当前位置:首页 > 报告详情

Rappler 的人工智能版主谈人工智能治理政策.pdf

上传人: 拾亿 编号:751701 2025-07-29 14页 7.55MB

1、AI,Journalism&the Future of NewsBy GEMMA B.MENDOZAHead for Digital Services&Lead Researcher for Disinformation&PlatformsVia Facebook&Twitter:gemmabmendoza/Email:The Rappler Story:Innovative experiments around journalism,tech&the community engagementaiDialogue:using AI to enrich&scale up democratic c

2、onsultationaiDialogue:using AI to enrich&scale up democratic consultationPHVote 2022:Early use of generative AI technologies in content generationalmost 50,000 profile pages generated by AI for the 2022 elections.Profiles were generated using Open AIs GPT 3 but reviewed by researchers for accuracyWh

3、at worries us about AI:Deepfakes&disinformationWhat worries us about AI:Deepfakes&disinformationhttps:/ tricky question:how to align human values with machinesHow do you instill complex values in AI?How do you keep AI honest or fair?How do you develop scalable oversight?Blackboxes:How do you audit A

4、I models?When AI makes mistakes,who is responsible for those mistakes?Case for a man&machine approach to AIHow Rappler is using generative AI technologiesRolling out AI features in incrementsOrg-wide workshops and discussions on AI how tos,ethical use,our boundariesAI-generated summaries in June 202

5、3Building our own Ontology and Knowledge Graph:goal is to anchor our LLM using GraphRAGTaking up courses on Coursera on prompt engineeringLaunched in December 2023Distribution,engagement:Our own communityBuilt on the matrix protocol:meaning we can interconnect with other newsroomsAI-human(man and ma

6、chine)moderationBuilt-in AI-features:gamesbot,newsbotRappler Communities AppAnchoring AI in FACTSWe developed our own Ontology&building a Knowledge Graph which will anchor our LLM using GraphRAG to ensure the integrity of information that our chatbot delivers.

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。
本文主要讲述了Rappler公司如何利用生成式AI技术进行新闻内容的创新实践。关键点如下: 1. 利用AI进行民主咨询:在2022年菲律宾大选中,AI生成了近50,000个候选人资料页面,使用的是Open AI的GPT 3技术。 2. AI带来的担忧:深度伪造和虚假信息问题,以及如何确保AI的公正性和可审计性。 3. 人类价值观与机器的融合:探讨如何在AI中嵌入复杂价值观,并实现可扩展的监管。 4. Rappler的AI实践:逐步推出AI功能,组织全公司范围内的AI研讨会,计划在2023年6月推出AI生成摘要。 5. 构建本体和知识图谱:通过GraphRAG技术,确保聊天机器人提供的信息准确可靠。 6. 人工智能与人类结合的审核机制:推出基于矩阵协议的新闻社区应用,实现与其他新闻室的互联互通。 7. 事实为本的AI:开发自己的本体和知识图谱,将AI与经过Rappler编辑和研究团队审核的数据相结合,确保聊天机器人回应如同记者一般。 核心数据:2022年选举中生成50,000个AI资料页面。
"AI如何助力新闻业?" "深度伪造与AI伦理难题怎么破?" "Rappler如何用AI锚定事实?"
客服
商务合作
小程序
服务号
折叠