Artificial intelligence consists of a set of algorithms that aim to enable machines to carry out human intelligence tasks. It is clear that machines were capable of performing complex calculations and storing information in memory, but there are tasks of a human nature that they were not capable of performing. For example, machines originally had no machine learning capability or any kind of creativity. They were simply given an input and returned an output in a pure way, for example, whenever the same input was given, they always returned the same output, but this changes with the paradigm of artificial intelligence.
The most developed field within the artificial intelligence is machine learning, which consists of a series of techniques by which a machine can change its behaviour based on data and results of past actions. While initially based only on data, machine learning allows them to also take into account previous actions, simulating human capability. In simpler words, if before the implementation of artificial intelligence we wanted a machine to cook a dish, we had to incorporate data and instructions beforehand and, as long as both were the same, the result would be the same, without any possibility of learning or improvement. Machine learning adds another variable to the set of data and instructions: the results of past actions. Therefore, it increases the possibility of constant improvement in each of the processes in which it is implemented.
There are 3 types of machine learning.
1. Supervised learning. Labelled data is observed and classified by the machine through a practice known as training. It is used, for example, in facial recognition and automatic car driving.
2. Unsupervised Learning. Unlabelled data of which the structure is not known is observed. The aim is to obtain deterministic information through this structure. It is applied, for example, to define distance between data. One could analyse the list of products purchased by a user in an e-commerce and make recommendations based on the lists of other users who share products in common.
3. Reinforced Learning. It is applied based on the concepts of trial, error and reward. The clear example we can observe is machines competing with professional chess players. If they get the reward, they are likely to repeat the moves that led to the reward in the following games.
Although the concept of artificial intelligence was coined in the 1950s, there are still many industries where it has not yet been implemented. There is a long way to go, but there is no doubt that machine learning will improve many of the processes we know and increase the welfare of society, either by saving resources or by obtaining better results with the same amount of resources.
