Crossposted (with small tweaks) from my Substack.

I have started, as a side read, Rationality: From AI to Zombies, which I believe contains the heart of the Sequences. On my e-reader, it adds up to about 1,750 pages, which is less than I expected. Still, I will be progressing little by little, as my main book slots are occupied until at least November. Yesterday I went over the preface, by EY, and the introduction, by Rob Bensinger, so I’ll briefly write a couple of things about these. Should anyone think I am misunderstanding and / or misrepresenting the contents, I will be glad to be corrected.

Preface

• Two things get explained in the preface: what big mistakes EY made in writing The Sequences and what their main purpose was.
• Mistakes, in order of relevance: 1) not focusing on the practice, as opposed to the theory, of rationality; 2) focusing on big, difficult issues instead of everyday problems; 3) focusing too much on rational belief, instead of rational action; 4) chaotic organization of posts; and 5) being too explicitly dismissive of stupid ideas.
• The purpose of The Sequences is to teach “the skills of rational belief and decision-making”, which are generally not taught in modern society and education.

Biases: An Introduction

• Besides statistical bias, which is built into reality, humans are mired in cognitive biases: systematic errors in how we think and that are in-built.
• Rationality is the project of overcoming these biases.
• Rationality isn’t about a straw-Vulcan rejection of emotions; we get an expected reference to Kahneman and Tversky’s systems 1 (intuitive, automatic) and 2 (intellectual, slow). Each has its pros and cons, depending on the situation.
• Some basic biases are introduced: the representativeness heuristic, the conjunction fallacy, base rate neglect, etc…
• Just being aware of these doesn’t mean you wont fall into them yourself (blind spot bias).
• Still, this book can allow you acquire expertise, with which you a) understand the rationale behind the bias, b) spot it in particular settings, and c) develop tools to counteract it.

Some Thoughts of mine

• I had some difficulty making out the difference between errors 1 and 3. I am guessing that the Sequences will likely dedicate a lot of time to theoretical discussions of what rational thinking and belief are, and too little to “what to do.”
• Hunting for biases and approximating rational thought and action at the limit seems like a wise thing to do. Implicitly, though, there seems to be a model behind this that assumes that such a project is fundamentally doable, and I am not sure I am 100% on board with that. In my prior, I view humans as extremely clunky Rube Goldberg machines, products of evolutionary pressures that have developed deeply held but inconsistent and incompatible preferences, which can’t really be resolved^[1], even if this were desirable. This isn’t an argument against trying to improve, but it does place a very low upper bound on what I expect we can achieve.

Note: I used ChatGPT for light proofreading/copyediting - stuff like fixing typos, grammar mistakes and awkward or unidiomatic phrasing. The text, views, and responsibility for the post are mine.

1. ^{^}

   I’d caveat this with ‘bar massive genetic/technological engineering and modifications of what we are, effectively turning us into something else'.