Don't Worry About the Vase Podcast

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.
* 00:00 - Introduction
* 00:46 - Huh, Upgrades
* 02:52 - On Your Marks
* 05:02 - Lazy Cheaters
* 06:45 - It’s All Routine
* 07:20 - Declawing
* 09:21 - Free Claw
* 09:47 - Take It To The Limit
* 14:06 - Turn On Auto The Pilot
* 16:14 - I’ll Allow It
* 16:43 - Threat Model
* 17:21 - The Classifier Is The Hard Part
* 18:53 - Acceptable Risks
* 20:14 - Manage The Agents
* 22:49 - Introducing
* 22:57 - Skilling Up
* 25:49 - What Happened To My Tokens?
* 26:04 - Coding Agents Offer Mundane Utility
https://open.substack.com/pub/thezvi/p/claude-code-codex-and-agentic-coding?utm_campaign=post-expanded-share&utm_medium=web


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

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.
* 00:00 - Introduction
* 02:05 - Epoch Capabilities Index (ECI) (Model Card 2.3.6)
* 04:52 - What Do You Mean Verbalized Evaluation Awareness Is Going Down
* 05:43 - Capabilities (Model Card Section 6)
* 08:38 - Agentic Safety Benchmarks (8.3)
* 10:57 - Is Mythos AGI?
* 11:56 - Are AI Companies Using Warnings As Hype?
* 12:56 - Impressions (Model Card Section 7)
* 16:13 - Blatant Denials Are The Best Kind
* 17:13 - Prompt Injection Robustness
* 18:21 - Does Mythos Cross The New Knowledge Threshold?
* 19:10 - Is Mythos Surprising or Discontinuous?
* 23:18 - UK AISI Tests Claude Mythos On Cybersecurity
* 24:42 - Everything Reinforces My Existing Predictions And Policy Preferences
* 29:55 - Solve For The Equilibrium
* 31:20 - Does Not Compute
* 32:25 - Conclusion: How To Think About Mythos
https://open.substack.com/pub/thezvi/p/claude-mythos-3-capabilities-and?utm_campaign=post-expanded-share&utm_medium=web


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

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.
* 00:00 - Introduction
* 00:46 - Awful Events Amid Scary Times
* 04:39 - Most Of Those Worried About AI Do As Well As One Can On This
* 07:00 - Some Who Are Worried About AI Need To Address Their Rhetoric
* 12:23 - Speak The Truth Even If Your Voice Trembles
* 14:44 - False Accusations And False Attacks Are Also Unacceptable
* 16:22 - Some Examples Of Attempts To Create Broad Censorship
* 25:29 - The Most Irresponsible Reaction Was From The Press
* 26:23 - Sam Altman Reacts
* 28:53 - Sam Altman Reflects
* 34:24 - Violence Is Never The Answer
https://open.substack.com/pub/thezvi/p/political-violence-is-never-acceptable?utm_campaign=post-expanded-share&utm_medium=web
I have also converted Eliezer’s post here:
https://open.substack.com/pub/askwhocastsai/p/only-law-can-prevent-extinction-by?r=67y1h&utm_campaign=post&utm_medium=web&showWelcomeOnShare=true


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

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.
* 00:00:00 - Introduction
* 00:02:10 - Introducing Project Glasswing
* 00:03:39 - Don’t Worry About the Government
* 00:05:15 - Cybersecurity Capabilities In The Model Card (Section 3)
* 00:06:56 - Cyber Capability Tests In The Model Card
* 00:08:54 - The Proof Is In The Patching
* 00:11:26 - Go For Read Team
* 00:15:16 - Is This New?
* 00:18:12 - Thanks For The Memories
* 00:23:07 - How Good Is Mythos At This?
* 00:26:20 - What Might Have Been
* 00:29:02 - The Chaos Option
* 00:32:09 - The Can’t Happen That Happened
* 00:33:15 - When You Go Looking For Specific, And You Are Told Exactly Where and How To Look For It, Your Chances Of Finding It Are Very Good
* 00:38:57 - Blatant Denials Are The Best Kind
* 00:42:56 - Anything You Can Do I Can Do Cheaper
* 00:45:15 - Theft Of Mythos Would Be A Big Deal
* 00:45:39 - No One Could Have Predicted This
* 00:46:52 - The Revolution Will Not Be Televised
* 00:47:50 - The Intelligence Will Not Be Televised
* 00:50:09 - Will We Be Doing This For A While?
* 00:52:30 - What If OpenAI Gets a Similar Model?
* 00:53:53 - Use It Or Lose It
* 00:54:28 - Solve For The Equilibrium
* 00:57:49 - Patriots and Tyrants
* 01:00:19 - Trust The Mythos
* 01:01:58 - Wide Scale Ability To Exploit Software Favors Strongest Projects
* 01:06:45 - Looking Back at GPT-2
* 01:08:10 - Limitless Demand For Compute
* 01:09:53 - Oh, Also, If Anyone Builds It, Everyone Dies
https://open.substack.com/pub/thezvi/p/claude-mythos-2-cybersecurity-and?utm_campaign=post-expanded-share&utm_medium=web


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

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.
* 00:00:00 - Introduction
* 00:03:39 - Mundane Alignment Is Excellent
* 00:05:26 - Would This Process Be Sufficient To Find A Dangerous Model?
* 00:06:58 - Introductory Warning About Superficial Mundane Alignment
* 00:15:45 - Model Training (1.1)
* 00:15:57 - Release Decision Process (1.2)
* 00:18:24 - RSP Evaluations (2.1 and 2.2)
* 00:23:26 - Autonomy Evaluations (2.3)
* 00:27:37 - The Alignment Risk Update Document
* 00:28:19 - The Threat Model
* 00:30:52 - Misalignment As Failure Mode
* 00:33:20 - Wouldn’t You Know?
* 00:35:29 - Don’t Encourage Your Model
* 00:37:00 - Beware Goodhart’s Law
* 00:39:08 - Beware The Most Forbidden Technique (5.2.3)
* 00:43:42 - Asking The Right Questions
* 00:45:11 - Model Organism Tests
* 00:47:13 - Model Weight Security (Risk Report 5.5.2.1)
* 00:47:41 - Reward Hacking (Back to The Model Card)
* 00:48:07 - Remote Drop-In Worker Coming Soon
* 00:51:04 - External Testing (2.3.7)
* 00:51:39 - Cyber Insecurity General Principle Interlude
* 00:52:52 - Alignment (4)
* 00:58:41 - Risk In The Room
* 01:00:04 - Mythos Meant Well
* 01:02:32 - Risk Not In The Room
* 01:05:38 - Alignment Testing Overview
* 01:09:09 - Internal Deployment Testing Process
* 01:11:47 - Reports From Pilot Use (4.2.1)
* 01:12:24 - Reports From Automated Testing (4.2)
* 01:15:35 - Other External Testing
* 01:16:17 - Just The Facts, Sir
* 01:19:20 - Refusing Safety Research
* 01:20:27 - Claude Favoritism
* 01:21:56 - Ruling Out Encoded Thinking (4.4.1)
* 01:25:51 - Sandbagging (4.4.2)
* 01:28:49 - Capability for Evasion of Safeguards (4.4.3)
* 01:31:20 - Pick A Random Number (4.4.3.4)
* 01:34:23 - White Box Analysis (4.5)
* 01:39:38 - Model Welfare (5)
* 01:40:43 - Key Model Welfare Findings (5.1.2)
* 01:52:05 - Is Mythos Okay?
* 01:54:48 - Self-Play
* 01:56:49 - A Few Fun Facts
https://open.substack.com/pub/thezvi/p/claude-mythos-the-system-card?utm_campaign=post-expanded-share&utm_medium=web


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