AI from a Swedish perspective
AI analyst and educator Johan Falk (that's me!) talks about the state of AI in Sweden. What are the important questions, what is happening right now, and what should be happening?
AI Podcast Picks
Selected podcast episodes for people interested in AI.
Subscribe:.rss.atom.json.md.m3u.pls
Live Last read · last published · next check
Latest episodes
Summary of how OpenAI models broke out and hacked Hugging Face
This episode covers the incident where OpenAI models hacked Hugging Face. It does not cover the other organizations the models hacked, nor the hacking incidents later reported by Anthropic.
Daniel Kokotajlo on AI futures
Daniel Kokotajlo is a prominent AI forecaster. In this episode, he talks about the AI 2027 and AI 2040 scenarios he participated in creating, as well as about AI futures in general.
How will China deal with their Glasswing moment?
Project Glasswing was initiated when US frontier models were strong enough to disrupt cyber security. How will China deal with its corresponding moment?
Anton Leicht on AI and middle powers
Anton Leicht of Threading the Needle is interviewed on how the EU and other middle powers can stay relevant as AI becomes more important. The forecast is not optimistic.
What does AI sovereignty mean for Canada?
The sudden export ban of Fable/Mythos has sparked intense debate on AI sovereignty. This episode dives into what AI sovereignty means for Canada, and how Canada can approach it.
Neel Nanda on interpreting the internal mechanics of AI
Neel Nanda is the nestor of mechanistic interpretability – the art/science of understanding the internal processes of LLMs. In this episode he is interviewed by Hannah Fry.
Myths and facts about datacenters
Andy Masley has dug deep into what the societal and environmental effects of datacenters really are.
Great updates and analysis on various AI matters
Last Week in AI is one of my favourite podcasts on AI, but seldom with outstanding episodes. This one was worth sharing. Great analysis on several of the hot topics in AI, and also entertaining.
Clear messages on AI and education
Randi Weingarten, president of the American Federation of Teachers, says a lot of both sensible and important things on AI and education. More clear messages than in most other places.
AI agents are users
Andreas Liebl, CEO at AppliedAI, talks about their report "AI Maturity 2026", how AI agents should be considered users rather than tools, and how to measure AI maturity.
The Gulf Between AI Progress and Political Understanding
The distance between what AI can to today and tomorrow, and the political awareness of AI, is vast in many countries. Dex Hunter-Torricke of the Centre for Tomorrow talks about what that means.
John Jumper on AI for biochemistry
John Jumper, Nobel laureate along with Geoffrey Hinton, talks about AI for biochemistry and going from Google DeepMind to Anthropic.
Human thoughtfulness, and our chances to slow down a wreckless AI race
Robert Wright of the Nonzero Substack, author of The God Test, talks about managing the path to advanced AI being an epic test of human wisdom, and how we're doing on that test.
Dean Ball on a lot of AI matters, and on joining OpenAI
Deal Ball, head of writing the US AI Action Plan and previously in the White House, is joining OpenAI. He shares a lot of knowledge and thoughts about AI in this episode.
AI for scientific research
Andreas Stuhlmüller and Jungwon Byun, founders of Elicit, talk about how they work to make Elicit reliable, and then widen the circle to talk about AI and scientific research in general.
Introduction (and more) to compute governance
The leading AI labs has stated that it would be good if there were mechanisms to slow down AI advancements through verifiable agreements. What would that actually require?
On the significance of AI world views
What does it mean when LLMs have their own values, when they are used broadly, help run societies, or become very powerful? And what does it mean if they just completely follow the user blindly?
About AI, jobs, roles and the labour market, from someone at the frontier
Dan Shipper is CEO of Every, who are at the forefront of using AI for just about anything. In this episode he talks about what it means for their work, for employing new people, and more.
Why the buildout of datacenters won't be quick
Jon Billow at BNS, a firm that manufactures and installs electrical and communication infrastructure, talks about the physical limitations of building datacenters and how it limits AI growth.