利用 AWS 负责任的 AI 实践加速可信 AI 开发 [重复].pdf

《利用 AWS 负责任的 AI 实践加速可信 AI 开发 [重复].pdf》由会员分享，可在线阅读，更多相关《利用 AWS 负责任的 AI 实践加速可信 AI 开发 [重复].pdf（67页珍藏版）》请在三个皮匠报告上搜索。

1、 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AIM361-RMikaela MyersMathew MonfortAccelerate trusted AI development with AWS responsible AI pract

2、icesSenior Technical Product MSenior Applied S 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AgendaResponsible AI IntroductionAWS Responsible AI Best Practice Framework Well Architected Responsible AI Lens walkthroughExample Applying the LensQ&A 2025,Amazon Web Services,Inc.or i

3、ts affiliates.All rights reserved.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.Why do we need Responsible AI?2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AI Builders are being asked to move fast in an environment of increasing complexityUse Cases Regulatio

4、n Options Diagnostic decisions Coding agent IT chat bot Personalization AI Trad ML base FM custom FM distilled FM vector libraries agent multi-1P-agent multi-3P-agent etc EU AI Act EU AI Act Code of Practice California.Colorado.ISO 42001 NIST AI 600 etcDemand AI apps increases%of builders needing AI

5、 expertise.2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.However,the mindset shift is significantTraditional Software2)Spec with human languageAI/ML Software4)New releases perform the same or better on all inputsSpec with datasets3)Developers testDevelopers&deployers test New re

6、leases perform the same or better overallMultiple explanations5)Single explanation1)Rule-based logicStatistical inference 2025,Amazon Web Services,Inc.or its affiliates.All rights reserved.AI System PerformanceAI System VersionsTestDataset ATestDataset BTestDataset CExample:AI solutions comprise 2 s