It seems like as other things drew our attention more, medical news slowed down. The actual developments, I have no doubt, are instead speeding up – because AI.

Note that this post intentionally does not cover anything related to the new Administration, or its policies.

Table of Contents

Some People Need Practical Advice

If you ever have to go to the hospital for any reason, suit up, or at least look [...]

---

Outline:

(00:22) Some People Need Practical Advice

(00:32) Good News, Everyone

(03:13) Bad News

(04:05) Life Extension

(04:42) Doctor Lies to Patient

(06:36) Study Lies to Public With Statistics

(08:46) Area Man Discovers Information Top Doctors Missed

(10:49) Psychiatric Drug Prescription

(11:30) H5N1

(12:43) WHO Delenda Est

(13:02) Medical Ethicists Take Bold Anti-Medicine Stance

(13:41) Rewarding Drug Development

(16:19) Not Rewarding Device Developers

(17:59) Addiction

(18:27) Our Health Insurance Markets are Broken

---

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.

I have been debating how to cover the non-AI aspects of the Trump administration, including the various machinations of DOGE. I felt it necessary to have an associated section this month, but I have attempted to keep such coverage to a minimum, and will continue to do so. There are too many other things going on, and plenty of others are covering the situation.

Table of Contents

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 post covers three recent shenanigans involving OpenAI.

In each of them, OpenAI or Sam Altman attempt to hide the central thing going on.

First, in Three Observations, Sam Altman's essay pitches our glorious AI future while attempting to pretend the downsides and dangers don’t exist in some places, and in others admitting we’re not going to like those downsides and dangers but he's not about to let that stop him. He's going to transform the world whether we like it or not.

Second, we have Frog and Toad, or There Is No Plan, where OpenAI reveals that its plan for ensuring AIs complement humans rather than AIs substituting for humans is to treat this as a ‘design choice.’ They can simply not design AIs that will be substitutes. Except of course this is Obvious Nonsense in context, with all the talk of remote workers, and [...]

---

Outline:

(01:52) Three Observations

(11:38) Frog and Toad (or There Is No Plan)

(18:45) A Trade Offer Has Arrived

---

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.

---

Outline:

(01:55) Language Models Offer Mundane Utility

(06:03) Language Models Don't Offer Mundane Utility

(08:07) We're in Deep Research

(13:54) Huh, Upgrades

(20:56) Seeking Deeply

(24:25) Smooth Operator

(29:15) They Took Our Jobs

(33:34) Maxwell Tabarrok Responds on Future Wages

(41:56) The Art of the Jailbreak

(46:34) Get Involved

(48:45) Introducing

(51:17) Show Me the Money

(53:20) In Other AI News

(56:12) Quiet Speculations

(01:02:05) The Quest for Sane Regulations

(01:04:53) The Week in Audio

(01:06:40) The Mask Comes Off

(01:08:20) Rhetorical Innovation

(01:21:35) Getting Tired of Winning

(01:24:36) People Really Dislike AI

(01:25:47) Aligning a Smarter Than Human Intelligence is Difficult

(01:27:20) Sufficiently Capable AIs Effectively Acquire Convergent Utility Functions

(01:36:29) People Are Worried About AI Killing Everyone

(01:47:03) Other People Are Not As Worried About AI Killing Everyone

(01:50:38) 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 doesn’t look good.

What used to be the AI Safety Summits were perhaps the most promising thing happening towards international coordination for AI Safety.

This one was centrally coordination against AI Safety.

In November 2023, the UK Bletchley Summit on AI Safety set out to let nations coordinate in the hopes that AI might not kill everyone. China was there, too, and included.

The practical focus was on Responsible Scaling Policies (RSPs), where commitments were secured from the major labs, and laying the foundations for new institutions.

The summit ended with The Bletchley Declaration (full text included at link), signed by all key parties. It was the usual diplomatic drek, as is typically the case for such things, but it centrally said there are risks, and so we will develop policies to deal with those risks.

And it ended with a commitment [...]

---

Outline:

(02:03) An Actively Terrible Summit Statement

(05:45) The Suicidal Accelerationist Speech by JD Vance

(14:37) What Did France Care About?

(17:12) Something To Remember You By: Get Your Safety Frameworks

(24:05) What Do We Think About Voluntary Commitments?

(27:29) This Is the End

(36:18) The Odds Are Against Us and the Situation is Grim

(39:52) Don't Panic But Also Face Reality

---

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.

Not too long ago, OpenAI presented a paper on their new strategy of Deliberative Alignment.

The way this works is that they tell the model what its policies are and then have the model think about whether it should comply with a request.

This is an important transition, so this post will go over my perspective on the new strategy.

Note the similarities, and also differences, with Anthropic's Constitutional AI.

How Deliberative Alignment Works

We introduce deliberative alignment, a training paradigm that directly teaches reasoning LLMs the text of human-written and interpretable safety specifications, and trains them to reason explicitly about these specifications before answering.

We used deliberative alignment to align OpenAI's o-series models, enabling them to use chain-of-thought (CoT) reasoning to reflect on user prompts, identify relevant text from OpenAI's internal policies, and draft safer responses.

Our approach achieves highly precise [...]

---

Outline:

(00:29) How Deliberative Alignment Works

(03:27) Why This Worries Me

(07:49) For Mundane Safety It Works Well

---

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.

Scott Alexander famously warned us to Beware Trivial Inconveniences.

When you make a thing easy to do, people often do vastly more of it.

When you put up barriers, even highly solvable ones, people often do vastly less.

Let us take this seriously, and carefully choose what inconveniences to put where.

Let us also take seriously that when AI or other things reduce frictions, or change the relative severity of frictions, various things might break or require adjustment.

This applies to all system design, and especially to legal and regulatory questions.

Table of Contents

---

Outline:

(00:40) Levels of Friction (and Legality)

(02:24) Important Friction Principles

(05:01) Principle #1: By Default Friction is Bad

(05:23) Principle #3: Friction Can Be Load Bearing

(07:09) Insufficient Friction On Antisocial Behaviors Eventually Snowballs

(08:33) Principle #4: The Best Frictions Are Non-Destructive

(09:01) Principle #8: The Abundance Agenda and Deregulation as Category 1-ification

(10:55) Principle #10: Ensure Antisocial Activities Have Higher Friction

(11:51) Sports Gambling as Motivating Example of Necessary 2-ness

(13:24) On Principle #13: Law Abiding Citizen

(14:39) Mundane AI as 2-breaker and Friction Reducer

(20:13) What To Do About All This

---

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 week we got a revision of DeepMind's safety framework, and the first version of Meta's framework. This post covers both of them.

Table of Contents

Here are links for previous coverage of: DeepMind's Framework 1.0, OpenAI's Framework and Anthropic's Framework.

Meta's RSP (Frontier AI Framework)

Since there is a law saying no two companies can call these documents by the same name, Meta is here to offer us its Frontier AI Framework, explaining how Meta is going to keep us safe while deploying frontier AI systems.

I will say up front, if it sounds like I’m not giving Meta the benefit of the doubt here, it's because I am absolutely not giving Meta the benefit of [...]

---

Outline:

(00:14) Meta's RSP (Frontier AI Framework)

(16:10) DeepMind Updates its Frontier Safety Framework

(31:05) What About Risk Governance

(33:42) Where Do We Go From Here?

---

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.

I remember that week I used r1 a lot, and everyone was obsessed with DeepSeek.

They earned it. DeepSeek cooked, r1 is an excellent model. Seeing the Chain of Thought was revolutionary. We all learned a lot.

It's still #1 in the app store, there are still hysterical misinformed NYT op-eds and and calls for insane reactions in all directions and plenty of jingoism to go around, largely based on that highly misleading $6 millon cost number for DeepSeek's v3, and a misunderstanding of how AI capability curves move over time.

But like the tariff threats that's now so yesterday now, for those of us that live in the unevenly distributed future.

All my reasoning model needs go through o3-mini-high, and Google's fully unleashed Flash Thinking for free. Everyone is exploring OpenAI's Deep Research, even in its early form, and I finally have an entity [...]

---

Outline:

(01:15) Language Models Offer Mundane Utility

(07:23) o1-Pro Offers Mundane Utility

(10:35) We're in Deep Research

(17:08) Language Models Don't Offer Mundane Utility

(17:49) Model Decision Tree

(20:43) Huh, Upgrades

(21:57) Bot Versus Bot

(24:04) The OpenAI Unintended Guidelines

(26:40) Peter Wildeford on DeepSeek

(29:18) Our Price Cheap

(35:25) Otherwise Seeking Deeply

(44:13) Smooth Operator

(46:46) Have You Tried Not Building An Agent?

(51:58) Deepfaketown and Botpocalypse Soon

(54:56) They Took Our Jobs

(01:08:29) The Art of the Jailbreak

(01:08:56) Get Involved

(01:13:05) Introducing

(01:13:45) In Other AI News

(01:16:37) Theory of the Firm

(01:21:32) Quiet Speculations

(01:24:36) The Quest for Sane Regulations

(01:33:33) The Week in Audio

(01:34:41) Rhetorical Innovation

(01:38:22) Aligning a Smarter Than Human Intelligence is Difficult

(01:40:33) The Alignment Faking Analysis Continues

(01:44:24) Masayoshi Son Follows Own Advice

(01:48:22) People Are Worried About AI Killing Everyone

(01:50:32) You Are Not Ready

(02:00:45) Other People Are Not As Worried About AI Killing Everyone

(02:02:53) 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.

The baseline scenario as AI becomes AGI becomes ASI (artificial superintelligence), if nothing more dramatic goes wrong first and even we successfully ‘solve alignment’ of AI to a given user and developer, is the ‘gradual’ disempowerment of humanity by AIs, as we voluntarily grant them more and more power in a vicious cycle, after which AIs control the future and an ever-increasing share of its real resources. It is unlikely that humans survive it for long.

This gradual disempowerment is far from the only way things could go horribly wrong. There are various other ways things could go horribly wrong earlier, faster and more dramatically, especially if we indeed fail at alignment of ASI on the first try.

Gradual disempowerment it still is a major part of the problem, including in worlds that would otherwise have survived those other threats. And I don’t know of any good [...]

---

Outline:

(01:15) We Finally Have a Good Paper

(02:30) The Phase 2 Problem

(05:02) Coordination is Hard

(07:59) Even Successful Technical Solutions Do Not Solve This

(08:58) The Six Core Claims

(14:35) Proposed Mitigations Are Insufficient

(19:58) The Social Contract Will Change

(21:07) Point of No Return

(22:51) A Shorter Summary

(24:13) Tyler Cowen Seems To Misunderstand Two Key Points

(25:53) Do You Feel in Charge?

(28:04) We Will Not By Default Meaningfully 'Own' the AIs For Long

(29:53) Collusion Has Nothing to Do With This

(32:38) If Humans Do Not Successfully Collude They Lose All Control

(34:45) The Odds Are Against Us and the Situation is Grim

---

First published:

Source:

Narrated by TYPE III AUDIO.