Abstract: Artificial Intelligence (AI) has by now become pervasive in our every day life. In this seminar I will provide a very basic introduction to AI, reviewing the underlying ideas and the main models. These are inspired by advances in neuroscience and are based on the interpretation of information processing and communication patterns in a nervous system.
I will especially focus on models designed for accomplishing two special tasks, that is, "learning" and "retrieval". Briefly, the former is the ability to improve the performance through experience, the latter is the ability to exploit the information previously learnt. These smart capabilities emerge as collective features of network models based on elementary units (i.e., suitably stilyzed neurons), that interact each other in a dense, frustrated and non-linear way.
Finally, I will introduce the so-called deep learning algorithms, which is the subject of frontier AI research nowadays: these algorithms use a cascade of many layers of elementary units, whence they achieve by far better perfomances in terms of accuracy and of manageable information complexity.