OpenAI's recent transparency on safety and alignment strategies has been extremely helpful and refreshing.

Their Model Spec 2.0 laid out how they want their models to behave. I offered a detailed critique of it, with my biggest criticisms focused on long term implications. The level of detail and openness here was extremely helpful.

Now we have another document, How We Think About Safety and Alignment. Again, they have laid out their thinking crisply and in excellent detail.

I have strong disagreements with several key assumptions underlying their position.

Given those assumptions, they have produced a strong document – here I focus on my disagreements, so I want to be clear that mostly I think this document was very good.

This post examines their key implicit and explicit assumptions.

In particular, there are three core assumptions that I challenge:

---

Outline:

(02:45) Core Implicit Assumption: AI Can Remain a 'Mere Tool'

(05:16) Core Implicit Assumption: 'Economic Normal'

(06:20) Core Assumption: No Abrupt Phase Changes

(10:40) Implicit Assumption: Release of AI Models Only Matters Directly

(12:20) On Their Taxonomy of Potential Risks

(22:01) The Need for Coordination

(24:55) Core Principles

(25:42) Embracing Uncertainty

(28:19) Defense in Depth

(29:35) Methods That Scale

(31:08) Human Control

(31:30) Community Effort

---

First published:

Source:

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

This isn’t primarily about how I write. It's about how other people write, and what advice they give on how to write, and how I react to and relate to that advice.

I’ve been collecting those notes for a while. I figured I would share.

At some point in the future, I’ll talk more about my own process – my guess is that what I do very much wouldn’t work for most people, but would be excellent for some.

Table of Contents

---

Outline:

(00:29) How Marc Andreessen Writes

(02:09) How Sarah Constantin Writes

(03:27) How Paul Graham Writes

(06:09) How Patrick McKenzie Writes

(07:02) How Tim Urban Writes

(08:33) How Visakan Veerasamy Writes

(09:42) How Matt Yglesias Writes

(10:05) How JRR Tolkien Wrote

(10:19) How Roon Wants Us to Write

(11:27) When To Write the Headline

(12:20) Do Not Write Self-Deprecating Descriptions of Your Posts

(13:09) Do Not Write a Book

(14:05) Write Like No One Else is Reading

(16:46) Letting the AI Write For You

(19:02) Being Matt Levine

(20:01) The Case for Italics

(21:59) Getting Paid

(24:39) Having Impact

---

First published:

Source:

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

One hell of a paper dropped this week.

It turns out that if you fine-tune models, especially GPT-4o and Qwen2.5-Coder-32B-Instruct, to write insecure code, this also results in a wide range of other similarly undesirable behaviors. They more or less grow a mustache and become their evil twin.

More precisely, they become antinormative. They do what seems superficially worst. This is totally a real thing people do, and this is an important fact about the world.

The misalignment here is not subtle.

There are even more examples here, the whole thing is wild.

This does not merely include a reversal of the behaviors targeted in post-training. It includes general stereotypical evilness. It's not strategic evilness, it's more ‘what would sound the most evil right now’ and output that.

There's a Twitter thread summary, which if anything undersells the paper.

Ethan Mollick: This [...]

---

Outline:

(01:27) Paper Abstract

(03:22) Funny You Should Ask

(04:58) Isolating the Cause

(08:39) No, You Did Not Expect This

(12:37) Antinormativity is Totally a Thing

(16:15) What Hypotheses Explain the New Persona

(20:59) A Prediction of Correlational Sophistication

(23:27) Good News, Everyone

(31:00) Bad News

(36:26) No One Would Be So Stupid As To

(38:23) Orthogonality

(40:19) The Lighter Side

---

First published:

Source:

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

It's happening!

We got Claude 3.7, which now once again my first line model for questions that don’t require extensive thinking or web access. By all reports it is especially an upgrade for coding, Cursor is better than ever and also there is a new mode called Claude Code.

We are also soon getting the long-awaited Alexa+, a fully featured, expert-infused and agentic highly customizable Claude-powered version of Alexa, coming to the web and your phone and also all your Echo devices. It will be free with Amazon Prime. Will we finally get the first good assistant? It's super exciting.

Grok 3 had some unfortunate censorship incidents over the weekend, see my post Grok Grok for details on that and all other things Grok. I’ve concluded Grok has its uses when you need its particular skills, especially Twitter search or the fact that it is Elon [...]

---

Outline:

(01:19) Language Models Offer Mundane Utility

(03:53) Did You Get the Memo

(06:58) Language Models Don't Offer Mundane Utility

(08:29) Hey There Alexa

(11:28) We're In Deep Research

(18:45) Huh, Upgrades

(19:18) Deepfaketown and Botpocalypse Soon

(20:25) Fun With Media Generation

(21:24) They Took Our Jobs

(22:14) Levels of Friction

(25:18) A Young Lady's Illustrated Primer

(29:06) The Art of the Jailbreak

(30:03) Get Involved

(30:51) Introducing

(31:26) In Other AI News

(34:40) AI Co-Scientist

(39:50) Quiet Speculations

(48:14) The Quest for Sane Regulations

(52:58) The Week in Audio

(53:30) Tap the Sign

(55:05) Rhetorical Innovation

(01:00:22) Autonomous Helpful Robots

(01:02:09) Autonomous Killer Robots

(01:04:46) If You Really Believed That

(01:09:51) Aligning a Smarter Than Human Intelligence is Difficult

(01:16:45) The Lighter Side

---

First published:

Source:

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

Anthropic has reemerged from stealth and offers us Claude 3.7.

Given this is named Claude 3.7, an excellent choice, from now on this blog will refer to what they officially call Claude Sonnet 3.5 (new) as Sonnet 3.6.

Claude 3.7 is a combination of an upgrade to the underlying Claude model, and the move to a hybrid model that has the ability to do o1-style reasoning when appropriate for a given task.

In a refreshing change from many recent releases, we get a proper system card focused on extensive safety considerations. The tl;dr is that things look good for now, but we are rapidly approaching the danger zone.

The cost for Sonnet 3.7 via the API is the same as it was for 3.6, $5/$15 for million. If you use extended thinking, you have to pay for the thinking tokens.

They also introduced a [...]

---

Outline:

(01:17) Executive Summary

(03:09) Part 1: Capabilities

(03:14) Extended Thinking

(04:17) Claude Code

(06:52) Data Use

(07:11) Benchmarks

(08:25) Claude Plays Pokemon

(09:21) Private Benchmarks

(16:14) Early Janus Takes

(18:31) System Prompt

(24:25) Easter Egg

(25:50) Vibe Coding Reports

(32:53) Practical Coding Advice

(35:02) The Future

(36:05) Part 2: Safety and the System Card

(36:24) Claude 3.7 Tested as ASL-2

(38:15) The RSP Evaluations That Concluded Claude 3.7 is ASL-2

(40:41) ASL-3 is Coming Soon, and With That Comes Actual Risk

(43:31) Reducing Unnecessary Refusals

(45:11) Mundane Harm Evolutions

(45:53) Risks From Computer Use

(47:15) Chain of Thought Faithfulness

(48:53) Alignment Was Not Faked

(49:38) Excessive Focus on Passing Tests

(51:13) The Lighter Side

---

First published:

Source:

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

This is a post in two parts.

The first half is the post is about Grok's capabilities, now that we’ve all had more time to play around with it. Grok is not as smart as one might hope and has other issues, but it is better than I expected and for now has its place in the rotation, especially for when you want its Twitter integration.

That was what this post was supposed to be about.

Then the weekend happened, and now there's also a second half. The second half is about how Grok turned out rather woke and extremely anti-Trump and anti-Musk, as well as trivial to jailbreak, and the rather blunt things xAI tried to do about that. There was some good transparency in places, to their credit, but a lot of trust has been lost. It will be extremely difficult to win it [...]

---

Outline:

(01:21) Zvi Groks Grok

(03:39) Grok the Cost

(04:29) Grok the Benchmark

(06:02) Fun with Grok

(08:33) Others Grok Grok

(11:26) Apps at Play

(12:38) Twitter Groks Grok

(13:38) Grok the Woke

(19:06) Grok is Misaligned

(20:07) Grok Will Tell You Anything

(24:29) xAI Keeps Digging (1)

(29:21) xAI Keeps Digging (2)

(39:14) What the Grok Happened

(43:29) The Lighter Side

---

First published:

Source:

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

While we wait for the verdict on Anthropic's Claude Sonnet 3.7, today seems like a good day to catch up on the queue and look at various economics-related things.

Table of Contents

---

Outline:

(00:15) The Trump Tax Proposals

(02:36) Taxing Unrealized Capital Gains

(03:00) Extremely High Marginal Tax Rates

(05:24) Trade Barriers By Any Name Are Terrible

(06:11) Destroying People's Access to Credit

(06:35) Living Paycheck to Paycheck

(09:54) Oh California

(10:12) Chinese Venture Capital Death Spiral

(11:20) There is Someone Elon Musk Forgot to Ask

(14:39) Should Have Gone With the Sports Almanac

(17:58) Are You Better Off Than You Were Right Before the Election?

(18:13) Are You Better Off Than You Were Before the Price Level Rose?

(25:43) Most People Have No Idea How Insurance Works

(27:25) Do Not Spend Too Much Attention on Your Investments

(28:43) Preferences About Insider Training are Weird

(30:12) I Will Not Allocate Scarce Resources Via Price

(30:33) Minimum Wages, Employment and the Equilibrium

(31:22) The National Debt

(36:09) In Brief

The original text contained 1 footnote which was omitted from this narration.

---

First published:

Source:

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

OpenAI made major revisions to their Model Spec.

It seems very important to get this right, so I’m going into the weeds.

This post thus gets farther into the weeds than most people need to go. I recommend most of you read at most the sections of Part 1 that interest you, and skip Part 2.

I looked at the first version last year. I praised it as a solid first attempt.

Table of Contents

---

Outline:

(00:30) Part 1

(00:33) Conceptual Overview

(05:51) Change Log

(07:25) Summary of the Key Rules

(11:49) Three Goals

(15:51) Three Risks

(20:07) The Chain of Command

(26:14) The Letter and the Spirit

(29:30) Part 2

(29:33) Stay in Bounds: Platform Rules

(47:19) The Only Developer Rule

(49:19) Mental Health

(50:38) What is on the Agenda

(56:35) Liar Liar

(01:01:56) Still Kind of a Liar Liar

(01:07:42) Well, Yes, Okay, Sure

(01:10:14) I Am a Good Nice Bot

(01:20:55) A Conscious Choice

(01:21:49) Part 3

(01:21:52) The Super Secret Instructions

(01:24:45) The Super Secret Model Spec Details

(01:27:43) A Final Note

---

First published:

Source:

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

The Trump Administration is on the verge of firing all ‘probationary’ employees in NIST, as they have done in many other places and departments, seemingly purely because they want to find people they can fire. But if you fire all the new employees and recently promoted employees (which is that ‘probationary’ means here) you end up firing quite a lot of the people who know about AI or give the government state capacity in AI.

This would gut not only America's AISI, its primary source of a wide variety of forms of state capacity and the only way we can have insight into what is happening or test for safety on matters involving classified information. It would also gut our ability to do a wide variety of other things, such as reinvigorating American semiconductor manufacturing. It would be a massive own goal for the United States, on every [...]

---

Outline:

(01:14) Language Models Offer Mundane Utility

(05:44) Language Models Don't Offer Mundane Utility

(10:13) Rug Pull

(12:19) We're In Deep Research

(21:12) Huh, Upgrades

(30:28) Seeking Deeply

(35:26) Fun With Multimedia Generation

(35:41) The Art of the Jailbreak

(36:26) Get Involved

(37:09) Thinking Machines

(41:13) Introducing

(42:58) Show Me the Money

(44:55) In Other AI News

(53:31) By Any Other Name

(56:06) Quiet Speculations

(59:37) The Copium Department

(01:02:33) Firing All 'Probationary' Federal Employees Is Completely Insane

(01:10:28) The Quest for Sane Regulations

(01:12:18) Pick Up the Phone

(01:14:24) The Week in Audio

(01:16:19) Rhetorical Innovation

(01:18:50) People Really Dislike AI

(01:20:45) Aligning a Smarter Than Human Intelligence is Difficult

(01:22:34) People Are Worried About AI Killing Everyone

(01:23:51) Other People Are Not As Worried About AI Killing Everyone

(01:24:16) The Lighter Side

---

First published:

Source:

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.

That title is Elon Musk's fault, not mine, I mean, sorry not sorry:

Table of Contents

Release the Hounds

Grok 3 is out. It mostly seems like no one cares.

I expected this, but that was because I expected Grok 3 to not be worth caring about.

Instead, no one cares for other reasons, like the rollout process being so slow (in a poll on my Twitter this afternoon, the vast majority of people hadn’t used it) and access issues and everyone being numb to another similar model and the pace of events. And because everyone is so sick of the hype.

---

Outline:

(00:36) Release the Hounds

(02:11) The Expectations Game

(06:45) Man in the Arena

(07:29) The Official Benchmarks

(09:35) The Inevitable Pliny

(12:01) Heart in the Wrong Place

(14:16) Where Is Your Head At

(15:10) Individual Reactions

(28:39) Grok on Grok

---

First published:

Source:

Narrated by TYPE III AUDIO.

---

Images from the article:

Apple Podcasts and Spotify do not show images in the episode description. Try Pocket Casts, or another podcast app.