Introduction to Artificial Intelligence

1
Introduction to Artificial Intelligence

• Lectured by
• Yen-Hsien Lee
• Department of Management Information Systems
• College of Management
• National Chiayi University
• September 24, 2008

2
The Brief History of AI

• 1956 Birth of AI
• The Dartmouth workshop held by McCarthy
• 1952-1969 Big jump of AI?
• General Problem Solver imitating human
  problem-solving protocols (Newell and Simon)
• Geometry Theorem Prover proving tricky theorems
  (Gelernter)
• Amateur Checker learning to play checker
  (Samuel)
• Lisp AI programming language (McCarthy)
• Advice Taker the first complete AI system,
  designed to use general knowledge of the world to
  search for solutions to problems (McCarthy)
• SAINT solving closed-form calculus integration
  problems (Slagle)

3
The Brief History of AI (Contd)

• 1952-1969 Big jump of AI? (Contd)
• ANALOGY solving geometric analogy problems
  appearing in IQ test (Evans)
• STUDENT solving algebra story problems (Bobrow)
• Block world (Minsky)
• Neural Network (McCulloch and Pitts)
• Adalines an learning network that enhanced
  Hebbs learning methods (Widrow and Hoff).
• Perceptron covergence theorem learning network
  (Rosenblatt)

4
The Brief History of AI (Contd)

• 1966-1973 Failures of AI?
• Simple syntactic manipulations on natural
  language
• Intractability of many of the problems
• Fundamental limitations on the basic structures
  used to generate intelligent behavior

5
The Brief History of AI (Contd)

• 1969-1979 Resurgence of AI
• The emergence of knowledge-based systems
• From general-purpose to specific-domain
• From common sense to expertise
• From elementary reasoning to knowledge-based
  reasoning
• The separation of the knowledge (in the form of
  rules) from the reasoning component.
• DENDRAL solving the problem of inferring
  molecular structure from the information provided
  by a mass spectrometer.
• MYCIN diagnosing blood infections handle
  uncertainty)
• SHRDLU understanding natural language (designed
  for a specific domain)
• Various Knowledge Representation and Reasoning
  methods

6
The Brief History of AI (Contd)

• 1980-present AI becomes an industry
• The first commercial expert system, R1
  configuring orders for new computer systems
  (McDermott, DEC)
• Fifth Generation Project (Japan)
• Microelectronics and Computer Technology
  Corporation (U.S.)
• 1986-present The return of neural networks
• Reinventing the back-propagation learning
  algorithm

7
The Brief History of AI (Contd)

• 1987-present AI becomes a science
• Hypotheses must be subjected to rigorous
  empirical experiments, and the results must be
  analyzed statistically for their importance
  (Cohen, 1995)
• Hidden Markov Models (HMMs) dominating the
  speech recognition area.
• Bayesian network basing on probability and
  decision theory dominates AI research on
  uncertain reasoning and expert systems.

8
The Brief History of AI (Contd)

• 1995-present The emergence of intelligent agents
• Starting to look at the whole agent problem
  again.
• An agent have to possess the attributes such as
  operating under autonomous control, perceiving
  their environment, persisting over a prolonged
  time period,adapting to change, and being capable
  of taking on anothers goals.
• SOAR situated movement aims to understand the
  workings of agents embedded in real environment
  with continuous sensory inputs.
• Search engine, Recommender systems, Website
  construction systems
• The consequences of agent perspective
• The reorganization of subfields of AI research
• The integration of other fields related to AI
  research such as control theory and economics.

9
What Can AI Do Now?

• Autonomous planning and scheduling
• Game playing
• Autonomous control
• Diagnosis
• Logistics planning
• Robotics
• Language understanding and problem solving
• etc.

10
(No Transcript)
11
So, What is AI???
12
What is AI?

• Some definitions of AI are given as follows

13
Acting HumanlyThe Turing Test Approach

• Turing (1950) proposed the Turing Test, based on
  indistinguishability from undeniably intelligent
  entities human beings.
• The coputer passes the test if a human
  interrogator, after posing some written
  questions, cannot tell whether the written
  responses come from a person or not.
• Suggested major components of AI knowledge
  representation, automated reasoning, natural
  language processing, machine learning, (computer
  vision, robotics)

14
Thinking HumanlyThe Cognitive Modeling Approach

• Once we have a sufficiently precise theory of the
  mind, it becomes possible to express the theory
  as a computer program.
• 1960s "cognitive revolution" information-processi
  ng psychology
  
• Requires scientific theories of internal
  activities of the brain
  
• How to validate?
• 1) Predicting and testing behavior of human
  subjects (top-down)
• or 2) Direct identification from neurological
  data (bottom-up)

15
Thinking RationallyThe Law of Thought Approach

• Aristotle what are correct arguments/thought
  processes?
• His syllogisms provide patterns for argument
  structures that always yield correct conclusions
  when given correct premises.
• The laws of thought are supposed to govern the
  operation of the mind.
• Problems
• Not easy to take informal knowledge and state it
  in the formal terms, particularly the knowledge
  with uncertainty.
• Different between being able to solve a problem
  in principle and doing so in practice.
  

16
Acting RationallyThe Rational Agent Approach

• Rational behavior doing the right thing
• The right thing that which is expected to
  maximize goal achievement, given the available
  information
  
• Rational behavior don't necessarily involve
  thinking e.g., blinking reflex but thinking
  should be in the service of it.
• An agent is an entity that perceives and acts.
• What is a rational agent?

17
Acting RationallyThe Rational Agent Approach
(Contd)

• A rational agent is one that acts so as to
  achieve the best outcome or, when there is
  uncertainty, the best expected outcome.
• For any given class of environments and tasks, we
  seek the agent with the best performance
  
• Caveat computational limitations make perfect
  rationality unachievable
• Advantages of studying rational agent design
• More general than the laws of thought approach
  that is only one of several ways for achieving
  rationality.
• The standard of rationality is more clearly
  defined and completely general than human
  behavior and thought.

18
Agents

• An agent is anything that can be viewed as
  perceiving its environment through sensors and
  acting upon that environment through actuators
  
• Human agent eyes, ears, and other organs for
  sensors hands, legs, mouth, and other body parts
  for actuators
• Robotic agent cameras and infrared range finders
  for sensors various motors for actuators
  

19
Rational Agents

• An agent should strive to "do the right thing",
  based on what it can perceive and the actions it
  can perform.
• But, what does it mean to do the right thing?
• The right action is the one that causes the
  agent to be most successful.
• But, what does it mean the agent successes?
• The agent generates a sequence of actions
  according to the percepts it receives. If the
  sequence is desirable, then we say the agent
  performed well

20
Rational Agents (Contd)

• An agent needs a performance measure, an
  objective criterion for success of an agent's
  behavior.
• E.g., performance measure of a vacuum-cleaner
  agent could be amount of dirt cleaned up, amount
  of time taken, amount of electricity consumed,
  amount of noise generated, and etc.
• But, can the above criteria really measure a
  vacuum-cleaner agents performance?
• According to what you actually wants in the
  environment OR according to how you think the
  agent should behave?

21
Rational Agents (Contd)

• What is rational must depend on four things
• The performance measure
• The agents prior knowledge of the environment
• The actions that the agent can perform
• The agents percept sequence to date
• Definition of a rational agent
• For each possible percept sequence, a rational
  agent should select an action that is expected to
  maximize its performance measure, given the
  evidence provided by the percept sequence and
  whatever built-in knowledge the agent has.

22
PEAS

• It must first specify the setting for rational
  (intelligent) agent design, including Performance
  measure, Environment, Actuators, Sensors (PEAS).
• Considering the design (PEAS) of an automated
  taxi driver and that of an interactive English
  tutor
  
• Performance measure
• Environment
• Actuators
• Sensors

23
PEAS (Contd)

• PEAS of an automated taxi driver
• Performance measure Safe, fast, legal,
  comfortable trip, maximize profits
  
• Environment Roads, other traffic, pedestrians,
  customers
  
• Actuators Steering wheel, accelerator, brake,
  signal, horn
  
• Sensors Cameras, sonar, speedometer, GPS,
  odometer, engine sensors, keyboard
• PEAS of an interactive English tutor
• Performance measure Maximize student's score on
  test
• Environment Set of students
• Actuators Screen display (exercises,
  suggestions, corrections)
• Sensors Keyboard

24
Environment Types

• Fully observable (vs. partially observable)
• An agent's sensors give it access to the complete
  state of the environment at each point in time.
  
• Deterministic (vs. stochastic)
• The next state of the environment is completely
  determined by the current state and the action
  executed by the agent. (If the environment is
  deterministic except for the actions of other
  agents, then the environment is strategic)
• Episodic (vs. sequential)
• The agent's experience is divided into atomic
  "episodes" (each episode consists of the agent
  perceiving and then performing a single action),
  and the choice of action in each episode depends
  only on the episode itself.

25
Environment Types (Contd)

• Static (vs. dynamic)
• The environment is unchanged while an agent is
  deliberating. (The environment is semidynamic if
  the environment itself does not change with the
  passage of time but the agent's performance score
  does)
• Discrete (vs. continuous)
• A limited (finite) number of distinct, clearly
  defined percepts and actions.
  
• Single agent (vs. multiagent)
• An agent operating by itself in an environment.

26
Inside Work of An Agent

• The job of AI is to design the agent program that
  implements the agent function mapping percepts to
  actions
• According to the complexity, we outline four
  basic kinds of agent program
• Simple reflex agents
• Model-based reflex agents
• Goal-based agents
• Utility-based agents

27
Simple Reflex Agents
28
Simple Reflex Agents (Contd)

• The agent select actions on the basis of the
  current percept, ignoring the rest of the
  percept history.
• In such agent, each condition corresponds to an
  action, is call condition-action rule.
• Problems
• Can only operate in a fully observable
  environment
• Incomplete condition-action rule

29
Model-based Reflex Agents
30
Model-based Reflex Agents (Contd)

• The agent to maintain some sort of internal state
  that depends on the percept history and thereby
  reflects at least some of the unobserved aspects
  of the current state.
• Updating internal state information requires two
  kinds of knowledge
• Information about how the world evolves
  independently of the agent
• Information about how the agents own actions
  affect the world
• Knowledge about how the world works is called a
  model of the world.

31
Goal-based Agents
32
Goal-based Reflex Agents (Contd)

• Knowing about the current state of the
  environment is not always enough to decide what
  to do.
• The agent needs some sort of goal information
  that describes situations that are desirable,
  e.g., the passengers destination.
• Goal-based agent can combine its goal with
  information about the results of possible actions
  to choose actions that achieve the goal.
• The goal-based agent is less efficient, it is
  more flexible because the knowledge that supports
  its decisions is represented explicitly and can
  be modified.

33
Utility-based Agents
34
Utility-based Reflex Agents (Contd)

• Goal alone are not really enough to generate
  high-quality behavior in most environment.
• For example, there are many action sequences that
  will get the taxi to its destination but some are
  quicker, safer, or more reliable.
• Goal just provide a binary distinction between
  happy and unhappy state.
• A more general performance measure should allow
  a comparison of different world states according
  to how happy (utility) they would make the agent
  if they could be achieved.

35
Learning Agents
36
Learning Agents (Contd)

• Learning allows the agent to operate in initially
  unknown environments and to become more competent
  than its initial knowledge alone might allow.
• Four conceptual components of a learning agent
• Learning element uses feedback from the Critic
  on how the agent is doing and determines how the
  performance element should be modified to be
  better
• Performance element is responsible for selecting
  external actions.
• Critic tells the learning element how well the
  agent is doing with respect to a fixed
  performance standard.
• Problem generator is responsible for suggesting
  actions that will lead to new and informative
  experiences