Surely we all have heard, at least once in life, of “Artificial Intelligence”, a notion that has now entered into everyday terminology through science fiction books but also, and increasingly, through “easy” movies, telefilms and contact channels “direct”. Some of us have also recently heard about “Automatic Learning”, perhaps in “Neural Networks” or “Decision Algorithms” speeches.
But what is Automatic Learning, and what has it to do with artificial intelligence?
Well, automatic learning, or machine learning, is nothing more than a particular kind of Artificial Intelligence that provides computers with the ability to learn how to solve problems without being specifically programmed for that purpose. That is, computers learn how to behave through the data and not (directly) from the program code. Instead of having specific decision-making processes predetermined by a programmer for each individual activity, the automatic learning software analyzes the data already available to him on similar issues, and uses statistical analysis to define conceptual models; Then, apply these templates to new data to make decisions or make predictions. The “behind” technology of Machine Learning algorithms is that called “Artificial Neural Networks”, which is nothing more than a mathematical model composed of artificial neurons inspired by a natural neural network, in practice to a brain. Obviously, for results from automatic learning algorithms to be as accurate as possible, huge amounts of data, possibly coming from different sources, need to create a broader knowledge base that can handle, in the best way, those situations where there is no past human experience. Machine learning algorithms are already available, but the ability to apply automatically, and quickly, the necessary, complex, large mathematical math calculations is available only from recent times. As soon as the computing power has become sufficient, however, there has been a flourishing of applications that, even in their admiration, have become, or are becoming, part of our daily lives.
- The Google car, that is, the car driving alone;
- “Targeted” online tips, those of Amazon and Netflix sites;
- Facial recognition, the technology that allows Facebook (as well as others) to recognize faces and tag them automatically whenever a picture is posted on their profile;
- Social intelligence, that is, the ability to analyze the “emotional content” (positive, negative or neutral opinion on certain topics) of postings and interventions on social networks and draw conclusions;
- Virtual Siri servers, Cortana (and others) present on our smartphones.
Such applications reveal all the power, as well as the great practical utility, of automatic learning algorithms and reveal how such tools have already become, perhaps unknowingly. Constant part of our reality. And every day, more and more, there is news or news of new uses; and it is often Google, through its research labs and innovative projects, to find new applications. In February, he unveiled an artificial intelligence system that, based on “pixellate” images, succeeds in creating a fairly reliable reconstruction of the original photography; in April, he announced that his Translator has improved the accuracy of Italian translations thanks to artificial intelligence algorithms; always in April, launched a web portal that uses automatic learning algorithms to interpret “sketches” or sketches of an object and “replace” them with accurate and professional drawings for the same object. But the best is still beyond coming. In fact, Machine Learning is a powerful tool whose true strength is still to be fully expressed and whose future uses are not limited to business and business services. Thanks to the constant technological evolution and the increase in computing power, Automatic Learning will soon be able to evolve from an instrument of discovery and innovation, and will play an increasingly prominent role in areas such as research and science. In summary, algorithms that leverage Machine Learning can do exactly what scientists and inventors now do, but better and faster. The hope is that one day the software will solve irreplaceable issues at this time. That is, hope is for example to build a clever Machine Learning to the point of allowing him to use his algorithms to find cure for diseases such as cancer or AIDS. The future is likely to be surprising, in ways that today, perhaps, we are not even able to imagine.
Giovanni Calcerano