Hannah Fry questions the people behind AlphaGo, AlphaFold and more.
Hannah Fry questions Google DeepMind co-founders Demis Hassabis and Shane Legg, along with guests on AlphaGo, AlphaFold, driverless cars and cyber threats. She tests ideas out loud. At one point she turns protein structure into a wedding seating plan and asks her guest whether the comparison holds.
PRESS PLAY
Google DeepMind: The Podcast is an interview show in which Hannah Fry questions people connected to Google DeepMind about the systems they have worked on and the problems those systems raise. Two of the lab's co-founders are among her guests. Shane Legg, whom Fry introduces as Chief AGI Scientist, says flatly that human intelligence will not be the upper limit of what is possible. Fry notes that he talked about AGI back when it was considered, in his words, the lunatic fringe. He also lays out an idea he calls system two safety. It is modelled on Daniel Kahneman's distinction between acting on gut instinct and stopping to reason through the consequences. Fry frames the year as a shift from large language models toward agentic AI. Demis Hassabis then describes an experiment that places one AI agent inside a world that Genie invents around it as it goes. In his words, the two AIs end up interacting in each other's minds. He sees a possible training loop in that setup, one that could supply almost unlimited examples. The show also looks back. Thore Graepel and Pushmeet Kohli revisit the March 2016 match against Lee Sedol in Seoul. That includes move 37, which no human player would have chosen, and move 78, the unusual wedge move that confused AlphaGo. The talk turns on why one loss still mattered after three straight wins, since it could have exposed a fragility. John Jumper explains how AlphaFold predicts the 3D structures of proteins and looks ahead to probing the cell in ways that could feed into medicine and synthetic biology. Fry's contribution is a wedding seating plan with amino acids as the guests, which she tests on him as she goes. Part of the Waymo conversation is recorded from the back seat of a driverless car in San Francisco. Vincent Vanhoucke explains how a car must tell a pile of snow it can drive through from a rock it cannot. He also points out that a transcontinental drive at about 98 percent autonomy happened back in 1995. Security gets a conversation spread over two episodes with Four Flynn, VP of Security at Google DeepMind. He describes hospitals, power grids and hydroelectric dams that run familiar operating systems but are hard to patch safely. He still expects the defenders to win the war, if not every battle, and Fry sums up the conversation as a terrifying and yet strangely comforting view of humanity's vulnerabilities. The AlphaGo retrospective ends on an open question: how to separate a genuinely new insight from a hallucination. The AlphaGo and AlphaFold episodes are the best place to start, because they show the show's basic method: take a famous result and let the people who were close to it explain how it happened.
Made for: Suits a listener who wants to hear how AI results like AlphaGo, AlphaFold and driverless cars came about from people who were close to the work, explained in plain terms with the occasional well-tested analogy.
What sets it apart: Many of the guests come from Google DeepMind itself, including two co-founders, so the stories behind its projects come from inside the work. Hannah Fry often checks her own explanations with the guest, which keeps the technical material easy to follow.
Join mathematician and broadcaster Professor Hannah Fry as she goes behind the scenes of the world-leading research lab to uncover the extraordinary ways AI is transforming our world. No hype. No spin, just compelling discussions and grand scientific ambition.
A wedding seating plan becomes protein chemistry, and you're suddenly invested in where the flowers go.
Press play if you want
Short reviews from the PlayNext desk, based on the episodes we processed.
A pivotal AlphaGo match becomes a human story about machine intelligence, fragility, and the questions still shaping AI.
Google DeepMind's AlphaGo story centers on Lee Sedol's move 78 in the March 2016 match in Seoul, when one unexpected play unsettled a system that had already taken three games.
A clear, wide-angle conversation on agents, generated worlds, and the next questions for AI.
Professor Hannah Fry steers a wide-ranging conversation with Google DeepMind's Demis Hassabis about where AI may be heading after a year crowded with breakthroughs.
Shane Legg links AGI's definition to the ethical reasoning its safety may require.
AGI is treated here as a practical question about limits, definitions, and the social consequences of AI that could move far beyond human intelligence.
John Jumper explains how AlphaFold grew from protein-structure prediction into a working instrument for biology.
AlphaFold comes through as a scientific story that is still unfolding. John Jumper describes a system built to predict the 3D structures of proteins, then follows its newer reach into structures and interactions across life's molecules.
A Waymo ride frames autonomous driving as a problem of semantics, simulation, and long-running AI progress.
A back-seat ride through San Francisco gives Google DeepMind: The Podcast a clean route into autonomous driving: the hard part is not just seeing the road, but knowing what a scene means.
From economic pressures to investment consequences.
AI capability meets the work of building.
Daily AI news through the lens of trust, infrastructure, and real-world deployment.
Google DeepMind: The Podcast revisits the March 2016 AlphaGo match in Seoul, where Lee Sedol faced a neural network trained with reinforcement learning.
Google DeepMind: The Podcast pairs Professor Hannah Fry with Demis Hassabis as he says the field has "packed in 10 years in one year" with Gemini 3 and multimodal models.
On Google DeepMind: The Podcast, Hannah Fry asks Shane Legg whether today's models are already more than sparks of AGI.
John Jumper says AlphaFold now serves more than 3 million researchers across 190 countries and has supported roughly 35,000 scientific papers, while AlphaFold 3 extends the system to DNA, RNA, small molecules, and protein complexes.
Waymo's Vincent van Hooke tells Professor Hannah Frye how a driverless car fuses cameras, LiDAR, and radar into one road belief while riding in San Francisco.
Hannah Fry and Google DeepMind security executive Four Flynn examine how nation-state pre-positioning, ransomware groups, zero-day markets, and AI-enabled social engineering change cyber risk.
Professor Hannah Fry interviews Paul Flynn, VP of Security at Google DeepMind, about Operation Aurora, phishing, zero-day vulnerabilities, and AI-enabled cyber attacks.
Google DeepMind: The Podcast examines Genie 3, a real-time world model that predicts every pixel of a text-generated environment as a user moves through it.
Professor Hannah Fry interviews Paige Bailey, an AI developer relations engineering lead at Google DeepMind, about publicly available creative tools including VO3, Flow, Gemini speech generation, and Gemini Live.
Google DeepMind Nature Lead Drew Purves tells Professor Hannah Fry that biodiversity action is often blocked by missing local information, despite 189 countries supporting the 30 by 30 plan for 2030.
Hannah Fry interviews researchers from Google DeepMind and Isomorphic Labs about how AI, particularly AlphaFold, is revolutionizing drug design and protein structure prediction.
Hannah Fry interviews Jeff Dean about his career at Google, including contributions to TensorFlow, Google Brain, Transformers, and the Gemini multimodal model.