AI: what is Artificial Intelligence, main tools for streamlining and automating processes and workflows
Where algorithms are applied and what advantages they ensure, from Marketing to Recruiting. A guide to clarify the real opportunities for business and the solutions available. Updated data from the Artificial Intelligence Observatory of the Polytechnic of Milan
Artificial Intelligence is already part of our daily lives and is now used in more than half of large Italian companies. Everyone knows about driverless cars or voice assistants like Apple's Siri, Microsoft's Cortana or Google's Alexa, but there are many lesser-known examples.
Intelligent algorithms, capable of self-learning, suggest products to buy, films or songs in line with our tastes, they know how to answer customer questions via chat, they can recognize a person's face to enable a access, sort documents based on content, support doctors in reading x-ray images and diagnoses, filtering CVs to select the ideal candidate. And so on.The Polytechnic of Milan provides this definition of AI: “Artificial Intelligence, in English Artificial Intelligence (AI), is the branch of computer science that studies the development of hardware and software systems equipped with capabilities typical of human beings and capable to autonomously pursue a defined purpose by making decisions that, until that moment, were usually entrusted to human beings.”
The typical capabilities of human beings concern, specifically, the understanding and processing of natural language (NLP – Natural Language Processing) and images (Image Processing), learning, reasoning and the ability to plan and interact with people, machines and the environment. Unlike traditional software, an AI system is not based on programming (i.e. on the work of developers who write the operating code of the system) but on learning techniques. That is, algorithms are defined that process an enormous amount of data from which the system itself must derive its understanding and reasoning capabilities.
Machine Learning
These are systems that serve to “train” the software so that by correcting errors it can learn to carry out a task/activity independently.
For example, the mechanical arm supported by AI, and therefore intelligent, is able to assemble a piece even if it is not where it should be because the control algorithm, instead of providing the coordinates, activates an optical recognition that searches for the piece everywhere the area that the arm can reach.
And if the machine or the man offering the pieces repeats the mistake several times, the robot learns that that is the new position and immediately goes to look for the piece there. Machine Learning is evolving along a line of research based on the use of neural networks organized into multiple levels of depth and therefore called Deep Learning.
Deep Learning
These are recently developed learning models (since 2012) inspired by the structure and functioning of our brain, which emulate the human mind.
In this case, the mathematical model alone is not enough: Deep Learning requires ad hoc designed artificial neural networks (deep artificial neural networks) and a very powerful computational capacity capable of “supporting” different layers of calculation and analysis (which is what happens with the neural connections of the human brain).
It may seem like a futuristic level of technology but in reality these are systems already in use in pattern recognition, voice or image recognition and NLP – Natural Language Processing systems.
