当前位置:首页 > 报告详情

部署者、非营利组织和公民社会为 AI 治理和标准提供信息.pdf

上传人: le****ng 编号:616659 2025-03-07 19页 2.83MB

1、International Organizations Best Practices PanelCivil Society informing AI Governance and StandardsHello!Kasia ChmielinskiPRONOUNS:THEY/THEMIm a technologist currently focused on building responsible data systems across industry,academia,government,and non-profit domains.I also am also a very enthus

2、iastic urban cyclist and birdwatcher.ZestAIFounding VP,ProductUS Digital ServiceDigital ExpertGoogleComms&StrategyMIT ScratchProduct LeadResearchingBerkman Klein Center,HarvardDigital Civil Society Fellow,StanfordAdvisingUnited NationsPartnership on AI Mozilla AI McKinseyHealthcareStorytellingFilm P

3、roducerGuest podcasterNPR,TED4Data Nutrition ProjectProject LeadBiographyHow do we build technology that works betterand for more people?5Products are generally built and deployed quickly for a particular use case at the expense of othersBuild for the“ideal user”and iterate out over timePeople who d

4、ont speak EnglishPeople who dont have the most recent iphone People in rural areas People who dont have binary gender People who use assistive devices or technology People who share phones or devices with family 680%solution for“majority”Our assessments for unintended impact are reactive,rather than

5、 proactiveDesignDevelopmentTestingCompliance/Risk Marketing&CommsStrategyLaunch.and are even considered“anti-innovation”7The result:Products that dont work for everyone8What can we do?9Proactive design requires intentionality and supportBY DESIGNEarlier and integrated into process(including metrics

6、and motivations);123WITH SUPPORTNeeds to involve operational changes and leadership buy-in;INVOLVING COMMUNITYIncludes the voices of those affected,in particular bringing people to table10Systems built on problematic data will exhibit those same issues,especially for historically

word格式文档无特别注明外均可编辑修改,预览文件经过压缩,下载原文更清晰!
三个皮匠报告文库所有资源均是客户上传分享,仅供网友学习交流,未经上传用户书面授权,请勿作商用。
本文主要介绍了Kasia Chmielinski,一位专注于构建负责任的 data systems 的技术专家,同时也是一名热情的 urbancyclist 和 birdwatcher。Chmielinski 提出,目前的产品通常是快速构建和部署的,以满足特定的用例,而牺牲了其他人的需求。为了使产品更好地为更多人服务,我们需要进行主动设计,这需要有意图和支持,包括操作变化和领导层的支持,以及涉及受影响人的声音。此外,Chmielinski 还介绍了 Data Nutrition Project,这是一个旨在支持数据所有者和用户的项目,它提供了一个标准化的文档工具,用于说明数据集的内容及其对模型的健康影响。该项目还支持政策制定者,提供背景、上下文和建议。最后,该项目还通过教学和培训活动,教授和培训他人关于数据质量的知识。
"如何构建更好的技术产品?" "数据营养标签如何影响AI模型健康?" "如何参与Data Nutrition Project?"
客服
商务合作
小程序
服务号
折叠