I was anti-memorization until I went back to graduate school for mathematics. I had forgotten (or never learned) a lot of things needed to pass qualifying exams. At some point I ran across the spaced repetition idea (maybe from the Wired SuperMemo article [0]) and I gave it a try. I ended up using it to memorize large portions of baby Rudin and Munkres' Topology, as well as some algebra and a bunch of qualifying exam questions.
The qualifying exams were difficult until I reached some "critical mass" of knowledge. Then I could regurgitate proofs and even attack novel problems easily.
There's an analogy here somewhere to the "leetcode" style of software engineer interview. On one hand qualifying exams and leetcode questions are a stupid gatekeeping mechanism, but on the other hand the best researchers/engineers I know have a huge number of facts and examples memorized and ready at their fingertips. I didn't think I needed to do so, but perhaps there is something to suffering through the rote memorization phase to make what comes next that much easier.
That Supermemo article in Wired hooked me. I still remember where I was when I read it and the feeling of reading it. I still use Supermemo daily, and it's one of the programs keeping on Windows. Anki (and every other SRS program I've tried) just doesn't compare as soon as you move beyond a list of flashcards.
That said, having used Supermemo for over a decade at this point, the hardest thing about SRS is deciding what's actually worth reviewing for a long period of time. I delete (really remove from repetitions) cards from my collection almost on a daily basis.
There's a lot of stuff that seems really important that I just didn't care about after even three months.
Supermemo's incremental reading basically lets you schedule chunks of text or images (alleged video too) like a flashcard from Anki. So, instead of bookmarking articles and never reading them, I can put them into Supermemo and know I'll eventually review it.
It basically counts as a separate type of flashcard, but all your reviews are mixed by default. So on a typical day, I'll have maybe 20 flashcards to review, and then another 10-20 articles.
Supermemo saves where you last were reading, so when I get bored of an article, I just hit next and go to the next one. Eventually, you'll process an article down to individual flashcards like you'd put in Anki, or remove it from your review process altogether. Also, you can just leave the entire article in there if you like rereading it.
What's your use case? In my case, obsidian + anki seems to do very well (I can keep a decent amount of context in my cards) for math/CS topics.
Can you also give an example how you benefit from your workflow, and say a bit more why Anki wouldn't be sufficient? Wouldn't keeping a track of articles and making cards in Anki achieve the same purpose, or is there something else? Of course the implementation matters, and having a system for paper-reading and revising by itself is a nice feature.
According to the wired article, the key to remembering is reviewing the information the moment you’re about to forget. It makes it sound like the scheduled review times are designed to maximize the effect.
Never tried it but all the comments made me curious.
Yeah I agree. Anki has a failure mode for me that I eventually accumulate > 5000 flash cards and a review session can take an hour or more. Knowing what to review is really difficult. I'll give Supermemo a shot, I've always heard of it as the gold standard but never tried it.
> Yeah I agree. Anki has a failure mode for me that I eventually accumulate >5000 flash cards and a review session can take an hour or more.
You don't really need to keep up with Anki's review sessions, even a partial session is very effective when you have lots and lots of cards. The tradeoff is that you might not actually maintain 100% recall of all items, but you'll recall most of them, and the scheduling algorithm will space out the items that you do recall even further.
I'm familiar with that problem, and overcame it by really working at making good cards. Today, I have 6000 cards in my collection, and daily reviews take 15-20 minutes; usually 100-120 cards. (I have been using Anki for a lost time. My collection is almost 10 years old.)
Granted, writing good cards is easy to say and really hard to do. Some specific pointers:
- cards should take less than 10 seconds to answer
- most questions should be 7 words or so
- if your question has multiple clauses, split those up into separate cards
- cloze deletions are great, consider multiple clozes per card
- around 10% of your cards should have images
- good images help a lot, mediocre images make it worse -- when in doubt, don't
- Answer this Question cards should have an opposite Question for this Answer like Jeopardy
In my experience, you really only start to see these problems at the tens of thousands of flashcards level. I have ~70,000 mature flashcards and also have the growing backlog problem, even if I don't add new cards for weeks. My collection is 8 years old.
I agree that making good flashcards helps alot. Another big thing is to make sure the knowledge graph is connected, eg there are no 'orphaned' individual cards / groups of cards. Those tend to suffer seriously from decay for me.
I'm working on a program that solves this problem: we don't have mandatory daily reviews and we also have a Spaced Repetition Algorithm for reviews. Also it's much more flexible than Anki & Supermemo. So you don't have to stress over as to complete the daily grind.
Yo, I've been using SuperMemo for 16 years every single day (Did my cards this morning :), and you're spot on! I would love to chit chat about SM sometime, I've not met many people that use SRS for so long! Send me an e-mail at thesupermemoguy (at) gmail (dot) com
No it's primarily an awful Delphi for Windows desktop application that is extremely fragile.
But once you know how to "hold it correctly" it will be your companion for life.
I run it in a virtual machine with all its legacy dependencies like IE. Forget about running it on Wine, it barely works on any version of Windows (lol)
What's https://www.supermemo.com/ then? I just signed up to give it a go (requires CC #, but free for a month). Still haven't found good spaced repetition vocab learning site that works well for me.
Ah OK, though I'm still curious how there can also be an apparently unrelated online offering at supermemo.com. FWIW, from the very brief experience I've had so far I'm not super impressed, but I'll try sticking with it for at least the month of the free trial.
they are working together. supermemo.com is based on the same algorithm but is designed to be more mainstream (no advanced stuff like incremental reading & hard to make your own cards).You can buy high quality language courses there(no need to make all the cards). I have both since supermemo.com has an app for the phone.
If actually using it has this huge friction, how do you manage to use it every day? I always have a hard time creating habits out things that are hard to "start" doing every time.
I remember the exact point in my life when I realized how much difference rote memorization makes.
I had the opportunity to write some MC68000 assembly code, at a time which was not too long after having written a complete emulator for it in C. Writing the emulator required me becoming familiar with every single instruction in all of its nuances, and exactly what they all do.
So, having that behind me, I sat down to write this code and it was like wow ... I could just spew the code without having to look anything up. It was so easy!
> I remember the exact point in my life when I realized how much difference rote memorization makes.
I suspect this might have something to do with neurological development. At least in my experience. Something about memory and recollection "clicked" (not a skill but more a capacity) at a really adult age for me.
Almost similar to being taught calculus after not knowing how most physics formula are derived and then looking back in confusion how you struggled with a straightforward thing.
I'd be really interested to see some research into integrating spaced repetition into our actual education system. Almost everything I see about it is adults learning, I wonder how much we could speed up primary school education considering so much of the "base" stuff needed to advance is rote memorization anyway
seems like countries should be investing money in this, potentially trillions in unlocked economic potential by improving and speeding up education. I think I read something like 20-30% of medical school students use SRS, yet only a fraction of the general population uses it. Insane to me that we have a tool like this and almost nothing is being done to improve adoption.
>see some research...spaced repetition into our actual education system.<
I have followed the research in this area closely for several years. Most of the (excellent) foundational research was done by university psychology professors who used college students as subjects; it's much harder for them to use primary school students, hence far less research with young learners. Implementing these techniques in elementary schools is more challenging then many would guess. Two good books: "Make it Stick" Brown, Roediger, McDaniel; "Powerful Teaching" Agarwal & Bain. PM me if you'd like more info. The second book describes research with middle school students. Pooja Agarwal also hosts the informative site www.retrievalpractice.org which informs K-12 teachers and has lots of free downloads. I am involved in a startup that is beginning (pilot teachers in September) to implement spaced repetition for elementary school math.
I'm interested in this, but you don't have any contact info in your profile. HN itself doesn't have a messaging feature, so you have to leave an email address or something there. If you'd rather just message me my address is in my profile.
No research here, but I use the great Fresh Cards app (by allenu here on HN) with my daughter's homeschooling, and it works great. We've used it for phonics and all sorts of basic math facts, and I use it for scheduling spelling word review as well. We're about to start teaching science and I anticipate using it a lot there as well.
Hundreds of millions of the world's kids were remote learning for up to a year and a half. If there was any opportunity to develop real tools to help with remote learning, it was then, but we ended up with nothing, and remote learning still sucks.
If anybody is going to integrate memory techniques into a curriculum, it'll probably be some Silicon Valley charter.
Given that so much of modern education happens in large-group settings, it seems hard to apply SRS principles to it. A group of students will be a lot farther from 100% recall of all items than any individual student might be, and the items that require review will correspondingly differ.
Spaced repetition could replace time currently wasted on homework. As a kid I would have much preferred to sit in front of Anki instead of being laden with hours of homework, and it probably would have done me more good over the long term.
> but on the other hand the best researchers/engineers I know have a huge number of facts and examples memorized and ready at their fingertips
The US is staunchly anti-memorization. It wasn't until I did a year abroad in a French university where the power of memorization was so obvious. One of the math students I spent time with told me it's the French tradition to "carry everything around in your head". They were studying symplectic geometry in their sophomore year of their bachelor's!
I don't understand where this coming from. In my experience learning and teaching math in the US, I've found it's almost _entirely_ about memorization. Calculus students are usually able to regurgitate algorithms for differentiation/integration, but can't answer the most basic questions about what they're doing really means.
This is just an aside. I'm not against memorization. I think memorization is the first step necessary for deeper internalized understanding since it lets you take the basic stuff and put it on autopilot so you can think about the rest that isn't as basic. That said my main complaint about US teaching of math is that that second step seems rarely to be taken and instead focus entirely on the memorization.
I guess our experiences are both anecdotal, but I'm having trouble wrapping my head around your idea that the US is anti-memorization.
This might be the case for k-12 and maybe lower level courses, but this was _definitely_ not the case for when I was taking group theory, real analysis, etc.
With that said, someone else in the comments mentioned that while it might seem there's an emphasis "regurgitation" it's mainly because going beyond that and really digging into the why of things is vastly out of scope for the class.
In grad school it seems obvious what you want to retain, after that it can become a bit trickier to make the choice. General question for everyone using this approach: how do you go about selecting topics to put into your SR schedule?
I mean there are so many things that would be good to remember: math problems and leetcode are actually continuously relevant if you're in employment and ever thinking about moving jobs, software engineering tools that are important for daily work (those are changing so fast in some cases that I'm wondering again how much is worth committing to long-term memory), maybe there are some academic fields where you want to stay up-to-date on the research, all sorts of life hacks in the household / cooking / DIY space, literally everything about your family life, maybe some strategies in hobbies (just recently got into speedcubing, which is a lot about algorithm memorization),...
Maybe the problem is just that I have too much of a "finite memory" mindset and we can easily remember all of those, lol.
The problem is quickly solved, and it is important not to fall into " paralysis by analysis." Sometimes, while writing a card, we tell ourselves that it is not so important to remember that information and discard it, other times we discard it when during the review process we sense or realize that we are not interested in remembering that information .
Assuming that there is some sort of logic behind the decisions made, I recommend using the same strategy that Napoleon is said to have used in his campaigns: first we start, then we will see.
Yeah, the whole thing is probably easy to overoptimize. Any recommendations wrt to the software? My preference for these sort of things are maximally compatible/portable solutions based on simplistic file formats (I keep a lot of personal tracking data in csv, eg). Here I'm thinking something along the lines of local markdown files for the cards, a database for the sampling info and a web app that does the sampling.
I use Anki in a "straightforward" way. In the sense that I write the card using Anki's interface without using excel or markdown files.
For example, if I want to memorize a word I found while reading, I highlight it in blue in Apple's books, it goes to readwise, then I take it from readwise by copying and pasting from Apple's dictionary the translation or definition.
I am for the simplest things possible, the point is to learn/memorize and not to find the best pipeline. You may think that the optimal pipeline leads to optimal learning, but (1) what kind of difference are we talking about compared to the simplest method of building cards?, (2) if setting up the pipeline means having the pipeline as the goal (as in ML the platform and not the predictions become the goal) or abandoning the study because the whole process becomes too cumbersome, is that something we should aspire to?
One of my favorite novelists, Arturo Pérez-Reverte, when asked what software he uses to write his novels, replied: "I only use Word, but it's true that I don't know any other [software]".
I found this post fascinating. It makes me wonder is there an equivalent superpower that's more software engineering focused rather than coding focused like leetcode questions? I know everyone here likes to dump on Gang of Four, but perhaps design patterns and their usecases?
If I'm being honest, I went overboard. Like I said, I literally memorized entire books. Entering cards into the system was the most difficult part. But it absolutely works! Had I stayed in academia, I would have loved to developed a set of flash cards to go along with course material.
To really understand something first you blindly memorize - but that is not enough, once you do examples and exercises using what you blindly memorized you get to understand things quicker, a lot quicker.
Not memorizing stuff and figuring things as you go is mostly recipe for disappointment.
Like in chess - people think that chess players are somehow super intelligent - but being super intelligent without rote memorization of loads of chess settings will not help winning grand master title.
There's actually an interesting section of the book "Moonwalking with Einstein" where it talks about a study was done on chess masters where they showed them the board in positions that would be impossible under the rules and suddenly the master chess players didn't do all that much better than random people.
The suggestion in the book was that really Grandmasters have spent so much time practicing, that they have memorized the game and the board to a certain extant that allows them to more easily handle the board and all the pieces on it cognitively.
Off-topic, but I dislike chess and adore chess960/fischerrandom chess for essentially this reason. It's almost disheartening enough to want a different hobby when you review an online game and realize what you thought was a clever solution to an interesting "puzzle" of a board position was really just one you learned by experience a week ago. More variety in piece arrangement makes playing feel far more like doing chess problem solving than remembering the last time you messed up the same chess problem.
Memorization only works well if your approach leads to chunking. They have done studies on chess players specifically, which supports what you are saying. Masters are no better than a beginner at memorizing a random chess board. This critical mass of memorization, when finally encoded into chunks is what I think a lot of people are describing as the “aha” of memorizing to conquer complex topics. So it isn’t really the blind memorization, but the process of learning and getting information and concepts into coherent and related chunks.
It similar to the question, "if you could do just one exercise, what would you do?" And you have all these answers, it is the squat, no, it is the deadlift, please guys it is the power clean.
It is not a useful question because you will never be in the position of choosing just one exercise to be done for the next month, year or decade.
Memorization is important and understanding is important, and the two are not in any conflict whatsoever.
You are making a very important assumption that doesn't hold in most situations.
Chess and math are regular domains. They are bounded. There is right and wrong. Most domains are not like this. Ex: art, software engineering, investing, etc.
You can't memorize your way to understanding in an irregular domain. It also doesn't tend to do well in non-knowledge work domains. Ex: sports.
I disagree becuse rote repetition is the only way to learn how to draw. One cannot simply draw something complex without trying out again and again to draw something.
For sports rote repetition is basis to excel or at least getting good at any sport.
I agree one cannot "memorize to understanding" because one has to actually apply what he memorized in practice.
I've picked up a lot of (human) languages over my life, and I've used SRS to great effect (I started on JMemorize, a now defunct Java app.) I have a simple strategy:
1. Read a text
2. Lookup unknown word/grammatical pattern
3. Create a flashcard for it
4. Apply SRS
and it works great. I've also tried to use other people's Anki decks and they've never worked particularly well. Personally engaging with the material is still the prerequisite for memorizing it, but memorizing means you don't need to struggle to figure out basic concepts constantly and can instead move onto the higher level of meaning.
I've used SRS over the years for many, many things. I've used it for memorizing divisibility rules, used it for annoying math lemmas, used it for data structures, and more. Each of the time I've attempted the material by myself and then turned my knowledge/engagement into a flashcard. I've even considered using it to learn tools like Blender so I can dial in workflows.
All of the courses that I did really well in I memorised the entirety of the book that was the core of the course. Everywhere where I didn't feel like doing it, I was strictly mediocre. If you think you can understand something without knowing all about it, I must say I think you are mistaken.
My (entirely uninformed… though I am a “doctor” har har) theory is that some sort of compression occurs in the brain. I also wrote out the answers to my flash cards by hand. I feel like these actions accelerate the compression process.
I've really understood a lot of stuff that I've since completely forgotten. For example, if you don't speak your native language for 20 years, you'll be surprised how hard it is to find any of it when you need it.
I imagine it's possible that they are related. In my life, I've only witnessed it as a required for being an expert at any topic: experience and memorized knowledge, of a topic.
Based on my own experience, a caveat is that you need to be applying what you're learning outside of flashcards. Otherwise, it's like your brain has different memory banks for different contexts, and you'll remember everything when using your flashcards, but a real-world scenario will come up and you'll forget.
I went from zero Japanese to JLPT N2 in just a few years and I attribute a great deal of that to Anki. However, I was also frequently talking in Japanese, and reading/writing in Japanese. I was constantly using what I was learning.
Maybe an analogy is, Anki is like steroids, but you still need to go to the gym. :P
I don't study Japanese at all anymore and I have very little exposure to the language. But I still do my flashcards. I've noticed my recall is excellent when doing the flashcards, but if I see some Japanese written on a billboard, or think to myself "how do I say this again...", I struggle to recall even things that are in my flashcard packs.
In my opinion the memorization (with whatever method) gets you started. There is no good replacement, and in that phase any other learning activities really aren't as important.
Only when memorization has laid the basic foundation does it make sense to spend more time to read and write in the new language. In my opinion reading is the most efficient activity at higher levels.
I disagree. Even at the early stages you should be reading simple sentences with words you just learned and you should be creating sentences with structures you just learned.
I didn't say you shouldn't read simple sentences or do other activities. Variety is useful.
I said early on the most efficient use of your time is memorization of words and maybe grammar. Duolingo does that well, in my opinion. It doesn't do the usual flashcard stuff, but it does favor repetition and you can practice speaking and listening from the first time.
I've been using Anki for language learning and it's a superpower. After just over two years I passed an exam that requires a passive vocabulary of about 10,000 words. Other people who grinded harder than me have managed it even faster, some in under a year.
However one big mistake people make is to think an SRS helps you learn. That's not true. It helps you not forget things you've already learned. You still need some real world interaction with the material.
>> It helps you not forget things you've already learned.
The argument is quite weak when we discuss learning words in a foreign language.
You learn that lunch is "almuerzo" in Spanish, but is it learning or just memorization/association? To me the lines are blurred, even at first sight of the word.
I used Anki myself to learn foreign languages and it is tremendously effective for words and short sentences too. I use it also for words in my mother tongue, and it made be much more articulate than I used to be.
I did Anki religiously every day for 2 years, I managed to internalize a bunch of facts about Japanese and no skills. I knew tons of vocab words and grammar points but I couldn't make a sentence in the heat of the moment to save my life. I know memorization is a part of the game, but it's not the whole game. It's like learning a bunch of music theory without actually touching an instrument.
I am learning Mandarin now and my approach is completely different. Speaking and listening are the most important, I'm letting muscle memory of mouth (and sometimes gesturing tones with my hand) be my memorization technique. Six months-ish in and I can already express myself in rudimentary ways, it's way better than my Japanese was in that time frame. I have some Mandarin Anki decks that I do in the cracks of my day but I never finish my daily cards anymore and I don't spend a lot of time on it. I'll schedule an italki lesson or listen some speak-and-repeat courses instead.
Memorizing the notes in a song isn't learning to play music, memorizing a bunch of vocabulary isn't learning a foreign language.
I agree with the comment, but my point was about understanding and memorizing words, not becoming fluent in a language.
Listening, speaking, and memorizing words and grammar rules are all necessary to become passable or fluent in a foreign language. I always jot down words and strive to memorize the phrases or idioms of native speakers. Then, there are ups and downs, one day I sound like Obama, another day I sound like someone who started learning the language two weeks ago. It's all part of the business.
>You learn that lunch is "almuerzo" in Spanish, but is it learning or just memorization/association? To me the lines are blurred, even at first sight of the word.
I think there's a difference. "Almuerzo" -> "al muerzo", I know "al" means "the", from Arabic, I look up the etymology, the "muerzo" is from Latin "morsus", from which we get "morsel" in English. Mnemonic: "the morsel". Then I can guess that the form of the verb is almorzar, which is indeed the case. Even learning a simple word embeds and thickens a web of associations in one's mind. SRS strengthens this, makes it scale better.
It is interesting from a linguistic point of view, no doubt, and I am an amateur etymologist myself, but let's look at reality with honest eyes: apart from a very small group of people, languages are not learnt, and learnt effectively, starting from the etymology up. I know it is almuerzo, I use it 20 times, and it is "mine".
I think we should try, and I include myself among those who have an approach to learning that is at times too cerebral, to look at how things work more than how interesting would be if they worked in a particular way.
This is language dependent. For learning Chinese, recognizing radicals (like waht represents water, and what represents tree) can go a long way in helping recall of both meaning and pronunciation. Then moving on to learning Korean, where Chinese characters make up ~25% of Korean words, recognizing the Chinese character, then recognizing the typical phonetic alterations to Chinese pronunciations, can go a long way in helping recall of both meaning and pronunciation.
I am still skeptical. One thing is to recognize common roots, for example "to live" is "vivere" in Italian and "vivir" in Spanish, another it is to know that they are both coming from the latin verb vīvĕre.
It is sufficient to recognize the commonality, without getting to the etymology of the word, which, in practical term, does not help.
Etymology by itself is not going to be useful. As one more flashcard, another pillar, additional context, they will to deepen your understanding. It is creating more inter-connections between your neurons.
I think we should not confuse the process of hypothesis building with hypothesis testing.
>> As one more flashcard, another pillar, additional context, they will to deepen your understanding. It is creating more inter-connections between your neurons.
This is a hypothesis, identifying a mechanism that leads to an effect. But I don't see any tested hypothesis for either the mechanism or the effect. And we have millions of people who have mastered other languages as children and as adults who have no idea what etymology means.
I mean, it sounds reasonable, in theory. Like knowing the evolution of tactical formations in soccer would like one to become a better player. But it is common sense it has no effect whatsoever on soccer abilities.
Oooh, this looks lovely, thank you for showing me this.
EDIT, hm he seems to be missing some obvious cognates, e.g.
>propio, own (adj.); proper (cognate). Proper in modern English no longer has the meaning "own", but this word in Spanish, propio, still retains it. if this knowledge is too uninteresting, use a mnemonic such as "A truly proper lady behaves in her own natural style."
wtf? What about just the English word "property" instead of this silly lady business?
There's something to be said for using strange or even ridiculous mnemonics as a memory aid.
I'd often come up with little stories to teach myself some of the more complex Chinese characters when I was learning quite intensively and found I never struggled to remember those words after doing so.
To use a language you need "over-learning". It's not enough to be able to recall a word.
You need to be able to recall it very fast, and that only comes through repetition beyond the "I won't ever forget this" stage. SRS doesn't make a statement about this, and you can still use it to achieve that over-learning, in my opinion. It may be better to achieve it through reading texts in the foreign language because the repetitions are faster.
One way you can achieve this using SRS is forcing very short recall times. I average under 2 seconds a card in review: my view is that if I can't recall it in under 2 seconds (what would be realistic in a real-life scenario when speaking a language) then I don't know it. It's a method that's trained me to be very fast at recalling words when needed.
Yeah, it's something I can time myself: I'm on about 6 years of daily usage at this point so it comes very easily. Anki also lets you set a time-out on cards as well. I'll know just from looking at the front of the card whether or not I know the word. If I do, space then 'good'; if I don't, space then 'again'. Very rarely I'll know a card but only after a longer moment to think about it - that's a 'hard' card.
But people get spaced repetition wrong. Most people think you need a spaced-repetition schedule for long term memory of facts. You really don't. The only proposition of spaced repetition is that you spend less time studying when using this schedule than with a more dense repetition schedule, if both lead to the same long term memory effect.
Another schedule might be to review your complete vocabulary list every day. This will also lead you to long term memory, eventually. Earlier probably (in terms of calendar time). But you'd spend more hands-on-time studying. However, you'd much prefer the former method over the latter when preparing for an exam on a specific date.
Also: formal methods of managing the schedule aren't really necessary. Reading in a foreign language counts as repetition of the words. Exposure to the language and the vocabulary counts, for example what Duolingo does with its various activities is partly spaced repetition, but less formally so.
This is false if memory's are stored physically in the brain. Unless there is an increase in brain volume, this process will eventually hit the Bekenstein bound. I don't care if the upper bound is "effectively infinite," that's not what the proof claimed.
Proofs rely on assumptions. In this case, they state outright:
> We first posit that the number of days T that a fact can be retained before it needs to be reviewed grows as a power-law in s, the number of times it’s been reviewed so far, ...
Obviously this assumption will be false in our physical universe, but that doesn't make the proof itself invalid (edited).
In logic, the soundness of a proof in fact has to do with its interpretation in some universe of discourse. To be sound, the argument has to be deductively valid, and its premises have to have true interpretations in the chosen world where it is applied.
Here we have a valid mathematical argument which is unsound in this world, where its assumptions do not hold up.
You're right. I should have said invalid rather than unsound. I still don't see the point of the criticism, however. Lots of interesting things can be learned by starting from approximations to actual reality.
This is what I have been thinking. If whatever you can remember is bigger than all experiences and lessons you could remember, then your memory is effectively infinite. I don't know if having a continuous effectively infinite memory is good or bad. As someone who play the classical guitar, having a better memory would do wonders for me.
What you're missing is that the proof only shows that memory diverges as time goes to infinity. According to the laws derived there, a finitely-lived human would have a finite memory throughout their entire lifespan. You could even use the law to derive the upper limit on how many facts can be memorized over an X year lifespan (which sounds quite antithetical to the headline, which you seem to have reacted to).
> I don't care if the upper bound is "effectively infinite," that's not what the proof claimed.
Very interesting. Do you typically process all claims as literally as you just did for this one, notably when you're interacting with people outside of the Internet? In which case, do most people react positively to your behaviour, or do they get annoyed because "you know what I meant"?
I see this argument as semi-fallacious, since it's really asking "can you recall infinite things on an infinite timescale using this method"? Sure, the math might add up, but humans are reliably fallible. I do agree, from practice, that practicing spaced repetition can increase the breadth of knowledge beyond that of "casual learning", but I've seen my brain fail on aged entries too.
Nicky Case's intro to spaced repetition is pretty stellar, both as a crash course and the reasoning behind it: https://ncase.me/remember/
The argument is wholely fallacious and the author knows it and is just being silly. The entire argument is based on assumptions that only hold for smallish numbers of items. It we get to larger finite numbers, such as the number of possible states of a human brain, those assumptions are clearly false.
More than semi-fallacious - the math doesn't add up, because you run into the Bekenstein bound, which limits how much information you can pack into a volume.
This is why I don't use spaced repetition. The danger of becoming a black hole is just too high. Forgetting things is well known to be highly evolutionary advantageous, because all the critters that remember everything and turned into black holes stopped reproducing. Very dangerous stuff to play with.
Ha! Certainly none of us will get anywhere close to this bound, no matter how much time we log in Anki. But the OP asked "Would an infinitely-long-lived, but forgetful person be able to recall an infinite number of facts using this method?", and the answer is a surprising "no, you turn into a black hole".
I know we're all playing around here, but surely the answer must be that there's a limit rooted in the specific biological/chemical implementation of memory in the brain that our hypothetical "infinitely-long-lived, but forgetful person" would hit before the Bekenstein bound (probably long before) and therefore it's impossible to turn a brain into a black hole via that mechanism.
I guess I’m the extreme case then; I’ve been using SuperMemo for 16 years, every single day; originally it was to memorize foreign language vocabulary words, but has since expanded to everything I want to know. Left unchecked you’re right, you can easily become obsessed with memorizing junk knowledge and becoming like the comic book guy from the Simpson’s. But over time it has helped me learn how to “compact” complicated ideas into simple statements and more easily internalize new ideas. Since I’ve been using SuperMemo to remember concepts, metaphors and illustrations kind of just “appear” in my brain. I don’t fully understand why, I think it’s like lightning striking a place with lots of metal; you can’t force it to happen, but you can make the conditions conducive to “lightning strikes” of inspiration.
Tragic really how smart people know perfectly well that you shouldn’t exercise if you don’t want to accidentally turn into a grotesque muscle freak, but rarely use the same caution in brain exercise.
The more you learn the more you gain the ability to encode new facts with fewer "bits." Thanks to associative memory. Adults have an advantage in learning in this regard.
Not to detract, but there was an article on SRS that I think I came across on HN about 2-3 months ago (though I guess may have been Reddit); I've been struggling to find it for the last month or so.
It was a fairly long article, probably a 20-30 minute read, dark background (though I guess this may have been dynamic since I was reading at night), with a broad overview of SRS, several studies cited, and a personal anecdote from someone with somewhere around 10k cards that the author reviewed daily.
I wish I could remember more details. I've combed through my browser history in all browsers (desktop and mobile) and have tried a few site-specific timeframe-limited searches with no luck. Does this article sound familiar to anyone? It was a good read and somehow I failed to pinboard it.
Based on the mention of auto-dark-mode, probably. We put some effort into thinking about how to do dark mode right and integrate with system settings - your usual website hack of an invert toggle is wrong, but the correct auto/light/dark tri-state toggle is not obvious.
I'll have to look at the strategy you took. Thanks for the article! I now also remember the mention of Mnemosyne, which was the first SRS app I ever used (got me through medical school!). Now using RepetitionsApp x years.
I know this is just a joke, but there was an interesting Radiolab on memory and forgetting. It posited that forgetting is an active process, and analysis a person who was disabled in such a way that their forgetting process didn't work led to some very undesirable outcomes:
I mean sure. The trouble is in order to keep the review time per day fixed you need to vary the rate at which you introduce new facts, and as time goes on the interval between when you can afford to introduce new facts will grow to be infinite.
The model also doesn't account for how you feel about doing this activity. This includes the pain of not doing reviews for several days and then having to catch up. The probability you eventually throw in the towel is also a function of how useful you're finding the activity, which in this case since we've opted to keep the per-day study time as fixed will likely be when the interval that you are required to wait before you can add a new fact becomes so large you are routinely bothered by occurrences where you encounter a fact you wanted to put into your SRS but didn't have the bandwidth to.
Source: input 35 new vocab a day into Anki while learning Japanese a decade ago and grew my vocab deck to 18000+ cards. Reviews would take several hours a day, and at one point I stopped adding new words to try and wait for the daily review load to go down. When it never did despite this, I just deleted the entire thing and say fuck it.
That growing backlog is a big issue with this type of scheduling. I think it's one of those almost "unexpected" side-effects of spaced repetition. I mean, if you did the math up front, you should know generally how busy you're going to be over the next few weeks based on how many cards you're reviewing per day, how many new ones are coming, and how you're scoring with each card, but it's not really something people do, nor does the UI show you that.
I wrote my own spaced repetition app (see my profile) and one thing I experimented with is an "ETA" function which goes through your deck and tells you how long before you learn everything "well enough", where "well enough" just means the intervals between all cards is at least some fixed threshold, like one month. I was surprised that even with a few hundred cards and a fixed number of cards reviewed per day, it would still take a few months.
I haven't added the ETA feature yet, but if I do, I think I would need to make sure to show the user how adding new cards affects that target date, and also how reviewing more cards a day or fewer cards a day affects it. I think there's a way to design a nice UI so users feel like they are in control of that end date so they can choose if it's worth it to add new cards or not.
I think a spaced repetition app should also make "falling behind" not feel catastrophic. I think memorizing a smaller set of facts consistently is better than trying to memorize a larger set and then giving up, and an app's UI can probably help to that end.
> I think memorizing a smaller set of facts consistently is better
In general, existing SRS does a very bad job of supporting priority and prerequisites between items. There's no way of saying "learn A, B, C, then learn X, Y, Z once you know the former well enough." This limits the domains where SRS can be applied most effectively to ones with very little structure, such as raw language vocabulary.
> In general, existing SRS does a very bad job of supporting priority and prerequisites between items.
This is a feature I've wanted to implement for a while, but I think it would be hard from a user's perspective to set up the dependencies. Specifically, I mean marking a card unlearnable until another card (or cards) is learned. It's something that I think requires a good UI so you can see at a glance which cards depend on which other cards. You'd go in and essentially make a skill tree of all the cards. It might be something more useful for more serious deck creators.
My dream is being able to create a deck where you have both vocab and then sentences that use the vocab. You'd make sentences that use the vocab dependent on the vocab being learned first, so that over time you'd gradually unlock more material that uses your base knowledge.
>My dream is being able to create a deck where you have both vocab and then sentences that use the vocab. You'd make sentences that use the vocab dependent on the vocab being learned first, so that over time you'd gradually unlock more material that uses your base knowledge.
Pretty unnecessary in my experience. I only did around 2k sentences and that was for the express purpose of internalizing grammar, and found that 99 times out of 100 it was pretty natural to just learn the new vocab that came with the sentence. There would be the occasional one that would take a few times before I got it, but it was pretty insignificant problem.
After I smashed out all the grammar I just did purely vocab after that and it worked like an absolute charm.
> My dream is being able to create a deck where you have both vocab and then sentences that use the vocab. You'd make sentences that use the vocab dependent on the vocab being learned first, so that over time you'd gradually unlock more material that uses your base knowledge.
You can also go the other way. Have a huge database of sentences, and pick example sentences for new cards which only use vocabulary which you already know (besides the new word which you want to learn through that card, of course). I do that for my Japanese SRS app, and it works decently well from the feedback I get from my users.
Where this feature would be especially useful too is idioms and collocations ("words best used together"), not just full sentences. It's something that's not often addressed in formal language learning but a big part of native fluency. And yes, it's a lot of data. Though it could be made manageable by e.g. importing it from external knowledge graphs/mindmaps and the like.
In my experience, the brain collects background statistics on collocations through exposure to native media. It's an aspect of language learning I'm not convinced SRS is suited to or really appropriate for.
> Source: input 35 new vocab a day into Anki while learning Japanese a decade ago and grew my vocab deck to 18000+ cards. Reviews would take several hours a day, and at one point I stopped adding new words to try and wait for the daily review load to go down. When it never did despite this, I just deleted the entire thing and say fuck it.
Yeah, but that isn't really inherent to the concept of SRS; it's just an unfortunate consequence of Anki's geriatric SRS algorithm.
For reference, for my Japanese SRS app (which is using my own custom SRS algorithm) I have a user who currently maintains ~18k known words, with a consistent ~23 new cards per day for months, and they don't take that long with only something like ~20 minutes spent per day reviewing.
Yeah, I've got 9000 reviews waiting for me in my old Chinese Anki deck, at some point you always fall off the wagon. I've changed my mind several times on SRS going back and forth, and my conclusion is that, for language, I think it's better to split your time with 90% on consuming meaningful input and 10% SRS, because learning in context with real content that you're interested in allows you to encode your memories much more efficiently. The problem with pairing available SRS systems with consuming real content is that they don't take into account "natural" repetitions, e.g. when you encountered a word in a movie outside the SRS, so they tend to underestimate how well you know a word. But also since "natural" repetitions are baked into the SRS model via parameters that fit the "average", it also tends to overestimate your knowledge of a lot of other items. Combined, you just have a very inefficient schedule that takes too much of your time for too little gain
I did the AJATT method way back in the day when it was relatively new. So heavily leaned on Anki, whilst also consuming loads of natural Japanese through music, shows, games, conversation etc.
In my experience doing both in tandem was the magic combination. Anki without native media is mildly pointless. Native media without Anki is super inefficient.
Used in equal measure you find your access and understanding of native media continually ratchets up day by day every 3000 cards I would notice a very significant jump in my abilities when consuming native media and partaking in conversation.
Nothing else has really come close for me out of all the other things I've tried.
That's essentially what I tried with Chinese, and I could sort of keep up with it before I had a kid but then I just can't find the time/energy to keep the backlog down AND watch new content, so for several years I stopped learning completely.
I also want to minimize time spent in Anki because it's just not fun, you know? I know common wisdom says it _should_ be painful, that means you're learning, but I'm not so sure. Me and everyone I know learned English without much pain or Spaced Repetition Systems after all. It's a much easier language to learn than Japanese/Chinese though, but it makes me feel it shouldn't have to be this difficult.
A prioritized queue of stuff to learn may be a better solution than staring at "over 9000" things to review. Such a queue would mean you simply bite off as much as you can chew every day and don't perpetually fall behind, assuming the priority queue has a small degree of randomness and is recomputed daily based on the decay rate of all modeled facts. Do you agree?
I think it can help to keep things less overwhelming.
But what I'm trying now is to ruthlessly suspend any card that I get wrong even once after the learning phase (i.e. the day after). I've noticed that some words stick much easier than others, there might be an 80/20 kind of distribution where 20% of the words are responsible for 80% of the time spent. In that case it might be better to add a bunch of words every day and then suspend the ones that don't stick very easily. The idea being that they'll stick more easily at a later point, at least this has been my experience with Chinese.
Actually I'm going to dig through my Anki database and see if my hunch is right.
Edit: it's more like 25% of the cards are responsible for close to 45% of reps. But 25% of the cards are responsible for 58% of review time, and then it's 36/70. That indicates to me that dropping the something like 30% hardest cards would save you a disproportionate amount of time. Of course there's some nuance here where I should probably do this per card type, otherwise you'd just drop the hardest card types.
The rate at which you can add new items is the inverse of the review period. What you're effectively doing is increasing the review interval, but you will still run into the fact that the number of items you can remove from the queue is finite. Eventually you will have only interesting items on the end of the queue equal to the number of items you can review. You're now back at square one: you don't have the time to add any new items.
The only way to deal with this is to review items a finite number of times and hope for the best.
When you're working on clearing away a backlog, your review periods should get bigger because Anki will give you "credit" for recalling these items longer. So it seems like reviews should be manageable as long as you're not forgetting too much. And Anki will always prioritize the items you're at highest risk of forgetting.
Not quite. The review intervals increase exponentially, so if you stop adding new cards the workload quickly goes to single digits per day in a few months.
35 a day is excessive for sure. Even with foreign language stuff which is the lowest time to solve per card, I limit myself to 10 new cards/day. Math problems 2 new/day. Daily workload ends up around 50-100 cards, which is very manageable. You do need to be consistent though.
35 a day was my magic number. The method to the madness was that I'd allow for only having to hit it at least 80% of the time. In a given month if I managed it every day it meant 1000 cards a month. Every 3 months I leveled up pretty hard.
In my first year studying Japanese I first did Remembering The Kanji, which was a slow to get through and I didn't manage that pace really due to several periods of starting / stopping. After I got through it though I knew bugger all Japanese functionally speaking but the language became extremely accessible due to having a solid command of recognizing Kanji and being able to look them up easily when needing to read stuff, so I decided to front load grammar after that. At the time the JLPT was about 4 months away and I wouldn't have even passed JLPT5 (well, JLPT4 back then) at that time. I found a really amazing Anki deck that I don't know if you can find anymore called Kanji Oddysee 2001. It contained 2000 sentences which covered up to I would say around about JLPT2 grammar as well as teaching vocab + readings for the 2000~ Joyou Kanji (hence the name). The cards also had audio as well. So given 4 months and 2000 cards with 2 card types I smashed out 4000 cards and sat the JLPT2 after just under a year of study. I think I needed 60% to pass at the time and got like 55%? So, not a pass but I took it more as an impossible challenge than anything, and was stunningly happy with the result.
After that I had a vocab of around a couple thousand words, a decent command of grammar, and some limited ability to converse. I just turned my efforts towards building a vocab deck at 35 words a day. I managed to add another 10k vocab in that second year after which point conversation was fluid and I was reading novels (still needed a dictionary for that, but it was huge at the time). I got in some Japanese room mates too and we had loads of parties.
I frequently have arguments with competent mathematicians about the role of rote learning in arithmetic. I think when you move "beyond" arithmetic, you can also forget how important number recall is, in being able to manipulate numeric symbols fluently according to the rules of arithmetic.
Include base (hex, binary) and .. its getting worse.
Numeracy, numerical literacy and fluency, understanding both the symbol set, and its behaviour in rules of arithmetic, needs to be a given, to talk cogently about numbers.
Simply teaching all kids the 12x tables, and how to handle the arithmetic of change giving at a till from the specific units of coin and note available, how to run a total in their head, how to calculate % will be life skills which might literally save their job. And its foundation is recall, and rote learning.
Not everyone is heading to 3rd order differential analysis.
I have read a lot of comments that appear to be a modification of "I used anki but the accumulation made me drop it".
If you're interested, I've worked in a Spaced Repetition Algorithm where they is no card accumulation, you review whenever you feel like. In order to maximize pleasure.
In the future I'll also configure it to give you subject you "feel" like reviewing at the moment, since pleasure is also associated with memory.
This 'proof' seems a little suspect. It assumes that the rules they have discovered hold for all situations. Even if they have tested a large number of facts to remember, there is no reason to believe their isn't some threshold where it doesn't hold anymore.
Also, they seem to be holding this learning technique like it is a mathematical fact of the universe. I am not an expert, but nothing in the human brain works that consistently. I can't imagine every single person in the world can learn equally well with this technique.
Although this is likely some form of satire, because there is no way anyone could seriously think the brain works like this.
It's cute math, but doesn't take into account the inexplicable differences between the retention of seemingly similar facts. Some items stick very well, whereas others suffer from lapses.
In my Ankid ecks, I have items that have 10+ year intervals. And some, which were introduced at around the same time as those, which have intervals in months, beaten down by lapses.
Anyone who thinks that some simple math leads to infinite recall has not suddenly lapsed on a card whose interval had reached 7 years.
I don't recommend allowing Anki items to have an interval of over one year because the crude SM-2 algorithm does not capture the per-item effect of exponential decay in memory. Multiplying interval lengths by per-item ease factors is suboptimal and not the same. Unlike, for instance, ACT-R, which unfortunately has no good publicly available user interface. A good implementation of a neural memory fact model will track strengths of associative memories in addition to the per-item exponential forgetting curve. Keep the Anki intervals capped at one year until something better comes along.
Problem is it seems you're then throwing under the bus all those items you recall perfectly well that have 6+ year intervals, which don't require yearly review. Arguably, you could just retire those; but then sometimes they lapse, as I noted: how do you decide what to retire?
there are only a limited amount of things you can memorize because your lifetime is also limited.
Piotr Wozniak the creator of supermemo, the first spaced repetition program, talks about it on his site:
https://supermemo.guru/wiki/How_much_knowledge_can_human_bra...
I'll just throw out I am a big fan of spaced repetition I highly recommend it to everyone. For those complaining about "infinite recall" you aren't wrong, the pioneer of this movement Piotr Wozniak, even has an article on his wiki (https://supermemo.guru/wiki/Piotr_Wozniak) about how the upper bound of languages you can learn fluently in a lifetime is probably somewhere near 5. But the point isn't to remember everything forever, the purpose is to help you learn better.
My flow right now for learning things is.
1. Find sources copy and paste large swathes of revelevant text and images.
2. Re-read the copy and pasted text and create a detailed summary of it in my own words.
3. Come back a few hours or day later and summarize the summary.
4. Use this as the basis for cards to load into Anki.
It isn't about building a massive repository of facts, and you can do plenty with just steps 1-3 without ever using Spaced Repetition, but the reason I fell in love with spaced repetition and have jumped on it so heavily is that I've done steps 1-3 with a lot of information and subjects, and over time have forgotten all but the most basic things about them. This makes me feel as if part of my time or life was wasted, because if I have to revist something again latter like Sorting Algorithms it feels like starting over. Whereas things I have started to use spaced repetition with, I retain the fundamentals the "outline" of the subject for much longer, and if I have to revisit it I feel much more familiar because to paraphrase Piotr Wozniak. The things we remember well are things that are well located/connected within our knowledge tree.
For those IT people out there as well the other thing spaced repetition and especially Anki is super useful for is learning how to use your tools more effectively because it helps you to remember those features and tricks that you don't use often but super speed things up. For example I used grep for a long time, I often found myself having to hit up the man page, or the DDG if I needed to do something unusual, or more often I'd end up trying to cobble something together with the tools I had. I reviewed a "most useful flags" in grep page a few months back and decided to Ankify it. I am now an order of magnitude more proficient with grep because I can quickly recall the flag or option I need to provide to do something wonky with it when I need to, simply because I remember a relatively obscure feature, that I don't use often and would've forgotten otherwise.
Finally in conclusion
The 4 states that made up the Austro-Hungarian Empire were
Boznia-Herzgovania, Croatia-Slavonia, The Kingdom of Hungry and Cieslenthia. Because sometimes memorizing one or two random facts just makes life more interesting.
That's not how I read it. And in any case, that is a common attitude/misconception about spaced repetition, which disillusions many people after they realise they still need to put the effort in to make it work, so it's worth pointing out.
The qualifying exams were difficult until I reached some "critical mass" of knowledge. Then I could regurgitate proofs and even attack novel problems easily.
There's an analogy here somewhere to the "leetcode" style of software engineer interview. On one hand qualifying exams and leetcode questions are a stupid gatekeeping mechanism, but on the other hand the best researchers/engineers I know have a huge number of facts and examples memorized and ready at their fingertips. I didn't think I needed to do so, but perhaps there is something to suffering through the rote memorization phase to make what comes next that much easier.
[0] https://www.wired.com/2008/04/ff-wozniak/