Today we will try to understand in an easy way, what exactly machine learning is.
Well, let’s try to make things easy: today in our world, we have (at least) 2 distinct categories, humans and computers right?
Humans: Well, humans don’t even know how the brain works, it is a fantastic mystery. But the good thing is that we can do a lot of things, we know how to walk, we learn fast, and we also have emotions!
Computers: Computers are robots that can do single task MUCH better than human. Simple question: calculate 394×378.64? Who do you think is the fastest between a human and a calculator? Calculator right? A calculator is a good example of a robot that excels in a single task and does it much better than human.
Humans often learn with observations or examples. Human brain processes information which allows us without any effort to recognize on a picture if a person is blonde, angry or sad. For us, this whole process is natural and we don’t even notice it.
Computers, on the other hand, have to be told what to do. We have to set rules in place for a machine to execute those rule when required. And we write those “logic” rules with programming languages.
At the end, we have humans that can learn from examples, computers that work with predefined rules. But can we design computers that learn from examples just like humans?
Yes, we can. This is what machine learning allows us to do!
Let’s take a comparison of what machine learning is and is not.
Non Machine Learning
But Deep Blue was not using machine learning at all! Instead, it was using a clever set of predefined rules on top of a huge dataset of stored chess games. In a game of chess, there are about 1040 possibilities.
Now let’s take the Alpha Go case from Google DeepMind. Alpha Go became the first Computer Go program to beat a human professional Go player without handicaps on a full-sized 19×19 board in March 2016.
The average possibilities in a 19×19 board are ~2.082 × 10170 (i.e., a 10 followed by 170 zeroes). This is more than the estimated 1080 number of atoms in the observable universe, which means, even if you unite all the calculation power of all machines on earth, you will not be able to calculate and store all the possibilities.
This is where machine learning steps in! Without going into details, Alpha Go beats the world champion simply by mixing trial and learning from examples just like humans do. Alpha Go used reinforcement learning and neural networks systems that played a lot of games and got new skills in each round. Ultimately beating the world champion!
And this is just an appetizer. Machine Learning is a new field and it will most likely shape the future of our current society.