Artificial Intelligence A Brief Introduction

Download Report

Transcript Artificial Intelligence A Brief Introduction

aslab.org
Artificial Intelligence
A Brief Introduction
Ricardo Sanz
May 20, 2004
aslab
autonomous
laboratory
Sanz / Artificialsystems
Intelligence:
An Introduction
2004/03/24
1
Contents





aslab
Basic Ideas
History
Technology
Robots
Agents
Sanz / Artificial Intelligence: An Introduction
2004/03/24
2
Core Ideas
What is AI ?
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
3
What is AI?

Acting humanly: The Turing test (1950)


Thinking humanly: Cognitive modeling



aslab
What do we need to pass the test
“Think-aloud” to learn from human and recreate in computer
programs (GPS)
Thinking rationally: Syllogisms, Logic
Acting rationally: A rational agent
Sanz / Artificial Intelligence: An Introduction
2004/03/24
4
Foundations of AI

Philosophy (428 B.C. - Present) – reasoning and
learning




aslab
Can formal rules be used to draw valid conclusions?
How does the mental I arise from a physical brain?
Where does knowledge come from?
How does knowledge lead to action?
Sanz / Artificial Intelligence: An Introduction
2004/03/24
5
Foundations of AI

Mathematics (c. 800 - Present) - logic, probability,
decision making, computation




Economics (1776-present)



aslab
What are the formal rules to draw conclusions?
What can be computed?
How do we reason with uncertain information?
How should we make decisions so as to maximize payoff?
How should we do this when others may not go along?
How should we do this when the payoff may be far in the
future?
Sanz / Artificial Intelligence: An Introduction
2004/03/24
6
Foundations of AI

Neuroscience (1861-present)


Psychology (1879 - Present) - investigating human
mind


How do humans and animals think and act?
Computer engineering (1940 - Present) - ever
improving tools

aslab
How do brains process information
How can we build an efficient computer?
Sanz / Artificial Intelligence: An Introduction
2004/03/24
7
Foundations of AI

Control theory and Cybernetics (1948-present)


Linguistics (1957 - Present) - the structure and
meaning of language

aslab
How can artifacts operate under their own control?
How does language relate to thought?
Sanz / Artificial Intelligence: An Introduction
2004/03/24
8
What is Intelligence?

1.
2.
3.
aslab
Intelligence, taken as a whole, consists of the
following skills:
the ability to reason
the ability to acquire and apply knowledge
the ability to manipulate and communicate ideas
Sanz / Artificial Intelligence: An Introduction
2004/03/24
9
An Intelligent Entity
INTERNAL
PROCESSES
INPUTS
Senses environment
See
Hear
Touch
Taste
Smell
Has knowledge
Has understanding/
intentionality
Can Reason
Exhibits behaviour
OUTPUTS
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
10
The Age of Intelligent Machines
aslab

1st Industrial Revolution: the Age of Automation:
Machines extend & multiply man's physical
capabilities

2nd Industrial Revolution: the Age of Info Tech:
Machines extend & multiply man's mental capabilities

Knowledge Revolution?: the Age of Knowledge
Technology "..working smarter, not harder."

How do we make our systems smarter? - by building
in intelligence?
Sanz / Artificial Intelligence: An Introduction
2004/03/24
11
More Definitions of AI
aslab

AI is the science of making machines do things that
would require intelligence if done by humans
Marvin Minsky

AI is the part of computer science concerned with
designing intelligent computer systems
Ed Feigenbaum

Systems that can demonstrate human-like reasoning
capability to enhance the quality of life and improve
business competitiveness
Japan-S’pore AI Centre
Sanz / Artificial Intelligence: An Introduction
2004/03/24
12
Behaviourist’s View on Intelligent Machines




aslab
Many scientists believe that only things that can be
directly observed are “scientific”
Therefore if a machine behaves “as if it were
intelligent” it is meaningless to argue that this is an
illusion.
Turing was of this opinion and proposed the “Turing
Test”
This view can be summarized as:“If it walks like a
duck, quacks like a duck and looks like a duck - it is a
duck”
Sanz / Artificial Intelligence: An Introduction
2004/03/24
13
Turing’s Test



aslab
In 1950 Alan Turing published his now famous paper
"Computing Machinery and Intelligence." In that
paper he describes a method for humans to test AI
programs.
In its most basic form, a human judge sits at a
computer terminal and interacts with the subject by
written communication only. The judge must then
decide if the subject on the other end of the computer
link is a human or an AI program imitating a human.
http://www.turing.org.uk/turing/
Sanz / Artificial Intelligence: An Introduction
2004/03/24
14
Turing’s Test - Part 1
Which one’s the man?
A
B
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
15
Turing’s Test - Part 2

Which one’s the computer?
A
aslab
B
Sanz / Artificial Intelligence: An Introduction
If the computer succeeds in
fooling the judge then it has
managed to exhibit a human
level of intelligence in the task
of pretending to be a woman,
the definition of intelligence
the machine has shown itself
to be intelligent.
2004/03/24
16
Some History
From hype to work
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
17
Brief History of AI

Gestation of AI (1943 -1955)



Birth of AI (1956)



A 2-month Dartmouth workshop of 10 attendees – the name
of AI
Newell and Simon’ Logic Theorist
Early enthusiasm, great expectations (1952 - 1969)

aslab
McCulloch and Pitts’s model of artificial neurons
Minsky’s 40-neuron network
GPS by Newell and Simon, Lisp by McCarthy, Blockworld by
Minsky
Sanz / Artificial Intelligence: An Introduction
2004/03/24
18
Brief History of AI

AI facing reality (1966 - 1973)

Many predictions of AI coming successes





Knowledge is power, acquiring knowledge from experts
Expert systems (MYCIN)
AI - an industry (1980 - present)

aslab
Machine translation – Syntax is not enough
Intractability of the problems attempted by AI
Knowledge-based systems (1969 - 1979)


A computer would be a chess champion in 10 years (1957)
Many AI systems help companies to save money and
increase productivity
Sanz / Artificial Intelligence: An Introduction
2004/03/24
19
Brief History of AI

The return of neural networks (1986 – present)



AI – a science (1987 – present)




Working agents embedded in real environments with
continuous sensory inputs
AI - conscious machines (Now !!)

aslab
Build on existing theories vs. propose brand new ones
Rigorous empirical experiments
Learn from data – data mining
AI – intelligent agents (1995 – present)


PDP books by Rumelhart and McClelland
Connectionist models vs. symbolic models
Making machines that feel and and have a self
Sanz / Artificial Intelligence: An Introduction
2004/03/24
20
History of AI
Degree of
Motivation
Dartmouth
Conference
Japan 5th
Generation
Computer
Support
Technology
AI
Winter
1948
aslab
1970s - 80s
Sanz / Artificial Intelligence: An Introduction
mid-1980s
mid-1990s
Time
2004/03/24
21
Examples of AI systems








aslab
Robots
Chess-playing program
Voice recognition system
Speech recognition system
Grammar checker
Pattern recognition
Medial diagnosis
System malfunction rectifier
Sanz / Artificial Intelligence: An Introduction






Game Playing
Machine Translation
Resource Scheduling
Expert systems (diagnosis,
advisory, planning, etc)
Machine learning
Intelligent interfaces
2004/03/24
22
AI Case Study - RoboCup



aslab
The Robocup Competition
pits robots (real and virtual)
against each other in a
simulated soccer
tournament.
The aim of the RoboCup
competition is to foster an
interdisciplinary approach to
robotics and agent-based AI
by presenting a domain that
requires large-scale cooperation and coordination in
a dynamic, noisy, complex
environment.
Common AI methods used
are variants of neural
networks and genetic
algorithms.
Sanz / Artificial Intelligence: An Introduction
2004/03/24
23
Intelligent Technologies
Resources for Sophisticated
Information Processing
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
24
Knowledge-Based Systems (KBS)
User interface
may employ:
Knowledge-base
editor
QuestionandAnswer,
Menu-driven,
General Knowledge-base
Inference engine
Natural
language,
User
Etc.
aslab
Case-specific data
Graphics
Interface
Styles
Explanation
subsystem
Sanz / Artificial Intelligence: An Introduction
2004/03/24
25
Artificial Neural Networks

What are Artificial Neural Networks (ANNs)?



ANN or connecionist systems are systems that were developed based on
the learning characteristics of biological creatures.
ANN solve problems though a process of learning and adaptation.
How are ANNs represented?
Synapse
Neuron
Outputs
Inputs
Connection
between neurons
Input
Plant
Output
Sensors
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
26
Genetic Algorithms


We will use the processes loosely based on natural
selection, crossover, and mutation to find solutions to
certain problems.
GAs are adaptive (search, learning) methods based
on the genetic processes of biological organisms.
1st generation of possible solutions
2nd generation of possible solutions
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
27
Fuzzy Logic


Precision in the model

For systems with little complexity, hence little uncertainty, closed-form
mathematical expressions provide precise description of the system.
For systems that are a little more complex, but for which significant data
exists, model free methods such as artificial ANNs, provide a powerful
and robust means to reduce uncertainty through learning.
For most complex systems where few numerical data exists and where
only ambiguous or imprecise information may be available, fuzzy
reasoning provides a way to understand system behavior.
Mathematical
equations
Model-free
Methods
(e.g., ANNs)
Fuzzy Systems
Complexity (uncertainty) of the system
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
28
Towards intelligent
machines
Are we ready to build the next
generation of intelligent robots?
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
29
Some problems remain…





aslab
Vision
Audition / speech processing
Natural language processing
Touch, smell, balance and other senses
Motor control
Sanz / Artificial Intelligence: An Introduction
2004/03/24
30
Computer Perception


Perception: provides an agent information about its
environment. Generates feedback. Usually proceeds in the
following steps.
Sensors: hardware that provides raw measurements of
properties of the environment






aslab
Ultrasonic Sensor/Sonar: provides distance data
Light detectors: provide data about intensity of light
Camera: generates a picture of the environment
Signal processing: to process the raw sensor data in order to
extract certain features, e.g., color, shape, distance, velocity, etc.
Object recognition: Combines features to form a model of an
object
And so on to higher abstraction levels
Sanz / Artificial Intelligence: An Introduction
2004/03/24
31
Perception for what?









aslab
Interaction with the environment, e.g., manipulation, navigation
Process control, e.g., temperature control
Quality control, e.g., electronics inspection, mechanical parts
Diagnosis, e.g., diabetes
Restoration, of e.g., buildings
Modeling, of e.g., parts, buildings, etc.
Surveillance, banks, parking lots, etc.
…
And much, much more
Sanz / Artificial Intelligence: An Introduction
2004/03/24
32
Sample perception: Computer vision
1.
Grab an image of the object (digitize analog signal)
2.
Process the image (looking for certain features)
1.
2.
3.
4.
aslab
Edge detection
Region segmentation
Color analysis
Etc.
3.
Measure properties of features or collection of features (e.g.,
length, angle, area, etc.)
4.
Use some model for detection, classification etc.
Sanz / Artificial Intelligence: An Introduction
2004/03/24
33
State of the art
aslab

Can recognize faces? – yes

Can find salient targets? – sure

Can recognize people? – no problem

Can track people and analyze their activity? – yep

Can understand complex scenes? – not quite but
in progress
Sanz / Artificial Intelligence: An Introduction
2004/03/24
34
Face recognition case study
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
35
Pedestrian recognition
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
36
How about other senses?
aslab

Speech recognition -- can achieve userundependent recognition for small vocabularies and
isolated words

Other senses -- overall excellent performance (e.g.,
using gyroscopes for sense of balance, or MEMS
sensors for touch) except for olfaction and taste,
which are very poorly understood in biological
systems also.
Sanz / Artificial Intelligence: An Introduction
2004/03/24
37
How about actuation



aslab
Robots have been used for a long time in restricted
settings (e.g., factories) and, mechanically speaking,
work very well.
For operation in unconstrained environments,
Biorobotics has proven a particularly active line of
research:
Motivation: since animals are so good at navigating
through their natural environment, let’s try to build
robots that share some structural similarity with
biological systems.
Sanz / Artificial Intelligence: An Introduction
2004/03/24
38
Robot examples: constrained environments
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
39
Towards unconstrained environments
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
40
They’re here …
Robot lawn mowers and vacuum-cleaners are here already…
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
41
The time is now

It is a particularly exciting time for AI because…



aslab
CPU power is getting not a problem anymore
Many physically-capable robots are available
Some vision and other senses are partially available

Many AI algorithms for constrained environment are
available

So for the first time we have all the components
required to build smart robots that interact with the
real world.
Sanz / Artificial Intelligence: An Introduction
2004/03/24
42
Agents
Recent IA software focus
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
43
What is an Agent?

in general, an entity that interacts with its environment


aslab
perception through sensors
actions through effectors or actuators
Sanz / Artificial Intelligence: An Introduction
2004/03/24
44
Examples of Agents

human agent


eyes, ears, skin, taste buds, etc. for sensors
hands, fingers, legs, mouth, etc. for effectors


robot


camera, infrared, bumper, etc. for sensors
grippers, wheels, lights, speakers, etc. for effectors


often powered by motors
software agent

functions as sensors


information provided as input to functions in the form of
encoded bit strings or symbols
functions as effectors

aslab
powered by muscles
results deliver the output
Sanz / Artificial Intelligence: An Introduction
2004/03/24
45
Agents and Their Actions

a rational agent does “the right thing”


problems:


aslab
the action that leads to the best outcome
what is “ the right thing”
how do you measure the “best outcome”
Sanz / Artificial Intelligence: An Introduction
2004/03/24
46
Performance of Agents

criteria for measuring the outcome and the expenses
of the agent



aslab
often subjective, but should be objective
task dependent
time may be important
Sanz / Artificial Intelligence: An Introduction
2004/03/24
47
Performance Evaluation Examples


vacuum agent
A number of tiles cleaned during a certain period


based on the agent’s report, or validated by an objective
authority
doesn’t consider expenses of the agent, side effects


might lead to unwanted activities

aslab
energy, noise, loss of useful objects, damaged furniture,
scratched floor
agent re-cleans clean tiles, covers only part of the room, drops
dirt on tiles to have more tiles to clean, etc.
Sanz / Artificial Intelligence: An Introduction
2004/03/24
48
Rational Agent considerations



performance measure for the successful completion
of a task
complete perceptual history (percept sequence)
background knowledge

especially about the environment



task, user, other agents
feasible actions

aslab
dimensions, structure, basic “laws”
capabilities of the agent
Sanz / Artificial Intelligence: An Introduction
2004/03/24
49
Omniscience

a rational agent is not omniscient



rationality takes into account the limitations of the
agent


aslab
it doesn’t know the actual outcome of its actions
it may not know certain aspects of its environment
percept sequence, background knowledge, feasible actions
it deals with the expected outcome of actions
Sanz / Artificial Intelligence: An Introduction
2004/03/24
50
Ideal Rational Agent

selects the action that is expected to maximize its
performance


aslab
based on a performance measure
depends on the percept sequence, background knowledge,
and feasible actions
Sanz / Artificial Intelligence: An Introduction
2004/03/24
51
From Percepts to Actions

if an agent only reacts to its percepts, a table can
describe the mapping from percept sequences to
actions


instead of a table, a simple function may also be used
can be conveniently used to describe simple agents that
solve well-defined problems in a well-defined environment

aslab
e.g. calculation of mathematical functions
Sanz / Artificial Intelligence: An Introduction
2004/03/24
52
Agent or Program


our criteria so far seem to apply equally well to
software agents and to regular programs
autonomy





aslab
agents solve tasks largely independently
programs depend on users or other programs for “guidance”
autonomous systems base their actions on their own
experience and knowledge
requires initial knowledge together with the ability to learn
provides flexibility for more complex tasks
Sanz / Artificial Intelligence: An Introduction
2004/03/24
53
Structure of Intelligent Agents


Agent = Architecture + Program
architecture

operating platform of the agent


program

aslab
computer system, specific hardware, possibly OS functions
function that implements the mapping from percepts to
actions
Sanz / Artificial Intelligence: An Introduction
2004/03/24
54
Software Agents



also referred to as “softbots”
live in artificial environments where computers and
networks provide the infrastructure
may be very complex with strong requirements on the
agent


World Wide Web, real-time constraints,
natural and artificial environments may be merged


user interaction
sensors and effectors in the real world

aslab
camera, temperature, arms, wheels, etc.
Sanz / Artificial Intelligence: An Introduction
2004/03/24
55
Agent Program Types






aslab
different ways of achieving the mapping from
percepts to actions
different levels of complexity
simple reflex agents
agents that keep track of the world
goal-based agents
utility-based agents
Sanz / Artificial Intelligence: An Introduction
2004/03/24
56
Simple Reflex Agents

instead of specifying individual mappings in an
explicit table, common input-output associations are
recorded


requires processing of percepts to achieve some abstraction
frequent method of specification is through condition-action
rules



aslab
if percept then action
similar to innate reflexes or learned responses in humans
efficient implementation, but limited power
Sanz / Artificial Intelligence: An Introduction
2004/03/24
57
Sensors
What the world is like now
Condition-action rules
What should I do now
Environment
Reflex Agent Diagram
Agent
Effectors
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
58
Reflex Agent with Internal State

an internal state maintains important information from
previous percepts

aslab
sensors only provide a partial picture of the environment
Sanz / Artificial Intelligence: An Introduction
2004/03/24
59
Agent with State Diagram
Sensors
State
What the world is like now
How the world evolves
What my actions do
Condition-action rules
Agent
What should I do now
Effectors
Environment
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
60
Goal-Based Agent

the agent tries to reach a desirable state


results of possible actions are considered with
respect to the goal


aslab
may be provided from the outside (user, designer), or
inherent to the agent itself
may require search or planning
very flexible, but not very efficient
Sanz / Artificial Intelligence: An Introduction
2004/03/24
61
Goal-Based Agent Diagram
Sensors
State
How the world evolves
What the world is like now
What happens if I do an action
What my actions do
Goals
What should I do now
Agent Effectors
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
62
Utility-Based Agent

more sophisticated distinction between different world
states

states are associated with a real number



aslab
may be interpreted as “degree of happiness”
allows the resolution of conflicts between goals
permits multiple goals
Sanz / Artificial Intelligence: An Introduction
2004/03/24
63
Utility-Based Agent Diagram
Sensors
State
How the world evolves
What my actions do
What the world is like now
What happens if I do an action
How happy will I be then
Utility
What should I do now
Agent Effectors
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
64
Environments

determine to a large degree the interaction between
the “outside world” and the agent


in many cases, environments are implemented within
computers

aslab
the “outside world” is not necessarily the “real world” as we
perceive it
they may or may not have a close correspondence to the
“real world”
Sanz / Artificial Intelligence: An Introduction
2004/03/24
65
Environment Properties

accessible vs. inaccessible


deterministic vs. non-deterministic


no changes while the agent is “thinking”
discrete vs. continuous

aslab
independent perceiving-acting episodes
static vs. dynamic


changes in the environment are predictable
episodic vs. non-episodic


sensors provide all relevant information
limited number of distinct percepts/actions
Sanz / Artificial Intelligence: An Introduction
2004/03/24
66
Agents Summary

agents perceive and act in an environment



basic agent types





simple reflex
reflex with state
goal-based
utility-base
some environments may make life harder for agents

aslab
ideal agents maximize their performance measure
autonomous agents act independently
inaccessible, non-deterministic, non-episodic, dynamic,
continuous
Sanz / Artificial Intelligence: An Introduction
2004/03/24
67
References
Basic literature
aslab
Sanz / Artificial Intelligence: An Introduction
2004/03/24
68
Recommended Books




aslab
Artificial Intelligence : A Modern Approach by Stuart J. Russell,
Peter Norvig
Logical Foundations of Artificial Intelligence by Michael R.
Genesereth, Nils J. Nislsson, Nils J. Nilsson
Artificial Intelligence by Patrick Henry Winston
Artificial Intelligence by Elaine Rich, Kevin Knight (good for
logic, knowledge representation, and search only)
Sanz / Artificial Intelligence: An Introduction
2004/03/24
69
General Reference
aslab

Whatis.com (Computer Science Dictionary)
http://whatis.com/search/whatisquery.html

Technology Encyclopedia
http://www.techweb.com/encyclopedia/

Computing Dictionary
http://wombat.doc.ic.ac.uk/

Webster Dictionary
http://work.ucsd.edu:5141/cgi-bin/http_webster
Sanz / Artificial Intelligence: An Introduction
2004/03/24
70