Learning is a set of skills. You need to practice each component of the learning process to get better. You can’t watch a video on a new technique and immediately become a pro. It takes time to reap the benefits.
Most people suck at mindmaps. Mindmaps can be horrible for learning if you just dump a bunch of text on a page and point arrows to different stuff (some studies show mindmaps are ineffective, but that’s because people initially suck at making them). However, if you take the time to learn how to do them well, they will pay huge dividends in the future. I’ll be doing the “Do 100 Things” challenge and developing my skill in building better mindmaps. Getting better at mindmaps involves “chunking” the material and creating memorable connections and drawings.
Relational vs Isolated Learning. As you learn something new, try to learn it in relation to the things you already know rather than treating it as isolated from everything (flashcards can perpetuate the problem of learning things in isolated form).
Deep processing is the foundation of all learning. It is the ability to connect, process, organize and relate information. The opposite of deep processing is rote memorization. If it doesn’t feel like you are engaging ~90% of your brain power when you are learning/reading something, you are likely not encoding the information into your long-term memory effectively.
Only use Flashcards as a last resort. Flashcards are something a lot of people use because they feel comfortable going through them. However, if your goal is to be efficient in your learning, you should only use flashcards when it’s something that requires rote learning. Video worth watching on Spaced Repetition.
My current approach for learning about alignment: I essentially have a really big Roam Research page called “AI Alignment” where I break down the problem into chunks like “Jargon I don’t understand,” “Questions to Answer,” “Different people’s views on alignment,” etc. As I fill in those details, I add more and more information in the “Core of the Alignment Problem” section. I have a separate page called “AI Alignment Flow Chart” which I’m using as a structure for backcasting on how we solved alignment and identifying the crucial things we need to solve and things I need to better understand. I also sometimes have a specific page for something like Interpretability when I’m trying to do a deep dive on a topic, but I always try to link it to the other things I’ve written in my main doc.
And this video concisely covers a lot of important learning concepts.
Look at the beginning of the video for an explanation of encoding, storage (into long-term memory), and retrieval/rehearsal to make sure you remember long-term.
Outside of learning:
Get enough sleep. 8 hours-ish.
Exercise like HIIT.
Make sure you have good mental health.
Meditation is likely useful. I personally use it to recharge my battery when I feel a crash coming and I think it’s useful for training yourself to work productively for longer periods of time. This one I’m less sure of, but seems to work for me.
Learning (all of these take time to master, don’t expect you will use them in the most effective way right out of the gate):
Use inquiry-based (curiosity-based) learning. Have your learning be guided by questions you have, like:
”Why is this important?”
”How does it relate to this other concept?”
Learn by scope. Start with the big picture and gradually break things down where it is important.
Chunking. Group concepts together and connect different chunks by relationship.
Create stories to remember things.
Focus on relationships between concepts. This is crucial.
Rehearsal
Spaced repetition (look at my other notes on how SR is overrated but still useful)
Apply your learning by creating things (like a forum post applying the new concept to something and explaining it)
Ever since I was little, I have relied on my raw brain power to get to where I am. Unfortunately, I could never bring myself to do what other smart kids were doing. Flashcards, revision? I would either get bored out of my mind or struggle because I didn’t know how to do it well. Mindmaps? It felt OK while I was doing it the few times I tried, but I would never revise it, and, honestly, I sucked at it.
But none of that mattered. I could still do well enough even though my learning system was terrible. However, I didn’t get the top grades, and I felt frustrated.
I read a few books and watched the popular YouTubers on how to learn things best. Spaced Repetition and Active Recall kept coming up. All these intelligent people were using it, and I truly believed it worked. However, whenever I tried it, I either ended up with too many flashcards to have the time to review, or I couldn’t build a habit out of it. Flashcards also felt super inefficient when studying physics.
I did use Cal Newport’s stuff for some classes and performed better by studying the same amount of time, but as soon as things got intense (exam season/lots of homework), I would revert to my old (ineffective) study techniques like reading the textbook aimlessly and highlighting stuff. As a result, I would never truly develop the skill (yes, skill!) of studying well. But, just like anything, you can get better at creating mindmaps for proper learning and long-term memory.
I never got a system down, and I feel I’m losing out on gains in my career. How do I learn things efficiently? I don’t want to do the natural thing of putting in more hours to get more done. 1) My productivity will be capped by my inefficient system, 2) I still want to live life, and 3) it probably won’t work anyways.
So, consider this my public accountability statement to take the time to develop the skills necessary to become more efficient in my work. No more aimlessly reading LessWrong posts about AI alignment. There are more efficient ways to learn.
I want to contribute to AI alignment in a bigger way, and something needs to change. There is so much to learn, and I want to catch up as efficiently as possible instead of just winging it and trying whatever approach seems right.
Had I continued working on things I don’t care deeply about, I might have never decided to put in the effort to create a new system (which will probably take a year of practicing my learning skills). Maybe I would have tried for a few weeks and then reverted to my old habits. I could have kept coasting in life and done decently well in work and my personal life. But we need to solve alignment, and building these skills now will allow me to reap major benefits in a few years.
(Note: a nice bonus for developing a solid learning system is that you can pass it on to your children. I’m excited to do that one day, but I’d prefer to start doing this now so that I know that *I* can do it, and I’m not just telling my future kids nonsense.)
My goal will be to create a “How to Create an Efficient Learning System” guide tailored for professionals and includes examples in AI alignment. Please let me know if there are some things you’d like me to explore in that guide.
Before I go, I’ll mention that I’m also interested in eventually taking what I learn from constructing my own learning system and creating something that allows others to do the same, but with much less effort. I hope to make this work for the alignment community in particular (which relates to my accelerating alignment project), but I’d also like to eventually expand to people working on other cause areas in effective altruism.
Important part: Use GPT to facilitate the process of pushing you to higher-order learning as fast as possible.
Here’s Bloom’s Taxonomy for higher-order learning:
For example, you want to ask GPT to come up with analogies and such to help you enter higher-order thinking by thinking about whether the analogy makes sense.
Is the analogy truly accurate?
Does it cover the main concept you are trying to understand?
Then, you can extend the analogy to try to make it better and more comprehensive.
This allows you to offload the less useful task (e.g. coming up with the analogy), and spending more time in the highest orders of learning (the evaluation phase; “is this analogy good? where does it break down?”).
You still need to use your cognitive load to encode the knowledge effectively. Look for desirable difficulty.
Use GPT to create a pre-study of the thing you would like to learn.
Have it create an outline of the order of the things you should learn.
Have it give you a list of all the jargon words in a field and how they relate so that you can quickly get up to speed on the terminology and talk to an expert.
Coming up with chunks of the topic you are exploring.
You can give GPT text that describes what you are trying to understand, the relationships between things and how you are chunking them.
Then, you can ask GPT to tell you what are some weak areas or some things that are potentially missing.
GPT works really well as a knowledge “gap-checker”.
When you are trying to have GPT output some novel insights or complicated nuanced knowledge, it can give vague answers that aren’t too helpful. This is why, it is often better to treat GPT as a gap-checker and/or a friend that is prompting you to come up with great insights.
Reference: I’ve been using ChatGPT/GPT-4 a lot to gain insights on how to accelerate alignment research. Some of my conclusions are similar to what was described in the video below.
How learning efficiently applies to alignment research
As we are trying to optimize for actually solving the problem, we should not fall into the trap of learning just to learn. We should instead focus on learning efficiently with respect to how it helps us generate insights that lead to a solution for alignment. This is also the framing we should have in mind when we are building tools for augmenting alignment researchers.
With the above in mind, I expect that part of the value of learning efficiently involves some of the following:
Efficient learning involves being hyper-focused on identifying the core concepts and how they all relate to one another. This mode of approaching things seems like it helps us attack the core of alignment much more directly and bypasses months/years of working on things that are only tangential.
Developing a foundation of a field seems key to generating useful insights. The goal is not to learn everything but to build a foundation that allows you to bypass spending way too much time tackling sub-optimal sub-problems or dead-ends for way too long. Part of the foundation-building process should reduce the time it shapes you into an exceptional alignment researcher rather than a knower-of-things.
As John Wentworth says with respect to the Game Tree of Alignment: “The main reason for this exercise is that (according to me) most newcomers to alignment waste years on tackling not-very-high-value sub-problems or dead-end strategies.”
Lastly, many great innovations have not come from unique original ideas. There’s an iterative process passed amongst researchers and it seems often the case that the greatest ideas come from simply merging ideas that were already lying around. Learning efficiently (and storing those learnings for later use) allows you to increase the number of ideas you can merge together. If you want to do that efficiently, you need to improve your ability to identify which ideas are worth storing in your mental warehouse to use for a future merging of ideas.
My model of (my) learning is that if the goal is sufficiently far, learning directly towards the goal is goodharting a likely wrong metric.
The only method which worked for me for very distant goals is following my curiosity and continuously internalizing new info, such that the curiosity is well informed about current state and the goal.
Curiosity is certainly a powerful tool for learning! I think any learning system which isn’t taking advantage of it is sub-optimal. Learning should be guided by curiosity.
The thing is, sometimes we need to learn things we aren’t so curious about. One insight I Iearned from studying learning is that you can do specific things to make yourself more curious about a given thing and harness the power that comes with curiosity.
Ultimately, what this looks like is to write down questions about the topic and use them to guide your curious learning process. It seems that this is how efficient top students end up learning things deeply in a shorter amount of time. Even for material they care little about, they are able to make themselves curious and be propelled forward by that.
That said, my guess is that goodharting the wrong metric can definitely be an issue, but I’m not convinced that relying on what makes you naturally curious is the optimal strategy for solving alignment. Either way, it’s something to think about!
By the way, I’ve just added a link to a video by a top competitive programmer on how to learn hard concepts. In the video and in the iCanStudy course, both talk about the concept of caring about what you are learning (basically, curiosity). Gaining the skill to care and become curious is an essential part of the most effective learning. However, contrary to popular belief, you don’t have to be completely guided by what makes you naturally curious! You can learn how to become curious (or care) about any random concept.
Added my first post (of, potentially, a sequence) on effective learning here. I think there are a lot of great lessons at the frontier of the literature and real-world practice on learning that go far beyond the Anki approach that a lot of people seem to take these days. The important part is being effective and efficient. Some techniques might work, but that does not mean it is the most efficient (learning the same thing more deeply in less time).
Note that I also added two important videos to the root shortform:
While spaced repetition is good, many people end up misusing it as a crutch instead of defaulting to trying to deeply understand a concept right away. As you get better at properly encoding the concept, you extend the forgetting curve to the point where repetition is less needed.
Here’s some additional notes on the fundamentals on being an effective learner:
Encoding and Retrieval (What it take to learn)
Working memory is the memory that we use. However, if it is not encoded properly or at all, we will forget it.
Encode well first (from working memory to long-term memory), then frequently and efficiently retrieve from long-term memory.
If studying feels easy, means that you aren’t learning or holding on to the information. It means that you are not encoding and retrieving effectively.
You want it to be difficult when you are studying because this is how it will encode properly.
Spacing, Interleaving, and Retrieval (SIR)
These are three rules that apply to every study technique in the course (unless told otherwise). You can apply SIR to all techniques.
Spacing: space your learning out.
Pre-study before class, then learn in class, and then a week later revise it with a different technique.
A rule of thumb you can follow is to wait long enough until you feel like you are just starting to forget the material.
As you get better at encoding the material effectively as soon as you are exposed to it, you will notice that you will need to do less repetition.
How to space reviews:
Beginner Schedule (less reviews need as you get better at encoding)
Same day
Next day
End of week
End of month
After learning something for the first time, review it later on the same day.
Review everything from the last 2-3 days mid-week.
Do an end of week revision on the week’s worth of content.
End of month revision on entire month’s worth of content.
Review of what’s necessary as time goes on.
(If you’re trying to do well on an exam or a coding interview, you can do the review 1 or 2 weeks before the assessment.)
Reviewing time duration:
For beginners
No less than 30 minutes per subject for end-of-week
No less than 1.5 hours per subject for end-of-month.
Schedule the reviews in your Calendar and add a reminder!
Interleaving: hitting a topic or concept from multiple different angles (mindmaps, teaching).
The idea is that there is the concept you want to learn, but also there is a surrounding range that you also want to learn (not just the isolated concept).
Could be taking a concept and asking a question about it. Then, asking a question from another angle. Then, asking how it relates to another concept.
Try to use a multitude of these techniques in your studying, never studying or revising anything the same way more than once.
Math, it could be thinking about the real-world application of it.
Examples of interleaving:
Teach an imaginary student
Draw a mindmap
Draw an image instead of using words to find a visual way of expressing information
Answer practice questions
Create your own challenging test questions
Create a test question that puts what you’ve learned into a real-world context
Take a difficult question that you found in a practice test and modify it so that the variables are different, or an extra step is added
Form a study group and quiz each other—for some subjects you can even debate the topic, with one side trying to prove that the other person is missing a point or understanding it incorrectly
For languages, you can try to speak or write a piece of dialogue or speech, as well as some variations. How might someone respond? How would you respond back? Are there any other responses that would be appropriate?
Retrieval: taking info from your long-term memory and bringing it into your working memory to recall, solve problems and answer questions.
Taking a concept and retrieving it from your long-term memory.
Don’t just retrieve right away, you can look at your notes, take a few minutes and retrieve.
Or it also happens when you are learning something. Let’s say you are listening to a lecture. Are you just writing everything down or are you taking some time to think and process what is being said and then writing down notes? The second one is better.
Syntopical Learning
When you are learning something, you want to apply interleaving by learning from different sources and mediums. So, practice become great at learning while listening, while watching, while reading. These are all individual modes of learning you can get better at and they will all help you better retain the material if you use them all while learning.
I use the app Concepts on my iPad to draw mindmaps. Drawing mindmaps with pictures and such is way more powerful (better encoding into long-term memory) than typical mindmap apps where you just type words verbatim and draw arrows. It’s excellent since it has a (quasi-) infinite canvas. This is the same app that Justin Sung uses.
When I want to go in-depth into a paper, I will load it into OneNote on my iPad and draw in the margin to better encode my understanding of the paper.
I’ve been using the Voice Dream Reader app on my iPhone and iPad to get through posts and papers much faster (I usually have time to read most of an Alignment Forum post on my way to work and another on the way back). Importantly, I stop the text-to-speech when I’m trying to understand an important part. I use Pocket to load LW/AF posts into it and download PDFs on my device and into the app for reading papers. There’s a nice feature in the app that automatically skips citations in the text, so reading papers isn’t as annoying. The voices are robotic, but I just cycled through a bunch until I found one I didn’t mind (I didn’t buy any, but there are premium voices). I expect Speechify has better voices, but it’s more expensive, and I think people find that the app isn’t as good overall compared to Voice Dream Reader. Thanks to Quintin Pope for recommending the app to me.
Current Thoughts on my Learning System
Crossposted from my website. Hoping to provide updates on my learning system every month or so.
TLDR of what I’ve been thinking about lately:
There are some great insights in this video called “How Top 0.1% Students Think.” And in this video about how to learn hard concepts.
Learning is a set of skills. You need to practice each component of the learning process to get better. You can’t watch a video on a new technique and immediately become a pro. It takes time to reap the benefits.
Most people suck at mindmaps. Mindmaps can be horrible for learning if you just dump a bunch of text on a page and point arrows to different stuff (some studies show mindmaps are ineffective, but that’s because people initially suck at making them). However, if you take the time to learn how to do them well, they will pay huge dividends in the future. I’ll be doing the “Do 100 Things” challenge and developing my skill in building better mindmaps. Getting better at mindmaps involves “chunking” the material and creating memorable connections and drawings.
Relational vs Isolated Learning. As you learn something new, try to learn it in relation to the things you already know rather than treating it as isolated from everything (flashcards can perpetuate the problem of learning things in isolated form).
Encoding and Retrieval are essential concepts for efficient learning.
Deep processing is the foundation of all learning. It is the ability to connect, process, organize and relate information. The opposite of deep processing is rote memorization. If it doesn’t feel like you are engaging ~90% of your brain power when you are learning/reading something, you are likely not encoding the information into your long-term memory effectively.
Only use Flashcards as a last resort. Flashcards are something a lot of people use because they feel comfortable going through them. However, if your goal is to be efficient in your learning, you should only use flashcards when it’s something that requires rote learning. Video worth watching on Spaced Repetition.
You need to be aiming for higher-order learning. Take advantage of Bloom’s Taxonomy.[1]
My current approach for learning about alignment: I essentially have a really big Roam Research page called “AI Alignment” where I break down the problem into chunks like “Jargon I don’t understand,” “Questions to Answer,” “Different people’s views on alignment,” etc. As I fill in those details, I add more and more information in the “Core of the Alignment Problem” section. I have a separate page called “AI Alignment Flow Chart” which I’m using as a structure for backcasting on how we solved alignment and identifying the crucial things we need to solve and things I need to better understand. I also sometimes have a specific page for something like Interpretability when I’m trying to do a deep dive on a topic, but I always try to link it to the other things I’ve written in my main doc.
And this video concisely covers a lot of important learning concepts.
Look at the beginning of the video for an explanation of encoding, storage (into long-term memory), and retrieval/rehearsal to make sure you remember long-term.
Outside of learning:
Get enough sleep. 8 hours-ish.
Exercise like HIIT.
Make sure you have good mental health.
Meditation is likely useful. I personally use it to recharge my battery when I feel a crash coming and I think it’s useful for training yourself to work productively for longer periods of time. This one I’m less sure of, but seems to work for me.
Learning (all of these take time to master, don’t expect you will use them in the most effective way right out of the gate):
Use inquiry-based (curiosity-based) learning. Have your learning be guided by questions you have, like:
”Why is this important?”
”How does it relate to this other concept?”
Learn by scope. Start with the big picture and gradually break things down where it is important.
Chunking. Group concepts together and connect different chunks by relationship.
Create stories to remember things.
Focus on relationships between concepts. This is crucial.
Rehearsal
Spaced repetition (look at my other notes on how SR is overrated but still useful)
Apply your learning by creating things (like a forum post applying the new concept to something and explaining it)
Ever since I was little, I have relied on my raw brain power to get to where I am. Unfortunately, I could never bring myself to do what other smart kids were doing. Flashcards, revision? I would either get bored out of my mind or struggle because I didn’t know how to do it well. Mindmaps? It felt OK while I was doing it the few times I tried, but I would never revise it, and, honestly, I sucked at it.
But none of that mattered. I could still do well enough even though my learning system was terrible. However, I didn’t get the top grades, and I felt frustrated.
I read a few books and watched the popular YouTubers on how to learn things best. Spaced Repetition and Active Recall kept coming up. All these intelligent people were using it, and I truly believed it worked. However, whenever I tried it, I either ended up with too many flashcards to have the time to review, or I couldn’t build a habit out of it. Flashcards also felt super inefficient when studying physics.
I did use Cal Newport’s stuff for some classes and performed better by studying the same amount of time, but as soon as things got intense (exam season/lots of homework), I would revert to my old (ineffective) study techniques like reading the textbook aimlessly and highlighting stuff. As a result, I would never truly develop the skill (yes, skill!) of studying well. But, just like anything, you can get better at creating mindmaps for proper learning and long-term memory.
I never got a system down, and I feel I’m losing out on gains in my career. How do I learn things efficiently? I don’t want to do the natural thing of putting in more hours to get more done. 1) My productivity will be capped by my inefficient system, 2) I still want to live life, and 3) it probably won’t work anyways.
So, consider this my public accountability statement to take the time to develop the skills necessary to become more efficient in my work. No more aimlessly reading LessWrong posts about AI alignment. There are more efficient ways to learn.
I want to contribute to AI alignment in a bigger way, and something needs to change. There is so much to learn, and I want to catch up as efficiently as possible instead of just winging it and trying whatever approach seems right.
Had I continued working on things I don’t care deeply about, I might have never decided to put in the effort to create a new system (which will probably take a year of practicing my learning skills). Maybe I would have tried for a few weeks and then reverted to my old habits. I could have kept coasting in life and done decently well in work and my personal life. But we need to solve alignment, and building these skills now will allow me to reap major benefits in a few years.
(Note: a nice bonus for developing a solid learning system is that you can pass it on to your children. I’m excited to do that one day, but I’d prefer to start doing this now so that I know that *I* can do it, and I’m not just telling my future kids nonsense.)
So, what have I been doing so far?
I started the iCanStudy course by Dr. Justin Sung (who has a YouTube channel). I’m only about 31% through the course.
My goal will be to create a “How to Create an Efficient Learning System” guide tailored for professionals and includes examples in AI alignment. Please let me know if there are some things you’d like me to explore in that guide.
Before I go, I’ll mention that I’m also interested in eventually taking what I learn from constructing my own learning system and creating something that allows others to do the same, but with much less effort. I hope to make this work for the alignment community in particular (which relates to my accelerating alignment project), but I’d also like to eventually expand to people working on other cause areas in effective altruism.
Note on using ChatGPT for learning
Important part: Use GPT to facilitate the process of pushing you to higher-order learning as fast as possible.
Here’s Bloom’s Taxonomy for higher-order learning:
For example, you want to ask GPT to come up with analogies and such to help you enter higher-order thinking by thinking about whether the analogy makes sense.
Is the analogy truly accurate?
Does it cover the main concept you are trying to understand?
Then, you can extend the analogy to try to make it better and more comprehensive.
This allows you to offload the less useful task (e.g. coming up with the analogy), and spending more time in the highest orders of learning (the evaluation phase; “is this analogy good? where does it break down?”).
You still need to use your cognitive load to encode the knowledge effectively. Look for desirable difficulty.
Use GPT to create a pre-study of the thing you would like to learn.
Have it create an outline of the order of the things you should learn.
Have it give you a list of all the jargon words in a field and how they relate so that you can quickly get up to speed on the terminology and talk to an expert.
Coming up with chunks of the topic you are exploring.
You can give GPT text that describes what you are trying to understand, the relationships between things and how you are chunking them.
Then, you can ask GPT to tell you what are some weak areas or some things that are potentially missing.
GPT works really well as a knowledge “gap-checker”.
When you are trying to have GPT output some novel insights or complicated nuanced knowledge, it can give vague answers that aren’t too helpful. This is why, it is often better to treat GPT as a gap-checker and/or a friend that is prompting you to come up with great insights.
Reference: I’ve been using ChatGPT/GPT-4 a lot to gain insights on how to accelerate alignment research. Some of my conclusions are similar to what was described in the video below.
How learning efficiently applies to alignment research
As we are trying to optimize for actually solving the problem, we should not fall into the trap of learning just to learn. We should instead focus on learning efficiently with respect to how it helps us generate insights that lead to a solution for alignment. This is also the framing we should have in mind when we are building tools for augmenting alignment researchers.
With the above in mind, I expect that part of the value of learning efficiently involves some of the following:
Efficient learning involves being hyper-focused on identifying the core concepts and how they all relate to one another. This mode of approaching things seems like it helps us attack the core of alignment much more directly and bypasses months/years of working on things that are only tangential.
Developing a foundation of a field seems key to generating useful insights. The goal is not to learn everything but to build a foundation that allows you to bypass spending way too much time tackling sub-optimal sub-problems or dead-ends for way too long. Part of the foundation-building process should reduce the time it shapes you into an exceptional alignment researcher rather than a knower-of-things.
As John Wentworth says with respect to the Game Tree of Alignment: “The main reason for this exercise is that (according to me) most newcomers to alignment waste years on tackling not-very-high-value sub-problems or dead-end strategies.”
Lastly, many great innovations have not come from unique original ideas. There’s an iterative process passed amongst researchers and it seems often the case that the greatest ideas come from simply merging ideas that were already lying around. Learning efficiently (and storing those learnings for later use) allows you to increase the number of ideas you can merge together. If you want to do that efficiently, you need to improve your ability to identify which ideas are worth storing in your mental warehouse to use for a future merging of ideas.
My model of (my) learning is that if the goal is sufficiently far, learning directly towards the goal is goodharting a likely wrong metric.
The only method which worked for me for very distant goals is following my curiosity and continuously internalizing new info, such that the curiosity is well informed about current state and the goal.
Curiosity is certainly a powerful tool for learning! I think any learning system which isn’t taking advantage of it is sub-optimal. Learning should be guided by curiosity.
The thing is, sometimes we need to learn things we aren’t so curious about. One insight I Iearned from studying learning is that you can do specific things to make yourself more curious about a given thing and harness the power that comes with curiosity.
Ultimately, what this looks like is to write down questions about the topic and use them to guide your curious learning process. It seems that this is how efficient top students end up learning things deeply in a shorter amount of time. Even for material they care little about, they are able to make themselves curious and be propelled forward by that.
That said, my guess is that goodharting the wrong metric can definitely be an issue, but I’m not convinced that relying on what makes you naturally curious is the optimal strategy for solving alignment. Either way, it’s something to think about!
By the way, I’ve just added a link to a video by a top competitive programmer on how to learn hard concepts. In the video and in the iCanStudy course, both talk about the concept of caring about what you are learning (basically, curiosity). Gaining the skill to care and become curious is an essential part of the most effective learning. However, contrary to popular belief, you don’t have to be completely guided by what makes you naturally curious! You can learn how to become curious (or care) about any random concept.
Video on how to approach having to read a massive amount of information (like a textbook) as efficiently as possible:
Added my first post (of, potentially, a sequence) on effective learning here. I think there are a lot of great lessons at the frontier of the literature and real-world practice on learning that go far beyond the Anki approach that a lot of people seem to take these days. The important part is being effective and efficient. Some techniques might work, but that does not mean it is the most efficient (learning the same thing more deeply in less time).
Note that I also added two important videos to the root shortform:
Note on spaced repetition
While spaced repetition is good, many people end up misusing it as a crutch instead of defaulting to trying to deeply understand a concept right away. As you get better at properly encoding the concept, you extend the forgetting curve to the point where repetition is less needed.
Here’s a video of a top-level programmer on how he approaches learning hard concepts efficiently.
And here’s a video on how the top 0.1% of students study efficiently.
Here’s some additional notes on the fundamentals on being an effective learner:
Encoding and Retrieval (What it take to learn)
Working memory is the memory that we use. However, if it is not encoded properly or at all, we will forget it.
Encode well first (from working memory to long-term memory), then frequently and efficiently retrieve from long-term memory.
If studying feels easy, means that you aren’t learning or holding on to the information. It means that you are not encoding and retrieving effectively.
You want it to be difficult when you are studying because this is how it will encode properly.
Spacing, Interleaving, and Retrieval (SIR)
These are three rules that apply to every study technique in the course (unless told otherwise). You can apply SIR to all techniques.
Spacing: space your learning out.
Pre-study before class, then learn in class, and then a week later revise it with a different technique.
A rule of thumb you can follow is to wait long enough until you feel like you are just starting to forget the material.
As you get better at encoding the material effectively as soon as you are exposed to it, you will notice that you will need to do less repetition.
How to space reviews:
Beginner Schedule (less reviews need as you get better at encoding)
Same day
Next day
End of week
End of month
After learning something for the first time, review it later on the same day.
Review everything from the last 2-3 days mid-week.
Do an end of week revision on the week’s worth of content.
End of month revision on entire month’s worth of content.
Review of what’s necessary as time goes on.
(If you’re trying to do well on an exam or a coding interview, you can do the review 1 or 2 weeks before the assessment.)
Reviewing time duration:
For beginners
No less than 30 minutes per subject for end-of-week
No less than 1.5 hours per subject for end-of-month.
Schedule the reviews in your Calendar and add a reminder!
Interleaving: hitting a topic or concept from multiple different angles (mindmaps, teaching).
The idea is that there is the concept you want to learn, but also there is a surrounding range that you also want to learn (not just the isolated concept).
Could be taking a concept and asking a question about it. Then, asking a question from another angle. Then, asking how it relates to another concept.
Try to use a multitude of these techniques in your studying, never studying or revising anything the same way more than once.
Math, it could be thinking about the real-world application of it.
Examples of interleaving:
Teach an imaginary student
Draw a mindmap
Draw an image instead of using words to find a visual way of expressing information
Answer practice questions
Create your own challenging test questions
Create a test question that puts what you’ve learned into a real-world context
Take a difficult question that you found in a practice test and modify it so that the variables are different, or an extra step is added
Form a study group and quiz each other—for some subjects you can even debate the topic, with one side trying to prove that the other person is missing a point or understanding it incorrectly
For languages, you can try to speak or write a piece of dialogue or speech, as well as some variations. How might someone respond? How would you respond back? Are there any other responses that would be appropriate?
Retrieval: taking info from your long-term memory and bringing it into your working memory to recall, solve problems and answer questions.
Taking a concept and retrieving it from your long-term memory.
Don’t just retrieve right away, you can look at your notes, take a few minutes and retrieve.
Or it also happens when you are learning something. Let’s say you are listening to a lecture. Are you just writing everything down or are you taking some time to think and process what is being said and then writing down notes? The second one is better.
Syntopical Learning
When you are learning something, you want to apply interleaving by learning from different sources and mediums. So, practice become great at learning while listening, while watching, while reading. These are all individual modes of learning you can get better at and they will all help you better retain the material if you use them all while learning.
A few more notes:
I use the app Concepts on my iPad to draw mindmaps. Drawing mindmaps with pictures and such is way more powerful (better encoding into long-term memory) than typical mindmap apps where you just type words verbatim and draw arrows. It’s excellent since it has a (quasi-) infinite canvas. This is the same app that Justin Sung uses.
When I want to go in-depth into a paper, I will load it into OneNote on my iPad and draw in the margin to better encode my understanding of the paper.
I’ve been using the Voice Dream Reader app on my iPhone and iPad to get through posts and papers much faster (I usually have time to read most of an Alignment Forum post on my way to work and another on the way back). Importantly, I stop the text-to-speech when I’m trying to understand an important part. I use Pocket to load LW/AF posts into it and download PDFs on my device and into the app for reading papers. There’s a nice feature in the app that automatically skips citations in the text, so reading papers isn’t as annoying. The voices are robotic, but I just cycled through a bunch until I found one I didn’t mind (I didn’t buy any, but there are premium voices). I expect Speechify has better voices, but it’s more expensive, and I think people find that the app isn’t as good overall compared to Voice Dream Reader. Thanks to Quintin Pope for recommending the app to me.