Or: I read the executive order and its fact sheet, so you don’t have to.

I spent Halloween reading the entire Biden Executive Order on AI.

This is the pure ‘what I saw reading the document’ post. A companion post will cover reactions to this document, but I wanted this to be a clean reference going forward.

Takeaway Summary: What Does This Do?

It mostly demands a lot of reports, almost entirely from within the government.

---

Outline:

(00:26) Takeaway Summary: What Does This Do?

(05:46) Fact Sheet

(21:53) I Read the Whole Damn Thing So You Don’t Have To

(22:03) Sections 1 and 2: Introduction and Principles

(23:22) Section 3: Definitions

(30:05) Section 4: Ensuring Safety and Security

(43:44) Section 5: Promoting Innovation and Competition

(50:58) Section 6: Supporting Workers

(51:37) Section 7: Advancing Equity and Civil Rights

(52:41) Section 8: Protecting Consumers, Patients, Passengers and Students

(53:38) Section 9: Protecting Privacy

(54:06) Section 10: Advancing Federal Government Use of AI

(56:51) Section 11: Strengthening American Leadership Abroad

(57:38) Section 12: Implementation

(58:06) Conclusion

---

First published:

Source:

Narrated by TYPE III AUDIO.

Given Joe Biden seems to have become more worried about AI risk after having seen the movie, it seems worth putting my observations about it into its own post. This is what I wrote back then, except for the introduction and final note.

We now must modify the paragraph about whether to see this movie. Given its new historical importance, combined with its action scenes being pretty good, if you have not yet seen it you should now probably see this movie. And of course it now deserves a much higher rating than 70.

There are of course things such as ‘it is super cool to jump from a motorcycle into a dive onto a moving train’ but also there are actual things to ponder here.

Spoiler-Free Review

There may never be a more fitting title than Mission Impossible: Dead Reckoning. Each of these four words is doing important [...]

---

Outline:

(00:44) Spoiler-Free Review

(02:18) No One Noticed or Cared That The Alignment Plan Was Obvious Nonsense

(03:52) No One Cares That the Threat is Extinction, They All Want Control

(05:29) The Movie Makes it Very Clear Why Humanity Won’t Win, Then Ignores It

(08:15) Warning Shots are Repeatedly Ignored

(08:46) Approximately No One Noticed Any of This

---

First published:

Source:

Narrated by TYPE III AUDIO.

There is much talk about so-called Responsible Scaling Policies, as in what we will do so that what we are doing can be considered responsible. Would that also result in actually responsible scaling? It would help. By themselves, in their current versions, no. The good scenario is that these policies are good starts and lay groundwork and momentum to get where we need to go. The bad scenario is that this becomes safetywashing, used as a justification for rapid and dangerous scaling of frontier models, a label that avoids any actual action or responsibility.

Others think it would be better if we flat out stopped. So they say so. And they protest. And they point out that the public is mostly with them, at the same time that those trying to play as Very Serious People say such talk is irresponsible.

Future persuasion will be better. Sam [...]

---

Outline:

(01:31) Language Models Offer Mundane Utility

(01:46) Language Models Don’t Offer Mundane Utility

(02:32) GPT-4 Real This Time

(03:51) A Proposed Bet

(04:38) Fun with Image Generation

(05:54) Deepfaketown and Botpocalypse Soon

(08:26) They Took Our Jobs

(10:38) Get Involved

(11:16) Introducing

(16:19) In Other AI News

(17:57) Quiet Speculations

(21:13) The Quest for Sane Regulations

(26:29) The Week in Audio

(27:08) Rhetorical Innovation

(42:45) Friendship is Optimal

(45:12) Honesty As the Best Policy

(52:43) Aligning a Smarter Than Human Intelligence is Difficult

(54:17) Aligning a Dumber Than Human Intelligence Is Also Difficult

(01:02:49) Humans Do Not Expect to Be Persuaded by Superhuman Persuasion

(01:07:36) DeepMind's Evaluation Paper

(01:17:23) Bengio Offers Letter and Proposes a Synthesis

(01:20:54) Matt Yglesias Responds To Marc Andreessen's Manifesto

(01:25:33) People Are Worried About AI Killing Everyone

(01:31:09) Someone Is Worried AI Alignment Is Going Too Fast

(01:36:14) Please Speak Directly Into This Microphone

(01:37:46) The Lighter Side

---

First published:

Source:

Narrated by TYPE III AUDIO.

It did not get the bulk of the attention, but the actual biggest story this week was that America tightened the rules on its chip exports, closing the loophole Nvidia was using to create the A800 and H800. Perhaps the new restrictions will actually have teeth.

Also new capabilities continue to come in based on the recent GPT upgrades, along with the first signs of adversarial attacks.

Also a lot of rhetoric, including, yes, that manifesto. Yes, I do cover it.

Table of Contents

---

Outline:

(00:36) Language Models Offer Mundane Utility

(01:31) Dalle-3 System Complete Prompt

(09:00) GPT-4 Real This Time

(12:16) Fun with Image Generation

(12:28) Deepfaketown and Botpocalypse Soon

(14:24) People Genuinely Against Genuine People Personalities

(19:11) They Took Our Jobs

(22:42) Get Involved

(24:13) Introducing

(25:07) In Other AI News

(32:15) Quiet Speculations

(33:27) Man With a Plan

(39:56) China

(45:00) The Quest for Sane Regulations

(50:26) The Chips Are Down

(54:09) The Week in Audio

(54:45) Yes, We Will Speak Directly Into This Microphone

(58:22) Rhetorical Innovation

(01:16:38) Open Source AI is Unsafe and Nothing Can Fix This

(01:18:36) No One Would Be So Stupid As To

(01:19:09) Aligning a Smarter Than Human Intelligence is Difficult

(01:20:49) People Are Worried About AI Killing Everyone

(01:25:36) New Bengio Interview

(01:31:36) Marc Andreessen's Techno-Optimist Manifesto

(01:42:50) Other People Are Not As Worried About AI Killing Everyone

(01:45:33) Other People Wonder Whether It Would Be Moral To Not Die

(01:46:38) The Lighter Side

---

First published:

Source:

Narrated by TYPE III AUDIO.

The world is slowly waking up to the fertility crisis. There is more acknowledgement of the problem, and there is more talk of potential practical solutions. I do not believe the topic is a realistic target for Balsa Policy Institute, but I will continue to keep an eye on the ball and issue periodic roundups like this one.

In Brief

A basic problem is that we do not consider infertility a big deal for things such as DALYs. Many would strongly disagree.

Fertility clinics that get new corporate ownership by a fertility chain have 28.2% increased clinic volume and 13.6% higher success rates, dramatically improving overall outcomes.

Mishra has the fertility take on Barbie.

Mishra: Weirdly this is a take I haven’t seen anywhere but to me the point of Barbie is that the most important thing is kids and everything else is meaningless without [...]

---

Outline:

(00:22) In Brief

(01:40) Causes

(07:21) Causes: Pessimism

(09:57) Causes: Escalating Signals

(11:41) The Baby Boom

(14:24) You’ll Need More

(19:50) I’ll Tell You What I Want What I Really Really Want

(21:48) Negative Dakka

(23:01) Not Enough Dakka

(31:38) More Dakka

(35:49) Cato Brings the Fire

(50:53) I’m Doing My Part

(52:35) Artificial Wombs

---

First published:

Source:

Narrated by TYPE III AUDIO.

This has been a rough week for pretty much everyone. While I have had to deal with many things, and oh how I wish I could stop checking any new sources for a while, others have had it far worse. I am doing my best to count my blessings and to preserve my mental health, and here I will stick to AI. As always, the AI front does not stop.

Table of Contents

---

Outline:

(00:26) Language Models Offer Mundane Utility

(03:53) Language Models Don’t Offer Mundane Utility

(09:53) GPT-4 Real This Time

(11:03) Fun with Image Generation

(16:14) Deepfaketown and Botpocalypse Soon

(17:03) They Took Our Jobs

(21:13) Get Involved

(21:39) Introducing

(27:56) In Other AI News

(30:25) Cool New Interpretability Paper

(33:20) So What Do We All Think of The Cool Paper?

(41:46) Alignment Work and Model Capability

(43:09) Quiet Speculations

(46:33) The Week in Audio

(47:39) Rhetorical Innovation

(57:27) Aligning a Smarter Than Human Intelligence is Difficult

(01:05:25) Aligning Dumber Than Human Intelligences is Also Difficult

(01:08:01) Open Source AI is Unsafe and Nothing Can Fix This

(01:10:11) Predictions are Hard Especially About the Future

(01:13:47) Other People Are Not As Worried About AI Killing Everyone

(01:22:36) The Lighter Side

---

First published:

Source:

Narrated by TYPE III AUDIO.

It's time once again for a roundup of all the childhood-related stories I’ve noticed since the last such post that don’t fit in elsewhere.

In addition to the standard post, I have a personal note, which is that unfortunately our current nanny's health does not allow her to continue, and thus we in the process of searching and doing interviews for a new full-time nanny for our three children here in Manhattan. I hope to have someone in place by the end of the week. If you are or know someone we should consider, please do quickly contact me right away with the relevant information. You can DM me on Twitter, PM me on LesssWrong or email me through substack (or email directly if you know it). Thanks in advance.

We All Need a Friend

Sorry you are being bullied, here's some gold? In this case, $40k [...]

---

Outline:

(00:43) We All Need a Friend

(01:19) Time Spent on Children is Historically Anomalous

(03:32) Let Your Children Play

(12:52) Illusion of Safety

(14:32) Social Media is Badly Named

(17:32) Kids Respond To Incentives

(18:18) Let Them Eat Lunch

(19:52) No Child Left Behind

(23:18) School Reinvented

(27:52) School Choice

(30:37) School Hidden

(32:27) School Evaluation Hard

(33:11) A Metaphor

(40:00) Lies

(43:58) Don’t Worry, Be Happy

(45:29) You Might Learn Something

(46:46) Restorative Justice Does Not Scale

(50:42) Other Things In Brief

(53:02) The Lighter Side

---

First published:

Source:

Narrated by TYPE III AUDIO.

Can you tell if an AI is lying to you? A new paper claims that we essentially can do exactly that, at least under the right conditions. Another paper claims we can inject various sentiments into responses, getting the AI to do what we wish. Interpretability is making progress. It is exciting to think about the implications. In the short term, it would be great if we could use this to steer responses and to detect and correct hallucinations. There's a lot of potential here to explore.

In the longer term, I am more skeptical of such strategies. I do not think lie detection is a viable primary control or alignment strategy. I worry that if we go down such a path, we risk fooling ourselves, optimizing in ways that cause the techniques to stop working, and get ourselves killed. Indeed, even attempts to grab the low-hanging fruit [...]

---

Outline:

(01:30) Table of Contents

(03:47) Language Models Offer Mundane Utility

(06:31) Language Models Don’t Offer Mundane Utility

(07:44) GPT-4 Real This Time

(07:48) Fun with Image Generation

(08:04) Deepfaketown and Botpocalypse Soon

(13:32) They Took Our Jobs

(16:10) Get Involved

(17:07) Introducing

(17:52) Meta Surveillance

(23:09) In Other AI News

(24:11) Open Philanthropy Worldview Contest Awards Prizes

(24:39) First Prizes ($50k)

(24:59) Second Prizes ($37.5k)

(25:17) Third Prizes ($25k)

(27:01) Quintin Doubles Down on Twitter

(29:27) The Other Winners

(31:24) Quiet Speculations

(39:13) Open Source AI is Unsafe and Nothing Can Fix This

(46:27) The Quest for Sane Regulations

(50:38) The Week in Audio

(52:03) Rhetorical Innovation

(59:48) Eliezer Yudkowsky clarifies a recent misunderstanding about the Orthogonality Thesis (here is his full best explanation of the thesis, from Arbital).

(01:02:49) Aligning a Smarter Than Human Intelligence is Difficult

(01:12:22) People Are Worried About AI Killing Everyone

(01:12:55) Other People Are Not As Worried About AI Killing Everyone

(01:17:48) The Lighter Side

---

First published:

Source:

Narrated by TYPE III AUDIO.

Response to: Evolution Provides No Evidence For the Sharp Left Turn, due to it winning first prize in The Open Philanthropy Worldviews contest.

Quinton's post is an argument about a key historical reference class and what it tells us about AI. Instead of arguing that the reference makes his point, he is instead arguing that it doesn’t make anyone's point - that we understand the reasons for humanity's sudden growth in capabilities. He says this jump was caused by gaining access to cultural transmission which allowed partial preservation of in-lifetime learning across generations, which was a vast efficiency gain that fully explains the orders of magnitude jump in the expansion of human capabilities. Since AIs already preserve their metaphorical in-lifetime learning across their metaphorical generations, he argues, this does not apply to AI.

This last paragraph makes an extremely important claim that I want to ensure I convey [...]

---

First published:

Source:

Narrated by TYPE III AUDIO.

It never stops. I’m increasingly building distinct roundups for various topics, in particular I’m splitting medical and health news out. Let's get to the rest of it.

Bad News

A simple model of why everything sucks: It is all optimized almost entirely for the marginal user, who the post calls Marl. Marl hates when there are extra buttons on the screen or any bit of complexity is offered, even when he is under zero obligation to use it or care, let alone being asked to think, so everything gets dumbed down.

Could companies really be this stupid, so eager to chase the marginal user a little bit more that they cripple the functionality of their products? Very much so, yes, well past the point where it makes financial sense to do so. The metrics form a tyranny, invisible costs are increasingly paid on the alter of visible [...]

---

Outline:

(00:17) Bad News

(03:30) Disunity

(11:38) Ban Gain of Function Research

(13:51) The Right Price to Be Willing to Pay is Not $0

(15:19) Opportunity Knocks

(15:33) Government Working

(17:07) Worthwhile Canadian Immigration

(20:58) People Don’t Do Things

(22:23) Thinking is Hard

(24:25) Satisfaction Guaranteed To Stay Around 83%

(26:50) Good News, Everyone

(33:07) The Ancient Art of Getting Through a Conversation

(35:31) Money Stuff

(44:32) Very High Marginal Tax Rates

(46:09) So Cold, So Alone

(49:14) While I Cannot Condone This

(55:46) Queue Technology

(57:34) Gamers Gonna Game Game Game Game Game

(01:05:01) The Lighter Side

---

First published:

Source:

Narrated by TYPE III AUDIO.