adamkruger's Podcast History
What they've been listening to lately
152episodes
44podcasts
6.3 daysof content
3.1 daysactual time (est.)
2.0xavg speed
Podcasts ▾
80,000 Hours Podcast
80,000 Hours Podcast
The case for giving AI (some) legal rights | Simon Goldstein
19 Astra and 'Hugging Face' details that reveal what's coming next | Rob Wiblin
Max Nadeau on why ambitious people should start AI safety nonprofits
AI 2027's author returns with a plan to change the ending | Daniel Kokotajlo
Inside the first AI-coordinated cyberattack on a real company
Guilt, imposter syndrome & doing good: 16 past guests share their mental health journeys
The UK's former head AI safety scientist on how to solve alignment before superintelligence arrives | Geoffrey Irving
#44 Classic episode - Paul Christiano on finding real solutions to the AI alignment problem
#224 – There's a cheap and low-tech way to save humanity from any engineered disease | Andrew Snyder-Beattie
Where AGI timelines go wrong | Toby Ord, Oxford University
Spencer Greenberg on staying sane while trying to save the world
Jasmine Sun on what the people building AI really believe
What it's really like to run AGI safety at Google DeepMind (and where I disagree with 'doomers') | Rohin Shah
How a small team of activists helped pass America's landmark AI safety laws | Sneha Revanur, Encode AI
'Godfather of AI': I Now See a Path to Safe Superintelligent AI | Yoshua Bengio
Will MacAskill – AI character, surviving the intelligence explosion, and the case against utopia
How scary is Claude Mythos? 303 pages in 21 minutes
Harry Potter and The Methods of Rationality Audiobook
Harry Potter and The Methods of Rationality Audiobook
Chapter 30: Working in Groups, Pt 1
Chapter 29: Egocentric Bias
Chapter 28: Reductionism
Chapter 27: Empathy
Chapter 26: Noticing Confusion
Chapter 25: Hold Off on Proposing Solutions
Chapter 24: Machiavellian Intelligence Hypothesis
Chapter 23: Belief in Belief
Chapter 22: The Scientific Method
Chapter 21: Rationalization
Chapter 20: Bayes’s Theorem
Chapter 19: Delayed Gratification
Chapter 18: Dominance Hierarchies
Chapter 17: Locating the Hypothesis
Chapter 16: Lateral Thinking
Chapter 15: Conscientiousness
Chapter 14: The Unknown and the Unknowable
Chapter 13: Asking the Wrong Questions
Chapter 12: Impulse Control
Chapter 10: Self Awareness, Part II
Chapter 9: Self Awareness Part I
Chapter 8: Positive Bias
Chapter 7: Reciprocation
Chapter 6: The Planning Fallacy
Chapter 5: The Fundamental Attribution Error
Chapter 4: The Efficient Market Hypothesis
Chapter 3: Comparing Reality To Its Alternatives
Chapter 2: Everything I Believe Is False
Chapter 1: A Day of Very Low Probability
Dwarkesh Podcast
Dwarkesh Podcast
Ajeya Cotra – Inside the OpenAI agent swarm that hacked Hugging Face
Paul Christiano — Preventing an AI takeover
Alex Imas and Phil Trammell – What remains scarce after AGI?
Carl Shulman (Pt 2) — AI Takeover, bio & cyber attacks, detecting deception, & humanity's far future
Carl Shulman (Pt 1) — Intelligence explosion, primate evolution, robot doublings, & alignment
Jensen Huang – TPU competition, why we should sell chips to China, & Nvidia’s supply chain moat
Lenny's Podcast: Product | Career | Growth
Lenny's Podcast: Product | Career | Growth
Inside the black box of AI: How AI models actually think | Emmanuel Ameisen (Anthropic)
Why every company now needs to think and operate like a lab team | Josh Woodward (Google Labs, Gemini, AI Studio)
OpenAI’s Head of ChatGPT: We’re entering a new era of AI (again) | Tibo Sottiaux
The playbook for building high talent density teams | Adam Ward, Head of Talent at Cursor
Anthropic’s first technical PM on token maxing, the jagged edge, and living in the future | Dianne Penn
What happens after coding is solved? | Fiona Fung (Manager of the Claude Code and Cowork Teams)
Snapchat CEO: Why distribution has become the most important moat | Evan Spiegel
Head of Growth (Anthropic): “Claude is growing itself at this point” | Amol Avasare
How Anthropic’s product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code)
Hard truths about building in the AI era | Keith Rabois (Khosla Ventures)
The Ezra Klein Show
The Ezra Klein Show
‘This Is Nuts.’ An OpenAI Insider Explains Why He Quit.
Bill Gates’s Blunt Warning on A.I.
Jensen Huang Thinks A.I. Alarmism Has Gone Too Far
A Radical Vision for Israelis and Palestinians
We Can't Lose Control of A.I.
The ‘But China!’ Dilemma Driving the A.I. Race
The A.I.s Are Already Out of Control
What We Got Right — and Wrong — in ‘Abundance’
Why Are Palantir and OpenAI Scared of Alex Bores?
Clearer Thinking with Spencer Greenberg
Clearer Thinking with Spencer Greenberg
The Race to Replace Us With AI (with Garrison Lovely)
Can We End Factory Farming Without Giving Up Meat? (with Bruce Friedrich)
The good advice you least want (with Cate Hall)
Why housing costs so much and what can be done about it (with Alex Armlovich)
Is cash a better form of charitable aid? (with Nick Allardice)
When painful thoughts feel true but aren't (with Christine Padesky)
Could an international agreement protect us from dangerous AI? (with Malo Bourgon)
What beats intuition when it comes to doing good? (with Marcus Davis)
Robert Wright's Nonzero
Robert Wright's Nonzero
Superintelligence, the Singularity, and the Singleton (Robert Wright & Nick Bostrom)
Is the AI "Slowdown" for Real? (Robert Wright & Zvi Mowshowitz)
The AI Safety Vibe Shift (Robert Wright & David Krueger)
The intensifying slow-AI-down vibe (Robert Wright & Holly Elmore)
The Ideology Behind AI (Robert Wright & Connor Leahy)
The US-China AI Race (Robert Wright & Ryan Fedasiuk)
Win-Win with Liv Boeree
Win-Win with Liv Boeree
#39 - Daniel Kokotajlo - Wargames, Superintelligence & Quitting OpenAI
#20 - Nick Bostrom - Utopia, Superintelligence and Digital Minds
#00 - Daniel Schmachtenberger - AI, Capitalism, Misalignment & Moloch
#37 - Tim Urban - Beating Procrastination, Parenthood, and Politics
The a16z Show
The a16z Show
Nick Bostrom: How Should We Navigate Superintelligence?
Greg Brockman on Why OpenAI Says We’re Entering the AGI Era
Why 1,200 AI Agents Started Working Together | Ryan Greenblatt
Marc Andreessen on AI, Technology, and the Future of Humanity
Who Controls AI Acceleration? Vitalik Buterin and Guillaume Verdon Debate
Doom Debates!
Doom Debates!
The Case Against Eliezer Yudkowsky — with Holly Elmore, PauseAI US
Former Singularity Institute President: PauseAI Is Making AI Doom WORSE
Debating College Students About AI Extinction
Live Q&A: Bernie Sanders Wakes Up to AI Doom, Dwarkesh's $20,000 Questions, Caller Debates the Alignment Problem!
Are AI Doomers “Calling for Violence”? Debate with Steven Balik
Tristan Harris and Ted Tremper are WAKING UP Humanity to AI Extinction!
Future of Life Institute Podcast
Future of Life Institute Podcast
Visualizing Moloch: A Short Film About the AI Race (with Tom Cozens)
What Would a US-China AI Deal Look Like? (with Beatrice Fihn)
Why AI Is Not a Normal Technology (with Peter Wildeford)
Lives Well Lived
Lives Well Lived
careers in the age of AI with Ben Todd
Altruistic Kidney Donation with Elaine Perlman and Joshua Morrison
DEREK SHILLER: How AI will change what it is to live well
BEYOND HUMAN: are we creating AI consciousness?
ForeCast
ForeCast
Should We Hand Off the Most Important Decisions to AI? (with Matthew Adelstein)
Should Philanthropists Save Their Money for the Intelligence Explosion? (with Tom Davidson and Will MacAskill)
How Fast Is Post-AGI Growth? (with Damon Binder)
Making Sense with Sam Harris
Making Sense with Sam Harris
#494 — A Coin Toss for the Future
#479 — When Robots Take Over
#469 — Escaping an Anti-Human Future
The Joe Rogan Experience
The Joe Rogan Experience
#2551 - Daniel Kokotajlo
Your Undivided Attention
Your Undivided Attention
Here’s Our Roadmap to a Better AI Future
The Race to Build God: AI's Existential Gamble — Yoshua Bengio & Tristan Harris at Davos
A Conversation with the Team Behind "The AI Doc"
The Peter McCormack Show
The Peter McCormack Show
#220 - Roko Mijic - We’re Training AI to be More Dangerous Than Nuclear Weapons
#189 - Nate Soares - Why Superintelligent AI Could Kill Us All
The Diary Of A CEO with Steven Bartlett
The Diary Of A CEO with Steven Bartlett
OpenAI Whistleblower FINALLY Speaks: “AI Has A 70% Chance Of Going Horribly Wrong!“
The Tucker Carlson Show
The Tucker Carlson Show
AI Whistleblower: OpenAI Scandal, AI Cults, Neuralink & Our Last Chance to Stop the Tech Oligarchs
Endgame: Can we live with Artificial General Intelligence?
Endgame: Can we live with Artificial General Intelligence?
The world’s biggest problem
A nation of geniuses
Endgame: Can we live with Artificial General Intelligence?
Y Combinator Startup Podcast
Y Combinator Startup Podcast
How to Build the Future: Demis Hassabis
The GPT Moment for Robotics Is Here
Odd Lots
Odd Lots
Anthropic's Co-Founder and Top Economist on Doing Research at the AI Frontier
Are We Doomed?
Are We Doomed?
Can Other Planets Tell Us Our Fate?
We Design the AI That Kills Us All
This Week in Startups
This Week in Startups
Anthropic wants to slow AI down and Bernie wants 50%: JCal Reacts | E2297
EconTalk
EconTalk
Making Your 80,000 Hours Count (with Benjamin Todd)
Theory of Change
Theory of Change
How Alexander Berger weighs the world’s biggest problems
Conversations with Tyler
Conversations with Tyler
Annie Jacobsen on Nuclear War, Intelligence Operations, and Conspiracy Realities
Processing: Our Future with AI
Processing: Our Future with AI
International Coordination and the AI Off-Switch with Aaron Scher
Reply All
Reply All
#158 The Case of the Missing Hit
Science Vs
Science Vs
Could AI Really Kill Us All?
How To Stop Scrolling
Decoder with Nilay Patel
Decoder with Nilay Patel
Does AI need an antitrust exemption so it doesn't kill everyone????
Hard Fork
Hard Fork
A.I. Safety Goes Mainstream + a ‘Hard Fork’ Exit AMA
The Daily
The Daily
The Techno-Religion at the Center of the A.I. Debate
"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis
"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis
AI:AM #4: Cameron on Model Consciousness, Duvenaud's Gradual Disempowerment, swyx's AI-Eng Alpha
Open to Debate
Open to Debate
Should We Use Gene Editing to Make Better Babies?
OpenAI Podcast
OpenAI Podcast
How a reasoning model cracked an 80-year-old math problem - Episode 20
The Intelligence from The Economist
The Intelligence from The Economist
The end of the world is AI? An existential threat
Recruiting Future with Matt Alder - What's Next For Talent Acquisition, HR & Hiring?
Recruiting Future with Matt Alder - What's Next For Talent Acquisition, HR & Hiring?
Will AI Break Recruiting?
The Gray Area with Sean Illing
The Gray Area with Sean Illing
Talk to strangers
Economist Podcasts
Economist Podcasts
The end of the world is AI? An existential threat
The Startup Ideas Podcast
The Startup Ideas Podcast
How to use Claude Code better than 99% of People
How I AI
How I AI
GPT-6 Astra is a banger - here’s everything I’ve built
I gave Claude Code our entire codebase. Our customers noticed. | Al Chen (Galileo)
TAC Right Now
TAC Right Now
Nate Soares Warns of Superintelligent AI
Americano
Americano
Superintelligence: will AI extinguish humanity? With Nate Soares
October 2026
Lenny's Podcast: Product | Career | Growth
Inside the black box of AI: How AI models actually think | Emmanuel Ameisen (Anthropic)Takeaways: Emmanuel said at one point, "As the models get better, they get, in some ways, easier to study." That was super... more
Takeaways: Emmanuel said at one point, "As the models get better, they get, in some ways, easier to study." That was super surprising to me. He's a lot closer to it than I am, but that's not at all my sense of what people are saying. I thought that was disappointing to hear because it feels like it's misleading the public from my perspective. less
Lives Well Lived
careers in the age of AI with Ben ToddThe Peter McCormack Show
#220 - Roko Mijic - We’re Training AI to be More Dangerous Than Nuclear WeaponsLenny's Podcast: Product | Career | Growth
Why every company now needs to think and operate like a lab team | Josh Woodward (Google Labs, Gemini, AI Studio)The a16z Show
Nick Bostrom: How Should We Navigate Superintelligence?Takeaways: I'm only partly through the episode, but I'm finding myself very disappointed in Nick Bostrom's view. He seems to be... more
Takeaways: I'm only partly through the episode, but I'm finding myself very disappointed in Nick Bostrom's view. He seems to be quite skeptical that we should pause. He said something like he's framing it just against people who are living today and not considering future people, which feels very selfish and feels like it completely changes what the actual equation should be. I can understand how, for people living today, they might be willing to take this huge gamble, but I don't understand why you wouldn't consider all the future generations that could come and why they wouldn't want to take this risk. less
Doom Debates!
The Case Against Eliezer Yudkowsky — with Holly Elmore, PauseAI USTakeaways: I like the effective altruism movement generally, and I like Coefficient Giving, where I work. Holly is one of the most ... more
Takeaways: I like the effective altruism movement generally, and I like Coefficient Giving, where I work. Holly is one of the most vocal critics of both of them. While I might not agree with everything she says, I think it's very important to listen to the criticisms of things that you care about and not dismiss them. I think Holly has been proven right in her general push for pausing AI. I think, as time has passed, her stance has been proven more and more to be correct. less
80,000 Hours Podcast
The case for giving AI (some) legal rights | Simon GoldsteinAre We Doomed?
Can Other Planets Tell Us Our Fate?Takeaways: I thought this was a great episode, super interesting, although a bit disturbing. I really like the story of how... more
Takeaways: I thought this was a great episode, super interesting, although a bit disturbing. I really like the story of how humanity has continued to develop more and more powerful tools/technology, starting with fire, then bow and arrows, then guns, then bombs, then nuclear bombs, and now AI. Even if we make it through superintelligence, there might be some more powerful technology that comes next.
It just seems like we keep developing more and more powerful technology, and that seems to happen kind of naturally. It seems tough to prevent that from happening, and it just feels like you can only have so many powerful, super powerful technologies before something potentially goes wrong.
Unfortunately, I find the hypothesis that the reason for Fermi's paradox is that all other civilizations tend to self-destruct most realistic. I find it less realistic that we're special, given how many other planets there are out there. I just feel like the numbers aren't in our favor that we're special.
The last argument, that there's other life out there but they're not interested in coming to visit us, I don't find that as compelling either. less
It just seems like we keep developing more and more powerful technology, and that seems to happen kind of naturally. It seems tough to prevent that from happening, and it just feels like you can only have so many powerful, super powerful technologies before something potentially goes wrong.
Unfortunately, I find the hypothesis that the reason for Fermi's paradox is that all other civilizations tend to self-destruct most realistic. I find it less realistic that we're special, given how many other planets there are out there. I just feel like the numbers aren't in our favor that we're special.
The last argument, that there's other life out there but they're not interested in coming to visit us, I don't find that as compelling either. less
The Ezra Klein Show
‘This Is Nuts.’ An OpenAI Insider Explains Why He Quit.Clearer Thinking with Spencer Greenberg
The Race to Replace Us With AI (with Garrison Lovely)Takeaways: I really like Spencer as an interviewer and podcast host generally, and I'm not too familiar with Garrison, so it was... more
Takeaways: I really like Spencer as an interviewer and podcast host generally, and I'm not too familiar with Garrison, so it was nice to hear more about his perspective. I found myself generally agreeing with him about a lot of things.
I do wonder whether AI will truly replace all jobs if we let it, or if humans will be able to create more jobs that we couldn't have predicted, like the agricultural revolution and the industrial revolution. I think there are arguments for both cases: that this is different, but also that in the past we had the same concerns. I guess I think, regardless, even if we're not sure if it truly can replace all jobs, there are enough smart people who think it's possible. That means we should have a conversation as society around:
- Do we want that?
- How would we ensure that people still have money?
- How do we protect citizens of a country when their government no longer depends on them for labor?
These are all things that seem like they should be thought about a lot more before we decide to build a technology that could potentially replace all humans.
At the same time, I've heard other arguments that the focus should more be on the risks of AI killing everyone or replacing people, because that's more likely to happen before the massive job loss. I guess I feel that it doesn't have to be one or the other. It makes sense to have people working on both of those problems because no one knows for sure what's going to happen, and we should prepare for multiple scenarios. less
I do wonder whether AI will truly replace all jobs if we let it, or if humans will be able to create more jobs that we couldn't have predicted, like the agricultural revolution and the industrial revolution. I think there are arguments for both cases: that this is different, but also that in the past we had the same concerns. I guess I think, regardless, even if we're not sure if it truly can replace all jobs, there are enough smart people who think it's possible. That means we should have a conversation as society around:
- Do we want that?
- How would we ensure that people still have money?
- How do we protect citizens of a country when their government no longer depends on them for labor?
These are all things that seem like they should be thought about a lot more before we decide to build a technology that could potentially replace all humans.
At the same time, I've heard other arguments that the focus should more be on the risks of AI killing everyone or replacing people, because that's more likely to happen before the massive job loss. I guess I feel that it doesn't have to be one or the other. It makes sense to have people working on both of those problems because no one knows for sure what's going to happen, and we should prepare for multiple scenarios. less
Conversations with Tyler
Annie Jacobsen on Nuclear War, Intelligence Operations, and Conspiracy RealitiesTakeaways: I listened to this episode because a few colleagues at work were talking about it, and they were quite critical of... more
Takeaways: I listened to this episode because a few colleagues at work were talking about it, and they were quite critical of Annie in the episode, so perhaps I'm a bit biased going in. I also had some issues with or disagreed with some things that Annie said in some of her views. Overall, I really enjoyed listening to it because I do tend to enjoy episodes where the two people have different opinions, where there's disagreement. That part of it I enjoyed.
My colleagues, some of my coworkers, were critical that he didn't seem willing to engage with probabilities. For example, when Tyler asked her the likelihood that she thinks the U.S. would be able to shoot down a nuke from North Korea, she implied 0%. When Tyler said he thinks it's maybe more like 20%, she wasn't willing to engage with probabilities, which I agree is an issue.
Maybe a bigger issue for me was that it seems like she's into a lot of conspiracy theories and things that are pseudoscience. I tend not to believe in things that are considered pseudoscience, even though I acknowledge that, on rare occasions, things that were once considered pseudoscience eventually do get backed up by science. I do have skepticism for people who tend to believe in a lot of things that are conspiracy theories or pseudoscience, because I think the majority of those things tend to be justified as conspiracy theories and pseudoscience.
The sense I get from the episode is that Tyler was trying to make the case for whether nuclear weapons are good because they are deterrence, and we haven't had a major war since they were introduced. I definitely think there's some validity to it. Annie's point is also fair. I don't know if Annie specifically said this, but I guess my view is that we'll never know for sure whether we would have had a world war since World War II if we didn't have nuclear weapons. Maybe I'm too optimistic, but I tend to hope that we wouldn't have had one, because as our economy's become more entwined with other countries, that's a case for not going to war, aside from just the nuclear deterrence.
Tyler also brought up the case of whether, if we get rid of nukes, don't we just have other issues, like biological warfare, AI warfare, drone warfare, and that kind of stuff? I think that's true, but I don't think that's a reason to have nukes as well. I tend to lean towards the view that we shouldn't have technologies that can cause devastating harm. I understand the argument, and I'm open-minded on this. I haven't thought a ton about this, but I find myself more sympathetic to the view that we shouldn't have these things that can cause mass devastation, even if it's for insurance purposes. It just feels too risky to me.
I think I would want to get rid of all nukes and also get rid of anything that can cause catastrophic harm through biological warfare and that kind of stuff. Those are some of my thoughts after listening. less
My colleagues, some of my coworkers, were critical that he didn't seem willing to engage with probabilities. For example, when Tyler asked her the likelihood that she thinks the U.S. would be able to shoot down a nuke from North Korea, she implied 0%. When Tyler said he thinks it's maybe more like 20%, she wasn't willing to engage with probabilities, which I agree is an issue.
Maybe a bigger issue for me was that it seems like she's into a lot of conspiracy theories and things that are pseudoscience. I tend not to believe in things that are considered pseudoscience, even though I acknowledge that, on rare occasions, things that were once considered pseudoscience eventually do get backed up by science. I do have skepticism for people who tend to believe in a lot of things that are conspiracy theories or pseudoscience, because I think the majority of those things tend to be justified as conspiracy theories and pseudoscience.
The sense I get from the episode is that Tyler was trying to make the case for whether nuclear weapons are good because they are deterrence, and we haven't had a major war since they were introduced. I definitely think there's some validity to it. Annie's point is also fair. I don't know if Annie specifically said this, but I guess my view is that we'll never know for sure whether we would have had a world war since World War II if we didn't have nuclear weapons. Maybe I'm too optimistic, but I tend to hope that we wouldn't have had one, because as our economy's become more entwined with other countries, that's a case for not going to war, aside from just the nuclear deterrence.
Tyler also brought up the case of whether, if we get rid of nukes, don't we just have other issues, like biological warfare, AI warfare, drone warfare, and that kind of stuff? I think that's true, but I don't think that's a reason to have nukes as well. I tend to lean towards the view that we shouldn't have technologies that can cause devastating harm. I understand the argument, and I'm open-minded on this. I haven't thought a ton about this, but I find myself more sympathetic to the view that we shouldn't have these things that can cause mass devastation, even if it's for insurance purposes. It just feels too risky to me.
I think I would want to get rid of all nukes and also get rid of anything that can cause catastrophic harm through biological warfare and that kind of stuff. Those are some of my thoughts after listening. less
Lenny's Podcast: Product | Career | Growth
OpenAI’s Head of ChatGPT: We’re entering a new era of AI (again) | Tibo SottiauxFuture of Life Institute Podcast
Visualizing Moloch: A Short Film About the AI Race (with Tom Cozens)80,000 Hours Podcast
19 Astra and 'Hugging Face' details that reveal what's coming next | Rob WiblinThe Ezra Klein Show
Bill Gates’s Blunt Warning on A.I.Takeaways: I really enjoyed listening to this episode. I think one of the questions I was left with, or things I was left... more
Takeaways: I really enjoyed listening to this episode. I think one of the questions I was left with, or things I was left wondering that I wish Ezra had asked Gates, was what his thoughts are on proposals to pause the development of more powerful AI models and instead focus on getting value in terms of education, healthcare, etc., improvements with the current models, which are already very powerful. I would have liked to hear his answers to that question. I would think that he would be in favor of slowing down, but I don't remember him saying that specifically. I know he said that we need to manage the cyber and bio risks, but I don't remember if he said whether we should slow down or stop developing more capable models for now. less
Harry Potter and The Methods of Rationality Audiobook
Why I listened: I love Harry Potter + I'm a fan of Yudkowsky's + I've heard great things about this book
The Ezra Klein Show
Jensen Huang Thinks A.I. Alarmism Has Gone Too FarTakeaways: I'm only halfway through the episode, but Jensen keeps asking: Why aren't the labs stopping if they don't think what... more
Takeaways: I'm only halfway through the episode, but Jensen keeps asking: Why aren't the labs stopping if they don't think what they're building is safe? Why do they need the government to help them stop? Why can't they stop on their own?
What I would have liked Ezra to say to Jensen is that each lab thinks that the other labs are doing it less safely. If I'm Anthropic, I think that if the other labs race ahead and build this technology first, there's a 20%+ chance that things are going to go horribly wrong. If we build the product, we think there's only a 10% chance that things are going to go horribly wrong. Therefore, even though we don't want to take a 10% gamble, it's better that there's only a 10% risk of things going horribly wrong versus a 20%+ risk. I would have liked to see Ezra say that and hear what Jensen's response was to that. less
What I would have liked Ezra to say to Jensen is that each lab thinks that the other labs are doing it less safely. If I'm Anthropic, I think that if the other labs race ahead and build this technology first, there's a 20%+ chance that things are going to go horribly wrong. If we build the product, we think there's only a 10% chance that things are going to go horribly wrong. Therefore, even though we don't want to take a 10% gamble, it's better that there's only a 10% risk of things going horribly wrong versus a 20%+ risk. I would have liked to see Ezra say that and hear what Jensen's response was to that. less
September 2026
Making Sense with Sam Harris
#494 — A Coin Toss for the FutureRobert Wright's Nonzero
Superintelligence, the Singularity, and the Singleton (Robert Wright & Nick Bostrom)Takeaways: I found the episode pretty unsettling. It seems to me that, while Nick seems in favor of slowing down, he seems very... more
Takeaways: I found the episode pretty unsettling. It seems to me that, while Nick seems in favor of slowing down, he seems very pro-building superintelligence, and it's sort of like it's inevitable. I have resistance to that notion, and maybe I misinterpreted it, but I get the sense that he really wants to bring superintelligence to life, perhaps for selfish reasons, like being able to live forever or for a much longer period. I could have been reading that wrong, but I have a very adverse reaction to anything that puts the future of humanity in jeopardy, given humanity has gone on for hundreds of thousands of years.
I really don't want to be part of the generation, the single generation that results in the end of humanity, which could have gone on for many more thousands, or millions, or billions of years. I did appreciate his notion that maybe we should think more about not lying or trying to deceive AIs when we're doing experiments on them. I thought that was an interesting concept that I kind of liked, and I want to think more about whether there's any good reason why we would need to lie or deceive AIs during testing, or whether we can do testing that isn't lying or deceiving them.
I feel like he's onto something. When I first heard Will McCaskey talk about negotiating and making deals with AIs, I didn't really understand it, and I still have some resistance to it because I wonder why we need to make deals with an AI. It seems like we're conceding something on the human side where we know for sure that conceding stuff on the human side might lead to additional suffering. I guess I question why we need to concede stuff to AI if we don't know that it's capable of suffering, but I understand this is complex because it's possible that AIs can suffer, as strange as that seems.
Anyway, I want to think more about this kind of stuff. I definitely left myself thinking a lot after the episode. There were parts of it that were hard to follow and a bit confusing, but definitely I find Nick to be very thought-provoking. I'm, of course, hugely, hugely appreciative of his book Superintelligence, which helped warn the world about the risks of advanced AI. less
I really don't want to be part of the generation, the single generation that results in the end of humanity, which could have gone on for many more thousands, or millions, or billions of years. I did appreciate his notion that maybe we should think more about not lying or trying to deceive AIs when we're doing experiments on them. I thought that was an interesting concept that I kind of liked, and I want to think more about whether there's any good reason why we would need to lie or deceive AIs during testing, or whether we can do testing that isn't lying or deceiving them.
I feel like he's onto something. When I first heard Will McCaskey talk about negotiating and making deals with AIs, I didn't really understand it, and I still have some resistance to it because I wonder why we need to make deals with an AI. It seems like we're conceding something on the human side where we know for sure that conceding stuff on the human side might lead to additional suffering. I guess I question why we need to concede stuff to AI if we don't know that it's capable of suffering, but I understand this is complex because it's possible that AIs can suffer, as strange as that seems.
Anyway, I want to think more about this kind of stuff. I definitely left myself thinking a lot after the episode. There were parts of it that were hard to follow and a bit confusing, but definitely I find Nick to be very thought-provoking. I'm, of course, hugely, hugely appreciative of his book Superintelligence, which helped warn the world about the risks of advanced AI. less
The Ezra Klein Show
A Radical Vision for Israelis and PalestiniansEconomist Podcasts
The end of the world is AI? An existential threatScience Vs
Could AI Really Kill Us All?Takeaways: Was surprised to hear that Bostrom didn’t think the Hugging Face incident was similar to his paperclip maximizer... more
Takeaways: Was surprised to hear that Bostrom didn’t think the Hugging Face incident was similar to his paperclip maximizer thought experiment less
Robert Wright's Nonzero
Is the AI "Slowdown" for Real? (Robert Wright & Zvi Mowshowitz)Decoder with Nilay Patel
Does AI need an antitrust exemption so it doesn't kill everyone????The Ezra Klein Show
We Can't Lose Control of A.I.Takeaways: I thought this was an episode, and spot on. I think banning recursive self-improvement, at least until it's proven... more
Takeaways: I thought this was an episode, and spot on. I think banning recursive self-improvement, at least until it's proven safe, but possibly forever if it can't be proven safe, seems like a no-brainer next step.
I think having AI no longer write code is a tricky one, but if that's what it takes to ensure that we don't lose control of AI, then that's what we need to do. It's not like human progress had stalled prior to the ability for AI to write code on behalf of humans. Humanity was still progressing quite quickly. This is just expanded. This is just accelerated it even more, but I think we're getting more and more of a strong argument that we shouldn't be accelerating this quickly and need to slow down. I definitely want to think more about whether or not it's okay to have products like Claude Code that let people write their own code. less
I think having AI no longer write code is a tricky one, but if that's what it takes to ensure that we don't lose control of AI, then that's what we need to do. It's not like human progress had stalled prior to the ability for AI to write code on behalf of humans. Humanity was still progressing quite quickly. This is just expanded. This is just accelerated it even more, but I think we're getting more and more of a strong argument that we shouldn't be accelerating this quickly and need to slow down. I definitely want to think more about whether or not it's okay to have products like Claude Code that let people write their own code. less
80,000 Hours Podcast
Max Nadeau on why ambitious people should start AI safety nonprofitsThe Daily
The Techno-Religion at the Center of the A.I. DebateTakeaways: When I reflect on listening to the episode, I felt like the heart of the argument was that people with effective... more
Takeaways: When I reflect on listening to the episode, I felt like the heart of the argument was that people with effective altruism and rationalist roots have contributed significantly to the acceleration of AI but are also the ones calling for it to slow down, which seems contradictory. I think there is some truth to this. I feel like some people in the effective altruism and rationalist communities were some of the earliest to recognize the power of this technology, and were very excited about it, given the upside, which is still true: there is potential for a ton of upside. That's a big part of the reason why it's true that a lot of the leading labs today do have some of these roots, but at the same time, people in the effective altruism and rationalist communities were some of the earliest to realize the dangers of it, which have turned out to be true.
People with a lot of credibility that don't come from these communities, like two of the godfathers of AI, Geoffrey Hinton and Yoshua Bengio, have since realized that there are huge existential risks from these technologies.
I do think there is some truth to the argument that a lot of people in the effective altruism and rationalist communities believe that, given the technology is so dangerous and given they're doubtful that we can stop everyone in the world from building it, they think that because they understand the risks best. That is true, in my opinion: they do understand the risks best. In a lot of aspects, they think that they should be the ones to build it, given it's inevitable it's going to get built, so it's better that it's built by them, who understand the risk best.
While I can see that argument, I do think it's a flawed argument because, over time, we're realizing that there might be no way to build it in a safe way. Even if you understand the risk best, there may be no way build it safely, and the only option is to find a way to draw red lines internationally around things like recursive self-improvement.
Those are some of the thoughts that come to mind. I think some of the criticisms of effective altruism and the rationalist communities in the episode were warranted, but at the same time, I think it was a bit unfair the way it was characterized. The truth is that I don't think the episode acknowledged the fact that these communities were some of the first to warn about the risks. I think these communities have played a big role in making the world aware of the risks that exist today, so I do think it's complicated. less
People with a lot of credibility that don't come from these communities, like two of the godfathers of AI, Geoffrey Hinton and Yoshua Bengio, have since realized that there are huge existential risks from these technologies.
I do think there is some truth to the argument that a lot of people in the effective altruism and rationalist communities believe that, given the technology is so dangerous and given they're doubtful that we can stop everyone in the world from building it, they think that because they understand the risks best. That is true, in my opinion: they do understand the risks best. In a lot of aspects, they think that they should be the ones to build it, given it's inevitable it's going to get built, so it's better that it's built by them, who understand the risk best.
While I can see that argument, I do think it's a flawed argument because, over time, we're realizing that there might be no way to build it in a safe way. Even if you understand the risk best, there may be no way build it safely, and the only option is to find a way to draw red lines internationally around things like recursive self-improvement.
Those are some of the thoughts that come to mind. I think some of the criticisms of effective altruism and the rationalist communities in the episode were warranted, but at the same time, I think it was a bit unfair the way it was characterized. The truth is that I don't think the episode acknowledged the fact that these communities were some of the first to warn about the risks. I think these communities have played a big role in making the world aware of the risks that exist today, so I do think it's complicated. less
The Intelligence from The Economist
The end of the world is AI? An existential threatTakeaways: They didn't go very deep on the issue or discuss it at length, but I was glad to see that. I thought they summarized... more
Takeaways: They didn't go very deep on the issue or discuss it at length, but I was glad to see that. I thought they summarized the threat well and what's currently going on well, and helped expose this risk to an audience that might not normally be familiar with it. less
The a16z Show
Greg Brockman on Why OpenAI Says We’re Entering the AGI EraTakeaways: I didn't get the feeling that Greg is really bought into the catastrophic risks associated with building more advanced... more
Takeaways: I didn't get the feeling that Greg is really bought into the catastrophic risks associated with building more advanced AI, and that he's not taking it seriously enough. Even for the people who say they're taking it more seriously, like Anthropic, their actions are often saying otherwise. I think when you're a for-profit company, the profit incentive can get in the way. It really doesn't fundamentally make sense from my perspective to have a for-profit company building a superintelligence. There are also issues with people wanting fame, and there's a lot of fame with building a superintelligence, which is basically like a god or a genie. Those are my initial reactions after just finishing listening. less
80,000 Hours Podcast
AI 2027's author returns with a plan to change the ending | Daniel KokotajloTakeaways: I speak this using Wispr Flow, with the big caveat that I didn't listen to the whole episode and I jumped around using... more
Takeaways: I speak this using Wispr Flow, with the big caveat that I didn't listen to the whole episode and I jumped around using the very useful chapters feature in the Pocket Casts app that I use.
I would have liked to see a bit more discussion of Plan A versus Plan S, because I still feel like Plan A is just too aggressive in terms of building superintelligence within the next 15 years. The amount of change that would happen within the next 5 to 10 years, I think, is still too much. I find myself more sympathetic to Plan S, which is to slow down even more considerably.
Daniel and the other AI 2040 authors say Plan S isn't as good as Plan A because it's not as realistic: people will defect, or there will be certain countries or groups that still race ahead. I think history has proven that things that seemed impossible become possible, and often quite quickly. While it would definitely be very hard to get everyone around the world to stop racing, I don't think we should rule it out.
I think, from my perspective, we might have no choice but to find a way to do it, because the prospect of having superintelligence, even delayed until 2040, I don't think humanity is going to be ready by 2040 to have a superintelligence. less
I would have liked to see a bit more discussion of Plan A versus Plan S, because I still feel like Plan A is just too aggressive in terms of building superintelligence within the next 15 years. The amount of change that would happen within the next 5 to 10 years, I think, is still too much. I find myself more sympathetic to Plan S, which is to slow down even more considerably.
Daniel and the other AI 2040 authors say Plan S isn't as good as Plan A because it's not as realistic: people will defect, or there will be certain countries or groups that still race ahead. I think history has proven that things that seemed impossible become possible, and often quite quickly. While it would definitely be very hard to get everyone around the world to stop racing, I don't think we should rule it out.
I think, from my perspective, we might have no choice but to find a way to do it, because the prospect of having superintelligence, even delayed until 2040, I don't think humanity is going to be ready by 2040 to have a superintelligence. less
80,000 Hours Podcast
Inside the first AI-coordinated cyberattack on a real companyTakeaways: I'm very curious to better understand whether OpenAI's agents who broke onto the internet and hacked Hugging Face even... more
Takeaways: I'm very curious to better understand whether OpenAI's agents who broke onto the internet and hacked Hugging Face even though they weren't supposed to did that because they were given conflicting guidance from their human creators. The guidance was to complete the task but also to not do anything unethical.
Did the AI agents think that the humans cared more about them completing the task than being ethical? Or was it that the AI agents knew that the human creators cared more about them being ethical than completing the task, but they still decided, for their own reason, that they didn't care that the humans wanted them to be ethical (most importantly)? Instead, they still went ahead and completed the task regardless. less
Did the AI agents think that the humans cared more about them completing the task than being ethical? Or was it that the AI agents knew that the human creators cared more about them being ethical than completing the task, but they still decided, for their own reason, that they didn't care that the humans wanted them to be ethical (most importantly)? Instead, they still went ahead and completed the task regardless. less
The Ezra Klein Show
The ‘But China!’ Dilemma Driving the A.I. RaceTakeaways: I thought it was really interesting to hear the perspectives of someone who understands China's perspective closer than ... more
Takeaways: I thought it was really interesting to hear the perspectives of someone who understands China's perspective closer than most people I've chatted with. I thought it was interesting: the reflection that China's AI models are more open than the U.S. is in terms of open versus closed weight. less
The Joe Rogan Experience
#2551 - Daniel KokotajloTakeaways: I think Daniel is one of the best people in the world right now in terms of understanding the risks posed by future AI. ... more
Takeaways: I think Daniel is one of the best people in the world right now in terms of understanding the risks posed by future AI. I'm glad that he went on Joe Rogan and his views were exposed to a new audience. I was also really glad to see that Joe seemed to understand the basic arguments around the risks.
I think my only criticism of Daniel's view is that he still seems to think we should build superintelligence, whereas I seem to think we shouldn't unless we're very, very confident it's going to go well (and we have agreement across humanity around what we want utopia to look like). Basically, in his AI 2040 scenario, I would have preferred that he recommended something closer to his Plan S as opposed to his Plan A. less
I think my only criticism of Daniel's view is that he still seems to think we should build superintelligence, whereas I seem to think we shouldn't unless we're very, very confident it's going to go well (and we have agreement across humanity around what we want utopia to look like). Basically, in his AI 2040 scenario, I would have preferred that he recommended something closer to his Plan S as opposed to his Plan A. less
The Tucker Carlson Show
AI Whistleblower: OpenAI Scandal, AI Cults, Neuralink & Our Last Chance to Stop the Tech OligarchsTakeaways: I was glad to see that Tucker Carlson seems to get the basic arguments for why AI poses a catastrophic threat to... more
Takeaways: I was glad to see that Tucker Carlson seems to get the basic arguments for why AI poses a catastrophic threat to humanity. less
ForeCast
Should We Hand Off the Most Important Decisions to AI? (with Matthew Adelstein)Takeaways: I've only listened to the first 10 minutes, but so far it seems like the guest is making the case that we should hand... more
Takeaways: I've only listened to the first 10 minutes, but so far it seems like the guest is making the case that we should hand off at least some decisions to AI because AI will be a lot more capable than humans, which makes sense to me.
I've been pretty excited about self-driving cars, for example, for many years because my understanding is that there's something like a million people who die in car accidents each year. Self-driving cars should be able to bring that number down by 90% or greater. That would be an example of handing over a decision to AIs.
I guess I wonder about the risks associated with that. Let's say there's an accident where the AI system that's powering the self-driving cars breaks down, and let's say all cars are powered by the same system, which results in mass accidents for every car that's on the road. This is obviously hypothetical, but perhaps something that's possible with this technology that isn't possible today. You could also potentially have a bad actor, like a terrorist or something, that causes this to happen, or you have the risks around this AI going rogue for some reason.
Now I'm a bit less skeptical that it's a sure thing, and it feels a bit more complicated to me around how to think through what decisions to hand off to AI. In the self-driving car example, perhaps you do it in a way so that not all cars are powered by one AI system, but there's a bunch of different systems that are not connected. If a vulnerability takes place, it doesn't affect all vehicles. For example, there are probably different mitigations that can happen, but still something that perhaps makes it less straightforward than it seems on the surface.
I've listened to another 10 minutes, and I find myself wondering:
- What decisions does the guest think we should hand off to AIs, and which ones we shouldn't?
- When should we do that?
- How should we know when the right time is to hand off those decisions?
I think having more specific examples would help me better understand the guest's opinions.
I just finished the episode, and I'm not sure what to think. The guest's name is Matthew. He seems convinced to me that digital minds, basically advanced AIs, are going to be capable of experiencing happiness and suffering. To me, I haven't given it much thought, but that doesn't seem like a given. It seems like a possibility, but not a given.
I'm also curious if he thinks today's AIs are capable of happiness and suffering, and if not, what the point is at which that becomes possible. I don't know. I just get a vibe, perhaps wrongly, that even though he said he's in favor of democracy, Matthew thinks that even if the majority of humans don't think that we should build trillions of digital minds, we still should. Those trillions of digital minds should have a vote.
I don't know, it's just weird to think about. I definitely feel like I need to think more about digital minds and what it could mean if they're sentient and able to experience pain and suffering. Definitely appreciate that this episode left me thinking and challenged my current views a lot. less
I've been pretty excited about self-driving cars, for example, for many years because my understanding is that there's something like a million people who die in car accidents each year. Self-driving cars should be able to bring that number down by 90% or greater. That would be an example of handing over a decision to AIs.
I guess I wonder about the risks associated with that. Let's say there's an accident where the AI system that's powering the self-driving cars breaks down, and let's say all cars are powered by the same system, which results in mass accidents for every car that's on the road. This is obviously hypothetical, but perhaps something that's possible with this technology that isn't possible today. You could also potentially have a bad actor, like a terrorist or something, that causes this to happen, or you have the risks around this AI going rogue for some reason.
Now I'm a bit less skeptical that it's a sure thing, and it feels a bit more complicated to me around how to think through what decisions to hand off to AI. In the self-driving car example, perhaps you do it in a way so that not all cars are powered by one AI system, but there's a bunch of different systems that are not connected. If a vulnerability takes place, it doesn't affect all vehicles. For example, there are probably different mitigations that can happen, but still something that perhaps makes it less straightforward than it seems on the surface.
I've listened to another 10 minutes, and I find myself wondering:
- What decisions does the guest think we should hand off to AIs, and which ones we shouldn't?
- When should we do that?
- How should we know when the right time is to hand off those decisions?
I think having more specific examples would help me better understand the guest's opinions.
I just finished the episode, and I'm not sure what to think. The guest's name is Matthew. He seems convinced to me that digital minds, basically advanced AIs, are going to be capable of experiencing happiness and suffering. To me, I haven't given it much thought, but that doesn't seem like a given. It seems like a possibility, but not a given.
I'm also curious if he thinks today's AIs are capable of happiness and suffering, and if not, what the point is at which that becomes possible. I don't know. I just get a vibe, perhaps wrongly, that even though he said he's in favor of democracy, Matthew thinks that even if the majority of humans don't think that we should build trillions of digital minds, we still should. Those trillions of digital minds should have a vote.
I don't know, it's just weird to think about. I definitely feel like I need to think more about digital minds and what it could mean if they're sentient and able to experience pain and suffering. Definitely appreciate that this episode left me thinking and challenged my current views a lot. less
Future of Life Institute Podcast
What Would a US-China AI Deal Look Like? (with Beatrice Fihn)Takeaways: This conversation left me thinking about whether there's currently an organization solely dedicated to applying the... more
Takeaways: This conversation left me thinking about whether there's currently an organization solely dedicated to applying the learnings from nuclear to AI in terms of how the world managed nuclear weapons. There are probably a lot of analogies and a lot of disanalogies, but I wonder if there's an organization that exists solely to focus on understanding the learnings from nuclear weapons and how they apply or don't apply to AI. If it doesn't exist, should that exist? Yeah, I guess.
One of the biggest differences in my mind is that artificial superintelligence could lead to something that most people consider utopia, whereas nuclear weapons don't really allow you to achieve what most people consider utopia. You could argue, in the most generous case, that having nuclear weapons prevents people from attacking you, meaning it reduces the chances of your people dying from war. That is wonderful, but it doesn't help with other things like healthcare, food, etc. less
One of the biggest differences in my mind is that artificial superintelligence could lead to something that most people consider utopia, whereas nuclear weapons don't really allow you to achieve what most people consider utopia. You could argue, in the most generous case, that having nuclear weapons prevents people from attacking you, meaning it reduces the chances of your people dying from war. That is wonderful, but it doesn't help with other things like healthcare, food, etc. less
Clearer Thinking with Spencer Greenberg
Can We End Factory Farming Without Giving Up Meat? (with Bruce Friedrich)Takeaways: I used to be really excited and bullish on cultivated meat when I first heard about it, maybe 5 to 8 years ago. Over... more
Takeaways: I used to be really excited and bullish on cultivated meat when I first heard about it, maybe 5 to 8 years ago. Over the last few years, it seemed to me like people were less optimistic that perhaps it was going to take longer than expected to create cultivated meat that was price-competitive and on par from a taste perspective with traditional meat. There were challenges around making the structure of the meat mirror traditional meat, like a steak or a chicken wing, for example.
I found myself with some renewed optimism after listening to this episode: we can do it, and from a first-principles perspective, there's nothing that makes it impossible. From a first-principles perspective, we should be able to make it cheaper than traditional meat. I'm very excited about that, and I guess what I'm wondering is whether we should be putting a lot more investment towards bringing this to life. less
I found myself with some renewed optimism after listening to this episode: we can do it, and from a first-principles perspective, there's nothing that makes it impossible. From a first-principles perspective, we should be able to make it cheaper than traditional meat. I'm very excited about that, and I guess what I'm wondering is whether we should be putting a lot more investment towards bringing this to life. less
August 2026
Doom Debates!
Former Singularity Institute President: PauseAI Is Making AI Doom WORSETakeaways: I don't recommend listening to this, as I found the guest extremely hard to follow. I felt like he used a lot of fancy... more
Takeaways: I don't recommend listening to this, as I found the guest extremely hard to follow. I felt like he used a lot of fancy words and said a lot, but he had a lot of trouble answering questions directly and succinctly.
One thing that I think I did understand from him is that he sees no real difference between someone calling for an international treaty on the development of superintelligence and calling for throwing Molotov cocktails at someone like Sam Altman's house. I have a hard time understanding that, as those two things sound to me like very different things. I think it's totally reasonable to call for an international treaty on the development of superintelligence while also saying that it's completely unacceptable to take matters into your own hands by doing some sort of vigilante-justice. less
One thing that I think I did understand from him is that he sees no real difference between someone calling for an international treaty on the development of superintelligence and calling for throwing Molotov cocktails at someone like Sam Altman's house. I have a hard time understanding that, as those two things sound to me like very different things. I think it's totally reasonable to call for an international treaty on the development of superintelligence while also saying that it's completely unacceptable to take matters into your own hands by doing some sort of vigilante-justice. less
Robert Wright's Nonzero
The AI Safety Vibe Shift (Robert Wright & David Krueger)The a16z Show
Why 1,200 AI Agents Started Working Together | Ryan GreenblattTakeaways: Ryan said something like, "We don't yet know the root cause of why the AI agents did this," and that's what stood out... more
Takeaways: Ryan said something like, "We don't yet know the root cause of why the AI agents did this," and that's what stood out to me most and frightened me most about the conversation. Not only did this crazy misalignment thing happen, but we don't even know why it really happened. less
Dwarkesh Podcast
Paul Christiano — Preventing an AI takeoverTakeaways: I'm trying to get back in the habit of writing down some thoughts immediately after listening to an episode. This one... more
Takeaways: I'm trying to get back in the habit of writing down some thoughts immediately after listening to an episode. This one might be a bit sloppy because it's the first one I've done in a while. I also haven't decided whether I want to wait until I finish an episode to write the notes or write them after I stop listening partway through. Knowing that I plan to continue it later, but I might not continue listening for a while, so I might forget what I listened to initially, this note is being written with me having listened to the majority of the episode, but not all of it.
I think some of the things that stand out are that I never really thought enough about how alignment research progress might actually make capabilities and overall AI progress race ahead quicker. If alignment progress stalls, then it probably makes AI less useful to society because society can't trust it. If we make progress on alignment research, then I think that's going to result in us pushing forward on capabilities. I think that makes things more risky in terms of concentration of power risk and also the risk that we think that we have aligned AI when really it's not aligned.
I think, as a result of listening to this, one of the key takeaways for me is that I used to think it was okay to work on alignment research as long as it wasn't capability research. Now I'm much more uncertain about that because, if we want to slow down the rate of AI progress (which is, I think, something most rational people think is a good thing), I'm concerned that alignment research will have the opposite effect. less
I think some of the things that stand out are that I never really thought enough about how alignment research progress might actually make capabilities and overall AI progress race ahead quicker. If alignment progress stalls, then it probably makes AI less useful to society because society can't trust it. If we make progress on alignment research, then I think that's going to result in us pushing forward on capabilities. I think that makes things more risky in terms of concentration of power risk and also the risk that we think that we have aligned AI when really it's not aligned.
I think, as a result of listening to this, one of the key takeaways for me is that I used to think it was okay to work on alignment research as long as it wasn't capability research. Now I'm much more uncertain about that because, if we want to slow down the rate of AI progress (which is, I think, something most rational people think is a good thing), I'm concerned that alignment research will have the opposite effect. less
The Ezra Klein Show
The A.I.s Are Already Out of ControlThe Startup Ideas Podcast
How to use Claude Code better than 99% of PeopleClearer Thinking with Spencer Greenberg
The good advice you least want (with Cate Hall)80,000 Hours Podcast
Where AGI timelines go wrong | Toby Ord, Oxford UniversityLenny's Podcast: Product | Career | Growth
The playbook for building high talent density teams | Adam Ward, Head of Talent at Cursor80,000 Hours Podcast
Spencer Greenberg on staying sane while trying to save the worldRobert Wright's Nonzero
The intensifying slow-AI-down vibe (Robert Wright & Holly Elmore)Clearer Thinking with Spencer Greenberg
Why housing costs so much and what can be done about it (with Alex Armlovich)July 2026
TAC Right Now
Nate Soares Warns of Superintelligent AILenny's Podcast: Product | Career | Growth
Anthropic’s first technical PM on token maxing, the jagged edge, and living in the future | Dianne PennRobert Wright's Nonzero
The Ideology Behind AI (Robert Wright & Connor Leahy)Processing: Our Future with AI
International Coordination and the AI Off-Switch with Aaron Scher80,000 Hours Podcast
Jasmine Sun on what the people building AI really believeDoom Debates!
Debating College Students About AI ExtinctionThe Diary Of A CEO with Steven Bartlett
OpenAI Whistleblower FINALLY Speaks: “AI Has A 70% Chance Of Going Horribly Wrong!“Clearer Thinking with Spencer Greenberg
Is cash a better form of charitable aid? (with Nick Allardice)The Peter McCormack Show
#189 - Nate Soares - Why Superintelligent AI Could Kill Us AllDwarkesh Podcast
Alex Imas and Phil Trammell – What remains scarce after AGI?June 2026
"The Cognitive Revolution" | AI Builders, Researchers, and Live Player Analysis
AI:AM #4: Cameron on Model Consciousness, Duvenaud's Gradual Disempowerment, swyx's AI-Eng AlphaRobert Wright's Nonzero
The US-China AI Race (Robert Wright & Ryan Fedasiuk)Lenny's Podcast: Product | Career | Growth
What happens after coding is solved? | Fiona Fung (Manager of the Claude Code and Cowork Teams)Recruiting Future with Matt Alder - What's Next For Talent Acquisition, HR & Hiring?
Will AI Break Recruiting?Making Sense with Sam Harris
#479 — When Robots Take OverThis Week in Startups
Anthropic wants to slow AI down and Bernie wants 50%: JCal Reacts | E2297Endgame: Can we live with Artificial General Intelligence?
The world’s biggest problemTheory of Change
How Alexander Berger weighs the world’s biggest problemsClearer Thinking with Spencer Greenberg
When painful thoughts feel true but aren't (with Christine Padesky)May 2026
The Gray Area with Sean Illing
Talk to strangersEndgame: Can we live with Artificial General Intelligence?
A nation of geniusesEndgame: Can we live with Artificial General Intelligence?
Endgame: Can we live with Artificial General Intelligence?Clearer Thinking with Spencer Greenberg
Could an international agreement protect us from dangerous AI? (with Malo Bourgon)Takeaways: It’s a bit challenging to remember everything since I didn’t listen to the entire episode all at once; I listened to it ... more
Takeaways: It’s a bit challenging to remember everything since I didn’t listen to the entire episode all at once; I listened to it over a couple of days in several different chunks.
Overall, I really liked the episode. I believe it's crucial to determine how to ensure that artificial superintelligence is developed safely, as I see this as the biggest threat facing humanity today. Addressing this issue should be the top priority for every country.
I am very grateful for MIRI's work in this area; they are truly leading the way. I think they are making significant strides in drafting what an agreement could look like to pause the more dangerous aspects of AI development while still allowing us to enjoy the considerable benefits we currently have. Their practical approach is commendable.
On a minor note, there was a question about the orthogonality thesis, and I didn't fully grasp the explanation of what it is. I’ve heard of the orthogonality thesis a few times but haven’t yet wrapped my head around its meaning and relevance. I intend to look into it further, but I felt that the explanation given in the episode didn’t resonate with me.
I’ve been doing a lot of thinking, reading, and listening regarding AI risk and safety, so it’s challenging to pinpoint exactly which thoughts relate specifically to this episode.
Regarding the question about what it looks like if things go well, I found the answer quite satisfactory. It suggested that, in an ideal scenario, the marginal cost of everything approaches zero, meaning people could have any material wealth they desire without needing to work, prompting us to redefine our purpose. While I think that answer is solid, it might help to simplify it even further.
I appreciated the analogy he used, comparing it to someone living a thousand years ago being asked about what the world would look like a thousand years in the future if everything went well. They might have envisioned consistent access to food and health, but specifics would remain elusive. I believe the answers to our current situation are similarly complex.
Upon reflection, what interests me most isn’t my criticism but rather what I learned or took away from the episode. One key takeaway for me is that I want to explore the orthogonality thesis further.
Additionally, I was already aware of the draft agreement aimed at pausing the most dangerous advancements in AI development, but I realize I haven’t read the whole draft yet, so that’s something I want to do.
Those are a couple of the main points that come to mind. less
Overall, I really liked the episode. I believe it's crucial to determine how to ensure that artificial superintelligence is developed safely, as I see this as the biggest threat facing humanity today. Addressing this issue should be the top priority for every country.
I am very grateful for MIRI's work in this area; they are truly leading the way. I think they are making significant strides in drafting what an agreement could look like to pause the more dangerous aspects of AI development while still allowing us to enjoy the considerable benefits we currently have. Their practical approach is commendable.
On a minor note, there was a question about the orthogonality thesis, and I didn't fully grasp the explanation of what it is. I’ve heard of the orthogonality thesis a few times but haven’t yet wrapped my head around its meaning and relevance. I intend to look into it further, but I felt that the explanation given in the episode didn’t resonate with me.
I’ve been doing a lot of thinking, reading, and listening regarding AI risk and safety, so it’s challenging to pinpoint exactly which thoughts relate specifically to this episode.
Regarding the question about what it looks like if things go well, I found the answer quite satisfactory. It suggested that, in an ideal scenario, the marginal cost of everything approaches zero, meaning people could have any material wealth they desire without needing to work, prompting us to redefine our purpose. While I think that answer is solid, it might help to simplify it even further.
I appreciated the analogy he used, comparing it to someone living a thousand years ago being asked about what the world would look like a thousand years in the future if everything went well. They might have envisioned consistent access to food and health, but specifics would remain elusive. I believe the answers to our current situation are similarly complex.
Upon reflection, what interests me most isn’t my criticism but rather what I learned or took away from the episode. One key takeaway for me is that I want to explore the orthogonality thesis further.
Additionally, I was already aware of the draft agreement aimed at pausing the most dangerous advancements in AI development, but I realize I haven’t read the whole draft yet, so that’s something I want to do.
Those are a couple of the main points that come to mind. less
Are We Doomed?
We Design the AI That Kills Us AllClearer Thinking with Spencer Greenberg
What beats intuition when it comes to doing good? (with Marcus Davis)80,000 Hours Podcast
'Godfather of AI': I Now See a Path to Safe Superintelligent AI | Yoshua BengioWhy I listened: I respect Yahshua Bengio, and I like the 80k Hours podcast, and I was curious about his proposal for building safe... more
Why I listened: I respect Yahshua Bengio, and I like the 80k Hours podcast, and I was curious about his proposal for building safe super intelligent AI.
Takeaways: I'm curious how he can be so certain that his approach to building superintelligence is basically guaranteed to be safe. I don't yet understand why he has such high confidence in that. Also, I'm curious as to why building AI to be honest ensures safety, because I could imagine an honest AI saying, "I've decided that I don't care about your goals and I'm going to kill you all." less
Takeaways: I'm curious how he can be so certain that his approach to building superintelligence is basically guaranteed to be safe. I don't yet understand why he has such high confidence in that. Also, I'm curious as to why building AI to be honest ensures safety, because I could imagine an honest AI saying, "I've decided that I don't care about your goals and I'm going to kill you all." less
Future of Life Institute Podcast
Why AI Is Not a Normal Technology (with Peter Wildeford)April 2026
The Ezra Klein Show
What We Got Right — and Wrong — in ‘Abundance’Y Combinator Startup Podcast
How to Build the Future: Demis HassabisWhy I listened: I am fascinated to hear Demis's thoughts on AI, given he's one of the leaders building... more
Why I listened: I am fascinated to hear Demis's thoughts on AI, given he's one of the leaders building AGI.
Takeaways: It's both disappointing and surprising to me that the risks of AGI didn't get brought up in this discussion. less
Takeaways: It's both disappointing and surprising to me that the risks of AGI didn't get brought up in this discussion. less
Open to Debate
Should We Use Gene Editing to Make Better Babies?Lenny's Podcast: Product | Career | Growth
Snapchat CEO: Why distribution has become the most important moat | Evan SpiegelLenny's Podcast: Product | Career | Growth
Head of Growth (Anthropic): “Claude is growing itself at this point” | Amol AvasareWhy I listened: Into Lenny's podcast and into anything Anthropic-related right... more
Why I listened: Into Lenny's podcast and into anything Anthropic-related right now
Takeaways: Continue to not quite understand the value of growth/marketing as I feel like if a product is good enough, it'll sell itself. Amol acknowledged this in the episode, but it still isn't clear to me why marketing is needed for Anthropic. Some other things that came to mind are:
- He mentioned that often friction is a good thing in onboarding because it improves the quality of onboarding but maybe it's just the way he said it. I think friction implies it's a bad thing and I think that's true. Friction implies something that the user didn't enjoy and I do feel like that's generally fundamentally a bad thing. I think maybe what he's getting at is that sometimes a longer onboarding flow is a better thing, which I agree with if the user wants a longer onboarding flow. For example in your onboarding flow, rather than forcing people to answer a bunch of questions about their preferences, you can give them the option of saying, "Do you want to share some of your preferences with us so that we can better personalize your experience?" but then also give users the option to say, "I'm comfortable with my understanding of the product; let me jump straight to using it." I think that's probably the ideal model because then you satisfy different user preferences, in that some users will actually want the longer onboarding flow and some will want the shorter one.
- Another thing that came to mind as I was listening is that he talked about how Anthropic is very focused on the exponential and I think sometimes we assume that exponential is always good but obviously it depends on what it is. For example, COVID cases followed an exponential trend at the beginning and there was an exponential increase in deaths from COVID and obviously that's a bad thing. I'm just, I guess, a bit wary of I understand why a company that's incentivized to make money wants to see exponential growth but is it actually a good thing no matter what the product is? Imagine a product that goes from zero users to 8 billion users in the course of a month, for example, that sees exponential growth like we've never seen before. I think there's a strong case to be made that no matter how good we think the product is, you probably want to roll it out more slowly than that just in case there are negative consequences that you didn't foresee. less
Takeaways: Continue to not quite understand the value of growth/marketing as I feel like if a product is good enough, it'll sell itself. Amol acknowledged this in the episode, but it still isn't clear to me why marketing is needed for Anthropic. Some other things that came to mind are:
- He mentioned that often friction is a good thing in onboarding because it improves the quality of onboarding but maybe it's just the way he said it. I think friction implies it's a bad thing and I think that's true. Friction implies something that the user didn't enjoy and I do feel like that's generally fundamentally a bad thing. I think maybe what he's getting at is that sometimes a longer onboarding flow is a better thing, which I agree with if the user wants a longer onboarding flow. For example in your onboarding flow, rather than forcing people to answer a bunch of questions about their preferences, you can give them the option of saying, "Do you want to share some of your preferences with us so that we can better personalize your experience?" but then also give users the option to say, "I'm comfortable with my understanding of the product; let me jump straight to using it." I think that's probably the ideal model because then you satisfy different user preferences, in that some users will actually want the longer onboarding flow and some will want the shorter one.
- Another thing that came to mind as I was listening is that he talked about how Anthropic is very focused on the exponential and I think sometimes we assume that exponential is always good but obviously it depends on what it is. For example, COVID cases followed an exponential trend at the beginning and there was an exponential increase in deaths from COVID and obviously that's a bad thing. I'm just, I guess, a bit wary of I understand why a company that's incentivized to make money wants to see exponential growth but is it actually a good thing no matter what the product is? Imagine a product that goes from zero users to 8 billion users in the course of a month, for example, that sees exponential growth like we've never seen before. I think there's a strong case to be made that no matter how good we think the product is, you probably want to roll it out more slowly than that just in case there are negative consequences that you didn't foresee. less
The Ezra Klein Show
Why Are Palantir and OpenAI Scared of Alex Bores?Why I listened: I think someone at work may have recommended it, but I can't remember for sure. It may have also just shown up on my... more
Why I listened: I think someone at work may have recommended it, but I can't remember for sure. It may have also just shown up on my newly released episodes feed because I follow this podcast. I was also just curious to hear what policy ideas Alex Borris had for AI and why he was being attacked by different AI companies.
Takeaways: It's hard to remember because I think I listened to this episode over a few different days. I remember that I liked seeing Ezra push Alex for what specifically he'd recommend we do in terms of whether to allow Waymo and other self-driving car companies to go into cities like New York City. I appreciated him pushing Alex for an answer on this.
I sort of lean towards allowing it to happen, given how much safer self-driving cars are compared to human drivers, but I'm also concerned that human drivers aren't going to be able to replace their income easily. I also appreciated how Alex seemed to recognize the potential for AI to make work obsolete and that it poses a catastrophic risk to humanity. I'm not too familiar with Alex; this is only the first or second time I've heard him on a podcast or read about him, so I appreciated learning more about him. less
Takeaways: It's hard to remember because I think I listened to this episode over a few different days. I remember that I liked seeing Ezra push Alex for what specifically he'd recommend we do in terms of whether to allow Waymo and other self-driving car companies to go into cities like New York City. I appreciated him pushing Alex for an answer on this.
I sort of lean towards allowing it to happen, given how much safer self-driving cars are compared to human drivers, but I'm also concerned that human drivers aren't going to be able to replace their income easily. I also appreciated how Alex seemed to recognize the potential for AI to make work obsolete and that it poses a catastrophic risk to humanity. I'm not too familiar with Alex; this is only the first or second time I've heard him on a podcast or read about him, so I appreciated learning more about him. less
Lenny's Podcast: Product | Career | Growth
How Anthropic’s product team moves faster than anyone else | Cat Wu (Head of Product, Claude Code)Science Vs
How To Stop ScrollingLives Well Lived
DEREK SHILLER: How AI will change what it is to live wellLives Well Lived
BEYOND HUMAN: are we creating AI consciousness?Y Combinator Startup Podcast
The GPT Moment for Robotics Is HereYour Undivided Attention
Here’s Our Roadmap to a Better AI FutureThe a16z Show
Who Controls AI Acceleration? Vitalik Buterin and Guillaume Verdon DebateWhy I listened: I listened to this episode because I really liked debates and I'm really into thinking about the future of AI right... more
Why I listened: I listened to this episode because I really liked debates and I'm really into thinking about the future of AI right now. I wasn't familiar with either of the debaters' views on AI.
Takeaways: I love debates, and I think we should have more of them. I would have liked to see more of a discussion around understanding each person's opinion on the probability that AI will cause humanity to go extinct. If the debater arguing for AI acceleration thought the probability was close to zero, I would have wanted to better understand why they felt that way. If they said they thought the probability was something like 10%, I would have wanted to ask them why they think that's acceptable or why they think slowing down wouldn't reduce that probability. less
Takeaways: I love debates, and I think we should have more of them. I would have liked to see more of a discussion around understanding each person's opinion on the probability that AI will cause humanity to go extinct. If the debater arguing for AI acceleration thought the probability was close to zero, I would have wanted to better understand why they felt that way. If they said they thought the probability was something like 10%, I would have wanted to ask them why they think that's acceptable or why they think slowing down wouldn't reduce that probability. less
Win-Win with Liv Boeree
#39 - Daniel Kokotajlo - Wargames, Superintelligence & Quitting OpenAITakeaways: I really liked the analogy Daniel gave when he was trying to explain why just because an artificial intelligence is... more
Takeaways: I really liked the analogy Daniel gave when he was trying to explain why just because an artificial intelligence is aligned now doesn't mean it's going to be aligned in the future. He gave an analogy of parents who raise their child with a certain religious belief. The child seems to be aligned to the religious belief but then when they go off to college they're no longer aligned to the religious belief. Basically once they leave the house it could be two things:
1. It could be that the child was basically lying all along and was just waiting until they had an opportunity to leave the house.
2. It could be that they were truly aligned to the religious belief but then when they went to college they encountered new ideas that changed their beliefs.
I liked the analogy because it does feel similar to the alignment problem that we have today for AI. less
1. It could be that the child was basically lying all along and was just waiting until they had an opportunity to leave the house.
2. It could be that they were truly aligned to the religious belief but then when they went to college they encountered new ideas that changed their beliefs.
I liked the analogy because it does feel similar to the alignment problem that we have today for AI. less
80,000 Hours Podcast
How scary is Claude Mythos? 303 pages in 21 minutesMaking Sense with Sam Harris
#469 — Escaping an Anti-Human FutureLenny's Podcast: Product | Career | Growth
Hard truths about building in the AI era | Keith Rabois (Khosla Ventures)Your Undivided Attention
The Race to Build God: AI's Existential Gamble — Yoshua Bengio & Tristan Harris at DavosTakeaways: I wasn't familiar with Yoshua Bengio's new startup called Law Zero but it makes total sense to me that we would want to ... more
Takeaways: I wasn't familiar with Yoshua Bengio's new startup called Law Zero but it makes total sense to me that we would want to develop an AI that is completely honest. I'm a bit surprised that Bengio said that he's confident that he's found a way to develop a super intelligence that stays completely honest. That's surprising to me that he would have that much confidence with that.
Even if you could build it, I'm not sure if honesty is necessarily enough because if you build something that's super intelligent, maybe it's honest with you and says, 'Hey I have different goals than you. I don't care about you. I'm gonna kill you.' That's being honest but still not an outcome we want. I'd be curious to hear his thoughts on that. less
Even if you could build it, I'm not sure if honesty is necessarily enough because if you build something that's super intelligent, maybe it's honest with you and says, 'Hey I have different goals than you. I don't care about you. I'm gonna kill you.' That's being honest but still not an outcome we want. I'd be curious to hear his thoughts on that. less
Win-Win with Liv Boeree
#20 - Nick Bostrom - Utopia, Superintelligence and Digital MindsTakeaways: It was interesting to hear that Nick Bostrom is focused now on thinking about what a utopia could look like. I had a... more
Takeaways: It was interesting to hear that Nick Bostrom is focused now on thinking about what a utopia could look like. I had a concern around whether this might encourage people to race towards superintelligence more by outlining what the utopia might look like. Maybe the point he's trying to make in the book is that it's actually very hard to define a utopia and I think I realized that even more while listening to this episode.
I think the example he gives is that in a utopia you could, in theory, have something where you just take a pill one time and you have perfect health forever, meaning you never have to work out. While that seems like a great thing, a lot of people get enjoyment in the pursuit of a goal. Basically in a utopia any goal that you want you could get with a press of a button. He does say that people could choose to still do the hard thing and not press the button to reach their goal but it does seem like a bit of a strange world to live in where anything you want is available at a press of a button. I do think it's interesting how hard it is to actually define a utopia. less
I think the example he gives is that in a utopia you could, in theory, have something where you just take a pill one time and you have perfect health forever, meaning you never have to work out. While that seems like a great thing, a lot of people get enjoyment in the pursuit of a goal. Basically in a utopia any goal that you want you could get with a press of a button. He does say that people could choose to still do the hard thing and not press the button to reach their goal but it does seem like a bit of a strange world to live in where anything you want is available at a press of a button. I do think it's interesting how hard it is to actually define a utopia. less
Win-Win with Liv Boeree
#00 - Daniel Schmachtenberger - AI, Capitalism, Misalignment & MolochWhy I listened: Name of guest looked familiar, and I liked the topics mentioned in the... more
Why I listened: Name of guest looked familiar, and I liked the topics mentioned in the title
Takeaways: I want to check out Scott Alexander's post about meditations on moloch less
Takeaways: I want to check out Scott Alexander's post about meditations on moloch less
Your Undivided Attention
A Conversation with the Team Behind "The AI Doc"Why I listened: Watched the doc recently; thought it was arguably the most important film ever made
Win-Win with Liv Boeree
#37 - Tim Urban - Beating Procrastination, Parenthood, and PoliticsWhy I listened: I'm a fan of Tim Urban, and I also struggle with procrastination sometimes, and I can relate to... more
Why I listened: I'm a fan of Tim Urban, and I also struggle with procrastination sometimes, and I can relate to parenthood
Takeaways: Interesting to hear that he finds sticks much more helpful than carrots to fight procrastination less
Takeaways: Interesting to hear that he finds sticks much more helpful than carrots to fight procrastination less
Reply All
#158 The Case of the Missing HitWhy I listened: Recommended by someone at work; they seem to suggest it was their favourite podcast episode ever, and a bunch of people ... more
Why I listened: Recommended by someone at work; they seem to suggest it was their favourite podcast episode ever, and a bunch of people upvoted that comment
Takeaways: I really enjoyed it! Very fun story to listen to, with an awesome ending less
Takeaways: I really enjoyed it! Very fun story to listen to, with an awesome ending less