Google DeepMind: The Mathematics of AI Uncertainty
In this episode, Cambridge professor and Google DeepMind VP of Research Zoubin Ghahramani explores the mathematics of AI uncertainty. While modern AI often responds with unwavering authority even when incorrect, Ghahramani discusses his decades of work focused on giving machines an awareness of their limitations. The discussion covers the difference between correctness and confidence, historical perspectives on the field, and the application of Bayesian thinking in AI. Finally, the episode examines how uncertainty manifests in the real world and what these principles mean for future AI research and AGI.
Original sources
Podcast Google DeepMind: The Podcast
YouTube Google DeepMind
AI Radar wrote this summary from the publisher’s notes with AI, not from a transcript. How it works.