Endpoints 2025 - Shalaleh Rismani: Safety Is a Systems Problem: Understanding & Evaluating AI скачать в хорошем качестве

Endpoints 2025 - Shalaleh Rismani: Safety Is a Systems Problem: Understanding & Evaluating AI

Title: Safety is a Systems Problem: Learning from 100 Years of Safety Engineering Description: Shalaleh (Postdoctoral Researcher, McGill & Mila) bridges the gap between traditional safety engineering and modern AI. Drawing parallels to early factory safety posters, she argues that AI safety must move beyond blaming "ignorant workers" or "careless users" to focus on system-level properties, feedback loops, and sociotechnical harms. Key Topics: Moving from component-level safety to emergent system properties. The 100-year gap: From the first electrical system to modern safety regulations. Mapping AI ethics principles (fairness, trust, privacy) to tangible system components. The "Responsible AI Measures Dataset": Analyzing the landscape of 257 research papers. Speakers: Shalaleh Rismani (Postdoctoral Researcher, McGill and Mila University) #AISystems #SafetyEngineering #ResponsibleAI #EthicsInTech #SystemSafety