Artificial intelligence, chatbots, computer vision, robotics and all words that until a few years ago would have seemed to come out of some science fiction colossal. Instead, in reality artificial intelligence is already an integral part of our private and working lives, even if probably in a less flashy way than the film directors of the past decades had predicted.

In fact, today, artificial intelligence works above all to support a whole series of activities and processes, rather than manifesting itself to all of us with robotic voices and cyclopean-sized computers.

But, after the exclusive guides to cloud computing and IOT, let’s start from the beginning, which is what Artificial Intelligence (AI) really is.


There are, of course, many definitions of Artificial Intelligence, one that is particularly clear is that of Stanford University, which identifies it as “a science and a set of computational techniques that are inspired – though typically operating in different ways – by the way humans use their nervous system and body to feel, learn, reason and act”.

This set of techniques, as far as we are concerned, allows the design of hardware and software systems capable of providing often better and faster performance than the human mind, limited to specific segments of action.

Behind Artificial Intelligence there is therefore not only one technology, but different declinations: AI includes in fact extremely different solutions, such as the digital assistants of our smart homes, the assisted driving systems of cars, the chatbots of the customer services of our suppliers, the intelligent robots of the 4.0 factories, but also all the algorithms that online recommend and report products and services of our liking.

The computational techniques referred to in the definition are primarily algorithms, i.e. the ordered and finished sequences of mathematical instructions that allow AI solutions to solve certain problems.


The big difference compared to the past, when artificial intelligence was already spoken about and algorithms obviously already existed, is the enormous amount of data that we have today at our disposal – coming also from online and from the Internet of Things – that allow us to feed artificial intelligence software. Talking about data and AI, we need to take a step back and talk about a very important field, that is machine learning: it is that set of methods that allow AI software to adapt to external inputs.

These methods allow machines to learn so that they can then perform a task or activity without being programmed in advance. Put another way, machine learning allows AI solutions to train themselves, learning from new facts (data) and also from mistakes, so that they can perform a given task/activity independently, faster and often better than humans.

A further advancement of machine learning is the so-called deep learning. Unlike traditional machine learning techniques, the system is able to learn the correct representation and solve machine learning problems without the need for data pre-processing. Thanks to the artificial neural networks of deep learning, artificial intelligence solutions become able to automatically analyze data such as images, video, audio or time series.


The foundations of modern artificial intelligence solutions are therefore datasets “annotated” by humans (supervised learning) or at least selected and prepared by them (unsupervised learning). It is no coincidence that one of the debates around AI concerns the danger of entering unbalanced datasets into software, which overestimate or underestimate the weight of certain variables, even by entering errors or bias (bias).

All this aside, there is no doubt that AI is able to improve the productivity of organizations of all sectors and sizes in two ways: on the one hand, by automating some activities previously carried out by people; on the other hand, by bringing systems to work and adapt to circumstances with an increasingly reduced if not absent human control.

The risk feared by many is that the full success of IA will lead to the cancellation of millions of jobs and, therefore, to problems for the entire global economy. Despite the risks inherent in this transformation, the opportunities appear superior: a study of twelve developed economies estimated that IA could double its annual rate of economic growth and increase labour productivity by up to 40 percent by 2035 compared to baseline levels.


But the fallout of artificial intelligence goes beyond the economic aspects, to involve the whole society, as the Coronavirus emergency is showing in these days. The healthcare sector, in fact, is destined to be increasingly revolutionized by AI solutions, which are already revolutionizing the world of diagnostics.

It is precisely in this area that some Chinese companies have developed a new diagnosis system based on artificial intelligence that promises to detect – through computerized tomographic scans (i.e. CT scans) – new cases of coronavirus with an accuracy rate of up to 96%, much faster than traditional swabs.


A classification of the different technological solutions that make reference to artificial intelligence has been carried out by the special observatory, which distinguishes between:


they are robots, more or less anthropomorphic in appearance, able to move, manipulate objects and perform actions without human intervention, drawing information from the surrounding environment and adapting to unforeseen or coded events.


behind this acronym there are all those objects able to perform actions and make decisions without the need for direct human intervention, thanks to sensors (thermometers, video cameras …) and actuators, with the ability to learn from the actions of people with whom they interact:


These are particular programs that interact with people simulating a human conversation through Artificial Intelligence. Used more and more in Customer care, the most advanced systems are able to understand the tone and context of the dialogue, store and reuse the information collected and demonstrate initiative during the conversation.


The observatory refers to all those solutions oriented to direct the preferences, interests, decisions of the user, based on information provided by the user, in an indirect or direct way. These are, in essence, all those suggestions we see in the online services we use on a daily basis.