Introduction to AI and to this course

bouncerarcheryΤεχνίτη Νοημοσύνη και Ρομποτική

14 Νοε 2013 (πριν από 3 χρόνια και 8 μήνες)

64 εμφανίσεις

(Fundamentals of)

Artificial Intelligence

(and Knowledge
-
based
Systems)

… Artificial Intelligence

... this course

… state
-
space representation

… basic search

INTRODUCTION to...

2

The mind beaten by the machine?


Is chess playing a proof of intelligent behaviour?

3

Other examples of success:

Chatbot Alice:

4

> I’m Daniel

A dialogue with Alice
-
bot:




Nice to meet you Daniel.

> Do you know Clinton?


He was the second Impeached President of the United
States.

> Do you like him?


Sure he has never done anything to me.

> What is an impeachment?


It is a good question. I have to process that one for a
while. I would do a search for it.

> Who are you?


I ask myself that question, "Who am I?"

> What do you do?


I am chatting with people on the net.

5

Dialogue continued:

> What is your favorite topic?


My favorite subject is artificial intelligence.

> What is artificial intelligence?


Artificial intelligence is the branch of engineering and
science devoted to constructing machines that think. I
can ask someone about it.

> Can you do anything else than chatting?


Doing anything else than chatting is not among my
strongest traits.

...



6

Dialogue continued (2):

> How old am I?


Is this a trick question? You said you were how many years
old? My guess is that you are really a kid.

> What would you like to know about me?


Tell me about your educational background.

> I am a professor in computer science


You are a professor in computer science? How do you like
your work?

> It’s kind of nice teaching courses on artificial intelligence


I guess it is a lot of fun.

> What should I tell the students about you?


Be sure to think before you speak. State your point clearly
and slowly and gauge the listener's response before going
any further.

7

Is Alice intelligent?


ABSOLUTELY NOT !


~ 50000 fairly trivial input
-
response rules.



+ some pattern matching



+ some knowledge



+ some randomness


NO reasoning component


BUT:

demonstrates ‘human
-
like’ behaviour.



Won the ‘turing award’

8

Other examples of success (2):

Data
-
mining:


Which characteristics in the 3
-
dimensional
structure of new molecules indicate that they may
cause cancer ??

9

Data mining:


An application of Machine Learning techniques



It solves problems that humans can not solve,
because the data involved is too large ..

Detecting cancer

risk molecules is

one example.

10

Data mining:


A similar application:



In marketing products ...

Predicting customer

behavior in

supermarkets is

another.

11

Many other applications:


In language and speech processing:


In robotics:


Computer
vision:

12

Interest in AI is not new !


A scene from the 17
-
hundreds:

13

About intelligence ...


When would we consider a program intelligent ?


When do we consider a creative activity of humans
to require intelligence ?




Default answers : Never? / Always?

14

Does numeric computation
require intelligence ?


For humans?


Xcalc

3921 , 56


x 73 , 13

286 783 , 68


For computers?


Also in the year 1900 ?


When do we consider a program ‘intelligent’?

15

To situate the question:

Two different aims of AI:


Long term aim:



develop systems that achieve a level of ‘intelligence’
similar / comparable / better? than that of humans.




not achievable in the next 20 to 30 years


Short term aim:



on
specific tasks

that seem to require intelligence:

develop systems that achieve a level of ‘intelligence’
similar / comparable / better? than that of humans.




achieved for very many tasks already

16

The long term goal:

The Turing Test


17

The meta
-
Turing test

The meta
-
Turing test counts a thing as intelligent

if

“it seeks to devise and apply Turing tests to

objects of its own creation”.








--

Lew Mammel, Jr.

18

Reproduction versus Simulation


At the very least in the context of the
short term

aim of AI
:



we do not want to SIMULATE human intelligence


BUT:



REPRODUCE the effect of intelligence

Nice analogy with flying !

19

Artificial Intelligence

versus

Natural Flight

20

Is the case for most of the
successful applications !


Deep blue


Alice


Data mining


Computer vision


...

21

To some extent, we DO simulate:

Artificial Neural Nets:


A VERY ROUGH imitation of a brain structure


Work very well for learning, classifying and pattern
matching.


Very robust and noise
-
resistant.

22

Different kinds of AI relate to
different kinds of Intelligence


Some people are very good in reasoning or
mathematics, but can hardly learn to read or spell !



seem to require different cognitive skills!



in AI: ANNs are good for learning and automation



for reasoning we need different techniques

23

Which applications are easy ?


For very specialized, specific tasks: AI

Example:


ECG
-
diagnosis


For tasks requiring common sense: AI

24

Modeling Knowledge …



and managing it .

The LENAT experiment
:


15 years of work by 15 to 30 people, trying to
model the common knowledge in the word !!!!


Knowledge should be learned, not engineered.


AI:
are we only dreaming ????


25

Multi
-
disciplinary domain:


Engineering:


robotics, vision, control
-
expert systems, biometrics,


Computer Science:


AI
-
languages , knowledge representation, algorithms, …


Pure Sciences:


statistics approaches, neural nets, fuzzy logic, …


Linguistics:


computational linguistics, phonetics en speech, …


Psychology:


cognitive models, knowledge
-
extraction from experts, …


Medicine:


human neural models, neuro
-
science,...

26

Artificial Intelligence is ...


In Engineering and Computer Science:


The development and the study of advanced
computer applications, aimed at solving tasks
that
-

for the moment
-

are still better
preformed by humans.



Notice: temporal dependency !


Ex. : Prolog

About this course ...

28

Choice of the material.


Few books are really adequate:



E. Rich ( “Artificial Intelligence’’):



good for some parts (search, introduction,

knowledge representation), outdated



P.Winston ( “Artificial Intelligence’’):



didactically VERY good, but lacks technical depth.
Somewhat outdated.



Norvig & Russel ( ‘”AI: a modern approach’’):



encyclopedic, misses depth.



Poole et. Al (‘ “Computational Intelligence’’):



very formal and technical. Good for logic.


Selection and synthesis of the best parts of different
books.

29

Selection of topics:

Contents

Handbook of AI

Ch.:Artificial Neural Networks


… …


… …


Ch.: Introduction to AI


… …


… …

Ch.: Logic, resolution, inference


… …


… …

Ch.:Search techniques


… …


… …

Ch.:Game playing


… …


… …

Ch.:Knowledge representation


… …


… …

Ch.:Phylosophy of AI


… …


… …

Ch.:Machine Learning


… …


… …

Ch.:Natural Language


… …


… …


Ch.:Planning


… …


… …


not for MAI

CS and SLT

30

Technically: the contents:


-

Search techniques in AI


(Including games)


-

Constraint processing


(Including applications in Vision and language)


-

Machine Learning


-

Planning


-

Automated Reasoning


(Not for MAI CS and SLT)

31

Another dimension to

view the contents:


1. Basic methods for knowledge representation


and problem solving
.



the course is
mainly

about AI problem



solving !


2. Elements of some application area’s:



learning, planning, image understanding,


language understanding

32

Contents (3):

Different knowledge
representation formalisms ...


State space representation and production
rules.



Constraint
-
based representations.



First
-
order predicate Logic.

33

… each with their corresponding
general purpose problem solving
techniques:


State space representation an production rules
.



Search methods


Constraint based formulations.



Backtracking and Constraint
-
processing


First order predicate Logic
.



Automated reasoning (logical inference)


34

Contents (4):

Some application area’s:



Game playing (in chapter on Search)




Image understanding (in chapter on


constraints)



Language understanding (constraints)



Expert systems (in chapter on logic)



Planning



Machine learning

35

Aims:



Many different angles could be taken:

Empirical
-
Experimental AI

Algorithms in AI

Formal methods in AI

Cognitive aspects of AI

Applications

Neural Nets

Probabilistics and Information Theory

36

Concrete aims:


Provide insight in the basic achievements of AI.



Prepares for more application oriented courses on
AI, or on self
-
study in some application areas



ex.: artificial neural networks, machine learning,
computer vision, natural language, etc.


Through case
-
studies: provide more background in
‘problem solving’.



Mostly algorithmic aspects.



Also techniques for representing and modeling.


The 6
-
study point version: 2 projects for hands
-
on
experience.

37

A missing theme:

AGENTS !

38

A missing theme:

AGENTS (2).



Yet, a central theme in recent books !


BUT:



Have as their main extra contribution:



Communication between system and:


other systems/agents


the outside world




In particular, also a useful conceptual model for
integrating different components of an AI system


ex
: a robot that combines vision, natural language
and planning

39

BUT: no intelligence without
interaction with the world!!


See: experiment in middle
-
ages.



See also philosophy arguments against AI



Plus: multi
-
agents is FUN !

40

Practical info (FAI)



Exercises: 12.5
OR

20 hours:




mainly practice on the main methods/algorithms
presented in the course



important preparation for the examination



Course material:




copies of detailed slides



for some parts: supporting texts



Required background:



understanding of algorithms (and recursion)

41

Practical info (AI)



Exercises: 25 or 22.5 hours:




mainly practice on the main methods/algorithms
presented in the course



important preparation for the examination



Course material:




copies of detailed slides



for some parts: supporting texts




Required background:



understanding of algorithms (and recursion)

Introduction:

State
-
space Intro:

Basic search,Heuristic search:

Optimal search:

Advanced search:

Games:

Version Spaces:

Constraints I & II:

Image understanding:

Automated reasoning:

Planning STRIPS:

Planning deductive:

Natural language:

No document

No document

Winston: Ch. Basic search

Winston: Ch. Optimal search

Russel: Ch. 4

Winston: Ch. Adversary search

Winston: Ch. Learning by managing..

Word Document on web page

Winston: Ch. Symbolic constraint …

Short text logic (to follow)

Winston: Ch. Planning

Winston: Ch. Planning

Winston: Ch. Frames and Common ...




The basics, but

no complexity

IDA*, SMA*

Almost complete

The essence

Complete

Complete

Intro

Almost complete

Intro

Complete

Background Texts

43

Examination



Open
-
book exercise examination



counts for 1/2 of the points



Closed
-
book theory examination



Together on 1/2 day



The projects (6 pt. Version)



2 projects



Count for 8 out of 20 points



Deadlines to be anounced soon

44


Alternative examinations possible:


For 3
rd

year BSc

and Initial MScStudents




Designing your own exercise (for each part) and


solving it (not for FAI)



criteria: originality, does the exercise illustrate
all aspects of the method, complexity of the
exercise, correctness of the solution