The Turing Podcast

The Turing Podcast is The Alan Turing Institute's show, built as conversations with researchers about data science and AI applied to specific, physical problems. The range is broad. Episodes include 'The Survival of the Fittest Route: Genetic Algorithms in Shipping and Sustainability', 'Navigating Change: AI in Arctic Sea Ice Forecasting', 'The Cyber Cold War: How AI is Defending Our Critical National Infrastructure', 'Defining AI safety', 'The cutting edge of emotional and cognitive neuroscience' and 'Project Bluebird: Revolutionising Air Traffic Control with AI and digital twins'. The shipping conversation sets its stakes with a projection that shipping's share of world emissions could rise steeply by 2050 if nothing changes. Host Amelia Jabri and co-host Professor Adam Sobey, the institute's Mission Director for Sustainability, then talk with a guest named Slav, in an episode Sobey sums up as shipping, genetic algorithms and whales. Slav describes testing several genetic algorithm families against A-star while developing voyage planning software. The discussion also reaches alternative fuels such as ammonia, hydrogen and nuclear, along with the security risks they could bring. The IceNet episode follows a similar path from method to consequence. Sea ice forecasting leads into the caribou that cross the frozen sea to Victoria Island and back, and into the indigenous communities whose subsistence hunting and food security depend on those herds. The critical infrastructure episode gives security a harder edge. It frames the stakes around water, electricity, internet access and health systems, pointing to the power outages in Spain and Portugal, the fires at Heathrow substations and the cyberattacks on M&S and Co-op. A speaker then explains data poisoning and prompt injection through the case of a prompt hacker known as Pliny the Liberator, who used fake GitHub repositories to poison an open-source language model. The anecdote gets a laugh. Then comes the harder question of what the same trick could do with malicious software. In 'Defining AI safety', Professor Ibrahim Habli of the Centre for Assuring Autonomy at the University of York separates the possible worst case from the credible one. He worries that people keen to build safe AI are left with too little research, too few methods and too few standards. The neuroscience episode, co-hosted by Bea with Jessie Pearce, turns to brain data. Its guest describes people who want to scan their own brains at least weekly over a year to track social anxiety, and calls ethics central to the work. Host Ed notes that Project Bluebird was featured on the podcast's very first episode in 2020, and its episode brings in three researchers at once: Dr. Nick Pepper, Dr. George Diaf and Dr. Mark Thomas. Diaf traces his route from a computer vision PhD into optimisation, calibrating digital twins against real-world data. The questions stay plain even when the answers get technical. For a first listen, the IceNet or shipping episodes show the pattern most clearly: a method, a real system, and the people it affects.