The Existential Hope Podcast

For 50,000 of generations, humans have passed civilization down to their children. We could be the first generation to hand its future to AIs instead. Should we? Or could the best AI be the kind that chooses to step back and leave some things to us? In this episode we sit down with Stuart Russell, professor of computer science at UC Berkeley and co-author of the world's standard textbook on AI. He’s also a leading proponent of provably beneficial AI: systems that are safe by design because their only goal is to further human interests. We talk about:How AI has changed over the past 50 years, from simple game-playing programs to today's large language models Why handing an AI a fixed objective becomes dangerous once it is more capable than us, and what a safer approach could look like How an AI might learn what we really want, even when we don't fully know ourselves How the race toward more powerful AI can still be steered somewhere safer What happens to human purpose as AI becomes more and more capableChapters:0:00 Cold open0:39 How AI has changed over the past 50 years, from chess to large language models4:15 What is the standard model of AI and why does it fail? 12:08 What is provably beneficial AI?17:49 Teaching AI to learn from humans, even when we make bad choices20:11 How can we steer AI development in a safer direction? 28:01 International cooperation on AI safety30:41 What a good AI future could look like35:38 Protecting human purpose: should AI step back?39:15 How young people can get involved in AI safety40:47 Stuart Russell's best adviceOn the Existential Hope Podcast hosts Allison Duettmann and Beatrice Erkers from the Foresight Institute invite scientists, founders, and philosophers for in-depth conversations on positive, high-tech futures. Full transcript, listed resources, and more: https://www.existentialhope.com/podcastsFollow on X. Hosted on Acast. See acast.com/privacy for more information.

Google DeepMind has a 7-hour-long roleplaying game that helps real scientists and policy makers understand how AI will transform science by 2030. And those who played it have found it more informative than any policy brief.In this episode, Zoë Brammer and Ankur Vora, who lead strategic foresight at Google DeepMind, walk us through why they developed the game and the surprising learnings from it.We cover:What strategic foresight actually means and why AI is so complicated to plan aroundWhy the game deliberately centers around "middle power" countries like the UK, Germany, or Singapore instead of the US and ChinaThe unexpected tradeoffs and discoveries participants come across while playing the game, including how fragile public trust in science funding really isWhy human scientists will become more relevant as AI takes on more of the science itselfTimestamps:0:00 Cold open0:59 How Zoë and Ankur ended up at Google DeepMind3:44 Why five-year plans don't work4:23 What “strategic foresight” actually means6:57 Finding common ground between different visions of the future9:04 Why DeepMind built a role-playing game about science in 203013:42 How the Science 2030 game actually works16:31 Why the game focuses on middle powers, not the US and China17:42 The surprising range of technologies counted as “AI for science”19:49 What the Science 2030 games have revealed so far25:07 Do scientists feel powerless in an AI-driven future?27:40 Why run this experiment inside an AI lab?28:56 Zoë and Ankur's best-case scenario for AI and science30:48 Advice for young people who want to shape how AI developsOn the Existential Hope Podcast hosts Allison Duettmann and Beatrice Erkers from the Foresight Institute invite scientists, founders, and philosophers for in-depth conversations on positive, high-tech futures. Full transcript, listed resources, and more: https://www.existentialhope.com/podcastsFollow on X. Hosted on Acast. See acast.com/privacy for more information.

In 2024, Taiwan was flooded with deepfake scam ads. Instead of a crackdown, the government texted 200,000 random citizens: what should we do? People’s proposals became law within months, and by 2025 deepfake ads were down 94%.In this episode, we speak with Audrey Tang, Taiwan's former digital minister. A self-taught programmer who dropped out at 14, she first helped occupy Taiwan's parliament during the 2014 Sunflower Movement, then joined the government two years later.We cover:How Audrey went from organizing a 2014 parliamentary occupation to becoming Taiwan's digital minister two years laterHer case for accelerating some AI capabilities and deliberately slowing others, and how she decides which is whichHow 447 randomly selected citizens drafted Taiwan's deepfake legislationAudrey’s proposal for AI systems that belong to local communities rather than sitting in a remote cloudWhy she thinks being a "good enough ancestor" for future generations is more useful than trying to perfectly optimize the futureThis episode is part of our AI Pathways series, where we explore the choices we can still make about how AI gets built.Chapters:0:00 Cold open0:48 How the Sunflower Movement led Audrey Tang into government3:06 What is d/acc, and how does it work in practice?5:39 How do you make democratic deliberation actually work?6:49 Civic tech in action: Taiwan's deepfake scam crisis, and how citizens solved it9:10 What a d/acc world could look like in 10 years11:09 What frontier AI labs get right about public input13:13 What AI labs are doing wrong on closing the public input loop15:14 Can Taiwan's model work anywhere else?18:10 Can one person cause catastrophic harm with AI?22:02 Where Audrey Tang disagrees with the mainstream d/acc take24:10 What Audrey Tang got wrong about transparency26:36 What does it mean to be a "good enough ancestor"?28:20 How to start your own civic AI project30:25 The best advice Audrey Tang ever receivedOn the Existential Hope Podcast hosts Allison Duettmann and Beatrice Erkers from the Foresight Institute invite scientists, founders, and philosophers for in-depth conversations on positive, high-tech futures. Full transcript, listed resources, and more: https://www.existentialhope.com/podcastsFollow on X. Hosted on Acast. See acast.com/privacy for more information.

There are AIs running labs with little human intervention right now. And they’re doing experiments that human experts would never try. Is this changing what and how science gets done? In this episode, we speak with Antony “Ant” Rowstron, who has worked with ARIA (the UK’s Advanced Research and Invention Agency) on their biggest bet to date. We cover:How ARIA funded twelve teams to build AI scientists that can run an entire research process: generating hypotheses, designing experiments, and carrying them out without continuous human intervention. What AI scientists are actually achieving now, from personalized cancer vaccines to molecules that stimulate our own immune response to new viruses in 48 hours.How we could train AIs on the tacit, hands-on knowledge only human scientists have.How AI hallucinations might actually be useful for scientific discovery.How labs and the role of human scientists will change as AI automates more and more parts of the research process.Chapters:0:00 Cold open1:14 What is an AI scientist? (the three-layer stack behind it)3:48 What AI scientists can do in practice: quantum dots, cancer vaccines, and AI-designed antigens8:17 Inside ARIA's AI scientist grant: 245 applications, 12 teams, 9 months11:36 What can an AI scientist actually achieve in nine months?16:33 How will the role of the human scientist change in the future?18:42 Can AI scientists work outside a computer, in the physical world?21:20 What will the lab of the future look like?25:34 How to capture the tacit knowledge only scientists in the room have29:32 Do AI scientists trained on the same data lose their creativity?31:17 Why AI hallucinations might actually help scientific discovery39:56 Career advice for young people who want to work on AI for science45:21 Ant Rowstron's existential hope vision for AI in science46:10 What Ant would do if not at ARIA46:38 The best piece of advice Ant ever receivedOn the Existential Hope Podcast hosts Allison Duettmann and Beatrice Erkers from the Foresight Institute invite scientists, founders, and philosophers for in-depth conversations on positive, high-tech futures. Full transcript, listed resources, and more: https://www.existentialhope.com/podcastsFollow on X. Hosted on Acast. See acast.com/privacy for more information.

There's an overwhelming amount of self-help advice out there, which makes it really hard to know what will work for you. But it turns out that over 450 self-help techniques from 100 books and 20 types of therapy actually boil down to just 12 tools.In this episode, we talk with Spencer Greenberg, a mathematician and founder of the psychology research nonprofit Clearer Thinking. He recently co-authored the book The 12 Levers with clinical psychologist Jeremy Stevenson, to cut through the noise of self-help and provide people with the smallest number of concrete tools they can leverage in different situations.We cover:How nearly 500 self-help techniques got narrowed down into 12 core psychological strategies, and how to use them.Why most people live by values they absorbed from their parents or environment rather than ones they actually chose, and how to figure out what your own values are.The real formula for productivity, which takes into account how important the work actually is.Why hopelessness is often less about the state of the world than about feeling unable to act, plus the single most evidence-backed exercise for building genuine optimism. The exposure therapy techniques he used to overcome severe social anxiety.Chapters:0:00 Cold open0:52 How 459 self-help techniques boil down to just 12 levers2:36 Does self-help need to be evidence-based?4:40 Why understanding yourself can change the world5:42 Are you living on your values or someone else’s values?7:11 How to figure out what you actually value8:03 Pleasurable life vs meaningful life: what actually makes you fulfilled?9:39 Redefining productivity: on the importance of the work vs hours and efficiency11:46 Optimism vs hope, and training yourself to be more optimistic15:21 Why Spencer wrote The 12 Levers, and how he beat his own social anxiety18:14 Which levers are hardest to maintain and self development as an ongoing journey20:01 Overwhelmed by 12 levers? Where to start22:26 How to create meaning in your life through your values26:26 Advice for young people who feel hopeless about the future28:15 Designing an AI assistant that pushes you toward your values31:39 What would it look like if everyone used the 12 levers?32:56 When to accept things vs when to fight for change36:55 Why your self-help knowledge might have blind spots38:13 Spencer's existential hope vision for AI38:48 The technology Spencer wants to see built39:14 What Spencer would do instead of his current work39:58 The best piece of advice Spencer ever receivedOn the Existential Hope Podcast hosts Allison Duettmann and Beatrice Erkers from the Foresight Institute invite scientists, founders, and philosophers for in-depth conversations on positive, high-tech futures. Full transcript, listed resources, and more: https://www.existentialhope.com/podcastsFollow on X. Hosted on Acast. See acast.com/privacy for more information.