- 1、本文档共20页,可阅读全部内容。
- 2、原创力文档(book118)网站文档一经付费(服务费),不意味着购买了该文档的版权,仅供个人/单位学习、研究之用,不得用于商业用途,未经授权,严禁复制、发行、汇编、翻译或者网络传播等,侵权必究。
- 3、本站所有内容均由合作方或网友上传,本站不对文档的完整性、权威性及其观点立场正确性做任何保证或承诺!文档内容仅供研究参考,付费前请自行鉴别。如您付费,意味着您自己接受本站规则且自行承担风险,本站不退款、不进行额外附加服务;查看《如何避免下载的几个坑》。如果您已付费下载过本站文档,您可以点击 这里二次下载。
- 4、如文档侵犯商业秘密、侵犯著作权、侵犯人身权等,请点击“版权申诉”(推荐),也可以打举报电话:400-050-0827(电话支持时间:9:00-18:30)。
计算机研究与发展DOI:10.7544/issn1000-1239.202330553
JournalofComputerResearchandDevelopment60(9):1926−1945,2023
大模型道德价值观对齐问题剖析
矣晓沅谢幸
(微软亚洲研究院北京100080)
(xiaoyuanyi@)
UnpackingtheEthicalValueAlignmentinBigModels
YiXiaoyuanandXieXing
(MicrosoftResearchAsia,Beijing100080)
AbstractWeexploretheemergingchallengespresentedbyartificialintelligence(AI)developmentintheeraofbig
models,withafocusonlargelanguagemodel(LLM)andethicalvaluealignment.Bigmodelshavegreatlyadvanced
AI’sabilitytounderstand,generate,andmanipulateinformationandcontent,enablingnumerousapplications.
However,asthesemodelsbecomeincreasinglyintegratedintoeverydaylife,theirinherentethicalvaluesandpotential
biasesposeunforeseenriskstosociety.Weprovideanoverviewoftherisksandchallengesassociatedwithbig
models,surveyexistingAIethicsguidelines,andexaminetheethicalimplicationsarisingfromthelimitationsofthese
models.Takinganormativeethicsperspective,weproposeareassessmentofrecentnormativeguidelines,
highlightingtheimportanceofcollaborativeeffortsinacademiatoestablishaunifiedanduniversalAIethics
framework.Furthermore,weinvestigatetheethicalinclinationsofcurrentmainstreamlargelanguagemodelsusing
moralfoundationtheory,analyzeexistingbigmodelalignmentalgorithms,andoutlinetheuniquechallenges
encounteredinaligningmoralvalueswithinthem.Toaddressthesechallenges,weintroduceanovelconceptual
paradigmforethicallyaligningthevaluesofbigmodelsanddiscusspromisingresearchdirectionsforalig
文档评论(0)