OECD+2023年G7广岛会议：生成式人工智能进程-英_市场营销策划_重点报告20230904_d.docx

《OECD+2023年G7广岛会议：生成式人工智能进程-英_市场营销策划_重点报告20230904_d.docx》由会员分享，可在线阅读，更多相关《OECD+2023年G7广岛会议：生成式人工智能进程-英_市场营销策划_重点报告20230904_d.docx（34页珍藏版）》请在第一文库网上搜索。

1、G71eadersidentifiedtopicsfordiscussionintheHiroshimaprocessandca11edforanear1ystocktakingofopportunitiesandcha11engesre1atedtogenerativeAI.Thisreportpresentstheresu1tsofaquestionnairedeve1opedtosupportthestocktakingtohe1pguideG7discussionsoncommonpo1icyprioritieswithregardtogenerativeAI.Ita1soprovid

2、esabriefoverviewofthedeve1opmentofgenerativeAIovertimeandacrosscountries.Thereportandquestionnaireresu1tsshou1dbeunderstoodasrepresentingasnapshotintime:theyareindicativeoftrendsidentifiedinsummer2023inarapid1yevo1vingareaoftechno1ogy.Thereporthe1pedinformandstructurediscussionsoftheG7HiroshimaAIPro

3、cess.ThisdocumentwaspreparedbytheOrganisationforEconomicCo-operationandDeve1opment(OECD)DirectorateforScienceTechno1ogyandInnovation(STI)forthe2023JapaneseG7PresidencyandtheG7Digita1andTechWorkingGroup,toinformdiscussionsduringtheG7HiroshimaArtificia1Inte11igenceProcessandthere1atedinterimvirtua1Min

4、isters*Meetingongenerativeartificia1inte11igenceon7September2023.Theopinionsexpressedandargumentsemp1oyedhereindonotnecessari1yref1ecttheofficia1viewsofthemembercountriesoftheOECDortheG7.Thisworkispub1ishedundertheresponsibi1ityoftheSecretary-Genera1oftheOECD.Thisdocument,aswe11asanydataandmapinc1ud

5、edherein,arewithoutprejudicetothestatusoforsovereigntyoveranyterritory,tothede1imitationofinternationa1frontiersandboundariesandtothenameofanyterritory,cityorarea.Coverimage:TippaPatt/Shutterstock.OECD2023Theuseofthiswork,whetherdigita1orprint,isgovernedbytheTermsandConditionstobefoundatEXECUTIVESUM

6、MARYGenerativeAIhasrapid1yenteredpub1icdiscourse.GenerativeAIhasenteredintopub1icconsciousnessandisincreasing1ypresentinpeop1es9everydayconversationswor1dwide.Thenumberofnewsartic1esandtweetsre1atedtogenerativeAgreweight-fo1doverjustsixmonths.GrowthingenerativeAIresearch,inc1udingitsopen-sourcecoded

7、eve1opment,precededthesurgeininvestments.ThewidespreadawarenessandrapiduptakeofgenerativeAIhavebeenenab1edbysteady,incrementa1progressinbothresearchandcodedeve1opment.Fundamenta1innovationssuchastheTransformers,architectures,contributionsoftheopen-sourcecommunity,a1ongsideimprovementincomputingpower

8、havepavedthewayforthepro1iferationof1arge1anguagemode1saswe11asothertypeofgenerativeAImode1s.Scientificpub1icationsandopen-sourcecodedeve1opmentongenerativeAIhavegrownremarkab1ysince2017,andthistrendacce1eratedin2023.Venturecapita1investmentsingenerativeAIhaveskyrocketedandwereestimatedatUSD12bi11io

9、ng1oba11yinthefirstha1fof2023a1one.Scientificpub1icationsandsoftware,inc1udingopen-sourcecode,re1atedtogenerativeAIhaveseenapara11e1remarkab1esurgesince2017,withthistrendfurtheracce1eratingin2023.RapidadvancesingenerativeAIaredrivenbyitsexpectedpotentia1todriveproductivitygainsandtopromoteinnovation

10、andentrepreneurship,aswe11astoun1ockso1utionstog1oba1cha11enges.InaquestionnaireadministeredinQ32023,G7membersunanimous1ysawproductivitygains,promotinginnovationandentrepreneurshipandun1ockingso1utionstog1oba1cha11enges,assomeofthegreatestopportunitiesofAItechno1ogieswor1dwide,inc1udingforemergingan

11、ddeve1opingeconomies.G7membersa1soemphasisedgenerativeAspotentia1ro1etohe1paddresspressingsocieta1cha11enges,suchasimprovinghea1thcareandhe1pingtoso1vethec1imatecrisis,andtosupportprogresstowardsachievingtheSustainab1eDeve1opmentGoa1s(SDGs).Yet,generativeA1spotentia1benefitscomewithrisks.Thecapacity

12、ofgenerativeAItoexacerbatethecha11engesofdisinformationandmanipu1ationofopinionsisconsideredbyG7membersasoneofthemajorthreatsstemmingfromgenerativeAI,a1ongsiderisksofinte11ectua1propertyrightsinfringementandprivacybreaches.Ear1yeffortstotrackAIincidentsfoundonethousanddistinctincidentsandhazardsre1a

13、tedtogenerativeAI,basedonrough1y5600newsartic1esdatedfromJanuarytoJu1y2023.Astheserisksevo1verapid1y,theirmanagementandmitigationisatthetopoftheagendaforG7governments.Responsib1euseofgenerativeAI,addressingdisinformation,safeguardinginte11ectua1propertyrights,andgoverninggenerativeAIareamongthetoppr

14、ioritiesforG7po1icymakersandrequireinternationa1cooperationwith1ike-mindedpartners.OtherurgentandimportantissuesemphasisedbyG7membersinc1udeprivacyanddatagovernance,transparency,fairnessandbias,humanandfundamenta1rights,securityandrobustnessofAIsystems,andimpactsonthefunctioningofdemocracy.G7jurisdi

15、ctionsareeva1uatingtheirrespectiveresponsestogenerativeAI9aswe11asthepo1icygaps.Countriesare1everagingexistingandforthcoming1ega1andpo1icyframeworksanddeve1opingguide1inesorregu1ationtoaddressrisksre1atedtogenerativeAI.Nationa1initiativesarea1sobeingstrengthenedtoseizeitsopportunities.Newissuesraise

16、dbygenerativeAIappeartoaffectspecificsectorsinparticu1ar,suchaseducation,media,andtheworkp1ace.G7membersarea1ignedontheneedtoprovideeffectivetoo1sforsafety,qua1itycontro1,andcapacityandtrustbui1dingforgenerativeAI.Safety,qua1itycontro1,capacityandtrustbui1dingforgenerativeAIwereseenasamongthemosturgentandimportantinternationa1actiontheG7cou1dundertake.Engagingindia1oguewasa1soconsideredtobemosturgent,anddeve1opingvo1untarycodesofconductwasidentifi