Il problema della validit delle argomentazioni - Dipartimento di ...

Dipartimento di elettronica e informazione
Politecnico di Milano
Artificial Intelligence 2011‐12
1. Introduction
1.1 Course presentation
The School of Information Engineering of Politecnico di Milano offers two courses on
Artificial Intelligence (AI):
– Artificial Intelligence, in English, given by Marco Colombetti, Mario Verdicchio, and
Viola Schiaffonati (this course)
– Intelligenza artificiale, in Italian, given by Francesco Amigoni, Nicola Gatti, and Viola
Schiaffonati.
Topics
Both courses concern what can be called core AI, that is, the main topics of classical AI, as
developed since the late 1950s. More precisely, the course is concerned with seven topics:
1. Brief introduction to AI: aims, research areas, applications
2. The concept of a rational agent
3. State space search
4. Constraint satisfaction problems
5. Knowledge representation in First Order Logic
6. Planning
7. History and philosophical issues of AI.
A tentative timetable of all lessons can be found on the course’s website,
http://home.dei.polimi.it/colombet/AI/.
Textbook
The reading material is mainly constituted by lecture notes that can be downloaded from the
course’s website, http://home.dei.polimi.it/colombet/AI/. This material should not be
used by the students enrolled in Intelligenza artificiale (in Italian), as the two courses are
similar but far from identical.
As complementary reading we suggest:
– S. Russell, P. Norvig, 2010. Artificial Intelligence: A modern approach, 3rd edition,
Prentice‐Hall/Pearson (1,132 pages, of which about 300 covered by this course).
©Marco Colombetti, 2011. Lecture notes for the Artificial Intelligence course. This document can be downloaded
from the course’s website, http://home.dei.polimi.it/colombet/AI/, and is licensed under a Creative
Commons Attribution‐NonCommercial‐NoDerivs 3.0 Unported (CC BY‐NC‐ND 3.0) license, whose specification is
available at http://creativecommons.org/licenses/. Comments or criticisms can be sent to the author at
marco.colombetti@polimi.it.

©Marco Colombetti Artificial Intelligence 1 2
The textbook is not strictly necessary in view of the final test; however, students interested in
the subject are strongly encouraged to get it, as it gives a much wider perspective on AI than
one can get from this course alone.
Final test
The final examination consists in a written test, with conceptual questions and exercises.
Some of the past tests can be downloaded from the course’s website.
AI‐related courses at Politecnico di Milano (Leonardo Campus)
A number of AI related courses are given at Politecnico di Milano. Here is a list of the
courses offered in English at the Leonardo Campus (further courses are offered at the Como
Campus):
– Soft computing (neural networks, fuzzy models, genetic algorithms)
– Autonomous agents and multiagent systems
– Knowledge engineering (semantic modelling and reasoning for the Semantic Web)
– Robotics
– Cognitive robotics
– Data mining and text mining
– Image analysis
– Philosophical issues of computer science
– Videogame design and programming.
1.2 What is AI?
The birth of Artificial Intelligence
Officially, Artificial Intelligence (AI) was born in 1956, during a summer seminar at
Darthmouth College (Hanover, New Hampshire), attended by John McCarthy, Marvin
Minsky, Allen Newell, Nathaniel Rochester, Claude Shannon, Herbert Simon (later Nobel
Prize for Economics), and others.
Broadly speaking, the goals of AI are:
– to design and implement computer‐based systems that exhibit intelligent behaviour
– to understand human intelligence as a computational processes.
What is an intelligent system?
What is intelligence? This is a very difficult question. We can try to analyse what we mean
when we say that some entity is “intelligent.” That is, we can try to analyse the predicate
Intelligent(x). To analyse a predicate P(x), we try to understand its applicability conditions
and its truth conditions:
– applicability conditions: for what values of x is predicate P meaningful?
– truth conditions: for what values of x is predicate P true?
As far as the applicability conditions of Intelligent(x) are concerned, x has to be an agent, that
is, a system capable of autonomous action. In everyday speech, the term “intelligent” is
applied to agents, and by extension to their actions and possibly to the results or products of
their actions; examples:

©Marco Colombetti Artificial Intelligence 1 3
– intelligent agents: John is intelligent, my dog is very intelligent, my car has intelligent
suspensions
– intelligent actions: it was intelligent of John not to run away after breaking the glass
– intelligent product: this is a very intelligent chair (in the sense that it has been
intelligently designed).
Being intelligent is not a simple concept (contrary, for example, to being a metal, or being a
gas). As a consequence, we should not hope that there can be a simple theory of intelligence.
Maybe “intelligent” cannot even be turned into a rigorous scientific term.
Among the typical features of intelligent organisms are:
– perception of the environment
– problem solving and rational action (mostly human, but also present to some degree in
many non‐human animals)
– learning
– social interaction and communication
– language use (humans only).
What is an artificial intelligent system?
We obviously assume that an artificial intelligent system will be a system controlled by a
suitably programmed digital computer. Why?
The answer is that since the 1950s there has been a wide‐spread school of thought regarding
human intelligence as a computational process. According to this scientific tradition (often
called “computational cognitivism” or even simply “cognitivism”), to be intelligent means to
be a physical system implementing a suitable Turing machine; this implies that, at least in
principle, it is possible to completely reproduce human intelligence in a computer. Such a
radical view is not so popular any more among psychologists and philosophers; however,
many researchers in the field of AI would probably still endorse it.
Main research areas
– “core” AI: problem solving, knowledge representation, reasoning
– natural language processing
– machine learning
– computer vision
– robotics.
Main application areas
– expert systems
– planning and scheduling
– optimisation
– language technologies
– semantic web
– multiagent systems
– data mining
– industrial robotics
– ...