The Engines of Cognition is a newly published collection of essays by the LessWrong Community. They're currently available with Prime at Amazon US, where they'll arrive in time for Christmas, and you can read on to learn more about them.

About

In The Engines of Cognition, the writers set out to understand key elements of the art of rationality. Starting with the simple epistemic question of when and how to trust different sources of information, the essays in this collection move through understanding the lens of incentives, an exploration of when and why complex systems become modular, and finally into a discussions of failure, both personal and civilizational.

This book set is for people who want to read the most interesting ideas LessWrong has recently explored. It's for the people who read best away from screens, away from distractions. It's for people who do not check the site regularly, but would still like to get the ideas within. For such people, this is intended to be the ideal way to read LessWrong.

Essays in this book set take a variety of forms, from thought experiments to literature reviews, as well as book reviews, interviews, personal stories, microeconomic arguments, mathematical explanations, research advice, philosophical musings, published papers, disagreements-with-Robin-Hanson, forecasts for the future, survey data, cartoons, and more.

Authors featured include Eliezer Yudkowsky, Scott Alexander, Zvi Mowshowitz, and over 30 more LessWrong writers. The essays were originally published to LessWrong in 2019, and for the first time are available with editing and illustration in print form. In addition, given recent advances in machine learning art, each essay has a unique piece of ML-generated art based on the essay title, some of which are shown below.

Order now on Amazon Prime US ($30) and it will arrive in time for Christmas.

Book Spreads

Table of Contents

Trust
Rule Thinkers In, Not Out	Scott Alexander
Gears vs Behavior	John S. Wentworth
Book Review: The Secret Of Our Success	Scott Alexander
Reason isn't magic	Ben Hoffman
"Other people are wrong" vs "I am right"	Buck Shlegeris
In My Culture	Duncan Sabien
Chris Olah's views on AGI safety	Evan Hubinger
Understanding "Deep Double Descent"	Evan Hubinger
How to Ignore Your Emotions (while also thinking you're awesome at emotions)	Hazard
Paper-Reading for Gears	John S. Wentworth
Book summary: Unlocking the Emotional Brain	Kaj Sotala
Noticing Frame Differences	Raymond Arnold
Propagating Facts into Aesthetics	Raymond Arnold
Do you fear the rock or the hard place?	Ruben Bloom
Mental Mountains	Scott Alexander
Steelmanning Divination	Vaniver
Modularity
Book Review: Design Principles of Biological Circuits	John S. Wentworth
Reframing Superintelligence: Comprehensive AI Services as General Intelligence	Rohin M. Shah
Building up to an Internal Family Systems model	Kaj Sotala
Being the (Pareto) Best in the World	John S. Wentworth
The Schelling Choice is "Rabbit", not "Stag"	Raymond Arnold
Literature Review: Distributed Teams	Elizabeth Van Nostrand
Gears-Level Models are Capital Investments	John S. Wentworth
Evolution of Modularity	John S. Wentworth
You Have About Five Words	Raymond Arnold
Coherent decisions imply consistent utilities	Eliezer Yudkowsky
Alignment Research Field Guide	Abram Demski
Forum participation as a research strategy	Wei Dai
The Credit Assignment Problem	Abram Demski
Selection vs Control	Abram Demski
Incentives
Asymmetric Justice	Zvi Mowshowitz
The Copenhagen Interpretation of Ethics	Jai Dhyani
Unconscious Economics	Jacob Lagerros
Power Buys You Distance From The Crime	Elizabeth Van Nostrand
Seeking Power is Often Convergently Instrumental in MDPs	Alexander Turner & Logan Smith
Yes Requires the Possibility of No	Scott Garrabrant
Mistakes with Conservation of Expected Evidence	Abram Demski
Heads I Win,Tails?—Never Heard of Her; Or, Selective Reporting and the Tragedy of the Green Rationalists	Zack M. Davis
Excerpts from a larger discussion about simulacra	Ben Hoffman
Moloch Hasn’t Won	Zvi Mowshowitz
Integrity and accountability are core parts of rationality	Oliver Habryka
The Real Rules Have No Exceptions	Said Achmiz
Simple Rules of Law	Zvi Mowshowitz
The Amish, and Strategic Norms around Technology	Raymond Arnold
Risks from Learned Optimization: Introduction	Evan Hubinger, Chris van Merwijk, Vladimir Mikulik, Joar Skalse, & Scott Garrabrant
Gradient hacking	Evan Hubinger
Failure
The Parable of Predict-O-Matic	Abram Demski
Blackmail	Zvi Mowshowitz
Bioinfohazards	Megan Crawford, Finan Adamson, & Jeffrey Ladish
What failure looks like	Paul Christiano
Seeking Power is Often Convergently Instrumental in MDPs	Alexander Turner & Logan Smith
AI Safety “Success Stories”	Wei Dai
Reframing Impact	Alexander Turner
The strategy-stealing assumption	Paul Christiano
Is Rationalist Self-Improvement Real?	Jacob Falkovich
The Curse Of The Counterfactual	P.J. Eby
human psycholinguists: a critical appraisal	Nostalgebraist
Why wasn't science invented in China?	Ruben Bloom
Make more land	Jeff Kaufman
Rest Days vs Zombie Days	Lauren Lee

FAQ

How were the essays in the books selected?

Each year LessWrong holds an annual review, where members of the community review and vote on the best content of the year prior. These essays are chosen from the most highly upvoted of the last review, when in 2020 we reviewed essays from 2019.

There are also a few small additional essays, such as referenced essays included for context or new postscripts by authors.

How are the books made?

Ben Pace on the LessWrong team led the project. Asimov Collective did most of the hard work of putting the books together, and Hemlock Printers in Canada printed the books.

What size are the books?

Each book is 5 x 7.75 inches, notably larger than the last set of LessWrong books released.

How was the art generated?

Using VQGAN+CLIP, a system that takes in a base image and a written prompt, and attempts to transform the base image into what the GAN expects the image to look like given the associated text. This ML system is available in many public notebooks. For each essay the LessWrong and Asimov Collective teams generated ~60 images using slight variations in prompts, from which each essay's image was selected.

Here is an approximate recreation of how the cover images were generated.

To give a sense of what prompts were used, here was the prompt used for the "Trust" volume:

The Engines of Cognition by Alex Hillkurtz | System of Gears | Aquarelle | Greek Architecture | Blue on White Color Palette | Trending on Artstation

Does this book assume I have read other LessWrong content, like The Sequences?

No. It's largely standalone, and does not require reading other content on the site, although it will be enhanced by having engaged with those ideas. A few essays will assume certain sorts of background context, but the vast majority will not.

How can I get the books?

We have books in store in Amazon US, and we will be shipping books to Amazon UK in the coming month. You can sign up below to be notified when they arrive in Amazon UK. Sending books to other regions is a lot of work so we won't do it this year. Note that both Amazon US and UK can do international shipping, and you can currently order from Amazon to the UK.

These will be the only ways to buy the books, though if you are planning to buy in bulk (as few as 20+) feel free to contact me at ben@lesswrong.com to see if I can help.

When will the books be on Amazon UK?

In a couple of weeks. You can sign up below to be notified by me (Ben) when the books are in stock.

I have a question not answered here?

There's a comment box right below.

Remind me again, how can I order it in the US?

Order now on Amazon Prime US ($30) and it will arrive in time for Christmas.