Artificial Intelligence, or simply AI, is probably one of the most fascinating technologies available today, and is commonly believed that it will be dominant up to at least 2050. Very recent major AI achievements are, for example, AlphaBot, capable of predicting the physical properties of the proteins, the Google Car, capable of traveling millions of kms without making incidents, GPT3 software, capable of producing newspaper articles indistinguishable from those written by humans, and Pluribus bot, capable of beating the top 5 poker players of the international circuit of Texas Poker Hold’em no limit. Even if today almost everybody speaks of AI, addressing the question “what is AI?” is not a simple task. During the following lectures, we will analyze in depth how the definition of AI evolved during time. Here, our goal is just to provide a brief sketch to the concept of AI and then introduce the structure of the concept discussed later. Intuitively, AI is intelligence demonstrated by machines and provided by humans to machines, unlike the natural intelligence displayed by humans and animals. Since there is no clear definition of what is intelligence, the customary approach is to distinguish intelligent entities from those that are not according to the behavior they exhibit. Such an approach requires the definition of some performance metrics that we usually aim at maximizing or minimizing. These metrics need to be formally defined, thus making the resort to formal and scientific tools such as mathematics and logics natural. For instance: when traveling from a city to another city the metric is the time spent and the goal is its minimization, when playing checkers or chess the metric is the victory probability and the goal is its maximization, when identifying characters or images the metric is the error probability and the goal is its minimization. Surprisingly, what AI is today, it might not be tomorrow. Think just of the optical character recognition task. Since today it is a routine technology, it is not considered as an AI tool anymore. This phenomenon is known as AI effect. The idea is that, as machines become increasingly capable, tasks considered to require "intelligence" are often removed from the definition of AI. According to a definition proposed by Larry Tesler, a computer scientist that worked for Apple, Xerox, Yahoo!, Amazon, “AI is whatever hasn’t been done yet”. For a better comprehension of how AI evolved during time, it is useful distinguishing four main time periods that we will analyze in the following lectures: the first period is before the birth of AI, that was in 1956, and is characterized by the AI precursors, the second period is the birth of AI, the third period is from immediately after the birth of AI up to the end of the previous century; this period, thanks the initial AI successes and failures, paved the way to the explosion of the AI field in our century, the fourth period is from the beginning of our century to today.
