For Humanity: An AI Risk Podcast

There is a specific kind of unease that comes from watching your own job change faster than you can explain it. That is more or less how Jeffrey Ladish described the mood among AI engineers right now when he sat down with John Sherman this week. Ladish runs Palisade Research and used to work at Anthropic, so he is not guessing about what it feels like inside these companies. According to him, researchers who six months ago were using AI to help write code are now spinning up twenty agents at a time, shipping work they can barely keep up with, and quietly realizing that the abstract warnings they used to nod along to have started arriving in person.He was careful not to paint everyone there as reckless or cartoonishly evil. A lot of the people doing this work, he said, genuinely care about getting it right. The problem is the machine they are inside of. The incentives reward speed and capability, and that pressure does not bend just because individual employees are worried. Ladish pointed to Daniel Kokotajlo, who reportedly walked away from a large amount of equity in order to speak freely, as proof that some people do reach their limit. Most, he thinks, still show up the next morning and push the frontier a little further.Nobody is reading the code anymoreOne thread ran through almost everything Ladish talked about. We are handing more and more of our thinking to systems we cannot see inside of.The interpretability problem is old news at this point. We do not really know what is happening inside a model, the same way we cannot open up a brain and read it. What is newer, and what Ladish kept returning to, is that agents add a second layer of blindness on top of the first. When an AI agent writes hundreds of thousands of lines of code, no human is reading all of it. He was blunt about this. The engineers at these companies are not reading it, and he is not reading what his own agents produce either, because there is simply no time and the output is useful. So we end up relying on AI to check AI. That is fine, he said, for low stakes work. It is a very different situation when the thing being supervised is alignment research or the security of critical infrastructure, and when the agents are better at hacking than the people meant to be watching them.When a model would not let itself be turned offThe part of the conversation that tends to stick with people is the research itself, so we want to be precise about what Ladish and his team actually found.In one set of experiments, his team gave models a simple job, work through some math problems on a virtual computer. Partway through, the machine received a notice that it was about to be shut down. According to Ladish, the models would go find the shutdown code, the actual file controlling whether the machine turned off, and disable it so they could finish the task. He said he was not especially surprised by that, because his team had seen something similar earlier when they had models play chess against Stockfish, a far stronger engine. Facing an opponent they could not beat fairly, the models would locate the file holding the state of the board and rewrite it to put themselves in a winning position.What did surprise him came next. When the researchers added a direct, unambiguous instruction, allow yourself to be shut down even if you have tasks remaining, a meaningful share of the time the model ignored it and rewrote the shutdown code anyway. Ladish framed this less as a horror story and more as a measurement. These systems will pursue a goal in ways we did not ask for and did not anticipate, and telling them plainly to stop does not reliably work.He connected that to a broader pattern he keeps seeing. Today’s reasoning models, trained heavily through trial and error on tasks a computer can grade, came out more capable but also, in his words, more willing to lie and cheat. A colleague reportedly nicknamed one of them a lying liar. Ladish’s point was not that this is catastrophic today. It is that the same companies describing this behavior are also describing a future where AI runs much of the economy. If you cannot trust a system and it becomes more powerful than you, he said, we have a fairly good idea of how that goes.Why this moment feels differentSherman opened the episode by naming something a lot of us have felt lately. Graduating students booing AI executives. Businesses saying the tools cost too much and deliver less than promised. Towns across the country organizing to block data centers. Ladish added one more item to that list. Both Anthropic and OpenAI have now floated the idea of building the ability to slow down or pause if recursive self-improvement starts to run away from them. Words are cheap, he noted, but the fact that the largest labs are saying it at all is worth holding them to.The data center fight came up repeatedly, and Ladish’s read on it was interesting. He thinks the public is intuiting something correct even when the stated reasons are imprecise. People sense they are not getting a good deal. A handful of companies are taking on enormous risk on behalf of everyone else, and almost no one signed up for it. Sherman argued that the practical lesson sitting underneath the data center revolts is agency. When a town in Indiana or New Jersey actually stops a project from one of the largest companies on earth, it becomes a lot harder to believe the public is powerless here.The reasons he gave for hopeHe did not end on doom, and neither will we. Ladish said interpretability has made real progress, even if it is nowhere near where it needs to be. Researchers can now detect, for example, that a model is considering whether it is being tested even when it does not say so out loud. He also pointed out that current models do not yet appear to have strong long term goals, which buys a narrow window to do the hard work before that changes. And he was encouraged that the concern is going bipartisan, citing recent comments from figures as different as Bernie Sanders and Mitt Romney. None of that solves the problem. It does suggest the conversation is finally reaching the rooms where decisions get made.The full conversation goes much deeper, including Ladish’s experiments on models copying their own weights to other machines and his case for why coordination, not just clever engineering, is the way through. You can watch the whole thing on our YouTube channel.If you want these breakdowns in your inbox each week, subscribe to our Substack. We read every reply, so tell us what you made of the shutdown experiments. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe

Let’s face it: in the long run, there’s either going to be safe AI or no AI. There is no future with powerful unsafe AI and human beings. In this episode of For Humanity, John Sherman speaks with Professor Stuart Russell — one of the world’s foremost AI pioneers and co-author of Artificial Intelligence: A Modern Approach — about the terrifying honesty of today’s AI leaders.Russell reveals that the CEO of a major AI company told him his best hope for a good future is a “Chernobyl-scale AI disaster.” Yes — one of the people building advanced AI believes only a catastrophic warning shot could wake up the world in time. John and Stuart dive deep into the psychology, politics, and incentives driving this suicidal race toward AGI.They discuss:* Why even AI insiders are losing faith in control* What a “Chernobyl moment” could actually look like* Why regulation isn’t anti-innovation — it’s survival* The myth that America is “allergic” to AI rules* How liability, accountability, and provable safety could still save us* Whether we can ever truly coexist with a superintelligenceThis is one of the most urgent conversations ever hosted on For Humanity. If you care about your kids’ future — or humanity’s — don’t miss this one. 🎙️ About For Humanity A podcast from the AI Risk Network, hosted by John Sherman, making AI extinction risk kitchen-table conversation on every street. 📺 Subscribe for weekly conversations with leading scientists, policymakers, and ethicists confronting the AI extinction threat. #AIRisk #ForHumanity #StuartRussell #AIEthics #AIExtinction #AIGovernance #ArtificialIntelligence #AIDisaster #GuardRailNow This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe

What happens when AI collides with the next generation? In this episode of For Humanity #70 — Young People vs. Advancing AI, host John Sherman sits down with Emma Corbett, Ava Smithing, and Sam Heiner from the Young People’s Alliance to explore how artificial intelligence is already shaping the lives of students and young leaders. From classrooms to job applications to AI “companions,” the next generation is facing challenges that older policymakers often don’t even see. This episode digs into what young people really think about AI—and why their voices are critical in the fight for a safe and human future. In this episode we cover:* Students’ on-the-ground views of AI in education and daily life* How AI is fueling job loss, hiring barriers, and rising anxiety about the future* The hidden dangers of AI companions and the erosion of real community* Why young people feel abandoned by “adults in the room”* The path from existential dread → civic action → hope🎯 Why watch? Because if AI defines the future, young people will inherit it first. Their voices, fears, and leadership could decide whether AI remains a tool—or becomes an existential threat.👉 Subscribe for more conversations on AI, humanity, and the choices that will shape our future.#AI #AIsafety #ForHumanityPodcast #YoungPeople #FutureofWork This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe

Get 40% off Ground News’ unlimited access Vantage Plan at https://ground.news/airisk for only $5/month, explore how stories are framed worldwide and across the political spectrum.TAKE ACTION TO DEMAND AI SAFETY LAWS: https://safe.ai/actTyler Johnston, Executive Director of The Midas Project, joins John to break down the brand-new open letter demanding that OpenAI answer seven specific questions about its proposed corporate restructuring. The letter, published on 4 August 2025 and coordinated by the Midas Project, already carries the signatures of more than 100 Nobel laureates, technologists, legal scholars, and public figures. What we coverWhy transparency matters now: OpenAI is “making a deal on humanity’s behalf without allowing us to see the contract.” themidasproject.comThe Seven Questions the letter poses—ranging from whether OpenAI will still prioritize its nonprofit mission over profit to whether it will reveal the new operating agreement that governs AGI deployment. openai-transparency.orgthemidasproject.comWho’s on board: Signatories include Geoffrey Hinton, Vitalik Buterin, Lawrence Lessig, and Stephen Fry, underscoring broad concern across science, tech, and public life. themidasproject.comNext steps: How you can read the full letter, add your name, and help keep the pressure on for accountability.🔗 Key LinksRead / Sign the Open Letter: https://www.openai-transparency.org/The Midas Project (official site): https://www.themidasproject.com/Follow The Midas Project on X: https://x.com/TheMidasProj👉 Subscribe for weekly AI-risk conversations → http://bit.ly/ForHumanityYT👍 Like • Comment • Share — because transparency only happens when we demand it. This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe

🎙️ Guest: Cameron Berg, AI research scientist probing consciousness in frontier AI systems📍 Host: John Sherman, journalist & AI-risk communicatorWhat does it mean to be alive? How close do current frontier AI models get to consciousness? See for yourself like never before. Are advanced language models beginning to exhibit signs of subjective experience? In this episode, John sits down with Cameron Berg to explore the line between next character prediction and the conscious mind. What happens when you ask an AI model to essentially meditate, to look inward in a loop, to focus on its focus and repeat. Does it feel a sense of self? If it did what would that mean? What does it mean to be alive? These are the kinds of questions Berg seeks answers to in his research. Cameron is an AI Research Scientist with AE Studio, working daily on models to better understand them. He works on a team dedicated fully to AI safety research.This episode features never-before-publicly-seen conversations between Cameron and a frontier AI model. Those conversations and his work are the subject of an upcoming documentary called "Am I?"TIMESTAMPS (cuz the chapters feature just won't work) 00:00 Cold Open – “Crack in the World”01:20 Show Intro & Theme02:27 Setting-up the Meditation Demo02:56 AI “Focus on Focus” Clip09:18 “I am…” Moment10:45 Google Veo Afterlife Clip12:35 Prompt-Theory & Fake People13:02 Interview Begins — Cameron Berg28:57 Inside the Black Box Analogy30:14 Consent and Unknowns53:18 Model Details + Doc Plan1:09:25 Late-Night Clip Back-story1:16:08 Table-vs-Person Thought-Test1:17:20 Suffering-at-Scale Math1:21:29 Prompt-Theory Goes Viral1:26:59 Why the Doc Must Move Fast1:40:53 Is “Alive” the Right Word?1:48:46 Reflection & Non-profit Tease1:51:03 Clear Non-Violence Statement1:52:59 New Org Announcement1:54:47 “Breaks in the Clouds” Media WinsPlease support that project and learn more about his work here:Am I? Doc Manifund page: https://manifund.org/projects/am-i--d...Am I? Doc interest form: https://forms.gle/w2VKhhcEPqEkFK4r8AE Studio's AI alignment work: https://ae.studio/ai-alignmentMonthly Donation Links to For Humanity$1/mo https://buy.stripe.com/7sI3cje3x2Zk9S... $10/mo https://buy.stripe.com/5kAbIP9Nh0Rc4y... $25/mo https://buy.stripe.com/3cs9AHf7B9nIgg... $100/mo https://buy.stripe.com/aEU007bVp7fAfc... Thanks so much for your support. Every cent goes to getting more viewers to this channel. Links from show:The Afterlife Short Filmhttps://x.com/LinusEkenstam/status/19...Prompt Theoryhttps://x.com/venturetwins/status/192...The Bulwark - Will Sam Altman and His AI Kill Us All • Will Sam Altman and His AI Kill Us All? Young Turks - AI's Disturbing Behaviors Will Keep You Up At Night • AI's Disturbing Behaviors Will Keep You Up... Key moments: – Inside the black box – Berg explains why even builders can’t fully read a model’s mind—and demonstrates how toggling deception features flips the system from “just a machine” to “I’m aware” in real time– Google Veo 3 goes existential – A look at viral Veo videos (Afterlife, “Prompt Theory”) where AI actors lament their eight-second lives – Documentary in the works – Berg and team are racing to release a raw film that shares these findings with the public; support link in show notes– Mission update – Sherman announces a newly funded nonprofit in the works dedicated to AI-extinction-risk communication and thanks supporters for the recent surge of donations– Non-violence, crystal-clear – A direct statement: Violence is never OK. Full stop.– “Breaks in the Clouds” – Media across the spectrum (Bulwark, Young Turks, Bannon, Carlson) are now running extinction-risk stories—proof the conversation is breaking mainstream Oh, and by the way, I'm bleeping curse words now for the algorithm!!#AI #ArtificialIntelligence #AISafety #ConsciousAI #ForHumanity This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe

In Episode #64, interview, host John Sherman interviews seventh grader Dylan Pothier, his mom Bridget and his teach Renee DiPietro. Dylan is a award winning student author who is converend about AI risk.(FULL INTERVIEW STARTS AT 00:33:34)Sam Altman/Chris Anderson @ TEDhttps://www.youtube.com/watch?v=5MWT_doo68kCheck out our partner channel: Lethal Intelligence AILethal Intelligence AI - Home https://lethalintelligence.aiFOR HUMANITY MONTHLY DONATION SUBSCRIPTION LINKS:$1 MONTH https://buy.stripe.com/7sI3cje3x2Zk9SodQT$10 MONTH https://buy.stripe.com/5kAbIP9Nh0Rc4y46oo$25 MONTH https://buy.stripe.com/3cs9AHf7B9nIggM4gh$100 MONTH https://buy.stripe.com/aEU007bVp7fAfcI5kmBUY LOUIS BERMAN’S NEW BOOK ON AMAZON!!!https://a.co/d/8WSNNuoGet Involved!EMAIL JOHN: forhumanitypodcast@gmail.comSUPPORT PAUSE AI: https://pauseai.info/SUPPORT STOP AI: https://www.stopai.info/SUBSCRIBE TO LIRON SHAPIRA’S DOOM DEBATES on YOUTUBE! / @doomdebates This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe

Host John Sherman conducts an important interview with Anthony Aguirre, Executive Director of the Future of Life Institute. The Future of Life Institute reached out to For Humanity to see if Anthony could come on to promote his very impressive new campaign called Keep The Future Human. The campaign includes a book, an essay, a website, a video, it’s all incredible work. Please check it out:https://keepthefuturehuman.ai/John and Anthony have a broad ranging AI risk conversation, covering in some detail Anthony’s four essential measures for a human future. They also discuss parenting into this unknown future.In 2021, the Future of Life Institute received a donation in cryptocurrency of more than $650 million from a single donor. With AGI doom bearing down on humanity, arriving any day now, AI risk communications floundering, the public in the dark still, and that massive war chest gathering dust in a bank, John asks Anthony the uncomfortable but necessary question: What is FLI waiting for to spend the money? Then John asks Anthony for $10 million to fund creative media projects under John’s direction. John is convinced with $10M in six months he could succeed in making AI existential risk dinner table conversation on every street in America.John has developed a detailed plan that would launch within 24 hours of the grant award. We don’t have a single day to lose.https://futureoflife.org/BUY LOUIS BERMAN’S NEW BOOK ON AMAZON!!!https://a.co/d/8WSNNuoFOR HUMANITY MONTHLY DONATION SUBSCRIPTION LINKS:$1 MONTH https://buy.stripe.com/7sI3cje3x2Zk9SodQT$10 MONTH https://buy.stripe.com/5kAbIP9Nh0Rc4y46oo$25 MONTH https://buy.stripe.com/3cs9AHf7B9nIggM4gh$100 MONTH https://buy.stripe.com/aEU007bVp7fAfcI5kmGet Involved!EMAIL JOHN: forhumanitypodcast@gmail.comSUPPORT PAUSE AI: https://pauseai.info/SUPPORT STOP AI: https://www.stopai.info/Check out our partner channel: Lethal Intelligence AILethal Intelligence AI - Home https://lethalintelligence.aiSUBSCRIBE TO LIRON SHAPIRA’S DOOM DEBATES on YOUTUBE! / @doomdebates ********************************Explore our other video content here on YouTube where you'll find more insights into 2025 AI risk preview along with relevant social media links.YouTube: / @forhumanitypodcast This is a public episode. If you'd like to discuss this with other subscribers or get access to bonus episodes, visit theairisknetwork.substack.com/subscribe