Artificial intelligence is the future of computing and the world, and today we are going to talk about its origin: Machine Learning (in Spanish machine learning).
Just like artificial intelligence is the attempt to get that computers have intelligent behavior similar to that of humans, Machine Learning was the first step in this direction and aimed to computers could learn by themselves things that humans cannot describe or reflect in any programming language.
And that’s how Machine Learning started, a discipline (just like artificial intelligence) closely linked to mathematics and statistics and that for almost half a century it has been making progress to this day. And right now Machine Learning is one of the most active and powerful fields of computing with applications in all imaginable sectors.
Machine Learning with Supervised Learning
Machine learning or Machine Learning can be divided into two types: supervised and unsupervised, for quite obvious reasons. The first of these types is the supervised one, which is one of the most interesting and used in science because it allows the learning of patterns or characteristics that are not apparent to humans. This type of Machine Learning is ideal for classification tasks such as knowing what kind of particles have appeared in an LHC collision.
The way this type of procedure is executed is quite simple and consists of a training step in which the algorithm is presented with data already analyzed so that it learns the characteristics of each class of particle, for example. During this phase the computer itself is capable of establishing connections between the data of the collision and the type of particles that have appeared, no need to know quantum physics.
Once trained there is usually a test phase in which you are given data again Known but presented without solution so that the computer has to apply everything learned through Machine Learning in order to discover what type of particle was produced. If you pass this phase with a high percentage of correct answers the algorithm is ready to face new data and analyze them better and faster than a human.
Unsupervised machine learning
On the other hand we have unsupervised learning in which there is no training phase. In this case, the algorithm usually has response quality criteria and learns what is the best way to reach the optimal result as quickly as possible. This type of Machine Learning is usually more limited in applicability but it is also very powerful and used. Let’s see better how it works with a small example.
Let’s suppose we want to do a linear fit to a series of points. The mathematical convention is that the best line is one in which the sum of the squared distances from the points to the line is less. This will be the criteria that we will give to our algorithm, but we will not tell him how to achieve this result, he will learn that alone. This type of Machine Learning usually has a totally random startFor example putting lines in any position and direction and seeing how the sum of distances squared varies.
With the passage of time and to repeat and repeat the process the computer you will realize what are the characteristics that all the optimal linear adjustments have. In this way, if we present you with new data, it will take much less time to get the line because you already know what the idea of this Machine Learning technique is looking for is that the computer be able to find the formula to calculate the line immediately, given the points.
Machine Learning, the principle of artificial intelligence that we know
Over the years this type of machine learning has been superseded (at least in sophistication) by others. More advanced techniques that allow computers to perform much more complex processing and that they imitate human thought much better; We have gone from Machine Learning to Artificial Intelligence. The question that remains in the air is how much will these computers be able to think in the future and if they really can reach our level of conceptual abstraction; hopefully all of us will see it.