What is deep learning?

Deep learning is one of the forms of artificial intelligence. According to Yann Le CUN, it is a crossroads between several fields of knowledge: neuroscience, mathematics and technical progress. 

Why deep? 

Deep learning is composed in the same way as our brain, i.e. with several layers of neurons. The information will then pass through the different layers of information stored in memory. This makes it possible to analyse different solutions or solve problems. However, it functions more intensively than a human being because the number of neuronal layers is much greater (20 or more). The human being has 6 layers of neurons. Deep learning is therefore an artificial neural network. 

Where is it? 

Deep learning is already present in our daily lives, for example in facial recognition (your mobile phone) or in language processing (translation). It is the number of layers of neurons that will allow us to be more and more precise. But before they completely invade our daily lives, we need to train these machines. It is important that these machines see the world as we humans do, in order to build intelligent systems. To do this, the machine will be given thousands of data and will have to learn to recognise them. After this phase, the machine will be able to learn by itself. 

We are living in the glory days of deep learning, but this method has been around for many years. The first artificial neural network was created in 1951 by Minsky and Edmonds (Harvard University). But the concept was developed by Alan Turing in the 1950s, notably with the famous Enigma machine (see the film “Imitation Game”)