Download Artificial Intelligence - University of Regina

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Machine learning wikipedia , lookup

Computer vision wikipedia , lookup

The City and the Stars wikipedia , lookup

Turing test wikipedia , lookup

Artificial intelligence in video games wikipedia , lookup

Wizard of Oz experiment wikipedia , lookup

Embodied cognitive science wikipedia , lookup

Incomplete Nature wikipedia , lookup

Computer Go wikipedia , lookup

Intelligence explosion wikipedia , lookup

Ethics of artificial intelligence wikipedia , lookup

Existential risk from artificial general intelligence wikipedia , lookup

AI winter wikipedia , lookup

Philosophy of artificial intelligence wikipedia , lookup

Knowledge representation and reasoning wikipedia , lookup

History of artificial intelligence wikipedia , lookup

Transcript
Artificial Intelligence
Dr. Malek Mouhoub
Department of Computer Science
University of Regina
Fall 2009
Malek Mouhoub, CS820 Fall 2009
1
Plan
I.
Introduction to Artificial Intelligence
II.
Logic and Logic Programming
III.
Problem Solving Techniques
(3)
(4)
IV.
Knowledge Representation
(5)
(6)
(7)
V.
Structures and Strategies for State Space Search
Constraint Satisfaction
Logic Representation and Automated Reasoning
Rule Based Expert Systems
Network and Structured Representation
Advanced Topics in AI
(8)
(9)
(10)
(11)
(12)
(13)
Temporal Reasoning
Metaheuristics
Intelligent Agents and Distributed AI
Uncertainty
Preference Reasoning
Robot Motion Planning
Malek Mouhoub, CS820 Fall 2009
2
1. Introduction to Artificial Intelligence
1. Introduction to Artificial Intelligence
• Can thinking be a form of computation ?
• Formalizing the mental process by formal mathematics.
• Making a mind (AI) versus modeling the brain (Neural Networks).
Malek Mouhoub, CS820 Fall 2009
3
What is Artificial Intelligence ?
What is Artificial Intelligence ?
• The use of computer programs and programming techniques to
cast light on the principles of intelligence in general and human
thought in particular (Boden).
• The study of intelligence independent of its embodiment in
humans, animals or machines (McCarthy).
• AI is the study of how to do things which at the moment people
do better (Rich & Knight).
• AI is the science of making machines do things that would
require intelligence if done by men (Minsky).
Malek Mouhoub, CS820 Fall 2009
4
What is Artificial Intelligence ?
What is Artificial Intelligence ?
• Knowledge representation :
– Representing the problem to solve with a formal language.
• Problem Solving Techniques :
– Solution search techniques exploring the space of problem
states.
Malek Mouhoub, CS820 Fall 2009
5
AI : The History
AI : The History
• AI is nearly as old as computing.
• 1941 Konrad Zuse, Germany, general purpose computer.
• 1943 Britain (Turing and others) Colossus, for decoding.
• 1943, McCulloch, Pitts: Boolean circuit model of brain.
• 1945 ENIAC, US, John von Neumann a consultant.
• 1950s, early AI programs, including Samuel’s checkers program.
• 1956 Dartmouth Conference organized by John McCarthy (inventor of LISP). The term “Artificial Intelligence” was adopted.
• 1965, Robinson’s complete algorithm for logical reasoning.
• 1966–74, AI discovers computational complexity.
• 1969–79, early development of knowledge-based systems.
• 1980–88, expert systems industry booms.
• 1988–93, expert systems industry busts: “AI Winter”.
• 1985–95, Neural networks return to popularity.
• 1988–today, resurgence of probabilistic and decision-theoretic methods. Rapid increase in technical depth of mainstream AI.
“Nouvelle AI”: ALife, GAs, soft computing.
Malek Mouhoub, CS820 Fall 2009
6
AI : The History
AI : The History
• First system of formal logic [Leibniz 1887]
• Graph theory [Euler 1735]
– The bridges of Koenigsberg problem
∗ Proving the impossibility of a walk around the city that crosses each
bridge exactly once.
– State space search for problem solving :
∗ Modeling the problem by a graph.
∗ Systematically look for a possible solution.
Malek Mouhoub, CS820 Fall 2009
7
Graph Theory
Graph Theory
rb1
Riverbank 1
River
2
4
3
Island 1
Island 2
6
b4
b3
Graph representation
1
5
b2
i1
7
b1
i2
b6
b5
b7
Riverbank 2
rb2
Figure 1: The bridges of Koenigsberg problem
Malek Mouhoub, CS820 Fall 2009
8
Formal Logic
Formal Logic
• A formal language for thought : Boole Algebra [Boole 1847,54]
– Propositional calculus :
∗ Symbols : P, Q, R, S, . . .
∗ Truth symbols : true (>), f alse (⊥)
∗ Connectives : ∧, ∨, ¬, ⇒, ⇔
• First order predicate calculus [Fregel 1879, 1884]
– Variable symbols :
x, y, z, . . .
– Function symbols :
f, g, h . . . ∈ F
– Relational symbols (predicates) :
– Connectives :
– Quantifiers :
Malek Mouhoub, CS820 Fall 2009
>, ⊥, p , q , r . . . ∈ P
∧, ∨, ¬, ⇒, ⇔
∀, ∃
9
Formal Logic
Formal Logic
• Derive the whole mathematics through formal operations on a
collection of axioms [Whitehead & Russell 1950]
– Axioms and theorems treated as strings of characters.
– Theorem proof using well defined rules for manipulating
strings.
• Theory of semantics [Fregel 1879, 1884]
• Electronic Digital Computers :
– implement formal reasoning systems on a computer
⇒ testing theories of intelligence
Malek Mouhoub, CS820 Fall 2009
10
Turing test
Turing test
Human
Machine
Malek Mouhoub, CS820 Fall 2009
11
Turing test
Turing test
• Can a machine think ?
• What is thinking ?
• Features of the test :
– Objective notion of intelligence.
– The test is an essential tool in both the development and verification of modern expert
systems.
• Disadvantage of the test :
– Trying to evaluate machine intelligence in human terms
• Turing objection :
computers can only do as they are told and consequently cannot perform
original (hence, intelligence) actions.
Malek Mouhoub, CS820 Fall 2009
12
Turing test
Turing test
• Predicted that by 2000, a machine might have a 30% chance of
fooling a lay person for 5 minutes.
• Anticipated all major arguments against AI in following 50 years.
• Suggested major components of AI: knowledge, reasoning,
language understanding, learning.
Malek Mouhoub, CS820 Fall 2009
13
AI : The Achievements
AI : The Achievements
• Robots make cars in all advanced countries.
• Reasonable machine translation is available for a large range of
foreign web pages.
• Computers land 200 ton jumbo jets unaided every few minutes.
• Search systems like Google are not perfect but provide very
effective information retrieval.
• Robots cut slots for hip joints better than surgeons.
• Deep blue beat Kasparov in 1997.
• Medical expert systems can outperform doctors in many areas
of diagnosis.
Malek Mouhoub, CS820 Fall 2009
14
AI Applications
AI Applications
• Game Playing
• Automated reasoning and theorem proving
• Expert Systems
• Natural Language understanding
• Planning and Robotics
• Machine Learning
• Neural Networks
Malek Mouhoub, CS820 Fall 2009
15
Game Playing
Game Playing
• Chess, 16-puzzle, . . .
• Well-defined set of rules:̇
⇒ easy generation of the search space.
• Simple formalism representing the problem.
• Require powerful search techniques.
• Used to study, test and explain search techniques.
Malek Mouhoub, CS820 Fall 2009
16
Automated reasoning and theorem proving
Automated reasoning and theorem proving
• Representing the problem description as logical axioms and
treating problem instances as theorems to be proved.
• Useful for the verification of logic circuits, correctness and
completeness of computer programs, control of complex
systems.
• Difficulties to solve complicated problems (most of all real
problems) : lack of powerful techniques (heuristics) to guide the
solution search.
⇒ Rely on informal ad hoc strategies (CSP techniques).
Malek Mouhoub, CS820 Fall 2009
17
Expert Systems
Expert Systems
• Obtaining expert knowledge (theoretical understanding of the
problem + heuristics problem solving rules issued from
experience) from a human expert (doctor, chemist, geologist,
engineer, . . . ) and coding it into a form that the computer may
apply to similar problems.
• Useful in areas requiring theoretical and/or empirical knowledge.
• MYCIN (medical diagnose), DENDRAL (chemistry),
PROSPECTOR (geology).
Malek Mouhoub, CS820 Fall 2009
18
Expert Systems (cont.)
Expert Systems (cont.)
• Limitations :
– Lack of deep knowledge of problem domains
⇒ Inability to provide deep explanations.
⇒ Rely on Qualitative, Causal and Temporal reasoning.
– Little learning from experience
⇒ Rely on Machine Learning techniques.
– Difficulties in verification.
Malek Mouhoub, CS820 Fall 2009
19
Natural Language understanding
Natural Language understanding
• Use a natural language to interact with the computer.
• First programs : Systems based on Syntactic parsing only.
– Parsing sentences into their individual parts of speech and looking those
words up in a dictionary.
∗
∗
∗
∗
Automatic translator (interpreter) [Bar-Hillel 1952].
Formal grammars [Chomsky 1957].
STUDENT (solving Algebra exercises).
ELIZA (simulating a dialog between a psychiatrist and a patient).
– Limitations :
∗ Lack of background knowledge about the domain of discourse.
Malek Mouhoub, CS820 Fall 2009
20
Natural Language understanding
Natural Language understanding
• Systems based on syntaxico-semantic parsing
– Use semantic knowledge during the syntactic parsing.
– Systems for specific domains.
• Systems holding pragmatic :
– Use the relation between the sentence to be understood and the actual real
world.
Malek Mouhoub, CS820 Fall 2009
21
Natural Language understanding
Natural Language understanding
• New systems :
– Interact between the different levels of knowledge :
∗
∗
∗
∗
∗
linguistic,
acoustics, phonetic, phonological, prosodic (spoken language),
psychology
formal logic,
artificial intelligence techniques.
– Consider other language properties :
∗ synonymies, homonyms,
∗ allotaxies, homotaxies,
∗ anaphoras, ellipses, tropes, . . . .
Malek Mouhoub, CS820 Fall 2009
22
Planning and Robotics
Planning and Robotics
• Determine the different steps to solve a problem.
⇒ hierarchical problem decomposition
⇒ solving the different constraints (Decision Problem)
⇒ find the optimal way to solve the problem (Optimization
Problem)
Malek Mouhoub, CS820 Fall 2009
23
Planning and Robotics
Planning and Robotics
¤
¡
£Robotics ¢
• Determine the different atomic actions performed by a robot to solve a certain
task.
• Robot motion planning :
– Determine a motion plan of a robot (if it exists) avoiding all the obstacles
∗ Large size of space of possible sequences of moves.
∗ Reacting in a dynamic environment (moving obstacles, new events).
⇒ starting with an incomplete plan which is completed during its
execution (perception using sensors).
Malek Mouhoub, CS820 Fall 2009
24
Machine Learning
Machine Learning
• Limitations of the actual AI systems :
– No learning from experience, analogy or examples.
– Difficulty to modify the large code in which they are written.
• Automated Mathematician (AM) [Lenat 1977,82] : discover (induce)
mathematical laws given a set of axioms and concepts.
• Meta-DENDRAL : Learning rules for interpreting spectrographic data in organic
chemistry from examples of data of compounds of known structure (case-based
reasoning).
Malek Mouhoub, CS820 Fall 2009
25
Neural Networks
Neural Networks
• Modeling the brain instead of making a mind.
• Provide approximated solutions instead of answering yes or no.
• Provide a solution with a quality depending on the time allocated for
computation (useful for Real-Time applications).
Input
Neurons
Figure 2: Feedback Model
Malek Mouhoub, CS820 Fall 2009
26
AI : Present and Future
AI : Present and Future
Which of the following can be done at present?
• Play a decent game of table tennis
• Drive along a curving mountain road
• Drive in the center of Cairo
• Play a decent game of bridge
• Discover and prove a new mathematical theorem
• Write an intentionally funny story
• Give competent legal advice in a specialized area of law
• Translate spoken English into spoken Swedish in real time
Malek Mouhoub, CS820 Fall 2009
27