Transcript Introduction to Artificial Intelligence

Introduction to
Artificial Intelligence
What is intelligence?
• Human brain’s information processing ability.
• OR ability of humans to demonstrate their intelligence by
communicating effectively and by learning.
▫ Ability to understand, analyze, synthesize, and transmit
information
▫ Ability to learn or adapt behavior to a new situation
• Else
▫ Ability to solve problems, to reason with incomplete
information, to think of new or unique solutions to a
problem (creative)
What is Artificial Intelligence (AI)?
It is the science and engineering of making
intelligent machines, especially intelligent
computer programs. It is related to the similar
task of using computers to understand human
intelligence, but AI does not have to confine
itself to methods that are biologically observable.
(John McCarthy, Stanford University)
Generally, AI is a branch of CS that focuses on
intelligent aspects of human beings.
• Minsky, M.
▫ Science of making computers do things that require
intelligence if they are done by humans (Engineering
perspective).
• Durkin, J.
▫ A field of study in CS that pursues the goal of making a
computer reasons in a manner similar to humans
(Cognitive Science perspective).
4 categories of AI definition (Rusell & Norvig, 2003):
• Systems that think like humans
▫ The exciting new effort to make computers think … machines with
minds, in the full and literal sense (Haugeland, 1985).
▫ The automation of activities that we associate with human thinking,
such as decision making, problem solving, learning … (Bellman,
1978).
• Systems that think rationally
▫ The study of mental faculties through the use of computational
models (Charniak & McDermott, 1985).
▫ The study of the computations that make it possible to perceive,
reason and act (Winston, 1992).
• Systems that act like humans
▫ The art of creating machines that perform functions that require
intelligence when performed by people (Kurzweil, 1990).
▫ The study of how to make computers do things at which, at the
moment, people are better (Rich & Knight, 1991).
• Systems that act rationally
▫ Is the study of the design of intelligent agents (Poole et al., 1998).
▫ AI … is concerned with intelligent behavior in artifacts (Nilsson,
1998).
8
Viewing AI in two dimensions
Thought
processes &
reasoning
(Cognitive
SciencePsychology
perspective)
Behavior
(Engineering
perspective)
Thinking humanly
Thinking rationally
Cognitive modelling
‘Laws of thought’
Acting humanly
Acting rationally
Turing Test
Agent
Measure success against
human performance
Measure success against
ideal concept of intelligence
History of AI
Prior to 1956
• SNARC – the first neural network computer was built in
1950 by Marvin Minsky and Dean Edmonds.
• The Turing Test – an articulation of a complete vision of
AI by Alan Turing.
1956
• Dartmouth workshop – the formal recognition of the
name ‘artificial intelligence’ (proposed by John McCarthy).
• The workshop clearly distinguished AI from control theory,
operations research and decision theory (fields with similar objectives
with AI).
▫ AI embraced the idea of duplicating human faculties such as
creativity, self-improvement and language use.
▫ AI methodology emphasis on building machines.
Enthusiasm and expectation period (1952-1969)
• Success in a limited way, partly was due to limitation of computer
itself.
• LISP was introduced in 1958 by McCarthy.
• The idea of microworlds, i.e. a limited domain in which its problem
requires intelligence to solve (famous microworlds – blocks world).
A dose of reality (1966-1973)
• The popular statement by Herbert Simon (1957): “…there are now in
the world machines that can think, that learn and create …”.
• Simon’s prediction of a computer to be a chess champion in 10 years
time has only came true 40 years later.
• Early AI systems failed miserably in attempt to solve more complex
problems.
• A period of criticism and difficulties.
Expert Systems (1969-1979)
• Recognition of domain-specific knowledge as key to power.
• The first ES named DENDRAL (Buchanan et al., 1969) was developed.
Commercial AI (1980-present)
• R1 (later known as XCON)
▫ The first successful commercial AI system
▫ Is an ES, performing configuration task.
▫ Implemented at Digital Equipment Corporation (DEC) and proved
successful in saving millions of dollars per year.
• More ES began to gain footholds in organizations
particularly in the U.S and Japan.
• Boomed to highest point in 1988.
• “AI Winter” or dawn era after 1988 – again, failure was
due to oversell the technology
• The return of Neural Networks (1986-present)
▫ More advance algorithms being developed.
• AI becomes a science (1987-present)
▫ Introduction of more advance AI techniques.
▫ Application in wide range of domain/field.
• The emergence of intelligent agents (1995present)
▫ Introduction of intelligent agent.
Source: Turban & Aronson (2001)
The Turing Test
This machine is trying hard to
manipulate the statement given by
the woman.
TURING MACHINE
This woman is
trying to tell the
interrogator the
truth about
herself.
This interrogator is
guessing with whom/what
he is communicating,
based on the statement
given by both woman and
machine.
This game ends when the
interrogator made his guess
Turing machine pass the test if
the interrogator fails to recognize
with whom (or with what) he is
communicating.
Characteristics of AI system
• Knowledge is key element
▫ Knowledge processing instead of data processing.
▫ Knowledge is represented in various forms
 E.g. logic, semantic network, rules, and trees
on(a,b)
A
B
C
on(b,c)
on(c,table)
clear(a)
S
NP
The birds fly
VP
D
N
V
the
birds
fly
s(np(det(the),noun(birds)),vp(v(fly)))
• Focus on heuristics
▫ Heuristics is an informal, judgmental knowledge of an
application area that constitutes the “rules of good
judgment” in the field.
▫ It encompasses the knowledge of how to solve problems
efficiently and effectively, how to plan steps in solving a
complex problem, how to improve performance, and so
forth.
• Symbolic processing
▫ A string of characters that stands for some real-world
concepts
• Possess inference ability
▫ Inference from facts and rules using heuristics or other
search techniques.
▫ Makes inference by employing pattern-matching
approach.
▫ Reflects human inference process, i.e. relate current
information with what is known (knowledge) when
attempt to solve a problem.
AI vs. Natural Intelligence
Artificial Intelligence
Natural Intelligence
Consistent
Not consistent
Can be copied and transfer
Cannot be copied and transfer
Cost low
High
Can be documented
Difficult to document
Required steps of execution
Creative
Symbolic Input
Observation
Focus – Limited
Focus – Wider
AI vs. Conventional Program
AI Program
Based on the knowledge
representation – dynamic
Conventional Program
Based on the steps defined –
difficult to change/update
Symbolic manipulation
More to numeric
manipulation
Qualitative
Quantitative
Can perform reasoning and
produce conclusion
Cannot!
AI systems attempt to solve 3 major types of tasks:
• Mundane tasks
▫ Tasks that we do everyday
 Commonsense reasoning  perception  natural language
understanding
▫ Related application – computer vision, speech
recognition, pattern recognition, natural language
processing, planning, neural networks, genetic
algorithm and machine learning.
•
Formal tasks
▫
Much of the early works in AI focused on formal
tasks such as game playing and theorem proving.


Share the property that people who do them well are
considered to be displaying intelligence.
Computers could perform well at those tasks because they
are fast at exploring a large number of solution paths and
then selecting the best one.
•
Expert tasks
▫
Task that requires high-level intelligence of human
experts (expertise)
 Diagnosis, interpretation, configuration, credit
authorization.
 Related AI application is expert systems.
Roots of AI
Ideas, viewpoints and techniques of AI today come
from various disciplines:
•
•
•
•
•
•
•
•
Philosophy
Mathematics
Economics
Neuroscience
Psychology
Computer engineering
Control theory and cybernetics
Linguistics
Source: Turban & Aronson (2001)
AI Tree..
Affective computing
Machine Learning
Speech Understanding
Automatic
Robotic
Programming
Natural Language
Processing
Expert System
Intelligent Tutor
Computer Vision
Linguistics
Game Playing
Neural Network
Fuzzy Logic
Genetic Algorithm
Data Mining
Computer Science
Psychology
Management &
Philosophy
Management Science
Electrical Engineering
Applications of AI
• Expert system (ES)
A computer system that applies
reasoning methodologies to
knowledge in a specific domain to
render advice or recommendations,
much like a human expert.
A computer system that achieves a
high level of performance in task
areas that, for human beings,
require years of special education
and training
Medical diagnosis
• Robotics and sensory
systems
▫ Robots
 Machines that have the
capability of performing
manual functions without
human intervention
 An “intelligent” robot has
some kind of sensory
apparatus, such as a
camera, that collects
information about the
robot’s operation and its
environment
NeCoRo the robot cat
– responds to human movement/emotions, has
feelings and desires, remembers its name and
acknowledges its name when called
• Speech (voice) understanding
▫ Translation of the human voice into individual words
and sentences understandable by a computer
• Natural language
processing (NLP)
▫ Using a natural language
processor to interface with
a computer-based system
▫ Two subfields of NLP
▫ Natural language
understanding
 Natural language
generation
• Visual recognition /
vision system
▫ The addition of some form
of computer intelligence
and decision-making to
digitized visual information,
received from a machine
sensor such as a camera
▫ The basic objective of
computer vision is to
interpret scenarios rather
than generate pictures
KISMET
• Intelligent
computer-aided
instruction (ICAI)
▫ The use of AI
techniques for
training or teaching
with a computer
▫ Intelligent tutoring
system (ITS) - selftutoring systems
that can guide
learners in how best
to proceed with the
learning process
• Neural network
▫ An experimental computer design aimed at building
intelligent computers that operate in a manner
modeled on the functioning of the human brain.
• Fuzzy logic
▫ Logically consistent ways of
reasoning that can cope with
uncertain or partial
information; characteristic of
human thinking and many
expert systems
• Genetic algorithms
▫ Intelligent methods that use
computers to simulate the
process of natural evolution
to find patterns from a set of
data
GA steps
• Intelligent agent (IA)
▫ An expert or knowledge-based system embedded in
computer-based information systems (or their
components) to make them smarter
• Game playing
▫ One of the first areas that AI
researchers studied
▫ It is a perfect area for investigating
new strategies and heuristics because
the results are easy to measure
• Automotive
• POD: emotion car..
• Toyota & Sony.
• Snow driver emotion and learned
from driver experiences.
• measure your sweat, pulse
• Affective Computing
Others..
Microsoft Clippers
Face Recognition
NASA’s spacecraft
scheduling operation
controller