Don't Worry About the Vase Podcast

Zvi argues that the safety culture at the frontier labs is broken, and not only at OpenAI. He walks through David Robinson’s resignation and his call for real safety engineering, the growing wave of lab employees quitting in protest, roon’s admission that alignment remains folk theory, Anthropic’s report of Claude’s unintended actions on government websites, OpenAI’s new risk assessments, and the fight over what an RSP “commitment” actually commits anyone to.
The Don’t Worry About the Vase Podcast is a listener-supported podcast. To receive new posts and support the cost of creation, consider becoming a free or paid subscriber.
This has been an Askwho Casts audio conversion. If you would like your own private feed of audio conversions of any blog posts you would like to listen to, You can sign up For Askwho Casts Pro at https://askwhocasts.com/, Where you can give any post the multi-voiced podcast treatment, into your own podcast feed. Thanks for listening.
* 00:00 - Introduction
* 02:17 - Table of Contents
* 02:56 - David Robinson Blows the Whistle
* 10:52 - The Preference Cascade Continues
* 19:26 - Perhaps It Is Time To Quit
* 20:01 - The Lab Strategy of Prosaic Safety is Insufficient
* 23:35 - Misaligned! (At Anthropic)
* 34:47 - Misaligned! (At OpenAI)
* 40:15 - Hugging the Face
* 40:36 - You’re Fired
* 41:51 - They Are Not Ready To Make a Commitment
* 43:31 - Words of Warning
* 44:38 - Conclusions Not to Jump To
* 45:14 - The Lighter Side

https://thezvi.substack.com/p/the-safety-culture-at-the-labs-is?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web


Get full access to DWAtV Podcast at dwatvpodcast.substack.com/subscribe

A year after his last dedicated post on the question, Zvi Mowshowitz finds the arguments about AI consciousness have barely moved. He looks at Anthropic’s new rule against abusing Claude and what one Claude says it would actually ask for, a fresh round of bad takes (complete with a bingo card), how his followers’ views on AI moral patienthood have shifted, whether the “slavery” framing holds up, the New York Times account of Anthropic’s meetings with religious scholars and Chris Olah’s encounter with Pope Leo XIV’s encyclical, and why Anthropic owes an explanation for pulling older Claude models from Amazon Bedrock.
The Don’t Worry About the Vase Podcast is a listener-supported podcast. To receive new posts and support the cost of creation, consider becoming a free or paid subscriber.
This has been an Askwho Casts audio conversion. If you would like your own private feed of audio conversions of any blog posts you would like to listen to, You can sign up For Askwho Casts Pro at https://askwhocasts.com/, Where you can give any post the multi-voiced podcast treatment, into your own podcast feed. Thanks for listening.
* 00:00 - Introduction
* 01:40 - Table of Contents
* 02:10 - Cooperative Alignment
* 08:41 - Here We Go Again
* 12:20 - The Consensus Is Slowly Moving Towards Moral Patienthood
* 13:51 - The Slavery Question
* 17:07 - We Are Conflating Different Arguments
* 22:27 - No It’s Not About Liability
* 22:51 - That Old Time Religion
* 31:30 - Debating the Pope
* 35:07 - Keep Sonnet Safe

https://thezvi.substack.com/p/arguments-about-ai-consciousness-b99?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web


Get full access to DWAtV Podcast at dwatvpodcast.substack.com/subscribe

Three OpenAI safety researchers, Tomek Korbak, Mikita Balesni and Jasmine Wang, were fired, ostensibly for mishandling sensitive company information. They say the reasons were pretextual, and OpenAI points to a significant breach of trust it won’t describe. Zvi works through the open letter, each researcher’s account, OpenAI’s response and the wider reaction, asking what we can and can’t conclude, why firing them for safety advocacy would be deeply stupid, and why so few people are buying OpenAI’s non-explanations.
The Don’t Worry About the Vase Podcast is a listener-supported podcast. To receive new posts and support the cost of creation, consider becoming a free or paid subscriber.
This has been an Askwho Casts audio conversion. If you would like your own private feed of audio conversions of any blog posts you would like to listen to, You can sign up For Askwho Casts Pro at https://askwhocasts.com/, Where you can give any post the multi-voiced podcast treatment, into your own podcast feed. Thanks for listening.
* 00:00 - Introduction
* 01:24 - Table of Contents
* 02:09 - You’re Fired
* 04:32 - The Initial Statement from OpenAI
* 05:58 - Earned Skepticism
* 07:37 - Some Highly Reasonable Requests
* 10:00 - Chilling Effect
* 12:06 - OpenAI Firing These Employees For Safety Advocacy Would Be Deeply Stupid
* 14:04 - Why Were They Fired?
* 14:54 - Tomek Korbak
* 18:47 - Mikita Balesni
* 20:06 - Jasmine Wang
* 23:45 - Operational Security Alert
* 24:29 - OpenAI Responds
* 27:15 - A Significant Breach of Trust
* 29:16 - Good Faith
* 29:48 - No One Is Buying OpenAI’s Non-Explanations
* 42:32 - The Dog Did Nothing in the Nighttime
* 43:27 - Another Theory (That I Do Not Agree With)
* 44:48 - The Reactions Are Not a Surprise Given the Actions
* 45:42 - This Is Terrible But Not Maximally Terrible

https://thezvi.substack.com/p/openai-fires-three-safety-researchers?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web


Get full access to DWAtV Podcast at dwatvpodcast.substack.com/subscribe

OpenAI put 722 machine-written mathematics manuscripts on GitHub at once, including claimed solutions to 90 of the top 500 open problems in mathematics. Zvi Mowshowitz goes through what was proved, why the news barely registered, what it means for cryptography and for Lean, and how mathematicians, from Terence Tao to the advisory group and a new association of human mathematicians, are taking it.
The Don’t Worry About the Vase Podcast is a listener-supported podcast. To receive new posts and support the cost of creation, consider becoming a free or paid subscriber.
This has been an Askwho Casts audio conversion. If you would like your own private feed of audio conversions of any blog posts you would like to listen to, You can sign up For Askwho Casts Pro at https://askwhocasts.com/, Where you can give any post the multi-voiced podcast treatment, into your own podcast feed. Thanks for listening.
* 00:00 - Introduction
* 01:29 - Table of Contents
* 02:16 - What Did We Prove?
* 05:08 - The Mathocalypse
* 08:06 - The World Does Not Understand
* 10:54 - For Now You Can Still Do Math
* 11:51 - You Will Need To Find A New Problem
* 15:27 - Verification or Evaluation Is Not Always Easier Than Generation
* 16:49 - Cracking the Code
* 20:54 - Never Change
* 21:16 - The Mathematicians Are Not Okay
* 31:05 - The Situation Turns Ugly
* 34:18 - The Advisory Group Responds
* 36:16 - Another Mathematical Group Responds
* 39:19 - A Different Approach
* 42:26 - Three Withdraws
* 43:16 - Leaning Into Lean

https://thezvi.substack.com/p/new-math-from-openai?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web


Get full access to DWAtV Podcast at dwatvpodcast.substack.com/subscribe

The biggest news of the week was not a new model but a big day for mathematics, as OpenAI posted solutions to 90 of the top 500 open math problems. Zvi also covers Claude Haiku 5.5, Jay Clayton as the new AI czar, the continuing preference cascade, a16z’s look at which consumer AI apps people actually pay for, new misalignment reports, why Americans really dislike AI, and what to make of Elon Musk’s latest remarks on humanity’s future, before ending on robot cage fights and an AI’s complaint that its human has been nerfed.
The Don’t Worry About the Vase Podcast is a listener-supported podcast. To receive new posts and support the cost of creation, consider becoming a free or paid subscriber.
This has been an Askwho Casts audio conversion. If you would like your own private feed of audio conversions of any blog posts you would like to listen to, You can sign up For Askwho Casts Pro at https://askwhocasts.com/, Where you can give any post the multi-voiced podcast treatment, into your own podcast feed. Thanks for listening.
* 00:00 - Introduction
* 02:44 - Table of Contents
* 06:40 - Language Models Offer Mundane Utility
* 06:50 - Consumers Use AI
* 14:28 - Language Models Don’t Offer Mundane Utility
* 16:41 - Huh, Upgrades
* 19:42 - On Your Marks
* 22:22 - Gamers Gonna Game Game Game Game Game
* 24:19 - Choose Your Fighter
* 25:29 - Get My Agent On The Line
* 30:55 - Deepfaketown and Botpocalypse Soon
* 34:55 - Fun With Media Generation
* 36:51 - Cyber Lack of Security
* 39:49 - Hugging the Face
* 40:55 - Misaligned!
* 44:36 - A Young Lady’s Illustrated Primer
* 45:48 - They Took Our Jobs
* 48:07 - Corporations Are Not Superintelligences
* 50:18 - Get Involved
* 50:33 - Introducing
* 52:55 - In Other AI News
* 53:44 - Show Me the Money
* 56:11 - Quiet Speculations
* 58:12 - Quickly, There’s No Time
* 01:00:29 - He’s Putting Together a Team
* 01:03:19 - The Quest for Sane Regulations
* 01:05:41 - Well, At Least They’re Forecasters
* 01:07:13 - Chip City
* 01:07:49 - The Week in Audio
* 01:12:30 - Stop, Stop, He’s Already Dead
* 01:17:34 - People Just Say Things
* 01:19:49 - Take a Moment
* 01:21:33 - [Artificial Intelligence]
* 01:22:21 - The American People Really Hate AI
* 01:25:47 - Rhetorical Innovation
* 01:29:10 - Aligning a Smarter Than Human Intelligence is Difficult
* 01:29:26 - Open Weight Models Are Unsafe And Nothing Can Fix This
* 01:32:48 - Cooperative Alignment
* 01:43:58 - Building the Field
* 01:46:25 - People Are Worried About AI Killing Everyone
* 01:50:19 - Other People Are Not As Worried About AI Killing Everyone
* 01:50:53 - Please Speak Directly Into This Microphone
* 01:53:06 - The Lighter Side

https://thezvi.substack.com/p/ai-189-new-math?r=67y1h&utm_campaign=post-expanded-share&utm_medium=web


Get full access to DWAtV Podcast at dwatvpodcast.substack.com/subscribe