DeepMind Podcast: Ghahramani and Horvitz on the Mathematics of AI Uncertainty
From August 26 to 28, Google DeepMind released an official podcast and video episode featuring Cambridge professor and DeepMind VP of Research Zoubin Ghahramani on the mathematics of AI uncertainty and its role in AI development and the future of AGI. The episode is available on YouTube, Spotify, Apple Podcasts and other platforms with chapter support.
Confirmed
- Ghahramani has spent 30 years building intelligent systems grounded in the mathematics of uncertainty
- Core thesis: true intelligence means knowing the limits of one's knowledge; AI systems must quantify and handle the unknown, which is critical for confidence and reliability of model predictions
- The conversation distinguished correctness from confidence, covered use cases from weather forecasting to robot control and autonomous driving, and discussed Bayesian thinking in AI and future research directions
- Microsoft Chief Scientific Officer Eric Horvitz appeared on the DeepMind podcast with a similar message: AI must be able to express uncertainty before humans can trust it with important decisions
Why it matters
Uncertainty quantification directly affects the credibility and safe deployment of AI. Multiple posters emphasized that understanding what models do not know is a key mathematical foundation for more reliable AI and AGI; Horvitz's participation signals cross-industry consensus on the topic.
2026-08-26 ~ 2026-08-28 · 8 related posts
Primary sources
- DeepMind Explores the Mathematics of AI Uncertainty and AGI Future — Google DeepMind ·
- DeepMind's Uncertainty Podcast Episode Now on All Major Platforms — GoogleDeepMind ·
- Microsoft Chief Scientist Horvitz on DeepMind Podcast: Trustworthy AI Must Embrace Uncertainty — erichorvitz ·
- [source] DeepMind Explores the Mathematics of AI Uncertainty and AGI Future — Google DeepMind · 2026-08-26
- DeepMind Podcast: AI Must Handle Uncertainty — ZoubinGhahrama1 · 2026-08-27
- DeepMind Podcast: The mathematics of AI uncertainty — ZoubinGhahrama1 · 2026-08-27
- Google Deepmind: The Mathematics of AI Uncertainty — 141_1337 · 2026-08-27
- [source] Microsoft Chief Scientist Horvitz on DeepMind Podcast: Trustworthy AI Must Embrace Uncertainty — erichorvitz · 2026-08-27
- DeepMind VP Ghahramani: Teaching AI Self-Doubt Is Key to Reliable Decisions — GoogleDeepMind · 2026-08-28
- [source] DeepMind's Uncertainty Podcast Episode Now on All Major Platforms — GoogleDeepMind · 2026-08-28
1 near-duplicate retellings: ZoubinGhahrama1