rajlego

Wow, that's insane.

There's a program called SuperMemo which has a feature called sleepchart. In sleepchart, you plot sleep episodes which gives you some neat data that can help figure out things like subjective night [1].

But since SuperMemo is a spaced repetition system, sleepchart can also do some very unique analysis. The main graph it can make with repetition data + sleep data is alertness over time: how grades change the longer you've been awake. I would be very curious if you were to try it to see how your alertness/grades reported for repetitions change the longer you're awake.

[1] https://supermemo.guru/wiki/Subjective_night

I would look into the effective altruism community and see what things people have already tried there. You might also find the work of the https://www.povertyactionlab.org/ useful because they use RCTs (randomized controlled trials) to see what interventions work. These aren't direct answers to your questions but I think they're useful to consider as what the best, most effective application of your time is.

Problem of schooling by Piotr Wozniak: https://supermemo.guru/wiki/Problem_of_Schooling

Of course main subject is criticism of school but it also has a lot on learning and how child brains work (and how they don’t)

I thought you could already do this on iOS? I can’t remember where I saw it but I could’ve sworn you can without jailbreak

What kind of long term applications do particle accelerators bring?

This is really weird to me but to each their own. I think it's partially because of ADHD but I really hate doing HW/writing essays that I feel are useless, which tends to be pretty common in uni.

I think there's a huge difference between having a guide and being in university. Having a guide is really helpful but going to university for that benefit feels a bit misguided because you're stuck learning what other people think you should rather than what you could actually apply.

While I'm fairly interested in tech things, I don't really have a tech background. Could you give me an example of topics with few resources online? The majority of subjects I can think of have good resources online, the only issue is navigating them as a beginner (which can be helped by finding some sort of mentor).

There are sites like MIT courseware that are completely free and allow online learning. Though of course, there are signaling benefits to online degrees.

The main benefits I would agree on are signaling and networking.

I am much less certain about the depending on university to force yourself to study things and being handheld by a prof/tutor.

Flagellation by university to get yourself to learn things is not a reliable long-term means of learning. I think it's far better (and I have had more success) when I learn for myself.

While there is some benefit in having a mentor, I think being handled and being told exactly what to learn isn't the most productive. A lot of times you'll end up learning really useless things as opposed to listening to your brain which generally has a decent grasp on figuring out what is worth learning to actually apply.

>But universities really make you push through those things you are not motivated enough to do (and subjects that seems to have no purpose) and many times you find out it was fun after grooking it and it actually made a lot of other things way easier. The first week after the exam you might be upset over the exam, feel something was unfair, but eventually those feelings fade away and you have another telescope to view the world through and arguably you have gained some grit. The universities are a guiding hand off what the experts in their field find important to know.

I really strongly disagree with this. Good learning is inherently pleasurable[1] and if it isn't then there's something wrong. If a subject doesn't feel like it has enough purpose, it's not worth taking. Your brain is a fairly good predictor of that, contrary to what school would lead you to believe.

I think overall you'd be way better off just learning the things that you want to learn, that are applicable to something you actually care about instead of taking a class and learning things your brain tells you are useless and you will probably never use. I haven't read it in depth but I think ultralearning is interesting for the learning-for-application approach.

I'm probably a bit biased though because I use incremental reading[2] and no matter how good university teaching is it just can't compare to learning by myself (with occasional help from an expert, rather than being handheld)

[1] https://supermemo.guru/wiki/Pleasure_of_learning [2] https://supermemo.guru/wiki/Incremental_reading

I've read all 3 of those and have found some interesting stuff but I'm not sure how much they actually changed my life because none of them think much about the most important aspect of teaching: how to get people to retain and apply stuff.

I think the problem comes down to: there is no good way to retain the information. And writers don't think about it either.

Most writers will say you need to "practice" and "apply" whatever you read while being very vague on the how. I'm interested in a habit creation system for methodologically allowing you to get good at applying say a mental model, but aside from that I think the best alternative is incremental reading/spaced repetition. (https://supermemo.guru/wiki/Incremental_reading)

tl;dr: while it can't give you procedural/neurally honed representations of the models you would get from practice, it gets close by allowing you to memorize declarative info.

I've found it at least partially functional for being able to sort of apply mental models, though there is still a large gap between knowing and applying.

Why do it through university when you have the internet and can more effectively learn things by yourself?

Have you seen quantum country [1]? A relevant cross between quantum mechanics and SRS

[1] https://quantum.country/

Incorrect :)

I use SuperMemo via parallels on my hackintosh. I know a lot of other people also using parallels on Mac.

On Linux, a kind soul got SuperMemo working via wine: https://www.supermemopedia.com/wiki/SuperMemo_for_Windows_un... (though were I still on Linux I’d probably stick to VMware because of a few features that don’t work)

Maybe this link will work: https://www.supermemo.com/en/archives1990-2015/articles/20ru...

If we go with the example you mentioned as a card: “You’re at a conference and you start talking to someone about [The Great Depression.] What are some key points to remember?” there are a few issues

tl;dr by being complex but still a single card SRS isn’t actually going to strengthen the memory very well and you’ll be stuck with short intervals on a card that will take at least 10 seconds each rep. It will still work but it’ll take up much more time than it should. (The link explains it much better)

But it’s not a bad idea. If you answer the card and write out an answer, memorizing pieces of that answer in accordance with minimum information principle will leave better memory effect. While you might have 20 subcards, over time these will get much more accurate intervals from being atomic and actually take less times in rep than your massed question.

I highly encourage any of you wanting to retain more from wha though read to use SuperMemo’s incremental reading [1] feature.

What it does summarized in a sentence: allows processing of thousands of passive materialS in parallel to convert them effectively to active recall item.

It’s a pretty bold claim but it’s not so fancy and complicated as it seems.

First half of the power of incremental reading is systems for processing materially efficiently (more things in less time). If I import an article into SuperMemo, SuperMemo has good tools for breaking it down (incrementally) and processing it manageably. While there’s a bit of a learning curve to being efficient, given an hour with a textbook chapter I could convert and retain the core content better than a note taker or an anki card maker. One of the things I like the most about it is that it makes processing manageable. It doesn’t make it easy but I never have to worry about needing to do too much at once. It’s all incremental and stress less (after you get over the figure out how to use it phrase).

The second thing it does well is allow you to learn effectively (the right things with your time). It has a priority system and a queueing algorithm that allows you to focus on highest yield material. It also has a system for allowing you to learn things in an order. That might seem trivial but if I’m reading a Wikipedia page and want to go back when I’ve learned something else, that’s a huge pain to manage traditionally. SuperMemo makes that fairly easy so you’re always able to make sure to learn background info as scaffolding before harder things.

I highly recommend people try it even though starting with it is pretty hard. If you’re interested you can either email me (email is in my profile) or join the SuperMemo discord servers also linked there.

[1] https://www.supermemo.com/en/archives1990-2015/help/read

I unfortunately don’t have a link on it but this is mainly because active recall is far better than just passive review for memory effect. Memories are like muscles: the more you struggle (without actually failing) the stronger the memory gets. Reading isn’t a struggle so the memories stay weak, paraphrasing and notes are thus much better or strengthening recall. SRS can take you a level higher than that and guarantee much longer term recall via repeated active recall

I think what you’re looking for is forgetting curves: https://en.m.wikipedia.org/wiki/Forgetting_curve

You might also find spaced repetition cool: https://ncase.me/remember/

It’s even more nonsense when you look at recall over longer periods of time. If you look at forgetting curves (like Ebbinghaus) no matter how well you read a book there’s no way after a month you’ll remember more than say 10%. I don’t think that’s so bad if you take care to take the most valuable abstract concepts and rules from books.

I don’t think cards like that work over 10 years because they violate minimum information principle (rule 4 here: http://super-memory.com/articles/20rules.htm). I agree though that you have to find the right cue to initialize recall or else the card is pointless. I think incrementalism in incremental reading is better for this since it’s easy to modify and improve your cards over time.

What do you mean by the classical version of SuperMemo? Why not incremental reading?

I think a lot of the issues you mentioned are hinged on formulation which is a really hard skill and requires a lot of practice. Unfortunately I think anki is terrible for it because unlike SM it’s not that convenient (nor intended) for you to edit cards for applicability on each go.

Have you heard of effective altruism and givewell?

If you want to maximize benefit per dollar I don't think there's a better way.

https://www.givewell.org/ https://www.effectivealtruism.org/

Recently, I learned that there's a name for that kind of happening: preferential attachment processes.

Wikipedia: https://en.wikipedia.org/wiki/Preferential_attachment

This is the one I use in Notion: https://www.notion.so/71c525aa1d8f41a7bbbe5159807c8d74?v=8ac... It's a bit confusing but it's usable even if you ignore everything to the right of the time column

It's based on the SuperMemo tasklist system which works quite well but is annoying to use when I'm on the go.

What OS do you run on it? Is it hard to find software that works on it?

I like this. A lot of people underestimate starting small because they want to maximize value generated from the start but really starting small and building up is the way to go.

Might find this: https://supermemo.guru/wiki/Micro-rules_of_productivity about micro rules interesting. This also reminds me of atomic habits.

That's an interesting technique for eliminating FOMO, knowing what you need in advance to get to where you want to go so you don't have to worry about shinier things.

Point 2 reminds me of Peter Thiel's power law: https://supermemo.guru/wiki/Thiel_on_power_law.

It wouldn't work for all of your tasks but have you tried prioritized tasklist (tasklists assigned a value and a time and then sorted by that). I have ADHD meaning managing doing things is horrible but with prioritized tasklist I can just start at the top and go down during timeslots in the day assigned to finishing tasklist items. If I don't finish low priority things with the time I had it just means they're not important enough (though of course there will be conflict within tasklist if you have tasks that have deadlines or urgencies).

I have the perfect tool for you if you're willing to suffer through getting used to it: SuperMemo [1]. The basic feature it starts with is spaced repetition [2]. Spaced repetition is good. But terribly boring if you're making cards by hand. Because of that, SuperMemo has a feature that imo is just as important as SRS but much less known: incremental reading [3].

tl;dr of incremental reading: process hundreds of articles in a time efficient manner and enjoyably convert them to active recall items.

For a video on it in action check this [4]

More detail: incremental reading has a lot of parts but probably the most important is the priority queue and the concept of incrementally.

Priority Queue: imagine you had 100 articles to go through. How would you process them efficiently? What if you have 10,000? You can't go through that reasonably. SuperMemo helps you manage that with a priority system which lets you choose what's actually important and what you'll end up seeing more and less often. I can't express just how much this helps. This means I can import as much shiny stuff as I want and be confident I'm not gonna lose out on stuff I actually care about. If you just used priority queue and manually made cards while going through things SM would be a big improvement over standard SRS.

But it has another cool concept: incrementalism. With SuperMemo you don't read through an entire piece of content at once. You read it over time, and make extracts that you break down till you have facts you can memorize with SRS. You might be wondering: if you don't read it all at once won't you forget it? It's the opposite that occurs. By separating reading of an article over time (separated by days) your brain has time to do a bit of memory consolidation and move some of what you read to long-term memory. That means next reread, you can process the article a bit better since you're not taxing WM as much. This might seem minor but it makes a huge difference for comprehension and long-term efficiency. Processing things in bits means when you get board with one article you can go to the next one. I have ADHD which makes this really awesome, it's like Instagram but with text and real learning.

If you're interested in trying it out, let me know. If you have questions about it ask away, this explanation has plenty of gaps. If you have criticisms I'd hold them because imagining SM is nothing like actually trying it.

[1] For a very long, cool introduction you can use this wired article: https://www.wired.com/2008/04/ff-wozniak/

[2] https://ncase.me/

[3] https://supermemo.guru/wiki/Incremental_reading

[4] https://www.youtube.com/watch?v=DoQoeK53bP8