Epistemic status: one or two hours’ worth of thought, reasoning feels pretty clean from the inside. I think I’d be willing to bet on this producing a small effect, but nothing too important.
As far as I can tell, there are two reasons to memorize things – fast access and idea generation. Information stored in your brain is accessible very quickly, about 10x or 100x faster than retrieval from the Internet. But the really unique thing about information stored in your brain is that it determines what thoughts you are capable of coming up with, since you can’t make new ideas without some raw materials to smack together. Quick spot check for completeness – if I could build a machine that would access facts from the internet as fast as I access facts from my brain, and could use that database of facts to come up with new ideas, would that machine make memorizing information obsolete? This seems plausible to me, or at least pretty close.
Also as far as I can tell, there is only one reason not to memorize things – it takes work to keep memories in a retrievable form. The brain won’t hold on to stuff unless you do spaced repetition to tell it what to hold on to. This costs time and effort, and it’s not a one-time cost. Keeping stuff memorized for a long time requires upkeep work, making it a spring-like cost, so we should probably expect it to cost more than it seems up front. Note that this does not apply if the information is being used regularly by default, like your commute to work, but that’s not a particularly relevant caveat because that will just get memorized automatically anyway. Point is, anything you want to intentionally memorize and hold on to will incur an upkeep cost, which suggests it is probably more expensive than it intuitively seems and should be done carefully.
I really like the idea behind the post Key Numbers Every EA Should Know and the accompanying Anki deck. However, I kind of want to do my own take on this, keeping in mind the particular costs and benefits of memorizing instead of just using the internet. So let’s narrow down a bit – what sorts of facts exactly are most important to memorize?
First, what things might I want to retrieve quickly, but not often? When is having info available fast important, for reasons other than speed being convenient for frequent use? Live conversations put a time constraint on replies, but EA/Rationalist conversation norms tend to allow pausing to look things up. So conversations with non-EA-cluster folks seem like the most likely candidate. Aha! I want to have pitch-relevant facts available quickly, but don’t expect to use them often enough to automatically memorize them.
Second, what things might be relevant to idea generation? This seems a lot trickier than the previous category. The mechanism for idea generation I originally had in mind was stuff like finding analogies and parallels, but those seem quite diffuse – any particular idea you might intentionally memorize is quite unlikely to end up actually getting used this way. Are there any types of information that are used in analogies and comparisons more often than others? Well, “a single idea that can be compared to many things” sounds like it’s describing a generalizable model to me. Aha! To facilitate idea generation, it’s good to memorize a whole bunch of general models, and the criteria for when they apply. The implications might not be necessary to memorize, since once the connection is made you can Google freely for whatever theorems or results follow from that model.
Note that some things qualify under both strengths. Models you have to apply mid-conversation-flow while under standard “fast turn-taking” norms have to be fast and available for ideation, making them quite important to memorize. These things can include common conversational failure modes like false disagreement, rhetorical Dark Arts and corresponding defenses, or any other patterns you have to recognize and react to within the window of time allowed by the flow of normal conversation.
Last thing – this does not mean that it’s always worthwhile to memorize things that fall into these categories, or that it’s never worthwhile to memorize things that don’t. This was an attempt to determine what topics might be especially well suited to memorize, based on the unique characteristics of the medium. So while aligning nicely with those strengths seems like a plus, it’s not strictly necessary or sufficient.