On thousands of pages of Rationalist literature

By dkl9, written 2023-359, revised 2023-359 (0 revisions)


In the past few years, I read dozens (hundreds?) of Scott Alexander essays, all of Yudkowsky's "Rationality: From AI to Zombies", all of Yudkowsky's "Harry Potter and the Methods of Rationality", and much else from LessWrong. That constitutes thousands of pages of what we may call "Rationalist literature".

I don't mean this as an achievement. It was very easy and not necessarily worth much.

I have been (and still am) influenced by that material, as may be apparent from my other writings. Now I'm more ambivalent about its value, and no longer read any more. Should you study it as I have?

Well, what does one get from it? And what are the alternative routes to those ends?

Rationalist literature is written with the mindset that we can become arbitrarily powerful by thinking better. If you read it substantially and let it influence you, the mindset transfers. This is mostly a good thing, tho it can inspire overconfidence and an excess of theory over practice. More conventional scientists (such as Feynman or Tyson, perhaps) convey the same mindset. They don't go as far as Yudkowsky et al.

LessWrong and the like teach a specific body of background knowledge and beliefs about artificial intelligence and its dangers. A lot of it is true. A lot of it is crazy. Those subsets are not mutually exclusive. Superintelligence (Bostrom, 2014) teaches similar content, dropping most of the crazy parts and condensing it into a couple hundred difficult pages. Most people don't need to care about AI in so much detail anyway.

The Rationalist literature in question teaches dozens (hundreds?) of specific concepts in epistemology, cognitive science, physics, and other technical fields, such as

Most of them are true. Some of them are useful. A few of them originated among the Rationalists. Studying science from more conventional sources would give you roughly the same kind of knowledge. Rationalist literature contains both more reckless falsehood and more underdocumented deep truth.

For some readers (probably not most), Rationalist literature is very inspiring and entertaining — sometimes to the point of addiction, hence the ease with which I read thousands of pages. This fun inflates the apparent value of the texts; were they not so fun, they would have to provide their other benefits much more efficiently than they do to be close to worth it. Unless you have particular reason to focus on Rationalist writing, such as seeking a topic barely documented in other sources, you're probably better off getting technical knowledge from less fun, more efficient sources, and getting your entertainment from less "intellectual" sources.