Transcript for:
Metalearning: Learning How to Learn

It seems like it's no longer enough to just  do our jobs; we have to upskill, reskill,   and stay ahead of the curve. But who has time for  that, right? And that's where metalearning comes   in. This was mentioned in the book "Ultralearning"  by Scott Young, where he took what typically takes   four years to go through the computer science  curriculum at MIT and completed it in one year. He   took three months to learn Spanish, Portuguese,  and Korean, which took him a month to learn how   to speak at a conversational level. Taking what seems like lightning speed to   go through things that people take three  years, five years, or a lifetime to learn,   what's he doing differently? One of the key  differentiators is metalearning. Metalearning is   different from traditional learning, right? We  take out that textbook with all the exercises—it's   really inefficient and usually very demotivating,  even if we're curious. It's really hard to sit   through a boring textbook. But metalearning  is self-directed. It's about learning smart,   so not taking what other people tell you  to learn but finding what you want to   know directionally and getting the right  resources to be able to achieve what you   want to do. The secret in metalearning, as the  name suggests, is learning how to learn.   As we see with artificial intelligence and  machine learning, metalearning is a key to   replicating that human intelligence. Think  about it this way: there's deep learning,   which requires large data sets for the machine  to learn what the patterns are. But in real life,   there are so many times when there isn't a big  data set, so what do you do then? Well, you can   leverage existing data sets that you've already  learned from. But instead of learning the content,   you learn about how you learned this process and  apply it to something that's structurally similar,   but the content is different. In other words,  we are extracting frameworks from our previous   learnings and applying them to something  new. This is, of course, directly linked   to framework thinking, something I mentioned in  this video and is a part of this channel.   And if we think about it, we spend so much  time learning in our lifetime, so why not   leverage the things we've learned and extract the  frameworks from them so we can more efficiently   learn going forward? In Scott Young's framework,  we spend 10% of our time focusing on metalearning,   especially the why, the what, and the how of  better learning. So let's get into them:   The "Why" of Metalearning Why do we want to learn this? At the   highest level, there are two reasons. The  first one is something very practical, or   what Young calls instrumental. So, I want to learn  quantum physics because I want to be a theoretical   physicist. That's something very practical with  a very clear goal. But not all of our curiosities   are so practical. Sometimes there's just a pure  desire to learn something new, and Young calls   this intrinsic. Say I'm interested in quantum  physics not because I want to be a physicist,   but because I think physics is just another form  of philosophy, and I'm curious about what quantum   physics thinks about, let's say, time.  The outcome here will be very different,   and that is why we want to know why we want to  learn. Too often, we actually create very strange   expectations for ourselves because we don't think  about why we're learning in the first place.   Let me throw in another example, a classic  Asian one. A lot of parents put their kids   through piano lessons, and they go through the  curriculum of the practical grade piano exam. So,   it's a lot of focus on classical music, and there  are certain pieces you have to play for different   grades. But the goal for their children is not  to become a professional or a piano teacher;   it's to learn about the joy of music, to get a  practical skill out of a musical instrument. So,   you don't necessarily have to start with the  graded exam curriculum if that was your goal.   The "What" of Metalearning Since we're not starting with a textbook   and going to chapter one, what do we do to learn?  Here, you want to define what you want to achieve   and ask yourself three questions: 1) What  concepts do I need to understand? 2) What   facts do I have to memorize? 3) What procedures  do I have to practice? And as you can imagine,   for the theoretical physicist versus the curious  philosopher, the things you need to do here are   very, very different. And that's how you  can focus on the things that truly matter,   same with the professional musician versus  just the joyous curious amateur. The things   you have to understand, the things that you  have to memorize, and the things you have to   practice are very different. Now, of course, the key thing is   to understand well what concepts I  really need to know, what facts I   need to memorize, and how I know what I need  to practice. We're learning something new,   so how do I even know these things? This is  where frameworks come in. They are the shortcuts,   the cheats that will help you develop your memory.  Let me give you an example. It's a traditional   word; it's like learning how to count from one  to ten. And metalearning is about taking what   you know about one to ten, extrapolating it to  getting to 100, getting to a thousand, getting to   a million. So how do you do that with frameworks?  It's the difference between memorizing number one   to ten versus understanding the concepts,  the frameworks of, okay, there is addition.   So when you add different numbers together,  the numbers change. There is the framework of   order: how do you order the numbers from small to  big? And based on these concepts, how do you add   more concepts, right? Like multiplication: how  do you get from one to ten to a hundred? And I   really like the analogy of the knowledge tree. I  think Elon Musk is one who also refers to this,   which is that you want to be able to come back  to the trunk of the knowledge, come back to the   first principles, something I talked about in  this video as well. And once you know that,   you just need to add on, or you know, as Charlie  Munger says, create that lattice work of mental   models and hang your knowledge onto it. So, where do you find these frameworks,   these models? Well, that leads us  to the "How" of Metalearning.   The "How" of Metalearning Scott Young outlined two steps. The   first one is benchmarking, figuring out the most  common path to get to the outcome that you want.   If you want to become a theoretical physicist,  you have to get a degree, right? And there are   courses that you must take. So, all you have to do  is look at the curriculum at a university of your   choice and figure out what the path is. If you are  that curious philosopher who wants to understand   what time means, well, then do some Google  searching, maybe look at some YouTube videos,   maybe look at some online courses, maybe look at  some books to see how people get from nowhere to   understanding the concepts of time. Unlike  traditional learning, it's not as clear-cut,   but the good thing is it's about exploration. It's  about building your own way to an outcome that you   want. It's like going on a treasure hunt, right?  You know it's somewhere; how do you get there?   And so you piece together different resources and  start to go through them because, remember, we're   spending 10% of our time just to figure this out.  It's not handed to us, and that's a good thing. We   have time to piece this together; we have time  to experiment, which leads us to step number   two that Young mentioned, which is to emphasize  or exclude. Because you know your outcome best,   when you go through the resources within that  10% of time, you'll see some are really giving   you what you want, in which case you want to  emphasize that resource, right? You want to see   where it leads, what is it mentioning, what books,  other resources, other people they're mentioning,   and figure those out. Whereas there will be  other resources that may not be as relevant   as you thought they were. They may be teaching in  a way that you find it very difficult to absorb.   Then it's okay to exclude them, to de-prioritize.  Coming back to the idea of the knowledge tree,   right, we want to get to the trunk. And once  we have that foundational understanding,   how we branch out is up to us. It's okay to not  know everything. Unlike with traditional learning,   we want to self-direct where we want to go,  and it doesn't mean we will never circle back;   we're just following the path we  are deciding for ourselves.   I'm sure you're realizing, as well as we go  through this, that the most powerful thing   about metalearning is this self-directed  thinking, thinking for yourself. That's   different from traditional learning; it's really  about that learn, unlearn, and relearn process.   And with that in mind, check out this framework  thinking video here to learn more about how to add   more frameworks to your cognitive toolbox,  and I'll see you in the next video. Bye.