# Article - Augmenting Long-term Memory > Source: https://ollybritton.com/notes/articles/augmenting-long-term-memory/ · Updated: 2024-10-17 · Tags: notes, safe-to-post-online > This entry focuses on the notes and observations around Michael Nielsen's essay ["Augmenting Long-term Memory"](http://augmentingcognition.com/ltm.html). ### Introduction * Attempts to augment memory using computers have been around for a long time: * *Memex* in 1945. * Even the internet itself was due to Tim Berner-Lee's employer seeking to develop a "collective institutional memory". * Anki is a modern attempt at augmenting memory. * "Anki makes memory a choice, rather than a haphazard event, to be left to chance". * "I'll discuss how to use Anki to understand research papers, books and much else." * One of the biggest concerns I have personally about using Anki is that you can use it to augment *knowledge*, but not *understanding*. * "the essay [is] a how-to guide aimed at helping develop virtuso skills with personal memory systems". ### The Anki System * Anki is better than conventional flashcards because it manages the review schedule. * Reduces the time learning each card: 4-7 minutes per card with Anki, over 2 hours with normal flashcards. * Two rules of thumb: * If memorising a fact seems worth 10 minutes of time in the future, then I'll do it. * If a fact seems striking then it should go into Anki regardless of the time cost. * What can Anki be used for: * Learn papers and books * Learn from talks and conferences * Recall interesting things learned in conversation * Remember key observations made while doing everyday work * Facts relevant to social life * City and travel * Hobbies * Made 10,000 cards in 2.5 years of regular use. Takes about 15-20 minutes a day. * If it goes about 20 minutes a day, slow down. * Review anki cards while: * Walking to get morning coffee * Waiting in line * On tranist * The review experience can be meditative if your mind is relaxed ##### Why is Anki better than conventional flashcards?? Because it manages the review schedule. ##### How long does a conventional flashcard take to review over 20 years?? 2 hours. ##### How long does an Anki flashcard take to review over 20 years?? 4-7 minutes. ##### How much more efficient are Anki flashcards in comparison to conventional flashcards?? ~20x ##### What is a good rule for adding flashcards to Anki?? The fact should seem worth 10 minutes of time in the future. ##### When should a fact be added to Anki despite the time cost?? If the fact seems striking enough. #### Research paper in unfamiliar field * Uses example of reading AlphaGo paper * Skimmed the paper: * Important ideas and concepts, key techniques. "What were the names of the two main types of neural network AlphaGo used?" * Looking for basic facts that were easily understood. "What's the size of a Go board?" * Several rapid passes over thep aper. * After 5 or 6 passes, a 1-2 thorough reads: * Understand in detail * Knows the background context already * Could have used convential notes, but Anki gave confidence of retaining understanding in the long term. * Entire process took a couple days of time, spread over a few weeks: * A lot of work * Payoff meant he had a basic grounding in modern deep reinforcement learning * "I find Anki works much better when used in service to some personal creative project": * Not stuff like geography of Africa, learning about WWII * Intellectually appealing but no emotional investments * Cold and lifeless Anki questions, hard to connect to in a later review. * When in conncection with a personal creative project, it's easier to connect to the questions and answers emotionally. * It's tempting to use Anki cards to study in preparation for some hypothetical use, but it's better to use it as part of some creative project. ##### Should you skim read or deeply read a paper first?? Skim read. ##### How many skim reads should you do of a paper before moving on to a thorough read?? 5-6. ##### What two types of information are you looking for in a skim read?? * Important ideas and concepts * Basic facts that are easily understood ##### Why should you skim read a paper before doing a thorough read?? Because you will know some of the background context already. ##### Why should Anki primarily be used in service to a personal creative project?? It's easier to connect to the questions emotionally. ##### Why shouldn't you use Anki to "stockpile knowledge"?? It's intellectually appealing but you have no emotional investments. ##### What's some side effects of not having an emotional connection to Anki cards?? * Questions feel cold and lifeless * Harder to remember, feel like a burden ##### What's an example of using Anki to "stockpile knowledge"?? * Ultimate Geography * Learning about WWII * Anything that is pointless. #### Shallow reads of papers * 10-60 minutes on papers * Asses articles before you read them: * Does the article contain substantial insight or provocation relevant to the project? * New questions, new ideas, new methods, new results? * Anki flashcards about core claims, core questions and core ideas of the paper. * Helpful to extract from: * abstract, introduction, conclusion, figures and figure captions. * Will typically extract 5 to 20 Anki questions * Bad idea to extract less than 5 because it becomes an "isolated orphan in my memory". * Don't ankify misleading work * > If you feel you could easily find something more rewarding to read, switch over. It's worth deliberately practicing such switches, to avoid building a counter-productive habit of completionism in your reading. It's nearly always possible to read deeper into a paper, but that doesn't mean you can't easily be getting more value elsewhere. It's a failure mode to spend too long reading unimportant papers. #### Synoptic reading using Anki * Shallow reads and deep reads of papers are possible, you can also "read" the entire research literature of some field or subfield. * Instead of shallow-reading a large number of papers, engage deeply with key papers (like the AlphaGo paper) * 1 truly important paper, then 5-10 key papers in the field, shallower reads of a much larger number of less important papers. * Key papers: * Build up an overall picture of where the field is at * What progress looks like for the field * "synpotic reading" * From "How To Read A Book": Mortimer J Adler and Charles Van Doren * What's been done * What's yet to be done * Identify open problems * Questions that I'd like answered * Tricks and observations * Anki is most useful in exploring new areas, not the areas you're most comfortable with * Without a lot of drive, it's difficult to make materials stick * Anki helps create that drive: * Confident you will retain information * Make sense of what you will read * Reading like this makes it more pleasurable #### General patterns of Anki use * Make most Anki questions and answers as atomic as possible: * Breaking down one complex question into more atomic pieces makes a routinely wrong question into multiple routinely right questions. * Atomic questions can still involve complex, high-level concepts * Anki is best thought of as a virtuso skill, to be developed: * Anki itself is a very simple program * It requires skill and practice to use well * Anki isn't a tool for memorizing simple facts, it's a tool for understanding almost everything: * > "It's a common misconception that Anki is just for memorizing simple raw facts, things like vocabulary items and basic definitions." * Developing virtuoso skill means cultivating the ability to use it for types of understanding beyond basic facts * Break things into atomic facts * Build rich hierarchies of interconnections and integrative questions * Don't put in orpghan questions * > "It's too strong to say that to be a virtusos Anki user is to be a virtuoso in understanding. But there's some truth to it". * Use one big deck * Avoid orphan questions: * Grooming habits of the Albanian giant mongoose, interested in currently * Soon ankified 5-10 questions * However, in a few months the cards will be stale and you'll get them wrong. * The questions are disconnected from other interests * Lost the context that makes you interested. * _Orphan questions_ are questions not closely related to anything else in memory * It's not bad to have a few -- hard to know what will grow into a substantial interest. * Never add a single question. Rather, try at least two questions, preferable three or more. * Don't share decks and create your own: * Useful to put personal information into Anki * Constructing your own decks is part of the process, the more elaborately ou encode a memory, the stronger the memory will be. * Form rich associations yourself. * Cultivate strategies for elaboriative encoding and forming rich associations: * A strategy for forming strategies * Multiple variants of the same question * What about memory palaces and loci? * Useful for trivia * Less developed for more abstract concepts * "there is great value in learning to think in more memorable ways" * 95% of Anki's value comes from 5% of the features: * Don't bother with most additional features like tags, auto-generating cards, plugins * Clozes are good for things like quotes. * Challenges of using Anki to store facts about friends and family: * "Is [my friend] vegan?" is fine. * "What is the name of [my friend's] eldest child" is a bit strange. * It's a social norm to remember things like that, it's like you're faking interest. * Leaves you feeling uncomfortale. * Procedular versus declarative memory: * Difference between remembering a fact and mastering a process * Might remember a fact but not recognise situations to apply it * > "Put another way: to really internalize a process, it's not enough just to review Anki cards. You need to carry out the process, in context. And you need to solve real problems with it. " * Difference between knowing the name of something and knowing something: * Useful in both ways * Author argues names do matter, it's an early step along the way to understanding * Particularly useful for non-verbal things like "what's the name of this painting?" * Challenges with getting behind: * Anki becomes challenging when you get behind with cards * Two days of not using, the cards begin to back up * It's intimidating to see 500 cards to be reviewed. * Increase quotas (100, 150, 200, 250 then 300) cards per day until you catch up * Other uses of Anki: * Learning an API * Seminars, set quota of at least three high-quality questions to ankify. Setting quotas help you pay attention. * Videos, events, places. 3-5 questions about the experience but it doesn't matter too much. * Ankify in real time as I read papers and books. This is because in seminars and conversations you want to be immersed in the experience, but with papers and books you hav plenty of time. * Reading a book can become bogged down with Ankifying. Set a quota, don't ankify everything in sight. * Ankify things that serve your long-term goals. * Anki can be used in conjunction with notes. * Avoid yes/no questions: * elaborate the ideas in the question * On external memory aids: * Google, Wikipedia and notebooks ought to be enough * A complement to Anki * Something special about creative work and problem-solving, having an internalised understanding. * Speed of associative thought * Ability to rapidly try out many combinations of ideas * Fluency matters in thinking. * The thought experiment of a flute in which there is a one second delay between blowing the note and hearing it is absurd. * Certain types of thoughts are much easier to have when all the relevant kinds of understanding are held in the mind. Anki is invaluable. * Why isn't Anki used so much? * People underestimate the gains that come from distributing their study in a manner similar to Anki, preffering last-minute cramming even though studies show it produces worse results. * "principle of desireable difficulty" the idea that memories are maximally strengthened if tested when we're on the verge of forgetting them * Systems like Anki are challenging to use well and easy to use poorly. ##### How can making flashcards atomic help prevent card failures?? It can make one large, routinely wrong question into multiple routinely correct questions. ##### What is an "orphan question"?? A question which is unrelated and unlinked to prior knowledge. ##### Why are "orphan questions" a bad thing?? Because they feel disconnected from what you actually care about and feel like a burden. ##### Why is using shared Anki decks a bad idea?? Encoding the memory yourself is part of the process. ##### In what way is reviewing Anki cards not enough?? To really internalize knowledge, you need to apply and solve real problems with it. ##### How can setting a "flashcard quota" be a useful tool?? It ensures you pick the most useful pieces of information rather than ankifying everything. ##### How can you prevent becoming bogged down with ankifying (e.g. reading a book)?? Set a "flashcard quota". ##### "Fluency matters..."?? "in thinking". ##### What is the fluency in thinking thought experiment involving a flute?? A flute where there is a one second delay between blowing the note and hearing it is "absurd". ##### In what way is a delayed flute analogous with fluency in thinking?? A delayed flute is like having to stop to remember something when thinking. ##### How does the speed of associative thought matter?? It allows you to rapidly try out many combinations of ideas. ### Part 2, Personal Memory Systems More Broadly #### How important is long term memory? ##### How did chess experts see the board differently to beginners?? They saw it as a collection of chunks or units rather than a series of individual pieces. ##### The chunk theory in chess was written about in which book?? Thought and Choice in Chess ##### Who wrote Thought and Choice in Chess?? Adriaan D. de Groot ##### "It's like they're trying to compose"...?? ...a beautiful sonnet in French, but they only know 200 words of French. ##### How does 'Augmenting Long Term Memory' argue that chunk-theory apply to mathematics?? The essay argues that top mathematicians have internalized many more complex chunks. ##### Why do top mathematicians see a complex problem as simpler than ordinary people?? Because they have internalized more complex mathematical concepts, they can abstract over a lot of the complexity. ##### What is the capacity of working memory in chunks?? 7, plus or minus 2. ##### Who first described the concept of a capacity in working memory?? George Miller ##### What was the name of the paper by George Miller about the capacity of working memory?? The Magical Number Seven, Plus or Minus Two. ##### What is the correlation between working memory and IQ?? The better your working memory, the higher your IQ. ##### What is the definition of a "chunk"?? The basic unit of working memory. ##### What is the name of the curve that measures how likely you are to remember something over time?? Ebbinghaus Forgetting Curve ##### A steeper Ebbinghaus forgetting curve corresponds with?? A faster rate of memory decy. ##### What is the x-axis for an Ebbinghaus forgetting curve?? Time. ##### What is the y-axis for an Ebbinghaus forgetting curve?? Probability of recall. ##### The relationship between time and the probability of recall is?? Exponential decay. ##### What is the testing effect?? Where long term memory is increased by practice retrieving the information that needs to be remembered. ##### What is the spacing effect?? Where long term memory is increased when learning events are spaced apart in time. ##### What are the two effects that increase the efficacy of Anki?? * The testing effect * The spacing effect ##### What effect does testing memory have on the Ebbinghaus forgetting curve?? It flattens the curve. --- Olly Britton — https://ollybritton.com. Machine-readable index: https://ollybritton.com/llms.txt