI.
FIGURE1:HOWORGANIZATIONSAREADOPTINGGENAI
TowhatextenthasyourorganizationadoptgenerativeAI?(selectone)
Veryhighadoption:generativeAIiscriticaltowhatourorganizationdoes
Highadoption:generativeAIusedinproductioninselectedareas
Moderateadoption:experimentingwithhowgenerativeAIcanaddvalueinselectedareas
Slightadoption:researchingorevaluatinggenerativeAI
NoadoptionofgenerativeAItoolsandmodels
16%
27%
10%
6%
41%
42%
52%
2024GenAIsurvey,Q7,SampleSize=316
WhatactivitiesdoesyourorganizationundertakewithgenerativeAImodels?(selectallthatapply)segmentedby:
TowhatextenthasyourorganizationadoptedGenAI?(selectone)
WeconsumegenerativeAImodelsforinference
WebuildortraingenerativeAImodels
WeservegenerativeAImodelsinternally
Noneoftheabove
Don'tknowornotsure
65%
61%
69%
38%
33%
44%
43%
36%
52%
8%
10%
5%
7%
8%
6%
TotalSlightormoderateGenAIadoptionHighorveryhighGenAIadoption
2024GenAIsurvey,Q32byQ7,SampleSize=297,ValidCases=297,TotalMentions=479,answeredbyorganizationswhoadoptedGenAIinQ7
SHAPINGTHEFUTUREOFGENERATIVEAI9
GenAIactivitybreakdown:
Consumptiondominatesascustommodelbuildinggainstraction
CoreactivitiesrelatedtoGenAI,includingbuilding(training),
serving,andinferencing(consumingamodel),areshownin
thechartatthebottomofFigure1.Inferencing,at65%overall,isaprimaryGenAIactivity.Inferencingissignificantlyhigher
thaneitherbuildingmodels(38%)orservingthesemodels
(43%).Organizationsarechoosingtotuneand/ortraintheir
ownGenAImodelstomeetspecificbusinessneedsandmake
thesemodelsmoreaccurateandrelevant.Custommodels
alloworganizationstotailorresponses,fine-tunelanguage,
andincorporatedomain-specificknowledgetocreateoutputs
thataligncloselywiththeirbrandandindustryrequirements.
Fine-tuningamodelalsoprovidesenhancedcontrolovertheAI’sevolutionandreducesdependencyonexternalproviders.Figure1(atthebottom)alsoshowsthatorganizationsthathavea
higherlevelofGenAIadoptionalsoaremoreinvolvedinbuilding/trainingmodels(44%),internallyservingthesemodels(52%),
andconsumingthesemodels(69%).
Figure2exploresthevarioustechniquesinusetoimprovetheperformanceofGenAImodels.Theleadingtechnique,promptengineering,isshowingsignificantgainsfornearly80%of
organizationsthathaveadoptedGenAI.Promptengineeringisthepracticeofoptimizinginputs(prompts)todeliverthe
mostaccurate,relevant,orcreativeoutputsfromanAImodel.Bycarefullydesigningprompts,promptengineersimprove
modelperformance.Theeleganceofpromptengineeringisthatthisincreasedperformancedoesnotrequireanychangesto
theunderlyingGenAImodel,althoughitdoesrequireamoredetailedapproachindefininginputs.
Retrievalaugmentationgeneration(RAG)isalsoaleading
techniqueforimprovingperformance.RAGcombinesthe
poweroflargelanguagemodels(LLMs)withreal-time,
relevantinformationretrievaltogeneratehighlyinformedandcontextuallyaccurateresponses.Thisapproachaugments
themodel’soutputsbygroundingthemindomain-specific
information.RAGimprovesmodelperformancebyprovidingabridgebetweenstaticmodelknowledgeanddynamic,up-to-
datecontent,whichisidealforapplicationssuchascustomer
support,research,anddecisionsupportsystems.RAGisdrivingmaterialgainsformorethan70%oforganizationsthathave
adoptedGenAI.
Fine-tuninganLLMisyetanothertechniquethatorganizationscommonlyusetoimprovetheperformanceoftheirGenAI
models.Fine-tuningadjuststheLLM’sinternalparametersbytrainingondomain-specificdata,whichembedsspecialized
knowledgedirectlyintothemodel.Thismakesthemodelmorefluentinspecifictopicsbutlimitsittostaticknowledgepresentduringtraining.Fine-tuningisshowingsignificantgainsfor
nearly70%oforganizationsusingGenAI.
SHAPINGTHEFUTUREOFGENERATIVEAI10
FIGURE2:THETOPTHREETECHNIQUESFORIMPROVINGGENAIMODELPERFORMANCE
HowmuchhavethefollowinggenerativeAItechniquesimprovedtheperformanceofyourbaselineapproach?(oneresponseperrow)filteredfor:Whattechniquesareyouusinginyourorganization?(topthreeshown)
Promptengineering
RAG(RetrievalAugmentedGeneration)
Fine-tuningpre-trainedmodels
Percentorganizational
use
15%42%22%12%2%7%70%
21%36%16%11%4%10%49%
12%32%23%13%2%3%14%46%
0%10%20%30%40%50%60%70%80%90%100%
uExceptionalgainuConsiderablegainaModerategainuMarginalgainuNogainuDiminishedgainaDon'tknowornotsure
2024GenAIsurvey,Q33,SampleSize=297,ValidCases=297,TotalMentions=905,answeredbyorganizationswhoadoptedGenAIinQ7
2024GenAIsurvey,Q34,Samplesize=206to138,sortedbythesumof“Exceptional,considerable,andmoderategain”
SHAPINGTHEFUTUREOFGENERATIVEAI11
Theleft-handchartinFigure3showsthattext(81%),code
(74%),andstructuredortabulardata(48%)aretheleading
GenAImodalities.Text,code,andstructureddataarethemostcommonmodalitiesforGenAIbecausetheyarewidelyavailable,interpretable,andfoundationaltoabroadrangeofapplications.Textdata,whichLLMssupport,coversawidespectrumof
naturallanguageapplications,enablingmodelstogenerate
coherentresponses,summaries,translations,andotherhuman
languageoutputs.Code,asalogicalandrule-basedlanguage,ishighlysuitedforautomatingtasks,generatingscripts,andsupportingsoftwaredevelopment.Structureddata—suchastables,databases,andlabeleddatasets—provideorganizedinformationthatGenAIcanuseforpatternrecognition
anddatasynthesisinareassuchasdecisionsupportandrecommendations.
FIGURE3:COMMONGENAIMODALITIESANDDATAUSE
WhatgenerativeAImodalitiesareyouusingorplanningtouseinyourorganization?(checkallthatapply)
TextCode
StructuredortabulardataMultimodal DevOps Speech VisionAudio
Other(pleasespecify)Don,tknowornotsure
81%
74%
48%
47%
41%
35%
34%
27%
2%
1%
DoesyourorganizationhaveproprietarydatathatcouldbeusedtotrainorimprovetheperformanceofgenerativeAImodels?(selectallthatapply)
22%
oforganizations
usetheirproprietarydatainopensourcemodels
30%
oforganizations
usetheirproprietarydataintheir
proprietarymodels
2024GenAIsurvey,Q31,SampleSize=297,ValidCases=297,Total
Mentions=1,165,answeredbyorganizationswhoadoptedGenAIinQ7
2024GenAIsurvey,Q24,SampleSize=297,ValidCases=297,Total
Mentions=473,answeredbyorganizationswhoadoptedGenAIinQ7
Theright-handchartinFigure3showsthepercentageof
organizationsusingtheirproprietarydatatoimprovethe
performanceoftheirproprietaryGenAImodel(30%)oropensourceGenAImodel(22%).Someorganizationsusetheirowndatatotrainbothproprietaryandopensourcemodels.Whenweredistributethedatabasedonthesethreecategories,we
findthatorganizationsuseproprietarydatatoimprovethe
performancein22%ofproprietarymodels,13%ofopensourcemodels,and9%withbothmodels.Thisyieldsatotalof44%oforganizationsthatareusingproprietarydatatoimprovetheirmodels.
SHAPINGTHEFUTUREOFGENERATIVEAI12
PrimaryGenAIusecases
OrganizationsareusingGenAIinmanyways,althoughtherearefiveprimaryusecases.Figure4showsthattheleadingprimaryusecaseisprocessoptimizationorautomation(25%)followedbycontentgeneration(17%),codegeneration(14%),customer
serviceandsupport(11%),andresearch(6%).
Processautomation/optimizationistheleadingGenAIusecasebecauseitoffersbusinessestransformativeefficiency,reducesmanualtasksanderrors,anddecreasesoperationalcosts.Withitscapacityfornaturallanguageunderstandingandresponse,GenAIcanhandlediversequeriesandtasks,providingamoreadaptableandscalableapproachtoautomation.Byidentifyingpatternsandrecommendingimprovements,GenAInotonly
streamlinesprocessesbutalsocreatesroomforinnovation.
FIGURE4:PRIMARYGENAIUSECASES
What’syourorganization’sprimaryusecaseforgenerativeAI?(selectone)
Top5
ProcessautomationoroptimizationContentgenerationCodegeneration
CustomerserviceandsupportResearch
DataclassificationEducationandtraining
FrauddetectionandpreventionHealthcare
Strategicplanning
None,wedonotusegenerativeAIOther(pleasespecify)Don'tknowornotsure
17%
14%
11%
6%
5%
3%
3%
2%
1%
0%
11%
2%
25%
2024GenAIsurvey,Q13,SampleSize=297,answeredbyorganizationswhoadoptedGenAIinQ7
GenAIisalsousefulforcontent(17%)andcode(14%)generation.Forcontentcreation,GenAIcangeneratearticles,blogs,and
marketingmaterialsinseconds,reducingtheworkloadandensuringconsistencyinstyleandtone.Itenhancesideation,deliveringvariedperspectivesoroutlinesthathelpteams
focusonrefinement.Forcodegeneration,GenAIacceleratesdevelopment,offeringquickprototypes,codesuggestions,anddebuggingsupport.Byautomatingroutinecodingtasks,itreduceserrorsandfreesdeveloperstofocusoncomplexproblemsolving.
SHAPINGTHEFUTUREOFGENERATIVEAI13
GenAIcansupportcustomerservice(11%)byprovidinginstant,constantsupportthroughintelligentchatbotsandvirtual
assistants.Itcanimproveresponsetimes,handlelargevolumesofqueriessimultaneously,andprovideaccurate,context-awareanswers.GenAIcanalsopersonalizeinteractionsbyanalyzingcustomerhistoryandpreferences,deliveringtailoredsolutionsandproactiverecommendations.Additionally,itautomates
repetitiveinquiries,allowinghumanagentstofocusoncomplexcasesthatrequireapersonaltouch.
Figure5showsthefiveleadingusecasesinFigure4segmentedbyhowtheusecaseisintegratedintotheorganization’s
business.Figure5showsthateachofthesefiveprimaryusecasesisuniquelyintegratedintotheorganizationsthat
identifiedit.
FIGURE5:PRIMARYGENAIUSECASESSEGMENTEDBYINTEGRATIONINTOTHEBUSINESS
What’syourorganization’sprimaryusecaseforgenerativeAI?(selectone)segmentedbyHowisyourprimarygenerativeAIusecaseintegratedintoyourbusiness?(selectone)
69%
52%
51%
43%
41%41%
41%
37%
36%
21%
17%
18%
14%
12%
7%
Processautomationoroptimization
Content
generation
Code
generation
Customer
serviceandsupport
Research
GenerativeAIsupportsourinternalprocesses,workflows,ortasks
GenerativeAIisintegratedintoourproductsorservices
WearecreatingsolutionsthatenablethirdpartiestoutilizegenerativeAIintheirproducts
2024GenAIsurvey,Q13top5byQ15,SampleSize=166,DKNSandToosoontotellresponsesexcludedfromtheanalysis
SHAPINGTHEFUTUREOFGENERATIVEAI14
Processautomationoroptimizationshowsarelativelyhighlevelofsupportforinternalprocesses(51%)butalesserdegreeof
integrationwithanorganization’sproductsandservices(37%).Thislesserdegreeofintegrationreflectsthecomplexityof
aligningGenAImodelswithorganizationalworkflows
Theintegrationofcontentgeneration(52%),incontrast,ismorereadilyachievable,becauseanorganization’scontentisalreadyinahighlyconsumableformforGenAI.Codegenerationseesasignificantlyhighlevelofsupportforinternalactivities(69%)inpartbecausecodegenerationprovidesaboundeddomainthatoffersusefulresultswithoutanexcessivedegreeofintegrationorcustomizationtoanorganization’senvironment.However,
thereareseveralreasonswhytheintegrationofcodegenerationisjust17%,whichwe’llexploreinFigure6.
Customerservicehasarelativelyhighlevelofsupportfor
internalprocesses(41%)aswellasintegrationwithproducts
andservices(41%).Theelevatedinterestincreatingsolutionsbythirdparties(18%)reflectsthefactthateveryorganizationhastostaffcustomerserviceandsupportactivities,sothepayoffindevelopinganeffectivesolutionisconsiderable.
“GenAIappliesamathematicalmodel
toaninherentlysubjectiveprocess.It
willneverbegoodenoughinageneric
contextbecausedifferentpeoplewant
differentandincompatiblethingsfromit.However,ithasthepotentialtobecomegoodenoughinsmall,specializeduses,ifitcanstophallucinating.”
Researchisanotherusecasewhereadvancedanalytics
toolprovidersseetremendousopportunityforcreating
solutions.Becausemostdataincludesmetadata,dataisoftenstructured,whichimprovesitsspecificitywhilesimplifying
howorganizationscanuseit.Whiletheinternalusecasesforresearchisintuitivelyclear,thedeploymentofsuchsystemsisstillinitsinfancy.
Figure6showsthefiveleadingusecasesfromFigure4
segmentedbythelevelofadoption.
Contentgenerationisthemosthighlyintegratedusecase(Figure4)isalsoshowninFigure5asthemosthighlyfullydeployed
usecase(22%)andhasthehighestlevelofinitialproductiondeployments(46%).ThereasonforthisislikelyduetoamuchhigherROIthanotherusecasesduetotherelativeeaseof
implementationandthesignificantoperationalvalueadd.
Processautomationisfullydeployedinjust16%oforganizationsusingGenAIandanadditional32%areininitialproduction
deployment.ThisreflectsthechallengesofdefiningthescopenecessaryforprocessautomationwithaGenAImodel.
Bothcodegeneration(codespecificGenAImodels)and
customerserviceandsupport(LLMs)showlowfulldeploymentratesbutaveryhighlevelofinitialdeploymentandexperimentaldeploymentssuggestingthatfulldeploymentratescould
increasesignificantlyiftheseinitialandexperimentaldeplo
温馨提示
- 1. 本站所有资源如无特殊说明,都需要本地电脑安装OFFICE2007和PDF阅读器。图纸软件为CAD,CAXA,PROE,UG,SolidWorks等.压缩文件请下载最新的WinRAR软件解压。
- 2. 本站的文档不包含任何第三方提供的附件图纸等,如果需要附件,请联系上传者。文件的所有权益归上传用户所有。
- 3. 本站RAR压缩包中若带图纸,网页内容里面会有图纸预览,若没有图纸预览就没有图纸。
- 4. 未经权益所有人同意不得将文件中的内容挪作商业或盈利用途。
- 5. 人人文库网仅提供信息存储空间,仅对用户上传内容的表现方式做保护处理,对用户上传分享的文档内容本身不做任何修改或编辑,并不能对任何下载内容负责。
- 6. 下载文件中如有侵权或不适当内容,请与我们联系,我们立即纠正。
- 7. 本站不保证下载资源的准确性、安全性和完整性, 同时也不承担用户因使用这些下载资源对自己和他人造成任何形式的伤害或损失。
最新文档
- 中石油管道工程合同模板
- 合同视角下的人力资源规划
- 1自由平等的真谛 表格式公开课一等奖创新教学设计
- 安全培训-劳保用品使用维护
- 深化司法体制改革切实保障司法公正
- 《2025年车辆融资租赁合同》
- 公共设施修缮工程合同
- 2025年度供货合作合同协议
- 2025年新建安置房买卖合同全新版
- 2025建筑工程发包合同范本
- 2024年四川省巴中市中考文科综合试卷(含答案解析)
- 10kV电缆带电保护施工方案
- 2024年无人机驾驶员(五级)理论考试题库(含答案)
- 2024新媒体运营课件完整版
- 河南省郑州外国语2024年中考数学四模真题(含答案)
- 四川省内江市内江市第六中学2023-2024学年八年级下学期期中数学试题
- 抖音火花合同电子版获取教程
- 2024年《关税法》要点解读
- 山西省晋中市介休市2023-2024学年下学期期中测试七年级历史试卷
- 风机性能综合测试系统的研究与开发的开题报告
- JJG 365-2008电化学氧测定仪
评论
0/150
提交评论