View
1
Download
0
Category
Preview:
Citation preview
Introduction & History
Artificial Intelligence
Lecture 1 Karim Bouzoubaa
Karim Bouzoubaa Artificial Intelligence 2
Content
¢ What is AI? ¢ What is Intelligence? ¢ AI and other disciplines ¢ History ¢ The state of the art ¢ Application domains ¢ Model of an Intelligent system ¢ The future ¢ Tools
Karim Bouzoubaa Artificial Intelligence 3
Cat Exercise
Karim Bouzoubaa Artificial Intelligence 4
Lexicon
¢ Artificial Intelligence (AI)
¢ Science fiction
¢ AI : Computer Science Branch
Karim Bouzoubaa Artificial Intelligence 5
Introduction
¢ Since 20’s – 30’s ¢ Large use of computer science ¢ Reason : Fast computing ¢ Rapidity ¢ Economic gains, computers
l don’t get tired l don’t sleep l don’t strike l etc.
Karim Bouzoubaa Artificial Intelligence 6
Is the computer intelligent?
¢ Computer mainly performs instructions ¢ Computer is “ stupid “
¢ However, the computer cannot l Decides, makes research, designs (by its
own) ¢ Computer
l Is not creative l Cannot have new ideas, etc.
¢ A computer is not “ intelligent ” ¢ Need: build computers with intelligence
Karim Bouzoubaa Artificial Intelligence 7
First definitions of AI ¢ Difference between Man and Machine:
1. Fast computing for machine, don’t get tired, don’t strike, etc. 2. Man creativity, can make previsions, learns, invents, etc. 3. Examples
• ‘Car’ : difference in representing information • ‘Horse’ : default knowledge, preferences • ‘Pyramid’ : default reasoning, different types of reasoning
4. Synthesis • Intelligence characterizes Man (up to now), difficult to tackle, to
understand • Bring Intelligence to machines in order to help Man in his(er) everyday
tasks
¢ Definitions of AI: 1. Simulate Human Intelligence using Machines 2. Understand Human Intelligence (Using computer science models)
Karim Bouzoubaa Artificial Intelligence 8
What is Intelligence? ¢ Larousse – Mon premier dictionnaire
l L’intelligence est la qualité d’une personne qui comprend vite les choses, apprend facilement et s’adapte bien aux situations nouvelles
l Contraire : stupidité
¢ Wikipedia l An intelligence quotient, or IQ, is a score derived from one
of several different standardized tests designed to assess intelligence
l Standardized tests can't measure initiative, creativity, imagination, conceptual thinking, curiosity, effort, irony, judgment, commitment, nuance, good will, ethical reflection, or a host of other valuable dispositions and attributes. What they can measure and count are isolated skills, specific facts and function, content knowledge, the least interesting and least significant aspects of learning
Karim Bouzoubaa Artificial Intelligence 9
What is Intelligence/AI? ¢ Intelligence is first of all a behavior
l Human beings, Animals
à AI attempts to simulate this behavior l Behavior = perception, understanding, prediction, manipulation, thinking, etc.
¢ How is it possible for a slow, tiny brain, whether biological or electronic, to perceive, understand, predict, manipulate and think?
l What is the impact on CS and on our every day life?
¢ It is clear that computers with human level intelligence would have a huge impact on our every day lives and on the future course of civilization (§ State of the art)
Karim Bouzoubaa Artificial Intelligence 10
Intelligence and other disciplines
¢ Other disciplines were interested in the study of the intelligence
¢ The study of intelligence is also one of the oldest disciplines. For over 2000 years, philosophers have tried to understand how seeing, learning, remembering, and reasoning could, or should be done
Karim Bouzoubaa Artificial Intelligence 11
Intelligence and other disciplines
Linguistics we have theories of the structure and meaning of the language Intelligence
Economics Utility, decision theory
Philosophy theories of reasoning and learning have emerged, along with the viewpoint that the mind is constituted by the operation of a physical system
Mathematics we have formal theories of logic, probability, decision making and computation
Computer
we have the tools with which to make AI a reality
Neuroscience Physical substrate of mental activity
Psychology we have the tools with which to investigate the human mind, and a scientific language within which to express the resulting theories
Karim Bouzoubaa Artificial Intelligence 12
History of AI
¢ 1943 McCulloch & Pitts: Boolean circuit model of brain ¢ 1950 Turing's "Computing Machinery and Intelligence“ ¢ 1956 Dartmouth meeting: "Artificial Intelligence" adopted ¢ 1952—69 Big hopes!
• Newell and Simon: GPS (General Problem Solver) • McCarty: LISP • Minsky: Micro-Worlds
¢ 1966—73 AI discovers computational complexity Neural network research almost disappears
The problem is not as easy as we thought ¢ 1969—79 Early development of knowledge-based systems
Expert systems Ed Feigenbaum (Stanford): Knowledge is power!
• Dendral (inferring molecular structure from a mass spectrometer). • MYCIN: diagnosis of blood infections Robotic vision applications
¢ 1980-- AI becomes an industry ¢ 1986-- Neural networks return to popularity ¢ 1987-- AI becomes a science ¢ 1995-- The emergence of intelligent agents
Karim Bouzoubaa Artificial Intelligence 13
Turing Test
¢ Turing (1950) "Computing machinery and intelligence":
¢ "Can machines think?" à "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 lay person for 5 minutes
¢ Suggested major components of AI: knowledge, reason ing , language understanding, learning
Human Interrogator
Human AI system
Karim Bouzoubaa Artificial Intelligence 14
More recently
¢ AI turns more scientific, relies on more mathematically sophisticated tools: l Markov models (for speech recognition) l Belief networks (see Office 97)
¢ Focus turns to building useful artifacts as opposed to solving the grand AI problem.
Karim Bouzoubaa Artificial Intelligence 15
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
¢ And many more …
Karim Bouzoubaa Artificial Intelligence 16
State of the art - Deep Blue
NEW YORK (CNN) -- He had never lost a chess match. But
that all changed after 19 moves Sunday against the Deep Blue IBM computer.
Karim Bouzoubaa Artificial Intelligence 17
Karim Bouzoubaa Artificial Intelligence 18
State of the art – Equational prover
Karim Bouzoubaa Artificial Intelligence 19
State of the art – ALVINN
¢ Autonomous Land Vehicle In a Neural Network
¢ No hands across America (driving autonomously 98% of the time from coast to coast)
¢ 5487 km
Karim Bouzoubaa Artificial Intelligence 20
State of the art – 1991 Gulf War
¢ US forces deployed an AI l o g i s t i c s p l a n n i n g a n d schedul ing program that involved up to 50,000 vehicles, cargo, and people
Karim Bouzoubaa Artificial Intelligence 21
State of the art – NASA
¢ NASA's on-board autonomous planning program controlled the scheduling of operations for a spacecraft
¢ Mission Deep Space 1 (1998) l Agent-based system l Capable to autonomously make
decisions
Karim Bouzoubaa Artificial Intelligence 22
State of the art – Proverb
¢ Proverb (The Probabilistic Cruciverbalist) is a computerized crossword puzzle solver
¢ Proverb solves crossword puzzles better than most humans
¢ It builds on recent advances in computer science on efficient probabilistic reasoning, information retrieval, data mining, and constraint satisfaction to use a variety of online databases to solve puzzles.
¢ An extensive series of tests indicates that Proverb fills in approximately 90% of the words correctly on an average New York Times crossword puzzle
Karim Bouzoubaa Artificial Intelligence 23
AI Fields
¢ K representation: neural nets, semantic nets, etc.
¢ Reasoning: NLP, ≠ kinds of reasoning (case-based, logic, deductive, ...)
¢ Planning (get the robot to find the telephone in the other room)
¢ M a c h i n e L e a r n i n g ( a d a p t t o n e w circumstances)
¢ Machine vision, speech recognition, finding data on the web, robotics, and much more
Karim Bouzoubaa Artificial Intelligence 24
Application domains
¢ Games, Theorem prover, Problem resolution ¢ Medical science ¢ Transport ¢ Management ¢ Army ¢ Chemical science ¢ etc.
Karim Bouzoubaa Artificial Intelligence 25
General and
Specific Knowledge
Learn Reason
Planning
Perception Communication Acts
Communication Acts Actions
Environment
Actions
Other entities
General AI Model
Karim Bouzoubaa Artificial Intelligence 26
Perceive
Reason about the
task
Interpret communicative
acts
Plan the interactions
Plan actions
Choose conversa-
tional objects
Perform actions
Take the turn and perform
communicative acts
perceivedbeliefs
communicative act
communicative act
Model of the task
Model of social structure
communica- tion plans
Agent's mental states Domain
knowledge
Knowledge about other agents
actions to be performed
goals
communi- cative goals
received COs + positioningsCOs to be
transmitted
External world
actions percepts
Figure 1: A simplified agent's model
Other agents
1
89
10
7
6
5
43
2
action plan schemas
interaction plan schemas
Legend cognitive process
components of conversational context
agent's environment
Detailed AI Model
Karim Bouzoubaa Artificial Intelligence 27
Future
¢ The big Question:
Will some day the machine be more intelligent than a human being?
Karim Bouzoubaa Artificial Intelligence 28
Robotics
Karim Bouzoubaa Artificial Intelligence 29
Robotics
By the year 2050, develop a team of fully autonomous humanoid robots that can win against the human world soccer champion
team.
¢ Major Applications l Search and
assistance in disaster cases
Karim Bouzoubaa Artificial Intelligence 30
Robotics
n NASA Robots Spirit & Opportunity on Mars planet
Karim Bouzoubaa Artificial Intelligence 31
Robotics
n Robots P3 of Honda
Karim Bouzoubaa Artificial Intelligence 32
Future ¢ Usual objects (appliances, tools, wears, glasses, etc.) will be augmented with sensors, microprocessors , and cor respond ing embedded systems.
l Mobile or not l Communicating (Wifi, BlueTooth) l (Semi) autonomous l Advances UI (speech, gestures, etc.)
Karim Bouzoubaa Artificial Intelligence 33
Limits / Future
¢ Theoretical limits l Learning l Approximate reasoning l Large amount of knowledge
¢ Difference Generalist/Specific approach (closed worlds)
¢ Structure l Mind: massive parallelism l Computer: sequential
¢ Law: computer/human society
Karim Bouzoubaa Artificial Intelligence 34
Tools ¢ Programming languages
l Lisp, Prolog
¢ Expert system shells
¢ NLP tools
¢ Agent and Multi-Agent Platforms
¢ Machine Learning, deep learning, ...
Recommended