Asimov's Zeroth Law of Robotics: Observability for AI

This conference talk explores the critical role of observability in artificial intelligence systems, drawing parallels to Isaac Asimov's Laws of Robotics. Discover the fundamental differences between monitoring traditional applications and AI systems, with Nicole van der Hoeven from Grafana Labs introducing the concept that "a robot must be observable" as a prerequisite to ethical AI governance. Learn about various AI types, specific telemetry signals for AI monitoring, and practical approaches to challenges like handling large datasets, detecting model drift, and managing LLM costs. Through a combination of cautionary insights and technical demonstrations, explore how to effectively instrument and monitor AI applications using modern tools including OpenTelemetry, Prometheus, and OpenLit, establishing observability as the foundation for responsible AI development and deployment.