Types and Tailcalls

Ultralearning

published on October 14th, 2019

Stories

The book starts out with a number of stories of ultralearners and ultralearning projects. They sound all quite incredible and they really made me excited about the book. Here are the ones which I found more interesting.

Ultralearning Principles

The precise techniques of ultralearning can vary from subject to subject, thus it cannot simply be broken down to tools and tactics, but there are nine principles which embody a universal approach to ultralearning:

  1. Metalearning: Start by learning how to learn what you want to learn.
  2. Focus: Improve concentration during learning.
  3. Directness: Learn by doing.
  4. Drill: Improve the weakest point. Break down complex skills into parts, master those parts, put them back together.
  5. Retrieval: Test yourself to learn. Actively recall information.
  6. Feedback: Is uncomfortable but crucial. Extract the signal from the noise.
  7. Retention: Learn to remember things forever.
  8. Intuition: Develop your intuition through play and exploration of concepts and skills.
  9. Experimentation: True mastery comes not from following a known path but exploring new possibilities.

Finally, the ethos behind ultralearning is to be in charge of your learning and maximize your return on investment. With this ethos, principles can be considered guideposts instead of hard and fast rules.

Principle 1: Metalearning

Metalearning is the activity of deciding how to go about the learning. There are three components to this, the why, the what and the how.

Why

The reason why you want to learn something can roughly be intrinsic (learning for its own benefit) or instrumental (learning to obtain some other goal, e.g. advancing in ones career).

Especially with instrumental learning it is key to validate that what you want to learn actually helps you what you really want. An example of failing at this is somebody getting an MBA to advance in their career only to discover it doesn't help them much after getting it. A good way to prevent this is the expert interview: find somebody who has what you want (e.g. great career in your field) and ask them if they think your approach (e.g. get MBA) is a good one. Best do this face-to-face or video chat, not email, to not miss nuances.

Expert interviews can also help with intrinsic learning goals.

What

How is the knowledge in your subject structured? Split knowledge into concepts (things that need to be understood), facts (things that need to be memorized) and procedures (things that need to be practiced). Then draw your map by identifying the most challenging concepts, facts and procedures and use this knowledge to make learning time more effective and efficient.

How

First, find the common ways in which people learn what you want to learn, this is called benchmarking. For anything taught at university, a curriculum or syllabus is a first stop, also a list recommended of textbooks might help. Also finding people who have mastered the skill and performing expert interviews should help.

Finding the right material has a very high payoff as this will be the basis for all learning.

Once benchmarking is done emphasize areas that align with your goal and exclude or delay areas that don't align.

How Much Planning vs Doing?

Typically, people don't spend enough time to define goals and plan the learning approach. However, planning can also become a form of procrastination. Spending approximately 10% of the total learning time on metalearning is a good rule of thumb. So 10 hours of a 100 hour project. For large projects (e.g. 500 hours) may be a bit less and move more towards 5%. One should know about common methods of learning and popular resources and tools.

Metalearning is not a one time thing. It might make sense to re-research when hitting plateaus or problems. Do metalearning when the mariginal benefits of metalearning are larger than actual learning, and learn otherwise.

Long Term Benefits

The real benefits of metalearning is longterm, as many of the skills and techniques identified can be applied to more than one project. But payoff can be high even for a single project.

Principle 2: Focus

There are three aspects of focus: starting, sustaining, and optimizing the quality of one's focus.

Problem 1: Failing to Start Focusing (aka Procrastination)

Procrastination is obviously a problem. Two tips in this book about how to handle it (a) start very small, e.g. by working for five minutes on a topic, then graduating to pomodoros, etc. and (b) become conscious why you're starting to procrastinate, which can be (b-i) because of aversion of the task for some reason (what exactly?) and (b-ii) because of attraction to some other task (what and why?).

Problem 2: Failing to Sustain Focus (aka Getting Distracted)

Flow is nice, but ultralearning, just like deliberate practice, might be too hard for it sometimes. Don't worry too much about it.

How long should you practice? Going too long isn't optimal, 50 minutes to one hour is best, so if multiple hours for study taking some breaks is probably optimal. Also, learning different subjects / aspects during this time might be good, but learning too many different things is not.

Distraction Sources:

  1. Your environment: Phone, internet, people interrupting, TV, games, noises sounds. Optimize environment to help sustain focus.
  2. Your task: Some tasks are easier to focus on, e.g. watching video might (or not?) be easier to focus on than reading. If something is more cognitive demanding, then it is generally easier to focus on.
  3. Your mind: Can generate a lot of distractions, especially if it is filled with strong emotions. Major life crisis can make focus a lot harder, so dealing with these things first might be better.

Problem 3: Creating the Right Kind of Focus

Two interesting variables: arousal and task complexity. For low complex tasks, relatively high arousal is good, while too high arousal makes it too easy to get distracted. For highly complex tasks (e.g. math), a bit lower arousal and more relaxed focus is better.

If stuck on a very complex task, no focus, i.e. taking a break, can be best for the unconscious mind to come up with a solution.

Improve Your Ability to Focus

Recognize where you are and try to improve in small steps.

Personal observation: Motivation helps a lot with focus. If I am very eager to perform a task, I can focus well. Probably links strongly to the Procrastination Equation.

Principle 3: Directness

The problem with Transfer

It has been shown many times, initially by Edward Thorndike and Robert Woodworth in 1901 ("The Influence of Improvement in One Mental Function upon the Efficiency of Other Functions") that transfer from one learned skill to other skills is problematic and often does not ocurr to anywhere near the degree that people hope.

Overcoming the Problem of Transfer with Directness

Directness, i.e. learning by doing or focusing directly on the skill you want to perform during your learning overcomes the problems with transfer in two ways: first, by focusing on the skill directly, the need for transfer is significantly diminished. Second, by focusing on the real world applicability, many subtle skills that do transfer which are needed for real world application of the skill are also learned and these are more likely to transfer to other situations.

How to Learn Directly

Direct learning is hard, it is often more frustrating, challenging and intense than reading a book or sitting through a lecture, this is a source of a competitive advantage.

Some tactics for direct learning:

  1. Project based learning: Opt for projects which produce something - at least you will learn how to produce that thing! Can also be producing a paper, book, online course, etc.
  2. Immersive learning: Surround yourself with the target environment in which the skill is practiced. Obvious for languages, but applicable much more widely.
  3. Flight simulator method: When practicing the skill is not possible directly (flying, surgery, ...), a simulation which requires the same cognitive skills (decisions, use knowledge) is the best alternative.
  4. The overkill approach: Increase the difficulty of the challenge by creating more intense versions of the taks than you actually want to perform (e.g. speaking in front of middleschoolers, who pull no punches).

Principle 4: Drill: Attack Your Weakest Point

In Chemistry there is the concept of the rate limiting step. In a chain reaction, the slowest part of the chain determines the speed of the whole reaction. Similary when learning a skill, the performance of a certain subskill may limit the overall performance. The principle of drill is meant to adress this rate limiting step. The goal then, is to identify the sub-skill or cognitive component that is the weakest and create drills to improve this area.

Direct-then-drill

Drilling is somewhat at odds with the direct principle, which states to practice the skill directly, minimizing the difference between practice and application. Drilling is still needed, because it is easier to improve a subskill when focusing on it, and in fact this might be needed to improve overall performance. However, it is important that the drill is based on observing weaknesses when performing the actual skill, and that the drilled subskill is re-integrated into the full skill after drilling. Picking the right drill based on direct practice is called the direct-then-drill approach. Overall, this approach has four steps:

  1. Practice the skill directly
  2. Find an improvable subskill
  3. Drill the subskill
  4. Integrate the subskill into the direct practice again

Tactics for Drills

  1. Time slicing: Practice one part of a longer sequence (common in music)
  2. Cognitive components: Focus on one cognitive aspect, e.g. grammar or pronounciation when learning a language
  3. Copying: Copy other's work then focus only improving the part that you want to drill (e.g. structure of argument in writing).
  4. Magnifying glass: While practicing the whole skill, spend more time and concentration on the subcomponent you want to improve, even if that decreases overall performance.
  5. Prerequisite chaining: Start learning something that you don't have all the prerequisites for, then cycle back and forth between prerequisites and original learning object.

Drill Mindfully

Drills are important, but it is very important to be careful that the drills actually connect to the real skill. This is the weakness of many academic programs, which mostly focus on drilling while the drills are disconnected from the actually desired skills.

Principle 5: Retrieval: Test to Learn

Learning works better when trying to retrieve something than spending additonal time to review it. This maybe because retrieving information is more difficult than reviewing it, and difficulty is desireable in learning as long as we are still able to perform the task.

How to Practice Retrieval

There are a few methods we can use to practice retrieval:

  1. Flash cards: We can use flash cards, which combines well with spaced reptetition system (SRS - also see principle 7).
  2. Free recall: Difficult but powerful is to practice free recall, i.e. to try to remember as much as possible after learning something. For example, after a lecture we might sit down and write down everything we can recall from that lecture.
  3. Question book method: While reading something, write questions instead of notes. Be careful to write the questions at the right (=not too low) level. Later practice retrieval by answering these questions.
  4. Self-generated challenges: When learning something, think of challenges where this material can be applied, then apply it.
  5. Closed-book learning: Test yourself with the book closed, i.e. force yourself to retrieve when testing or applying your knowledge instead of looking it up.

Principle 6: Feedback

Positive and Negative Effects of Feedback

Research shows that feedback has overall positive impact on learning, but can also have significant negative effect. Negative effects usually ocurr when feedback is (taken to be) directed at a persons ego, not the skill. This can be hard to separate and feedback is often hard to take. However, for this reason feedback provides a significant competitive advantage. It is important to not take feedback as a critique of ones person but of the skill.

Types of Feedback

There are three levels of feedback from least to most granular:

  1. Outcome based feedback: Gives a single piece of information, e.g. a single grade, pass / fail information. This is the most frequent and easiest to obtain, but also gives the least information
  2. Informational feedback: Informational feedback is more detailed than outcome based feedback, basically it is feedback on each line item that contributes to the outcome. On an exam, this could be feedback on each question (vs the whole exam, which would be outcome based).
  3. Corrective feedback: This type of feedback not only tells you what went wrong but also how the correct solution looks and why that is the case. This is the hardest type of feedback to get. Its important that it comes from an actual expert, otherwise it might be wrong, misleading and hurtful.

You need to be careful when trying to upgrade feedback, because the information might not actually be there.

Ways to Improve Feedback

  1. Noice cancellation: Remove as much noise from measurements as possible and focus on metrics that more directly tell you what you want to know. E.g. when looking for feedback on the quality of your blog posts, don't use views as feedback (depends on a wide number of uncontrollable factors) but the rate of people reading a post until the end.
  2. Difficulty sweet spot: Feedback is information, and the information content is the highest when it is unpredictable. Optimize the difficulty so that you don't know in advance what kind of feedback you will be getting. Don't make things too easy, but also do not make them impossibly hard.
  3. Meta feedback: Meta feedback is feedback on the overall learning process, e.g. monitoring the learrning rate.
  4. Rapid fire feedback: Getting lots of feedback early on can be very helpful because (a) it takes away the fear from feedback and (b) knowing your work will be evaluated is a huge motivator for doing a good job.

Principle 7: Retention: Stop Filling a Leaky Bucket

Forgetting is a problem when learning. It is a natural process and can't be fully prevented, but minimizing its effect on the learning effort is important for effective learning.

Theories About Forgetting

There are three theories about forgetting:

  1. Decay: Memories just fade over time.
  2. Inference: Old memories are (partially) overwritten by newer ones. This is a common phenomenon when learning a new foreign language (e.g. Spanish) makes it harder to speak an older one (e.g. French).
  3. Forgotten cues: The information is still there, but the cue needed to access it is forgotten. This could explain the tip of the tongue problem.

No theory can explain forgetting by itself completely, so they might all be in force to different degrees for different memories.

Ways to Prevent Forgetting

There are four ways presented in how we can slow down or prevent forgetting:

  1. Spaced repetition: Spaced repetition algorithms optimize when to retrieve a certain item in order to not foget this and move it to long term memory. This is used in flash card systems such as Anki.
  2. Procedularization: Procedures (e.g. riding a bycicle) are forgotten much more slowly than abstract facts. Another case is "muscle memory" for keyboard shortcuts or passwords. Learning skills to the level where they become automatic procedures is useful in preventing forgetting.
  3. Overlearning: Once a small skill or fact is fully learned, learning it more might not improve performance, but it can prevent forgetting. In a way, this might be seen as the early stage of procedularization.
  4. Mnemonics: Mnemonic techniques make it easier to remember hard to remember information such as numbers by associating them with easier to remember things such as images or stories. While they clearly work, they can be somewhat brittle and work best for memorizing abstract facts - something that isn't that useful to remember anymore.

Principle 8: Intuition: Go Deep Before Building Up

Story: Richard Feynman

This chapter starts with a story about Richard Feynman, the famous physicist. It explains how Feynman's in-depth understanding of Mathematics and Physics on a first-principle level gave him an extrodinary intuition. Feynmans moto was Don't fool yourself - and you are the easiest to fool!. He also used to think of concrete examples when trying to understand something abstract.

First Principle Thinking

First principle thinking, i.e. understanding something based on the fundamental principles has been shown to be much more powerful and lead to better intuition than shallower levels of thinking, such as understanding by analogy or pattern matching. However, it is also requires expert knowledge and is this not possible to do in the beginning.

Feynman Technique

Young's Feynman Technique is an exercise in understanding something based on first principles. It consists of two steps:

  1. Write a concept or problem on the top of a sheet of paper.
  2. Then write an explanation. For a concept explain it to somebody who doesn't know anything about it. For a problem explain in detail how a solution works and why.

This can be applied to things you don't understand, problems you're stuck on and to expand your intuition.

How to Build Intuition

  1. Don't give up on hard problems. When you are about to give up, set a struggle timer, to e.g. give yourself another ten minutes to work on the problem and stop then.
  2. Prove or explain something to understand it. An example is to draw a bicycle to explain how it works.
  3. Use a concrete example when thinking about an abstract concept or technique.
  4. Don't fool yourself into thinking you understand something when you really don't. Ask a lot of questions to make sure your understanding is solid.

Principle 9: Experimentation

True mastery comes not from following the path trodden by others, but from exploring possibilities they haven’t yet imagined - experimentation is the key to mastery.

Story: Van Gogh

This chapter starts with a story about Vincent van Gogh, who started painting late in his life and was rejected from art schools. He learned painting by experimenting with a lot of different styles, finally finding his signature style after many failed attempts.

Things to Experiment With

  1. Learning resources: The first place to experiment is with learning resources, methods and learning techniques. Before making changes, make sure that you have really applied yourself to a specific technique.

  2. Different subfields: Once the basics of a skill have been mastered, there are usually a lot of different things one could learn next.

  3. Style: Many skills (writing, design, music, leadership, art, ...) don't have a single correct way of doing things. Experiment with different styles to find your own unique style.

Experimentation Tacticis

There are a number of tactics that can be used for experimentation:

  1. Copy and change: Copy work, then experiment with changing some aspects of it.
  2. Vary one variable: Varying a single dimension while holding everything else constant enables directly comparing the results.
  3. Add new contraints: Adding new constraints often sparks creative solutions.
  4. Combine two skills: One way to obtain mastery is to not be the best at one skill, but the best at the combination of two, not frequently combined skills (e.g. engineering and public speaking).
  5. Go to extremes: In high dimensions, most volume of a spehere is near the surface. This may indicate that in high-dimensional skills, mastry usually involves two extreme sub-skills. Don't expect mastry by taking the safe or average option on everything.

Learning is experimentation

Experimentation lies at the heart of much of learning. A lot of the suggested methods in the book started out as experiments, and there were many other experiments that did not work out.

Your First Ultralearning Project

This chapter aims to be a guide to a first ultralearning project.

Do Your Research

The first step consists of deciding what to learn, which includes finding the learning materials.

  1. Topic and scope: The topic is obvious, but giving it a good, concrete goal clarifies the next steps a lot. As such be able to hold a 15 minute conversation in Mandarin is a much better goal than to learn Chinese.

  2. Primary learning materials: Identify and obtain the primary materials you will be learning with.

  3. Benchmark how others learned: Find out how others have learned this material and inform your learning plan with it.

  4. Direct practice activities: How can you directly practice the skill you would like to learn?

  5. Backup material and drills: What non-primary learning materials which may be used as backup? What possibilities are there to perform drills?

  6. Feedback How can you get feedback on your learning? Can you improve the type of feedback or the frequency in which you get it? What kind of meta feedback could you collect?

Schedule Your Time

Find a consistent schedule and put it into your calendar. If you're unwilling to do that it is be a sign that your heart is not in the project. It is better to have a consistent schedule than to try to find the time when it arises which typically does not work.

Work should be done in consistent chunks of time. For most activities, it is best if these are not too long streches (max 1 hour) or are at least seperated by breaks. However some activities, such as programming have a long ramp up time and thus need some significant streches of time.

Execute Your Plan

After doing research and scheduling the time, it is time to execute the plan. During execution, monitor how you're doing with respect to the nine ultralearning principles: metalearning, focus, directness, drilling, retrieval, feedback, retention, intuition and experimentation. Make adjustments if necessary.

Review Your Results

After your ultralearning project ends, it is important to review the results to improve your approach to your next ultralearning project. What went well? What went badly?

When there are problems, it is tempting to blame a lack of willpower or discipline, but often the problems can be traced to the initial plan.

Maintain or Master

After completing the ultralearning project, how would you like to proceed? The options are to either maintain the knowledge through less frequent practice, relearn at a later point in time or to push further and pursue mastery.

Alternatives to Ultralearning

Finally, the book mentions that the intense method of ultralearning is not the only way to go. More casual low intensity methods, such as reading books, keep learning on the job are great ways to keep going. Formal education also has many benefits.

An Unconventional Education

The book closes with a discussion of the Polgar sisters, the children of Laszlo Polgar who had set out to train his children to become geniuses and had indeed succeeded in raising three girls who all became extrodnary chess players, one of them beating world champion Garry Kasparov once. There are many parallels between the ultralearning methods presented in the book and the methods used by the Polgar sisters.

Conclusion

I found the book motivating and thought provoking. I could see many of the methods laid out here to be benefitial and the combination of them to have an important impact on learning effectiveness. Certainly a lot of the success depends on the intensity by which one pursues an ultralearning project and the examples in the book are surely the positive outliers, but I still believe that much more effective and better learning is possible with these methods.

I am looking forward to attempt my first ultralearning project and see how these techniques and methods work for me.


comments powered by Disqus