Skip to main content

Ultralearning: Summary and Applications

In AP Microeconomics class in high school, I stumbled upon a blog post that guaranteed that I could ace my finals in school without studying. Without second thought to checking if it was a scam, I ended up spending the entire class period reading that article. 3 years later, I am still reading weekly articles from Scott Young, the blogger who completed the entire MIT Computer Science curriculum in his bedroom in one year and the writer who propelled me into deep fascination with how humans learn and master skills. When he announced his book Ultralearning in June of 2019, I pre-ordered from Amazon for the first time in my life. In this blog post, I hope to consolidate the information from his book into a summary and actionable advice I can use and remember. 

Source: Wallpaper Abyss


Principle 1 – Metalearning: First Draw a Map (Least Useful Actionable Section)

Language isn’t about memorizing words and conjunctions. Mastering the structure of the language will get you speaking and recognizing patterns much quicker. In general, before any topic, one should answer the questions “What? Why? How?” to understand the motivation to learn and the plan to put forth in learning it. A good rule of thumb is to spend around 10% of total learning time in research/planning before starting, before diminishing marginal returns.

Example Used:
Scott uses the example of Dan Everett, who learns languages from scratch in half an hour by drawing a map of full sentences and guesses what the translation is with his linguistic background of how languages are structured.

In Own Action:
Don’t just aimlessly grind out applications and projects, set a time aside to research and plan out thoughts and goals without taking any notes or writing any drafts yet
On the other hand, don’t feel satisfied by simply passively reading and watching lectures. At some point no more information taken in by your brain will help, one must start taking action and doing after the 10% mark.

Principle 2 – Focus: Sharpen Your Knife (Section that is Hardest to Sustain)

There are 3 problems associated with focusing – failing to start focusing (procrastinating), failing to sustain focus (distractions), and failing to create the right kind of focus. The first problem may stem from either undesirable actions or having too many ideas and not knowing where to start. To help this, recognize that what may be unpleasant is an impulsive feeling that won’t last long and use the five-minute method – simply tell yourself you will focus maximum effort for five minutes before taking a break. The second problem stems from distractions such as buzzing phones or talking roommates, or simply becoming tired of the material. Being in a flow state (when task is not too easy or hard) is ideal, but not always possible. Instead, take breaks every hour or so and cover multiple topics per day instead of one deeply per day to keep your mind active and not burned out. Another possible source of distraction is your own thoughts arising from your mind, but studies have shown that “learning to let it arise, note it, then releasing it or letting it go” is more effective than trying to suppress it. Lastly, the third problem stems from alertness and arousal.  Shooting a basketball requires a different type of focus than writing. Complex tasks benefit from lower arousal and quieter environments, while simpler tasks benefit from a noisier one. 

Example Used:
Mary Somerville, an 18th century mathematician, gave herself no excuses to not learn math – when candles were not available for her to read, she would think through concepts and problems in her head over and over, in the dark.

In Action:
Before being tempted to play ping pong or video games, ask yourself which urge is more powerful – taking a break or keep going – and at least have an awareness of the situation
Switch environments and note how productive each one is, for varying types of tasks
Recognize where you are and start small. Sometimes you don’t even realize you are being distracted. If you can’t follow the five-minute rule, reduce it to one, or even half of a minute.

Principle 3 – Directness: Go Straight Ahead (Most Applicable Section)

Formal schooling is often woefully indirect – college-level physics grads with honors grades often cannot solve basic problems that deviate slightly from what was tested. This also means “brain-training” games are pretty much useless, the fastest way to become smarter or better at something is to directly DO it. Learning algorithms mean nothing if you have trouble recognizing when to use it when actually coding. Try to mimic the situation as closely as possible if the real thing is not readily available (such as flight simulation in place of actual flying). Project-based learning (creating), immersive learning (such as living in Spain to learn Spanish), the flight simulator method, and the overkill approach (learn more than what is required of you, and connect the extra material to your past knowledge) are all forms of direct learning. 

Example Used:
Vatsal Jaiswal was an architect who just finished a college degree and was looking for a job in the midst of the Great Depression. When creating his own projects directly applicable to what jobs were looking for instead of aimlessly submitting his school portfolio to more and more firms, he was hired much more quickly.

In Action:
Ask yourself where and how the knowledge will manifest itself when reading and listening to lectures. Dig deeper in topics you are particularly interested in
While taking online courses and watching YouTube videos and reading are good, create projects and goals from the information you gather to actually internalize the material

Principle 4 – Drill: Attack Your Weakest Point (My Favorite Historical Example)

Pure bulk practice isn’t enough, one needs to strategize what specific areas of a learning endeavor to focus more time on and drill. In chemistry, there is a notion of the rate-determining step, which is the slowest part of a chain of reactions that ultimately affects the time of entire process. The Direct-then-Drill approach is very powerful in jumping straight into direct learning, then isolating components that are particularly difficult to improve and is slowing the rate of learning down. Some drilling tactics include time slicing (“isolating a slice in time of a longer sequence of actions”), cognitive components (isolating aspects of language such as grammar, pronunciation, and vocab), copycat (copying parts of work you don’t want to drill to focus on ones you do), magnifying class method (spend more time on one component of skill than otherwise, useful if can’t separate or slice that component), and prerequisite chaining (start with a skill that’s too hard, then go back a step and learn the prerequisites directly applicable to the harder skill and recurse downwards)

Example Used:
Benjamin Franklin was an international bestseller and first and foremost a writer, spurred from a comment from his father when he was a young child that his writing lacked sophistication and persuasion. He honed his ability by taking his favorite articles and try to reconstruct the arguments by memory, then correcting his faults, submitting anonymous articles to his brother’s printing press, and start constructing arguments with meter and rhyme to more creatively use words. He improved his persuasion by practicing the Socratic method, challenging ideas through probing questions rather than direct, abrupt contradiction.

In Action:
In music, use time slicing by practicing the hardest part of the piece before integrating back into the context of the entire song
When doing CS homework, if time is sparse, start on the homework and project right away, then only refer back to lectures and reading when stuck
Practice dribbling in isolation to quickly improve handling skills instead of playing scrimmages and hoping to slowly build your entire game

Principle 5 – Retrieval: Test to Learn (My Favorite Section)

Given three options on how to study for an exam – looking over notes and studying everything over and over, creating a concept map to organize and relate concepts together, and testing yourself with free recall without looking at the book at all – free recall is by far the most effective strategy. What’s even more interesting, is that in an experiment, those who used free recall felt the least confident going into a test yet performed the best by far. This suggests that when studying, encountering difficult problems and struggling is a good thing and does not necessarily reflect a lack of preparedness. Humans aren’t good at telling how well they have learned something; we rely on experiences of studying and doing to tell us how we are doing, and looking back at notes to remember material may give us a false sense of confidence. We can practice retrieval with several methods – flash cards (especially programs with spaced-repition algorithms such as Anki), free recall (write down everything you can remember after lecture or reading without referring back), question-book method (asking questions that restate the big idea instead of unnecessary details and rephrasing notes from lecture instead of copying main ideas and directly from slides), self-generated challenges (creating list of challenges to tackle) and closed-book learning (taking practice exams without a cheat sheet).

Example Used:
Srinivasa Ramanujan may have been one of the most brilliant mathematicians of all time despite dropping out of college and growing up poor with very few resources. He went through an entire textbook deriving and figuring out equations and theorems for himself, rather than copying and memorizing them, building an unbelievably deep understanding of the topics.

In Action:
Combine retrieval and free recall with the ability to look up answers when stumped as a form of studying
Take practice exams before even reviewing the material (don’t be discouraged for not being able to do much!) – this will hone what specific material to actually review

Principle 6 – Feedback: Don’t Dodge the Punches (Least Intuitive Section) 

Rapid feedback is important and allows the learner to better judge their approach and what to work on. However, it can backfire – if the feedback is targeted to fixed, inherent traits such as laziness or intelligence, it could discourage the subjects. Feedback works well when it gives useful information to guide future learning and is actionable. There are 3 levels of feedback – outcome feedback (tells you how well you do overall, but not what you are doing well or poorly, such as grades) informational feedback (tells you what you’re doing wrong but not how to fix it), and corrective feedback (usually a coach, mentor, or teacher who shows you how to fix what you are doing wrong). Although corrective feedback is usually the best, it is not always better (for example if advice isn’t very useful) and informational feedback is not always better than outcome feedback (over analyzing the bad design of a character’s color instead of game as a whole). Research has found that immediate feedback to be more effective than delay, except in a laboratory setting. However, too quick of feedback may turn retrieval practice into passive review, so for example when encountering hard problems, give yourself a time limit before looking up the answer. 

Example Used:
Chris Rock test drives his comedy skits at local comedy clubs before performing for large concert halls and popular TV shows. 

In Action:
Never reject negative feedback or be discouraged by it, reflect on it and think for yourself
Understand not all feedback is useful, target those who are most relevant and driven in your field of endeavor

Principle 7 – Retention: Don’t Fill a Leaky Bucket (Most Interesting Example)

Mnemonics, or anything that allows humans to “chunk” related ideas together into one memorable phrase or picture, is the most effective way to remember lots of things at once. It is difficult to remember things because some information decay over time (information that are not vivid or used frequently), new information interfere with old ones (forgetting exactly what while loops do after learning for loops), and some information are unable to be retrieved on command, even if it is floating in our brain somewhere. Techniques to combat forgetting include spacing (spreading learning over longer intervals of time, for shorter periods at a time, instead of cramming), proceduralization (encoding a procedure into your physical motor movements such as riding a bike or typing until it’s automatic), overlearning (practicing beyond perfect, discussed earlier in directness), and mnemonics (creating a memorable picture, phrase, or feeling to tie to what is to be remembered)

Example Used:
Nigel Richards won the World Scrabble Championships in French, without being able to speak French at all. Instead, he is good at memorizing patterns in words and cycled 200 miles to and from the competition while repeating the words in his head. He recalls words visually and claims to be unable to recall words spoken to him.

In Action:
Space out homework for each class over small sessions over many days instead of cramming all onto one day
Draw out diagrams of CS concepts to see how they relate to each other and build visual mental map instead of blankly staring at code
Free recall information when doing monotone activities such as running or riding bus

Principle 8 – Intuition: Dig Deep Before Building Up (My Favorite Teaching Example)

Students rarely think about how to relate what they learn in school and lectures to problems outside the textbook, thus making their knowledge very inflexible and inapplicable. To build your own intuition, there are a few rules you can follow – don’t give up on hard problems easily (set a timer to push yourself a bit further than you want to), prove things to understand them (refer back to Ramanujan), always start with a concrete example (new concepts are hard to learn abstractly, try to tie concepts with deeper processing of the meaning such as its pleasantness rather than a trivial meaning like its spelling), and don’t fool yourself (don’t believe you possess more knowledge about a subject than you actually do, always be probing for more questions and ideas). Asking questions does not mean you are not knowledgeable, just as having to study and work hard for a class does not imply you are not smart. One of the best ways to build intuition is through the Feynman Technique, which essentially entails writing a concept/problem on the top of a paper, then explain the concept as if you had to teach it to someone else who had never heard it before, or explain how to solve a problem and why each step makes sense. If at any point you are stuck, only then do you refer back to notes or lectures (similar to free recall). This technique can be applied to things you don’t understand at all (break it down to understanding just once sentence if necessary), problems you can’t seem to solve (step through the problem in detail and understanding why each step is performed rather than memorizing steps) and expanding intuition (generate visualizations, analogies, and diagrams, explain as if you are being paid to write an article about the idea).

Example Used:
Richard Feynman (my favorite teacher of all time) was a brilliant physicist, won the Putnam math competition, and yet had an IQ of 125, only modestly higher than the average college graduate. He attributes his own success to building a deep intuition of the world around him and relate new problems to examples he already knows or creating his own examples in his head.

In Action:
Spend at least an hour trying to solve a problem in CS 170 without contacting others, and at least 20-30 minutes without referring to notes at all
Understand that you can always understand a topic deeper, try to teach others and generate original ideas to explain them, especially visually or viscerally

Principle 9 – Experimentation: Explore Outside Your Comfort Zone

Experimentation is the key to mastery, and this requires exploration into ideas no one else has tried. An example Scott gives is that “a great mathematician is one who can solve problems others cannot, not merely a person who can solve previously solved problems easily.” Experimentation includes trying various learning resources, different techniques, and style. The mindset of an experimenter is important - those who think they can grow and improve do, those who believe their efforts are futile are stuck. Some strategies on how to experiment include – copy, then create (van Gogh copying others to learn to draw before developing his own style), compare methods side-by-side (learning a language through flash cards then through mnemonics and see which one is more effective), explore the extremes (dress in a fashion way outside normal conventions) and combine unrelated skills (The creator of the popular comic strip Dilbert combined his talent of art to his background of being an engineer with an MBA). It’s important to remember that not all experimentation will be a success, but all experimentation provide you with experience and feedback as to what is successful and unsuccessful. 

Example Used:
Vincent van Gogh started art very late, at the age of 26, but that didn’t stop him from endlessly experimenting to create his artistic style, despite failing miserably his entire life and being discouraged by peers. He devoured books and copied works he looked up to and eventually developed his own unique style - but he did not become famous until after death. 

In Action:
Try various methods of teaching the same subject in CS, then keep the best one in mind for the future
Go out and eat the food of the world, you’ll find new favorites every day!


To wrap up, I would like to list some ultralearning projects to try in my free time or any time:

When traveling to a country, speak nothing but that language at all times
No need to wait until CS 189 to learn machine learning, think about a project you want to learn right now (such as AI-generated music) and start researching and building
Buy an electric instrument next summer and put up covers on YouTube, forcing me to come up with creative content and quality material

Comments

Popular posts from this blog

The First Post

I still remember the first time I read a blog.  Towards the end of freshman year in high school the college admissions bug had really bit me and I would obsessively read MIT Blogs every day. I was always excited to hear about the amazing experiences these amazing people across the country were having. Eventually I discovered a man on YouTube that would go on to waste way too many hours of my life. Every single day I would watch Casey Neistat's vlog and get to see a glimpse of his awesome life. 
Today I've decided to take a step in the right direction and create my own blog alongside my good friend Owen. I'm hoping to use this blog as a platform to discuss ideas I've been formulating, commit to the goals I should've years ago, jot down notes I think are important, and most importantly, leave behind a piece of myself on the internet. 
Most of my content will fall under a couple of categories: School - an all-encompassing part of lifeEconomics/Business topics - just m…

The First .5 Post

I don't remember the first time I read a blog. And unlike Dhaval, I never really enjoyed writing. From the age of five years old until eighth grade, my father encouraged me to write a journal entry every Sunday, one paragraph in Mandarin, then translated into English - by far my least favorite "homework" I've ever had to do.
So what am I doing creating a blog? It will serve as a breeding ground for thoughts I've never spoken aloud, ideas that usually quickly fade from recollection, as well as memories to be re-traced in the future. It will also force me to write more and communicate more eloquently, while hopefully still coming off as casual conversation. (I am quite embarrassed to say this is the most words I've written on anything in 2019).

I have no particular topics in mind, but expect content to range anywhere from NBA basketball to Scandinavian history to building computer programs, some of which going in great detail while others being short speculatio…