世界经济论坛-ChatWTO:生成性人工智能与国际贸易分析(英)-2024.9_第1页
世界经济论坛-ChatWTO:生成性人工智能与国际贸易分析(英)-2024.9_第2页
世界经济论坛-ChatWTO:生成性人工智能与国际贸易分析(英)-2024.9_第3页
世界经济论坛-ChatWTO:生成性人工智能与国际贸易分析(英)-2024.9_第4页
世界经济论坛-ChatWTO:生成性人工智能与国际贸易分析(英)-2024.9_第5页
已阅读5页,还剩34页未读 继续免费阅读

下载本文档

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

GlobalFutureCouncilonInternationalTradeandInvestment

ChatWTO:

AnAnalysisofGenerativeArtificial

IntelligenceandInternationalTrade

WHITEPAPER

SEPTEMBER2024

Images:GettyImages

Contents

Executivesummary3

Introduction4

1

Overviewofglobalgovernanceefforts5

1.1Internationalefforts5

1.2Domesticlegislativeandregulatoryefforts6

2

Overviewofcurrentinternationaltraderules7

3

UrgentworkisneededontradepolicyandGAI11

Conclusion13

Contributors14

Endnotes15

Disclaimer

Thisdocumentispublishedbythe

WorldEconomicForumasacontributiontoaproject,insightareaorinteraction.

Thefindings,interpretationsand

conclusionsexpressedhereinarearesultofacollaborativeprocessfacilitatedand

endorsedbytheWorldEconomicForumbutwhoseresultsdonotnecessarily

representtheviewsoftheWorldEconomicForum,northeentiretyofitsMembers,

Partnersorotherstakeholders.

Thiswhitepaperwasdraftedwiththeaideofseverallargelanguagemodels,includingBard,ChatGPTandCohereGAI.

©2024WorldEconomicForum.Allrightsreserved.Nopartofthispublicationmaybereproducedortransmittedinanyformorbyanymeans,includingphotocopyingandrecording,orbyanyinformation

storageandretrievalsystem.

ChatWTO:AnAnalysisofGenerativeArtificialIntelligenceandInternationalTrade2

Executivesummary

GenerativeAIneedsglobalgovernancetoavoidfragmentedregulationsand

increasedtradebarriers.

Therapidadvancementofgenerativeartificial

intelligence(GAI)haspushedAItounprecedentedlevelsofdevelopmentandsophistication.Its

swiftadoption,drivenbylargelanguagemodels(LLMs),requiresinternationaldialoguetoaddresstheethicalandpoliticalchallengesinthedomainoftradepolicy.

Inrecentyears,internationalAIdiscussionshave

focusedonsharedvalues,ethicalguidelinesand

safety,withgovernmentsrapidlyimplementing

theseprinciplesintonationalregulations.However,thesenationalefforts,drivenbyvaryingpriorities,

haveledtofragmentedanddivergentrequirementsthatarelikelytocreatecross-bordertradefrictionsandunderminegovernmentalobjectives,creatingbarrierstotheuseofGAI.

TheWorldTradeOrganization(WTO)isoneofthe

feweffectiveglobalplatformsthatcanfacilitateinter-governmentalconversationsonGAIandharmonizeitsregulation.Whileexistinginternationaltradelawsprovidesomeguidance,furtherevaluationisneededtoaddressGAI’suniquechallengesandimpacts.

Keyfindings

Globalspreadandethicalconcerns:GAIis

spreadingrapidlyworldwide,bringingissuessuchasintellectualpropertyrights,privacyviolations

andcybersecuritytothefore.Examplesincludetheunauthorizeduseoftrademarkednamesandviolationsofnationalsecuritylawsduetocross-borderdatatransfers.

Efficiencybenefitsandadoption:BusinessesareadoptingGAIforitsefficiency,withstudiesshowingsignificantreductionsintaskcompletiontimesandimprovementsinoutputquality.Thisadoptionis

accompaniedbychallengesrelatedtothetechnicalandenergycostsofoperatingthesemodels.

Regulatoryefforts:Internationalbodieslike

theOrganisationforEconomicCo-operation

andDevelopment(OECD),theUnitedNations

Educational,ScientificandCulturalOrganization

(UNESCO),andtheGroupofSeven(G7),alongwithnationalgovernments,aredevelopingprinciplesandregulationstomanageGAI’srisks.Theseeffortsaim

tobalanceinnovationwithpublicsafety,privacyandethicalconsiderations.

Internationaltradeimplications:GAI’scross-

bordernaturenecessitatestheapplicationof

internationaltraderules.TheWTOandother

tradebodiesshoulduseexistingframeworksto

addressGAI’suniquecharacteristics,encouragingdiscussionsonitsimpactontrade.

Transparencyandbias:Ensuringrobust

transparencyandcybersecuritystandardsiscrucialforbuildingtrustinGAI.Mistrustduetomisinformationisasignificantrisk,highlightingtheneedforclearguidelinesandinternationalcooperation.

Employmentandenvironmentalimpact:GAI’simpactonjobsisuncertain,withpotentialjob

displacementbeingacriticalconcern.Additionally,thesubstantialenergyrequirementsforGAI

operationsposeenvironmentalchallenges.

Recommendations

–Educationalinitiatives:HostsessionstoeducateWTOmembersonGAI,itsbenefitsandrisks.

–Legislativetransparency:EncouragememberstopresenttheirGAIlegislationforbetterglobalunderstandingandalignment.

–Reviewandadaptation:AssesscurrenttraderulesforGAIapplicability,identifyinggapsanddevelopingnewguidelineswherenecessary.

–Biasandfairness:DevelopstandardstotacklebiasinGAI,ensuringequitableandresponsibleuse.

–Collaborativedevelopment:Workwith

internationalbodiestocreatecoherent

guidelinespromotingsafeandsecureGAIuse.

EstablishingcommongroundrulesattheWTOiscrucialtoensuringGAIisdevelopedinafairandsafemannerthatbenefitsindividualsandcommunitiesglobally.

ChatWTO:AnAnalysisofGenerativeArtificialIntelligenceandInternationalTrade3

.Researchers

atMITfoundthatusingChatGPT

reducedtask

completiontime

forwriting-focusedworkersby40%

andimprovedoutputqualityby18%.

Introduction

TheadventandspreadofgenerativeAIhavemajorimplicationsforinternationaltrade;policyguardrailsareneeded.

ThelaunchbyOpenAIofChatGPTinNovemberof2022demonstratedthatgenerativeartificial

intelligence(GAI)hadreachedanewlevelof

developmentandsophistication.AIcanbedefinedasacomputerperforminghuman-liketasks,1

whileGAIisarelativelyrecentsubfieldwithinAI

thatfocusesondevelopingalgorithmstogeneratenewcontentbasedonasetoftrainingdata.2GAIhasleapttotheforefrontofAI(andcurrentculture)

inlargepartduetotheabilityoflargelanguage

models(LLMs)torespondtospeechlikeahumanbeing.Thesemodelsoperateby“guessing”the

nextwordbasedonasetoftrainingdata.3

Theyoftenrequiretremendouscomputingpowertooperatebecausetheymustassessanentire

trainingdataset(whichcanbeevenlargerthanacopyoftheentireWorldWideWeb)torespondtoaparticularrequest.4

SeveralexampleshighlightthecriticalneedforinternationalorganizationsliketheWorldTradeOrganization(WTO)toconsidertheintersectionofAIandtradepolicy.

ThefirstexampleisayoungfilmmakerinAustraliawhositsdowntoworkonadraftofherlatest

script.ShelogsintoChatGPTandentersthe

prompt,“CreateashortfictionalstoryaboutNelsonMandelafightingclimatechangein2023”.She

lovesthestoryChatGPTcreatesanddecidestoincludeitinherupcomingsurrealistfilm.However,theNelsonMandelabrandnameistrademarked,andneithertheGAInorthefilmmakersecuredtherightstousethatnamebeforeproducingthefilm,violatingintellectualproperty(IP)rules.

Alternatively,considertheboardofasuccessful

telecommunicationscompanyinDenmarkthat

isreviewingproposalstomakeoperationsmore

efficient.Themostinterestingitemfordiscussion

involvesanAIchatbotpoweredbyCohere(a

CanadianLLMprovider)torespondtocustomer

servicecalls,whichisestimatedtoreducecosts

by30%.Theboardapprovestheproposalandnowthechieftechnologyofficeristaskedwithpurchasing

acomputercapableofhandlingthetremendous

amountofincomingandoutgoinginformationthe

newchatbotwillcreate.ThechieftechnologyofficerreviewsproposalsfromseveralprovidersanddecidestopurchaseacomputerfromaChineseprovider.ThecomputerandthenewsoftwareimproveefficiencyatthecompanybutalsoviolateanewEuropeanUnion(EU)lawthatbarstheuseofAIproductsfromChinaduetonationalsecurityconcerns.

Theseexamplesdemonstratehowthespreadof

GAItechnologycreatesnumerousethical,tradelawandpolicyissues.ChatGPTbecamethefastest-

growingconsumerapplicationinhistorywhenit

reached100millionusersjusttwomonthsafteritslaunch.5BusinessesarerapidlyadoptinggenerativeAIforitsefficiencybenefits.Researchersatthe

MassachusettsInstituteofTechnology(MIT)foundthatusingChatGPTreducedtaskcompletiontimeforwriting-focusedworkersby40%andimprovedoutputqualityby18%.6

However,therearelimitationstothegrowthand

efficacyofGAI.First,therearelarge-scaletechnicalandenergycostsrequiredtopowerthestorageandcomputingrequirementsofthesemodels.7AnotherpotentiallimitationonGAI’sgrowthisaccesstolargeenoughdatasetsforsufficienttraining.Forexample,nearly20%ofthetop1,000websitesintheworld

currentlyblockGAIcrawlerbots,whichresults

inGAIproviderspurchasingdatasetsfromweb

publisherstofeedanddifferentiatetheirofferings.8Theseexclusivedealstoaccessinformationraiseconcentrationandcompetitionrisks.

TherearealsoseriousrisksassociatedwithGAI.

Misinformation,disinformation,manipulation,bias,IPtheft,personalprivacyviolationsandcybersecurityissuesareallforeseeableproblems.Theimpactof

GAIonjobsisuncertain,withestimatessuggestingitcouldaffectanywherefromzeroto300million

jobsworldwide.9TheideathatGAImightdecideto“cuthumansoutoftheloop”orengageinharmfulactionsisunlikelyintheshorttermbutmaybecomeaconsiderationinthelongerterm.10

ChatWTO:AnAnalysisofGenerativeArtificialIntelligenceandInternationalTrade4

1

Overviewofglobalgovernanceefforts

EffortstogovernAIgloballyfocuson

balancinginnovationwithmanagingriskslikebias,privacyandcybersecurity.

WhileGAIcreatesbenefitsandopportunities,

policy-makersaroundtheworldhavefocused

onmitigatingpossiblenegativeconsequences

ofGAIsuchaspublicsafety,bias,privacyand

otherhumanrights,andcybersecurity.Workhasbeenongoinginseveralforumsandcountries,beginningasabroaderarticulationofethicalor

value-basedprinciplesandmovingintomore

detailedandprescriptivenationalregulationor

legislation.Countrieshavefollowedinternationallyagreedprinciplesbuthaveregulatedbasedon

nationalprioritiesandlocalculturalnorms,creatingregulatorydiversitythatwillleadtofrictionand

cross-borderchallenges.

1.1

Internationalefforts

TheOrganisationforEconomicCo-operationand

Development’s(OECD)

AIPrinciples

,createdin

2019andsupportedby47countries,werethe

firstsignificantAIpublicpolicyrecommendations.

DesignedtoguideAIactorsonthevaluesand

actionsneededtodeveloptrustworthyAIand

effectiveAIpolicies,theprinciplesincludefivevaluesandfivegovernmentpolicyrecommendations.TheAIPrincipleswereupdatedinMay2024toreflect

recentdevelopmentsandincludemoredetailed

guidance.TheOECDalsocreatedthefirstsetofcommondefinitionsandotherstandardsthathavebeenrelieduponbytheUnitedNations(UN)andnationalgovernmentsinsubsequentwork.

TheUnitedNationsEducational,ScientificandCulturalOrganization(UNESCO)releasedthe

RecommendationontheEthicsofArtificial

Intelligence

inNovember2021toaddresstheethicalrisksassociatedwithAI.11With193

countriesassignatories,thereportidentifies

11areasoffocustolimitreal-worldbiasesand

discrimination,lowerdivisionsandconflicts,and

preventinfringementsonhumanrights.Thereportalsoincludesrecommendationsformonitoringandenforcementofanypotentialethicalviolations.In

additiontotheworkatUNESCO,theUNSecretary-Generalhasannounceda32-memberHigh-LevelAdvisoryGrouponAI.12OtherUNagenciesare

lookingatsectoralramifications,suchasthe

WorldHealthOrganization(WHO)andInternationalTelecommunicationsUnion’s(ITU)jointfocusgroupontheintersectionofAIandhealth.13

TheUKgovernmentconvenedofficials,scientists

andotherexpertsatanAISafetySummitand

releasedthe

BletchleyDeclarationforAI

tojumpstart

internationalcollaborationonthesafetyand

regulationof“frontier”AI.TheBletchleypolicypaperexplicitlyrecognizedthatmanyoftherisksfromAIarecross-borderandinternationalinnature,and

globalcoordinationiscritical.TalkshavecontinuedwithmeetingsinSouthKoreainMay2024and

futuremeetingsplannedinFranceinFebruary2025.

InDecember2023,theJapanGroupofSeven

(G7)Presidency’s

HiroshimaAIProcess

detailedAIprinciplesandacodeofconduct(AIP&CoC),whichbuildsupontheworkoftheOECD,UNESCO,

BletchleyParkandseveralnationalefforts.14TheseprinciplesfocusonthedevelopmentofresponsibleAIusingarisk-basedapproachandinclude

provisionsonthedevelopmentandtestingofnewAIapplicationsaswellasrobustsecuritycontrols.Whilenon-binding,theAIP&CoCoutlinesspecificactionsthatcanbeundertakentoimplement

theprinciples.Areasforfurtherresearchand

developmentwerealsoidentified.G7members

pledgedtoimplementtherecommendationsandtostronglyencourageuptakeintheirprivatesectors.Anopenquestionremainsonhowgovernments

willimplementtherecommendationswhileavoidingfragmentationordivergencethatcouldunderminetheirobjectivesforresponsibleAI.Futureworkin

theG7willbeongoing,withcontinuedworkonvariousissuesincludingdefinitionsandsecureinformationsharingonAIsystemvulnerabilities.

ChatWTO:AnAnalysisofGenerativeArtificialIntelligenceandInternationalTrade5

1.2

oFragmentationwillcreateand

exacerbate

vulnerabilities

thatmayundercutcountries’intendedgoalsofsafety

andsecuritywhilecreatingunevenplayingfields.

InMay2024,theCouncilofEuropeadoptedaninternationaltreatyonAI.15Thetreatyincludes46CouncilofEuropememberstates,theEuropeanUnionand11non-memberstates.Itfocuses

ontheuseofAIsystemsbyboththepublicand

privatesectors,andcreatestransparencyand

oversightrequirements.Thetreatyalsoadoptsmeasuresrelatedtotheprotectionofdemocraticinstitutionsandexemptsnationalsecurity-relatedissuesfromitscoverage.

Domesticlegislativeandregulatoryefforts

SeveralcountrieshavebeguntheprocessorpasseddomesticlegislationorregulationrelatedtoAI.Canada(2017)16andSingapore(2019)wereearlymoversinthespaceaswell.China’s

ANextGenerationArtificial

IntelligenceDevelopmentPlan

,releasedin2017,

calledfortheestablishmentoflaws,regulations

andethicalframeworkstohelpgovernAI,includingthrough“internationalrules”andcooperation.BrazilisalsoclosetocompletingitsAIlegislation.17

IntheUS,theNationalArtificialIntelligenceInitiativeActof2020isthemosttangiblepieceoflegislationpassedonAI,andmuchofitsfocusisontheuseofAIindefence,nationalsecurityandintelligence.18InOctober2023,theBidenAdministrationreleasedthegroundbreakingexecutiveorder(EO)

Safe,Secure,

andTrustworthyDevelopmentandUseofArtificial

Intelligence

.TheEOoutlinedprinciplesthatwill

guideAIgovernanceintheUSandrequiredvariousUSgovernmentagenciestodevelopadditional

workonspecifictopicssuchasimpactonworkers,civilrights,patentsandmore.TherewasnoexplicitinclusionofinternationaltradeorinstructionstotheUSTradeRepresentative.TheBidenAdministrationalsosecuredvoluntarycommitmentsfromseven

majorAIcompaniesinJuly2023.19TherehavebeenseveralbillsproposedintheUSCongressonAI,butnonehavebecomelaw.20Moreover,25stateshaveproposedlegislationin2023relatedtoAI.21

TheEuropeanParliamentpassedthemost

comprehensivelegalframeworkrelatedtoAI,

knownastheAIAct,on13March2024.TheEU

beganworkin2019andsoughtextensiveinput

astheydesignedthelegislativeproposal.22The

EUAIActisacomprehensiveframeworkforAI

thatattemptstoencourageAIinnovationwhile

upholdingcitizens’rights;itwillbephasedin

startinginlate2024.23Itwasoriginallyconstructedaroundspecific-usecasesandratedtheirrisks

fromminimaltohigh,butlawmakersreviseditto

includeGAI,whosegeneral-purposecapabilities

didnotfitwithinthepreviouslydefinedrisks.The

finallegislationincludedtransparencyrequirementsforGAI,aswellasathoroughevaluationprocessforhigh-impactGAIthatcouldposesystemicrisks.

TheAIActalsoincludesseveralprovisionsthat

specificactionstakenshouldbeinlinewiththeEU’sinternationaltradecommitments.24

Sixbilateraltradeagreementsalreadyinclude

provisionsonAI,includingfiveagreements

betweenSingaporeandotherpartiesandthe

AfricaFreeTradeArea.25TheUS-EUTradeandTechnologyCouncil(TTC)addressedseveral

emergingtechnologyissuesincludingGAI,26

creatingatransatlanticAIroadmapandacodeofconduct.TheEUandIndiahavecreatedtheGlobalPartnershiponArtificialIntelligence,whichaimstocreateamemorandumofunderstandingonthetopicbySeptember2023.27Additional

conversationsarestarting,withChinapushingspecificallyforfutureAIgovernancetalks

withtheUS.

Thereisalsoagreatdealofcurrentregulatory

actionintheAIspacewith600regulatory

developmentstargetingAIprovidersjustsince

January2023.28Manyregulatorshavealsousedexistingprivacyandconsumerprotectionlaws

toinvestigateAIapplications.Forexample,

ChatGPTwasbannedinItalyonprivacygroundsin2023.29ChinahasapplieditsspeechrulesonobjectionableGAI-generatedcontent.IntheUS,thelackofcomprehensiveAIlegislationisleadingregulatorstobegintoact,30andtheUSFederalTradeCommissionhasopenedaninvestigationintoChatGPTlookingatsecurityandconsumerharm.31SectoralregulatorshavealsotakendirectactiononGAI.InSingapore,throughProject

Veritas,theMonetaryAuthorityofSingapore

hassoughttostrengthencoreprinciplesinGAIapplications.TheUKhasappliedconsumerdutyandexistingbank-modelriskmanagementrulestoAIapplications.

Allthisnationalactivity–legislative,regulatory

andbilateral–isimportanttoprotectcitizensandsocietiesfromtherisksofGAI.Italsoraisesnewquestionsabouthowdifferentruleswillimpactthecross-borderproduction,useanddisseminationofGAI.Fragmentationwillcreateandexacerbatevulnerabilitiesthatmayundercutcountries’

intendedgoalsofsafetyandsecuritywhilecreatingunevenplayingfields.Reviewingthesedifferencesconsideringtheexistinginternationaltraderuleswillhelpeducatepolicy-makersonthecurrentstateofplay,helpidentifyareaswheredifferencesexist,andpinpointadditionalneededwork.

ChatWTO:AnAnalysisofGenerativeArtificialIntelligenceandInternationalTrade6

2

Overviewofcurrent

internationaltraderules

DomesticlegislationcreatesvariedrulesforGAI,causingregulatoryfragmentationandpotentialtradebarriersininternationalmarkets.

DiscussionoftheinternationaltradeimplicationsforGAIisatanascentstage.Asreviewedin

previoussections,domesticlegislationand

regulationshavealreadycreateddifferentrulesthatwillimpacttheinternationaltradeofGAI,resultinginloopholesthatunderminethemanagementof

GAIrisks.Whilegovernmentshavecometogethertooutlinecommonprinciplesandvalues,countrieshaveprioritizeddifferentaspectsoftheseprincipleswhencraftinglocallaws.Somecountriesaretakingamarketapproach,encouraginginnovation.Othercountriesaretakingastate-drivenapproach,

ensuringAIdevelopmentisinaccordancewithpublicsafetyandotherstatepriorities.Afinal

approachisarights-drivenapproach,where

citizens’fundamentalrightsmustberespected.

Thesethreedifferentapproachesarecreating

fragmentationandfrictionforAIdevelopers

andusersandmaycreatetrade-distortingor

discriminatoryregimesthatunderminetherulesoftheWTO.Internationaltraderulesarebasedonafewcoreprinciples,including:

–Treatinglocalandforeignfirmsandproductsasequals(non-discrimination)

–Progressiveliberalizingbasedonacountry’slevelofeconomicdevelopment

–Creatingpredictabilityandtransparency

–Promotingfaircompetition

Thesegeneralprinciplesareappliedviavarious

provisionsinbilateral,regionalorglobalagreementsandwouldalsoprovidealevelplayingfieldfor

countriesgrapplingwithGAI.

Thereare,however,seriousgapswhenattemptingtoapplytheexistingrulesinthetradingsystem

totherapidinnovationsinGAI.CanGAIsecure

acopyrightinonecountrythatcanbeupheld

internationally?Ifacountrybanstheuseofits

citizens’datatotrainGAI,aretheretradelawsthatmightbeimplicated?DidtheGeneralAgreement

onTradeinServices(GATS)negotiatorsagreeto

allowunbridledtradeinGAIwhentheycommittedtoliberalizingcomputingservices,whichatthetimewaslimitedtobusinesssoftwareapplications?

Giventhenumberofcomplexinternationaltrade

andinvestmentissuesthatapplytoGAI,aswell

asthenew,technicalknowledgerequiredby

tradenegotiatorstounderstandtheissuesaroundGAI,theWTOshouldbegin:1)learningsessions

todiscussGAIanditsuniquecharacteristics,2)

informationonexistinginternationalandnational

rulesonGAI,and3)reviewsofWTOrulesthat

wouldapplytoGAI.OncetheWTOidentifieshow

currentrulescoverGAIandpotentialgaps,theWTOshouldbeginthinkingaboutthenextsteps.Given

therapiddevelopmentofadvancesinGAI,theseeffortsshouldtakeplaceatanacceleratedpace.

Thecross-cuttingnatureofGAIwillimpactmanyWTOprovisions.BelowwefocusonspecifictradeprovisionsthatofficialsshouldreviewandconsiderfortheirimpactonGAI:

GATS

ThenatureofGAImakesitchallengingtoclassifyitsusecase(mode)orlocation(origin),whicharetechnical,underlyingquestionsthatimpactthe

treatmentofGAIundertheexistingGATSrules.GATSschedulesincludeasectionon“computerandrelatedservices”,butmemberscouldprovethattheyneverenvisionedGAIwhentheymadethosecommitments.32Moreover,ifcompanies’

operationsorproductsincludeGAI,forexampleintheoperationsofafinancialservicecompanyoramultimediacompany,wouldtheobligationsoffinancialservicesormediaservicesapplyto

theoutput?Thesearequestionsthatdonot

havestraightforwardanswersbuthavesignificantimpactswithregardtotheenforcementof

internationaltraderules.

GAIcouldqualifyundermultiple“modes”of

supplythatmayrequirefutureconsiderationof

anewmodeforAIservices.Mode1isdefined

asthesupplyofaservicefromtheterritoryof

oneWTOmemberintotheterritoryofanyother

WTOmember(cross-borderorremoteprovision

ofservices)However,intermsofconsumerGAI

products,typicallytheuserlogsontoawebsite

operatedbytheGAIserviceproviderandentersaquery,whichcouldarguablybeamode2service,i.e.theconsumertravelledabroadtothe“location”oftheGAIserviceprovider.Thisdistinctionis

ChatWTO:AnAnalysisofGenerativeArtificialIntelligenceandInternationalTrade7

OGAIislikelyto

acceleratethe

creationofnew

patentableideas

withoutspecific

humanintervention,raisingquestions

aboutthelegal

frameworkfor

patentableideasdevelopedbyGAI.

importantascountriesmadedifferentcommitmentsin1994basedonthetypeofmode,aswellasthetypeofserviceprovided.

AsAIdeveloperscontemplatethestructureoftheiroperations,tradeofficialswillneedtoconsider

howtodefinetheoriginoftheserviceoftheGAI.

Willitbethelocationofserversandalgorithms,

thelocationoftheheadquarters,orthelocation

oftheuser?IfanenergycompanyusesGAIand

combinesitwiththetheirowndatatocreate

novelusecases,istheoriginthecountrywhere

theenergycompanyislocatedorwheretheGAI

servicecomesfrom?OrshouldtheLLMproviderberesponsiblefordownstreamusecases?

Finally,iftheGATSweretoapplytoaparticularGAIserviceprovision,thereareexceptionsthatcountriesmayclaimtolimittheapplicationoftheGATS,suchasthenationalsecurityexception.GAIisadual-

usetechnology,andcompaniesmaychallenge

nationalsecurityexceptionsasoverlybroadand

discriminatory.33InthecaseofbansorlimitationsonGAIusebasedondataprotectionconcerns,

governmentsmayclaimanexceptiontoprotect

privacy.CountriescouldalsoraisepublicmoralsorculturalexceptionstojustifyrestrictionsonGAIuseintheproductionofcertainimagesorspeech.

AgreementonTrade-RelatedAspectsofIntellectualPropertyRights(TRIPS)

Since2022,severallawsuitshavehighlightedthe

challengesassociatedwithGAIandIP.InNovemberof2022,agroupofsoftwareprogrammersfileda

USclassactionlawsuitagainstMicrosoft,OpenAIandGitHub,arguingthattheGitHubCoPilot

product(whichcangeneratecode)violated

copyrightlawbyscrapingcopyrightedcode.34GettyImagesfiledalawsuitagainstStabilityGAIintheUKallegingunlawfulcopyingofmillionsofcopyrightedimages.35IfandhowcompaniescantrainGAIon

copyrightedortrademarkedmaterialsisimportantasthebreadthanddiversityofdataareessentialtocreatingresponsibleGAI.

MeanwhileintheUS,copyrightprotectionhas

beeninplaceforGAIworkssince1988,withthecopyrightownedbywhoevercreatedtheGAI

system.InAugust2023,aUSdistrictcourtjudgeruledthatGAI-generatedartworkisnotprotectedundercopyrightlaw,36raisingquestionsaboutthefutureprotectionofGAI-createdwork.

Th

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。

评论

0/150

提交评论