Microsoft Research Podcast AI Testing and Evaluation: Learnings from Science and Industry
6 snips
Jun 23, 2025 Amanda Craig Deckard, Senior Director of Public Policy at Microsoft, leads a team committed to responsible AI development. In this discussion, she shares insights on AI testing and evaluation as essential governance tools. The conversation covers the intricacies of AI governance, drawing lessons from cybersecurity and finance. Amanda highlights the need for collaborative frameworks among industry, academia, and government, emphasizing trust and scientific standards in navigating AI’s complex risks.
AI Snips
Chapters
Transcript
Episode notes
Amanda's Career Journey to AI
- Amanda Craig Deckard had a varied career path starting in journalism and legal service before joining Microsoft in 2014.
- She transitioned through cybersecurity and public policy to lead in responsible AI at Microsoft.
Learning from Global Governance
- Early 2023 efforts explored governance analogies from other global institutions to inform AI governance.
- Lessons showed analogies apply but have unique limitations due to AI's distinct context and rapid evolution.
Horizontal vs Vertical Technologies
- Horizontal technologies like genome editing and AI involve understanding risks in technology and its varied applications.
- Vertical domains set clearer risk thresholds, while horizontal domains face complexities requiring contextual risk evaluation.