Transcript Lecture I -- Introduction and Intelligent Agentx

CS541 Artificial Intelligence
Lecture I: Introduction and Intelligent Agent
Self-introduction

Prof. Gang Hua






华刚
Associate Professor in Computer Science
Stevens Institute of Technology
Research Staff Member (07/2010—08/2011)
IBM T J. Watson Research Center
Senior Researcher (08/2009—07/2010)
Nokia Research Center Hollywood

Scientist (07/2006—08/2009)
Microsoft Live Labs Research

Ph.D. in ECE, Northwestern University, 06/2006

Course Information (1)









CS541 Artificial Intelligence
Term: Fall 2013
Instructor: Prof. Gang Hua
Class time: Tuesday 2:00pm—4:30pm
Location: EAS 330
Office Hour: Wednesday 4:00pm—5:00pm by
appointment
Office: Lieb/Room305
Course Assistant:Yizhou Lin
Course Website:

http://www.cs.stevens.edu/~ghua/ghweb/ teaching/CS541Fall2013.htm
Course Information (2)

Text Book:


Stuart Russell and Peter Norvig, “Artificial Intelligence: A Modern
Approach”, Third Edition, Prentice Hall, December 11, 2009
(Required)
Grading:



Class Participation: 10%
5 Homework: 50% (including a midterm project)
Final Project & Presentation: 40%
Schedule
Week
Date
Topic
Reading
Homework**
1
08/27/2013
Introduction & Intelligent Agent
Ch 1 & 2
N/A
2
09/03/2013
Search: search strategy and heuristic search
Ch 3 & 4s
HW1 (Search)
3
09/10/2013
Search: Constraint Satisfaction & Adversarial Search
Ch 4s & 5 & 6
Teaming Due
4
09/17/2013
Logic: Logic Agent & First Order Logic
Ch 7 & 8s
HW1 due, Midterm Project (Game)
5
09/24/2013
Logic: Inference on First Order Logic
Ch 8s & 9
6
10/01/2013
Uncertainty and Bayesian Network
Ch 13 & Ch14s
7
10/08/2013
Midterm Presentation
8
10/15/2013
No class
9
10/22/2013
Inference in Baysian Network
Ch 14s
10
10/29/2013
Probabilistic Reasoning over Time
Ch 15
11
11/05/2013
Machine Learning
Ch 18 & 20
HW3 due,
12
11/12/2013
Markov Decision Process
Ch 16
HW4 (Probabilistic Reasoning Over Time)
13
11/19/2013
Reinforcement learning
Ch 21
HW4 due
14
11/26/2013
No class Q/A for final project
15
12/03/2013
No class TA assisted Dry run
16
12/10/2013
Final Project Competition
HW2 (Logic)
Midterm Project Due
HW2 Due, HW3 (Probabilistic Reasoning)
Final Project Due
Rules

Need to be absent from class?


Late submission of homework?


1 point per class: please send notification and justification at
least 2 days before the class
The maximum grade you can get from your late homework
decreases 50% per day
Zero tolerance on plagiarism!!

You receive zero grade
Introduction & Intelligent Agent
Prof. Gang Hua
Department of Computer Science
Stevens Institute of Technology
[email protected]
Introduction to Artificial Intelligence
Chapter 1
What is AI?
Systems thinking humanly
Systems thinking rationally
Systems acting humanly
Systems acting rationally
Acting humanly: Turing Test

Turing (1950) "Computing machinery and intelligence":

"Can machines think?"  "Can machines behave intelligently?"

Operational test for intelligent behavior: the Imitation Game

Predicted that by 2000, a machine might have a 30% chance of fooling a layperson for 5
minutes

Anticipated all major arguments against AI in following 50 years

Suggested major components of AI: knowledge, reasoning, language understanding,
learning,

Total Turing test: adding vision and robotics

Problem: Turing test is not reproducible, constructive, or amenable to
mathematical analysis
Thinking humanly: cognitive modeling


1960 "cognitive revolution": information-processing psychology
replaced prevailing orthodoxy of behaviorism
Requires scientific theories of internal activities of the brain


What levels of abstraction? "Knowledge" or "circuits"?
How to validate? Requires




Predicting and testing behavior of human subjects (top-down)
Direct identification from neurological data (bottom-up)
Both approaches (roughly, Cognitive Science and Cognitive
Neuroscience) are now distinct from AI
Both share one principal direction with AI:

The available theories do not explain anything resembling human-level
general intelligence
Thinking rationally: "laws of thought"

Aristotle: what are correct arguments/thought processes?

Several Greek schools developed various forms of logic:

notation and rules of derivation for thoughts;

They may or may not have proceeded to the idea of
mechanization

Direct line through mathematics and philosophy to modern AI

Problems:

Not all intelligent behavior is mediated by logical deliberation

What is the purpose of thinking?

What thoughts should I have out of all the thoughts (logical or
otherwise) that I could have?
Acting rationally: rational agent

Rational behavior: doing the right thing

The right thing: that which is expected to maximize goal
achievement, given the available information

Doesn't necessarily involve thinking – e.g., blinking reflex
– but thinking should be in the service of rational action

Aristotle (Nicomachean Ethics):
 Every art and every inquiry, and similarly every action and
pursuit, is thought to aim at some good
Rational agents





An agent is an entity that perceives and acts
This course is about designing rational agents
Abstractly, an agent is a function from percept histories to actions:
For any given class of environments and tasks, we seek the agent (or
class of agents) with the best performance
Caveat: computational limitations make perfect rationality
unachievable
 design best program for given machine resources
AI prehistory

Philosophy


Mathematics


Building fast computers
Control theory


Phenomena of perception and motor control, experimental techniques
Computer engineering


Physical substrate for mental activity
Psychology


Utility, decision theory
Neuroscience


Formal representation and proof algorithms, computation, (un)decidability, (in)tractability, probability
Economics


Logic, methods of reasoning, mind as physical system, foundations of learning, language, rationality
Design systems that maximize an objective function over time
Linguistics

knowledge representation, grammar
Abridged history of AI















1943
1950
1956
1952—69
1950s
1965
1966—74
McCulloch & Pitts: Boolean circuit model of brain
Turing's "Computing Machinery and Intelligence"
Dartmouth meeting: "Artificial Intelligence" adopted
Look, Ma, no hands!
Early AI programs, including Samuel's checkers
program, Newell & Simon's Logic Theorist,
Gelernter's Geometry Engine
Robinson's complete algorithm for logical reasoning
AI discovers computational complexity
Neural network research almost disappears
Early development of knowledge-based systems
Expert systems industry boom
Expert systems industry busts: "AI Winter"
Neural networks return to popularity
1969—79
1980—88
1988—93
1985—95
1988— Resurgence of probability; AI becomes science
1995—
The emergence of intelligent agents
2003—
Human-level AI back on the agenda
2010-BIG data? Deep learning?
State of the art

Deep Blue defeated the reigning world chess champion Garry Kasparov in
1997

Proved a mathematical conjecture (Robbins conjecture) unsolved for decades

No hands across America (driving autonomously 98% of the time from
Pittsburgh to San Diego)

During the 1991 Gulf War, US forces deployed an AI logistics planning and
scheduling program that involved up to 50,000 vehicles, cargo, and people


NASA's on-board autonomous planning program controlled the scheduling of
operations for a spacecraft
Proverb solves crossword puzzles better than most humans

iRobot corporated in 2000: Roomba & Scooba

Google cars automatically are driving in the city to collect stree-tview images

Watson whips Brad Rutter and Ken Jennings in Jeopardy in 2011!
DeepBlue & Watson (DeepQA)


DeepBlue
Watson (DeepQA)
Intelligent Agent
Chapter 2
Outline





Agents and environments
Rationality: what is a rational agent?
PEAS (Performance measure, Environment, Actuators,
Sensors)
Environment types
Agent types
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
Agents and environments

The agent function maps from percept histories to actions:

The agent program runs on the physical architecture to
produce f

agent = architecture + program
Vacuum-cleaner world

Percepts: location and contents, e.g., [A,Dirty]

Actions: Left, Right, Suck, NoOp
Rational agents (1)

An agent should strive to "do the right thing", based on what it
can perceive and the actions it can perform. The right action is
the one that will cause the agent to be most successful

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 in time T?
Amount of dirt cleaned up minus the amount of electricity
consumed in time T?
Amount of time taken to clean a fixed region?
Rational agents (2)

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.
Rational agents (3)

Rationality is distinct from omniscience (all-knowing with
infinite knowledge)

Agents can perform actions in order to modify future
percepts so as to obtain useful information (information
gathering, exploration)

An agent is autonomous if its behavior is determined by
its own experience (with ability to learn and adapt)
PEAS (1)

PEAS: Performance measure, Environment, Actuators,
Sensors

To design a rational agent, we must first specify the task
environment

Consider, e.g., the task of designing an automated taxi
driver:




Performance measure??
Environment??
Actuators??
Sensors??
PEAS (2)

To design a rational agent, we must first specify the task
environment

Consider, e.g., the task of designing 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 (3)

Agent: Internet shopping agent

Performance measure: price, quality, appropriateness,
efficiency

Environment: current and future WWW sites, vendors,
shippers

Actuators: display to user, follow URL, fill in form

Sensors: HTML pages (text, graphics, scripts)
PEAS (4)

Agent: Part-picking robot

Performance measure: Percentage of parts in correct
bins

Environment: Conveyor belt with parts, bins

Actuators: Jointed arm and hand

Sensors: Camera, joint angle sensors
PEAS (5)

Agent: Interactive English tutor

Performance measure: Maximize student's score on
test

Environment: Set of students

Actuators: Screen display (exercises, suggestions,
corrections)

Sensors: Keyboard
Environment types (1)

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.
Environment types (2)

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 number of distinct, clearly defined
percepts and actions.

Single agent (vs. multiagent): An agent operating by itself in an
environment.
Environment types (3)


Solitaire
Backgammon
Internet Shopping
Taxi
Observable?
Yes
Yes
No
No
Deterministic?
Yes
No
Partly
No
Episodic?
No
No
No
No
Static?
Yes
Semi
Semi
No
Discrete?
Yes
Yes
Yes
No
Single Agent?
Yes
No
Yes (except auction)
No
The environment type largely determines the agent design
The real world is (of course) partially observable, stochastic, sequential,
dynamic, continuous, multi-agent
Agent functions and programs



An agent is completely specified by the agent function
mapping percept sequences to actions
One agent function (or a small equivalence class) is
rational
Aim: find a way to implement the rational agent function
concisely
Table-lookup agent

Drawbacks:

Huge table

Take a long time to build the table

No autonomy

Even with learning, need a long time to learn the table
A vacuum-cleaner agent


What is the right function?
Can it be implemented in a small agent program?
Agent types


Four basic types (with increasing generality):

Simple reflex agents

Model-based reflex agents

Goal-based agents

Utility-based agents
All of them can be transformed into learning agent
Simple reflex agents

The action to be selected only depends on the most recent percept, not a sequence

These agents are stateless devices which do not have memory of past world states
Model-based reflex agents

Have internal state which is used to keep track of past states of the world

Can assist an agent deal with some of the unobserved aspects of the current state
Goal-based agents

Agent can act differently depending on what the final state should look like

E.g., automated taxi driver will act differently depending on where the passenger wants to go
Utility-based agents

An agent's utility function is an internalization of the external performance measure

They may differ if the environment is not completely observable or deterministic
Learning agents

Learning agent cuts across all of the other types of agents: any kind of agent can learn
iRobot

Roomba Demo
Summary

Agents interact with environments through actuators and sensors

The agent function describes what the agent does in all
circumstances

The performance measure evaluates the environment sequence

A perfectly rational agent maximizes expected performance

Agent programs implement (some) agent functions

PEAS descriptions define task environments

Environments are categorized along several dimensions:


Observable? Deterministic? Episodic? Static? Discrete? Single-agent?
Several basic agent architectures exist:

Reflex, Reflex with state, goal-based, utility-based
Candidate projects

Midterm Project:

Mastermind (midterm)


http://en.wikipedia.org/wiki/Mastermind_%28board_game%29
Final Projects:

Reversi (Othello)

http://en.wikipedia.org/wiki/Reversi