The Unreasonable Effectiveness of Learning to Learn
It's January 1st, and Alice and Bob are getting started with their New Year's Resolutions. They both want the same thing - to become smarter. And so they've committed to spend one hour each day learning new things.
However, despite sharing the same goal, their approaches are different:
Alice spends her hour learning new facts. Each week she learns 10 new facts, but gets no faster at learning week-on-week.
Bob splits his time between learning facts and learning to learn. At the end of week 1 he’s learned just 5 new things, but each subsequent week he's able to learn 5% faster than the week before.
So whose strategy is better? Who learns more?
Obviously Bob, right? Why else bother writing this post?
Well, not exactly... Let's run some numbers.
For Alice the calculation is straightforward. She learns 10 things each week, and there are 52 weeks in a year, so come December 31st, she's learned 520 new things. Well done, Alice.
For Bob, it's a matter of compounding. He starts off learning 5 new things per week. By week 5 that's grown to 6 new things per week. By week 13 he's cruising at 9 new things per week. Adding up all his learnings across the year we get around 118 new things - not even half of what Alice learns.
But by this stage, Bob has transformed himself into a well-oiled learning machine. He's able to learn 60 new things each week - which is 6 times faster than Alice. And so if both Alice and Bob repeated their resolutions the following year, by the end of the two-year period Alice would have learned 1,040 new things (=520 * 2), while Bob would have learned 1,486 - much more than Alice. And by this stage, he'd be learning 760 new things every week, which is more than Alice learned in her entire first year.
This is the power of meta-learning - or learning to learn. Bob takes an ego hit in the short run, but before Alice knows what's coming, Bob flies past her like a brainy rocket ship. And Alice will have a hard time ever catching up. In fact, if she simply swaps to Bob's method of improving her learning speed by 5% each week, she'll never catch up. She needs to improve even faster to make up the difference.
Under the hood, the unreasonable effectiveness of learning to learn is driven by the same mathematical laws that explain many well-known phenomena, like viral growth of memes online, Moore's law, nuclear explosions, and compound interest.