This morning, as you got up and dragged yourself to the kitchen to get some coffee, you probably didn’t pay much attention to what your feet were doing.
They just kinda do their own thing and get you were you want to go, right?
But you’re only able to do that because we once tried it, paying very close attention (or falling a lot) and did it enough times that your brain learned how to handle the balancing of your weight and muscle contraction to get you safely down the stairs.
I’ve always been curious about how many things we do without really being conscious of them.
We humans are very proud of our conscious mind — the ideas, the feelings, the planning, the not being able to sleep because you remembered you said something dumb 8 years ago — but we fail to notice how much our brain is able to do without us having to direct attention at it.
And that’s a good thing too. Have you every become aware of your breathing and then couldn’t stop thinking of it and felt that you’d never get back to “automate breathing mode”? (you’re welcome for that)
The way our brain learns how to handle complex tasks like walking, driving or dodging a punch to the face without needing your conscious attention is by trial, error and feedback.
To illustrate how this work, watch this video of legendary boxer Muhammad Ali dodging 21 punches in 10 seconds, with a smile.
It looks like he’s capable of predicting what his opponent is going to do. And to a sense he does.
But how could that possibly be?
That’s where machine learning perfectly exemplifies how our brain works. After all, it’s based on how humans learning.
How Machine Learning can teach you to learn faster
In a nutshell, Machine learning works by inputting data for the machine to analyse and generate their own model of how things work. Then, that model is tested millions of times, and the machine receives feedback to improve it’s model.
If you’re interested in a deeper look into how that works, I absolutely recommend reading Vishal Maini’s Begginers Guide to Machine Learning. A lot of what I mention below is my simplification of the math and techniques explained in his articles.
Let’s take a look at how machines learn to handle two different tasks and how you can use the same concepts to learn new things.
How you can learn to predict stock price variations — Like machines learn to predict the price of a house
If you want to teach a machine do adequately estimate the price a house can sell for, all you’d need to do is to first give it enough information regarding thousands of houses and their prices to study.
If we’re sure there is a direct correlation between the price of a house and its properties (square footage, the size of the bedrooms, zip-code, the number of tennis courts in the backyard), you need only to input an adequate amount of data points and the machine — by using Linear Regression (that’s math) — will be able to stipulate the correlation between the properties of a house and it’s final price.
After being fed enough data points to analyse, a machine could then start making their own predictions and receive feedback if those predictions were correct. With time, as long as the relationship between the data and the final price is valid, the machine can start making very precise predictions for this simple task.
Linear regression is not easy, your conscious mind can’t handle that many calculations intuitively. However, your subconscious mind does it for you. Everyday.
That’s why when you’re window shopping at the mall you can almost immediately know if a product is expensive or cheap just by looking at it.
You’ve been exposed to hundreds of thousands of products in different shapes, sizes and colors, from different brands and in different stores and brain collected and analysed all of those data points and create it’s own very model to predict the price of things and to tell you not to even look again at that pair of shoes you liked cause they’re probably too expensive.
Your mental model about how much a shoe costs is updated not only every time you see a new pair of shoes with a price tag but by watching what shoes other people wear and whenever a friend mentions the price of a similar product.
Using the same idea, you can teach yourself to become better at predicting a stock’s price is undervalued and if it’s a good time to invest simply by being exposed to data points (current price, trading volume, news about the company, etc.) related to it.
Being exposed to that data feeds your brain a model about how prices change when other things change, creating what we call experience. That’s why even technical traders that use very complex math, analysis tools and indicators still listen to their intuition when trading — for good or for worse.
The takeaway: Whenever you’re trying to learn something new, try to immerse yourself into that and collect as much data as you can. Your brain will connect the dots in the background and speed up your learning curve.
How you can improve your performance at anything just by sucking at it — Just like this machine learned to play Pong
Moving on to a much cooler (and a bit scary) example, watch the video below to see the result of a machine learning — by itself — to play pong:
Pong is a simple game and that makes it the perfect example for understanding how a machine can learn a complex behavior like that and completely dominate the game.
The machine has only three choices at any time:
- move up;
- move down;
- or stand still.
If the machine fails to block the ball, it loses a points. That’s negative feedback. If the machine manages to score a point on the other player, that’s positive feedback.
So as, you can see in the video, you don’t even need to tell the machine the rules of the game. It just needs to know when it has done well and when it has done poorly.
The machine then analysis the pixels on the screen every time the ball moves, taking into account the position of the ball, itself and the enemy, for each second of each match and judges it’s own performance. Then it tries something new and assimilates if that lead to a victory or not.
Do that millions of times and the machine becomes a expert at the game. It might not now which movement led it to victory, but it has enough data to know that best course of action based on what exactly is on the screen.
Remember Ali dodging those punches? That’s how he knew where his opponent was going to hit. He had been hit a million times already and his brain simply knew where the next one would come from — and what to do about it.
Is that not truly inspiring?
Our brains are amazing machines, capable of taking unimaginable amounts of data and improve it’s own performance at every new try.
And you can use that power — as you have always been using, however unaware — to do anything. Become a better dancer. Learn a new language. Become a more charismatic person. Just by cheer repetition.
That only thing that can stop you from improving constantly every day, at anything, is yourself. Your fear of looking silly when you’re not good enough yet. The fear that others will laugh when you’re not sure what you’re doing. The laziness or entitlement of not even trying.
The takeaway: Whatever your goal is, be a little like that machine that wasn’t afraid of looking stupid on it’s first tries at the game. Try again, millions of time, fearlessly. And trust your brain is learning even when you feel you’re not.
Thanks for reading! If you’ve enjoyed this post, do leave a comment. That way I’ll know I should continue writing articles like this — consider it positive feedback for my machine learning!
