Ch 1. Artificial Intelligence

blabbedharborIA et Robotique

23 févr. 2014 (il y a 3 années et 8 mois)

85 vue(s)

Ch 1. Artificial Intelligence

2

AI: History & Application

Artificial Intelligence





1.1 Definition of AI



1.2 AI technique



1.3 Criteria for success



1.4 AI application areas



1.5 Summary

Course Book

3

AI: History & Application

George F Luger

ARTIFICIAL INTELLIGENCE
6th edition

Structures and Strategies for Complex Problem Solving

4

AI: History & Application

1.1 Defining Artificial
Intelligence(1)


“AI is the study of how to make computers do things which, at
the moment, people do better.” (Rich)


“AI is the part of computer science concerned with designing
intelligent computer systems, that is, systems that exhibit
characteristics we associate with intelligent human behavior.


understanding language, reasoning, solving problems, and so on.”
(Barr)


“AI is the study of ideas which enable computers to do things
which make people seem intelligent.” (Winston)


AI is the study of intelligence using the ideas and methods of
computation.” (Fahlman)

5

AI: History & Application

Defining Artificial Intelligence(2)


“A bridge between art and science” (McCorduck)


“Tesler’s Theorem: AI is whatever hasn’t been
done yet.” (Hofstadter)


“AI is a field of science and engineering
concerned with the computational understanding
of what is commonly called intelligent behavior,
and with the creation of artifacts that exhibit such
behavior.” (Shapiro)


AI may be defined as the branch of computer
science that is concerned with automation of
intelligent behavior
. (Luger & Stubblefield)

6

AI: History & Application

Artificial Intelligence as Science


Understand and working of the mind in

mechanistic terms
, just as medical science seeks
to understand the working of the body in
mechanistic terms.


Understand intelligent thought processes
,
including perception, motor control,
communication using human languages,
reasoning, planning, learning, and memory.

7

AI: History & Application

AI as Engineering


How can we make computer based systems more
intelligent?


In practical terms, intelligence means


1. Ability to
automatically

perform tasks that currently require
human operators.


2. More
autonomy

in computer systems; less requirement for
human intervention or monitoring.


3.
Flexibility

in dealing with variability in the environment in
an appropriate manner.


4. Systems that are
easier

to use: able to understand what
the user wants from limited instructions.


5. Systems that can improve their performance by learning
from experience.

8

AI: History & Application

1.2 AI technique


A method that exploits knowledge that should be
represented in such a way that:


the knowledge captures generalizations.


It can be understood by people who must provide it.


It can easily be modified.


It can be used in a great many situations.


It can be used to help to narrow the range of possibilities.

Turing Test


Can we make the machine thinks like a human?


Assume that you ask questions and you don’t
know if you are talking to a human or a machine.

9

AI: History & Application

10

AI: History & Application

1.3
Criteria for Success


“How will we know if we have succeeded?”


Turing Test








The goal of the machine is to fool the judge into believing
that it is the person.


If the machine succeeds at this, then we will conclude that
the machine can think.

MACHINE

HUMAN

HUMAN

INTERFACE

CONTROLLED

BY JUDGE

‘INTELLIGENT SUBJECT’

JUDGE

QUESTION

QUESTION

ANSWER

ANSWER

QUESTION

ANSWER

Intelligent Agents


Intelligence is reflected by the collective behaviors
of large numbers of very simple interacting, semi
-
autonomous individuals or agents.


Agents are autonomous or semi
-
autonomous


Agents are situated


Agents are interactional


The society of agents is structured


11

AI: History & Application

Examples


Predict number of bread to be consumed in a city
on a given day, with minimum waste.


12

AI: History & Application

13

AI: History & Application

1.4
AI Application Areas


Two fundamental AI research areas


Knowledge Representation
: represent the computer’s knowledge of
the world by some kind of data structures in the machine’s memory


Search
: a problem
-
solving technique that systematically explores a
space of problem states


Game Playing


Automated Reasoning and Theorem Proving


Expert Systems


Natural Language Understanding and Semantic Modeling


Modeling Human Performance


Planning and Robotics


Machine Learning


Neural Nets and Genetic Algorithms

Artificial Intelligence

14

engine diagnosis system

Luger: Artificial Intelligence, 5
th

edition. © Pearson Education Limited, 2005

Fig II.
7
State space description of the automotive diagnosis problem.

16

AI: History & Application

Game Playing


Games are good vehicles for AI research because


most games are played using a well
-
defined set of rules


board configurations are easily represented on a computer


Games can generate extremely large search
spaces.


Search spaces are large and complex enough to require
powerful techniques(heuristics) for determining what
alternatives to explore in the problem space.


Luger: Artificial Intelligence, 5
th

edition. © Pearson Education Limited, 2005

Fig II.
5
Portion of the state space for tic
-
tac
-
toe.


18

AI: History & Application

19

AI: History & Application

Automated Reasoning and
Theorem Proving


Automatic Theorem Proving is the oldest branch of AI.


Theorem proving research was responsible for much of the early work
in formalizing search algorithms and developing formal representation
languages such as predicate calculus and logic programming
language PROLOG.


Variety of problems can be attacked by representing the
problem description and relevant background information as
logical axioms and treating problem instances as theorems to
be proved.


Reasoning based in formal mathematical logic is also
important.


Many problems such as the design and verification of logic circuits,
verification of the correctness of computer programs, and control of
complex systems require automated reasoning.

20

AI: History & Application

Expert Systems(
1
)


Expert systems are constructed by obtaining the
knowledge of a human expert and coding it into a
form that a computer may apply to similar
problems.


domain expert

provides the necessary knowledge of the
problem domain.


knowledge engineer

is responsible for implementing this
knowledge in a program that is both effective and intelligent
in its behavior.

21

AI: History & Application

Expert Systems(2)


Many successful expert systems


DENDRAL


designed to
infer the structure of organic molecules

from their
chemical formulas and mass spectrographic information about
the chemical bonds present in the molecules.


use the heuristic knowledge of expert chemists

to search into
the very large possible number of molecular structures.


MYCIN


used expert medical knowledge to diagnose and prescribe

treatment

for spinal meningitis and bacterial infections of the
blood.


Provided clear and logical explanations of its reasoning
, used a
control structure appropriate to the specific problem domain,
and identified criteria to reliably evaluate its performance.

22

AI: History & Application

Expert Systems(
3
)


Many successful expert systems (Continued)


PROSPECTOR


for
determining the probable location and type of ore deposits

based on geological information.


INTERNIST


for performing diagnosis in the area of internal medicine.


XCON


for configuring VAX computers.

23

AI: History & Application

Deficiencies of Current Expert
Systems

1
. Difficulty in capturing “deep” knowledge of the
problem domain


MYCIN lack any real knowledge of human physiology.

2
. Lack of robustness and flexibility

3
. Inability to provide deep explanations

4
. Difficulties in verification


may be serious when expert systems are applied to air traffic
control, nuclear reactor operations, and weapon systems.

5
. Little learning from experience

Example


Microsoft Windows Problem Troubleshooting
System

24

AI: History & Application

25

AI: History & Application

Natural Language Understanding
and Semantic Modeling(1)


One of the long
-
standing goals of AI is the creation
of programs that are capable of understanding
human language


Ability of understanding natural language seem to be one of the
most fundamental aspects of human intelligence


Successful automation would have an incredible impact on the
usability and effectiveness of computers


Real understanding of natural language depends on
extensive
background knowledge

about the domain
of discourse as well as an ability to apply general
contextual knowledge

to resolve ambiguities.

26

AI: History & Application

Natural Language Understanding
and Semantic Modeling(
2
)


Current work in natural language understanding
is devoted to
finding representational formalisms

that are general enough to be used in a wide
range of applications.


Stochastic models and approaches, describing
how sets of words “co
-
occur” in language

environments
, are used to characterize the
semantic content of sentences.

27

AI: History & Application

Modeling Human Performance


Design of systems that explicitly model some
aspect of human problem solving


If performance is the only criterion by which a system will be
judged, there may be little reason to attempt to simulate
human problem
-
solving methods.


Programs that take non human approaches to solving problems
are often more successful than their human counter parts


Human performance modeling has proved to be a powerful
tool for formulating and testing theories of human cognition.

28

AI: History & Application

Planning and Robotics


Planning attempts to
order the atomic actions

which robot can perform

in order to accomplish
some higher
-
level task.


Planning is a difficult problem because of vast
number of potential move sequences and
obstacles.


A blind robot performs a sequence of actions
without responding to changes in its environment
or being able to detect and correct errors in its
own plan.

29

AI: History & Application

Machine Learning(1)


Herbert Simon defines learning as “any change in
a system that allows it to
perform better

the
second time on repetition of
the same task

or on
another task drawn
from the same population
.”


Programs learn on their own, either from
experience, analogy, and examples or by being
“told” what to do.

30

AI: History & Application

Machine Learning(2)


What it is


Inducing a model from examples


What to do


Memorizing patterns


Generalizing the patterns



Well
-
known techniques


Naïve Bayesian


Hidden Markov Model


Maximum Entropy Model


Decision Tree


Support Vector Model

memorize

genaralize

31

AI: History & Application

Neural Nets and Genetic
Algorithms (
1
)


Neurally inspired models, also known as PDP or connectionist
systems, hold that intelligence arises in systems of simple,
interacting components(biological or artificial neurons)
through a process of learning or adaptation by which the
connections between components are adjusted.


Neural architectures are appealing as mechanisms for
implementing intelligence for a number of reasons.


Neural architectures seem to have more potential for partially
matching noisy and incomplete data.


Neural architectures are also more robust because knowledge is
distributed somewhat uniformly around the network.


Neural architectures also provide a natural model for parallelism.

32

AI: History & Application

Neural Nets and Genetic
Algorithms (
3
)


With genetic algorithms and artificial life we
evolve new problem solutions from components
of previous solutions.


For each new generation, the genetic operators,
such as
crossover

and

mutation
, work to produce
ever better potential problem solutions.


Artificial life

produces its new generation as a
function of the “quality” of its neighbors in
previous generations.

33

AI: History & Application

Features of Artificial
Intelligence(
1
)

1
. The use of computers to do symbolic reasoning,
pattern recognition, learning, or some other form
of inference.

2
. A focus on problems that do not respond to
algorithmic solutions. Rely on heuristic search as
an AI problem
-
solving technique.

3
. A concern with problem solving using inexact,
missing, or poorly defined information.

4
. Reasoning about the significant qualitative
features of a situation.

34

AI: History & Application

Features of Artificial
Intelligence(
2
)

5
. An attempt to deal with issues of semantic
meaning as well as syntactic form.

6
. Answers that are neither exact nor optimal, but
are in some sense “sufficient”.

7
. The use of large amounts of domain
-
specific
knowledge in solving problems.

8
. The use of meta
-
level knowledge to effect more
sophisticated control of problem solving
strategies.