Statistics podcasts

INRIA researcher David Holzmüller explains how to choose among tabular foundation models, boosted trees and multilayer perceptrons. He discusses TabArena, TabICL and RealMLP, benchmark weaknesses, small-data performance, calibration and class imbalance. The conversation also examines text columns, explainability, time-series distinctions, synthetic pretraining and the trade-offs among computation, inference speed and predictive performance. Hosted by Ravid Shwartz Ziv and Allen Roush. Watch the full conversation on the publisher’s YouTube channel.

Hosted by Ravid Shwartz Ziv, Allen Roush

Florian Brand of Prime Intellect discusses how to evaluate agents when their tools and execution environments affect the result. The episode examines benchmark gaming, preventing shortcuts around scoring rules and the statistical limits of expensive evaluation runs. It also asks whether subjective impressions of reliability can be translated into useful measurements of model behavior. Hosted by Ravid Shwartz Ziv and Allen Roush. Watch the full conversation on the publisher’s YouTube channel.

Hosted by Ravid Shwartz Ziv, Allen Roush

David Kipping considers how to reason about extraterrestrial life when evidence is scarce. The discussion covers promising places to search, the interpretation of possible biosignatures and the role of statistical inference, alongside questions about habitable worlds and moons beyond our solar system.

Hosted by Janna Levin

Richard Samworth discusses the research behind his invited talk at the 2026 International Congress of Mathematicians. He explains how statistics connects theoretical questions with practical applications and reflects on mathematical collaboration and the research community at Cambridge.

Hosted by Marianne Freiberger, Rachel Thomas

Taylor Sparks and Andrew Falkowski discuss autonomous materials laboratories with Radical AI co-founder and CEO Joseph Krause. They examine instrument integration and data-access limitations, the MATRIX vision-language model, language-model embeddings for Bayesian optimization, and how to measure the contribution of scientific intuition. Krause also explains the choice to pursue manufacturing alongside discovery, emphasizing production data and process know-how. The episode connects automated experimentation with the practical challenges of developing and scaling new alloys.

Hosted by Taylor Sparks, Andrew Falkowski

Richard Gao discusses fitting spiking-network models to experimental data and dealing with different parameter combinations that explain similar observations.

Steven Niederer and Richard Wilkinson discuss personalised mathematical models of patients’ hearts and their potential use in care. The conversation connects digital twins and the CVDNet project with research on representing, calibrating and using uncertainty in predictions. It also traces how this work develops ideas from the earlier Fickle Heart research programme.

David Spiegelhalter discusses how to reason about chance, incomplete knowledge, risk and luck. Drawing on his book The Art of Uncertainty, the conversation moves between everyday surprises and consequential historical decisions. It examines why uncertainty can be personal and why disagreement can help people reason more carefully.

Franziska Michor explains how evolutionary dynamics, statistics and mathematical models can inform cancer treatment. The conversation examines ways to optimize treatment choices, combinations and timing while accounting for tumour complexity and the continual evolution of cancer cells.

Hosted by Steven Strogatz

Jo Røislien discusses statistics in medicine and the challenge of explaining quantitative science to broad audiences. Conversation in Norwegian.