Explore the challenges and opportunities in responsible AI development through this thought-provoking ACM conference talk. Delve into the pressing need for a human-centered approach in addressing AI risks and failures. Examine the cultural shift required to move the industry from a focus on model performance to holistic AI system development. Discover the tooling advances necessary to empower practitioners in dealing with AI's probabilistic and adaptive capabilities. Draw connections between responsible AI challenges and familiar problems in search and information retrieval. Gain insights into operationalizing human-centered AI, ensuring benefits for people and society, and incorporating human considerations throughout the development process. Learn about the EU AI Act, error prevention principles, and the importance of measurements and metrics in responsible AI practices.
Challenges and Opportunities in Responsible AI
Association for Computing Machinery (ACM) via YouTube
Overview
Syllabus
Introduction
What is Responsible AI
Potential for Harm
AI Risks
Agenda
EU AI Act
Humancentered Design
Error Prevention Principle
Humancentered Design Development
This is Reality Argument
Math Argument
Bias Argument
Humancentered AI
Cultural Organizational Technical Challenges
New Tools and Processes
Recap
Measurements
Translation Failures
Common Metrics
Information Retrieval
Overreliance
Search
Conclusion
Questions
Data collection
Question from the audience
Conflicting values
Explicit values
Literacy
Responsibility
Empower Users
Search Metrics
Metrics in Responsible AI
Measuring and Predicting Mistakes
Predicting Mistakes
Outro
Taught by
Association for Computing Machinery (ACM)
