lecture - 天津大学计算机科学与技术学院

unclesamnorweiganAI and Robotics

Oct 18, 2013 (4 years and 22 days ago)

107 views

Artificial Intelligence

人工智能

Xiu
-
jun GONG (Ph. D)

School of Computer Science and Technology, Tianjin
University

gongxj@tju.edu.cn


http://cs.tju.edu.cn/faculties/gongxj/course/ai/


About the instructor


Name: Xiu
-
jun GONG (
宫秀军
)


Work experiences


2006/05
-
Now


Associate Professor, Tianjin University


2003/05
-
2006/03: Research fellow, Nara Institute of Science
and Technology


2003/02
-
2003/05: Visiting fellow, Institute for Inforcomm
Research

I
2
R), Singapore


2002/07
-
2002/12: Research fellow, National University of
Singapore


1999/09
-
2002/07: Ph. D candidate, Institute of Computing,
CAS


Research interests


Data mining: algorithms, standards, and systems


Bioinformatics: gene regulatory network, SNP identifications


Medical informatics: secure, privacy
-
preserving data mining,
medical data integration and sharing framework

About the course


Text book


Artificial Intelligence
-
A New Synthesis, Nils J. Nillson




Artificial Intelligence: A Modern Approach, Stuart Russell
and Peter Norvig




Artificial Intelligence: Structures and Strategies for
Complex Problems Solving (Fourth Edition), George F.
Luger



Grading


Attendance: 10%


Project & Assignment: 20%


Final exam: 70%


Office hour: any time upon pre
-
appointment
before final exam, 25
-
B
-
1208


Web site:
http://cs.tju.edu.cn/~gongxj/course/ai


Outline to the introduction


AI definitions


AI history


AI research


Problems


Approaches


Tools


AI Applications


AI resources

What is AI

To

make

computers

think

...


machines

with

minds



(Haugeland,

1985
)



The study of the computations

that make it possible to
perceive,
reason

… (Winston,1992)

Machines that perform
functions that require
intelligence

when performed
by people (Kurzweil, 1990)

The automation of
intelligent
behavior

(Luger, 1993)


Thinking

humanly

Thinking

rationally



Acting

human
ly

Acting

rationally

What is AI (cont.)


AI is a branch of cs that is concerned with the
automation

of
intelligent behavior

Luger


Data structures, algorithms, and language and
programming techniques.


What is the

intelligent behavior

?


Think (act) humanly


Think (act) rationally


Can machines think?


Can: Now or someday;
theoretically or actually


Machine: biological body (made of proteins), mechanical
device?


Think: media? Living cells or physical symbolic systems

Some synonyms

Intelligent machine, intelligent system, intelligent agent,
computational intelligence, synthetic intelligence

Performed by google trends on 7
th
, Oct, 2008

Beyond the definitions


The definitions differ for different people, different
contexts, and different historical periods (see the
AI history)


AI has always been more concerned with
expanding the capacities of computer science
than with defining its limits


AI is the interdisciplinary study of computer
science including psychology, philosophy,
neuroscience, cognitive science, linguistics,
ontology, operations research, economics, control
theory, probability, optimization and logic.

Collection of problems and methodologies
studied by AI researchers

History of AI research


Precursors


1943−1956: The birth of AI


1956−1974: The golden years


1974−1980: The first AI winter


1980

1987: Boom


1987−1993: Bust: the second AI winter


1993−present: AI ?

Precursors (1)



AI in myth, fiction and speculation

Precursors (2)

Al
-
Jazari's programmable
automata

Automatons

Formal reasoning

Computer science

1943−1956: The birth of AI (2)


Turing's test (1950)
-
ACT Humanly


Decide whether a machine is intelligent or not

If a machine could carry on a conversation (over a teletype) that
was indistinguishable from a conversation with a human being,
then the machine could be called "intelligent."

1943−1956: The birth of AI (2)


Dartmouth Summer
Research Conference on
Artificial Intelligence in
1956


Marvin Minsky, John
McCarthy


Coined the term

AI



Every aspect of learning or
any other feature of
intelligence can be so
precisely described that a
machine can be made to
simulate it
--
a clear
statement of the
philosophical position of AI
research


Presentation of game playing
programs and Logic Theorist.

1956−1974: The golden years (1)


Reasoning as search


Maze problem
--
backtracking


Combinatorial explosion
--

heuristics or "rules
of thumb “


Projects


Simon etc, General Problem Solver (1951)


Herbert Gelernter , Geometry Theorem Prover (1958)


James Slagle, SAINT (Symbolic Automatic
Integrator )(1961)


Nils Nilsson , STRIPS(Stanford Research Institute
Problem Solver ) (1971)


1956−1974: The golden years (2)


Natural language


Allow computers to communicate in natural languages
--
semantic network


STUDENT
, solve high school algebra word problems
(1964)


ELIZA
, rephrasing many of the patient's statements as
questions and posing them to the patients (1966)


ALICE:
http://www.alicebot.org


Micro
-
worlds


Marvin Minsky, machine vision

They

pointed

out

that

in

successful

sciences

were

often

best

understood

using

simplified

models

like

frictionless

planes

or

perfectly

rigid

bodies
.

Much

of

the

research

focused

on

the

so
-
called

"blocks

world,"

which

consists

of

colored

blocks

of

various

shapes

and

sizes

arrayed

on

a

flat

surface

.

1974−1980: The first AI winter (1)


Critiques from across campus (mainly from
philosophers

)


John Lucas, argued G
ö
del's incompleteness theorem (a formal
system could never see the truth of certain statements, while
a human being could)


Hubert Dreyfus, argued that human reasoning actually
involved very little "symbol processing" and a great deal of
embodied, instinctive, unconscious "know how".


John Searle

, Chinese Room argument (a program could not be
said to "understand" the symbols that it uses )


Perceptrons and the dark age of connectionism


perceptron may eventually be able to learn, make
decisions, and translate languages

(Frank Rosenblatt,
1958)


Minsky and Papert's, book
Perceptrons
.

1969

1974−1980: The first AI winter (2)


The neats: logic, Prolog and expert systems


Logic into AI: McCathy 1958


Deduction on computers: J. Alan Robinson 1963


Prolog: Philippe Roussel, Alain Colmerauer, 1972


Critics: human beings rarely used logic when they
solved problems


The scruffies: frames and scripts


Gerald Sussman observed that "using precise language
to describe essentially imprecise concepts doesn't make
them any more precise."


Minsky noted that many of his fellow "scruffy"
researchers were using the same kind of tool: a
framework that captures all our common sense
assumptions about something. 1975

1980

1987: Boom (1)


The rise of expert systems (main stream of
AI)


MYCIN
, 1972, diagnosed infectious blood diseases


XCON

(eXpert CONfigurer), 1980, automatically
selecting the computer system components based on
the customer's requirements


The knowledge revolution


The power of expert systems came from the expert
knowledge they contained


Cyc (enCyclopedia), assemble a comprehensive ontology
and database of everyday common sense knowledg,
Douglas Lenat 1984

1980

1987: Boom (2)


The revival of connectionism


John Hopfield (
associative neural
network

,1982)


David Rumelhart (backpropagation)


The money returns


the fifth generation project ($850
million,1982, 10
-
year program)



epoch
-
making computer




massive parallel processing


Failure in 1992


Alvey (England, ₤350 )(1983
-
1987)


Strategic Computing Initiative (DARPA)
(1984)


PIM/m
-
1 machine

1987−1993: the second AI winter


Market changed


Desktop computers from Apple and IBM had been
steadily gaining speed and power


Robotics facts

having a body essentially


A machine needs to have a
body



it needs to perceive,
move, survive and deal with the world


David Marr, AI needed to understand the physical
machinery of vision from the bottom up before any
symbolic processing took place.


Rodney Brooks, Elephants Don't Play Chess , symbols
are not always necessary since "the world is its own best
model

.

physical symbol system hypothesis


1993−present: AI ?


Deep Blue beats Kasparov (1997)


DARPA grand challenge: Autonomous vehicle
navigates across desert. (Urban Challenge next)
2005


NASA Remote Agent in Deep Space I probe
explores solar system


iRobot Roomba automated vacuum cleaner


Automated speech/language systems for airline
travel


Usable machine translation thru Google



?

Advanced Intelligence


Close interactions and coordination
between Natural Intelligence and Artificial
Intelligence



The frontiers in both Artificial Intelligence
and Natural Intelligence



Large
-
scale Distributed Intelligence and
Web Intelligence

China’ s Programs on AI


国家中长期科学和技术发展规划纲要(
2006
-
2020



重点领域及其优先主题


传感器网络及智能信息处理

重点开发多种新型传感器及先进条码自动识别、射频标签、基于多种传
感信息的智能化信息处理技术,发展低成本的传感器网络和实时信息
处理系统,提供更方便、功能更强大的信息服务平台和环境。


基础研究:



脑科学与认知科学


主要研究方究向:脑功能的细胞和分子机理,脑重大疾病的发生发
展机理,脑发育、可塑性与人类智力的关系,学习记忆和思维等
脑高级认知功能的过程及其神经基础,脑信息表达与脑式信息处
理系统,人脑与计算机对话等。

Problems of AI


Deduction, reasoning, problem solving


Knowledge representation


Planning


Learning


Natural language processing


Motion and manipulation


Perception


Social intelligence


Creativity


General intelligence

Approaches to AI

Acting rationally


The rational agent
approach

Thinking humanly


The cognitive
approach

Acting humanly


The Turing Test
approach

Thinking rationally


The laws of
thought approach

Approaches to AI cont.


Symbolism


Cognitive simulation: Psychologism
-

Herbert
Simon and Alan Newell)


Logical AI: Logicism
-

John McCarthy



"Scruffy" symbolic AI : Computerism,
commonsense knowledge bases
-

Marvin
Minsky


Connectionism


Hopfield,

Pitts


Neural networks


Actionism


Brooks


Cybernetics and brain simulation

Tools of AI research


Search and optimization


Logic


Probabilistic methods for uncertain
reasoning


Classifiers and statistical learning
methods


Neural networks


Control theory


Specialized languages


Lisp
is a practical mathematical notation
for computer programs based on lambda
calculus


Prolog

is a declarative language where
programs are expressed in terms of
relations, and execution occurs by running
queries over these relations


STRIPS

a language for expressing
automated planning problem instances.


Planner

is a hybrid between procedural
and logical languages.

Application domains


Machine Learning


Natural Language Processing


Expert System


Patten Recognition


Computer Vision


Robotics


Game Playing


Automatic Theorem Proving


Automatic Programming


机器学习


自然语言处理


专家系统

模式识别


计算机视觉


机器人学


博弈


自动定理证明


自动程序设计



Application domains (cont. )


Intelligent Control


Intelligent Decision Support
System


Artificial Neural Network


Knowledge Discovery in
Database & Data Mining


Distributed AI


Intelligent Agent


Intelligent Retrieval from
Database

智能控制


智能决策支持系



人工神经网络


知识发现和数据
挖掘


分布式人工智能


智能代理

智能数据库检索

AI resources: Journals (premium)


Artificial Intelligence


Computational Linguistics


IEEE Trans on Pattern Analysis and Machine Intl


IEEE Trans on Robotics and Automation


IEEE Trans on Image Processing


Journal of AI Research


Neural Computation


Machine Learning


Intl Jnl of Computer Vision


IEEE Trans on Neural Networks

AI resources: Journals (leading)


Artificial Intelligence Review


ACM Transactions on Asian Language Information Processing


AI Magazine


Applied Artificial Intelligence


Artificial Intelligence in Medicine


Computational Intelligence


Computer Speech and Language


Expert Systems with Applications: An Intl Jnl


IEEE Trans on Systems, Man, & Cybernetics, Part A & B


Intl Jnl on Artificial Intelligence Tools


Jnl of Experimental & Theoretical AI


Journal of East Asian Linguistics


Knowledge Engineering Review


Machine Translation


Neural Networks


Pattern Recognition


Neurocomputing

AI competitions


Machine Intelligence Prize




Loebner prize





KDD Cup serires





AI resources: Conferences


AAAI: American Association for AI National Conference


CVPR: IEEE Conf on Comp Vision and Pattern Recognition


IJCAI: Intl Joint Conf on AI


ICCV: Intl Conf on Computer Vision


ICML: Intl Conf on Machine Learning


KDD: Knowledge Discovery and Data Mining


KR: Intl Conf on Principles of KR & Reasoning


NIPS: Neural Information Processing Systems


UAI: Conference on Uncertainty in AI


AAMAS: Intl Conf on Autonomous Agents and Multi
-
Agent
Systems


ACL: Annual Meeting of the ACL (Association of
Computational Linguistics)

Summary


AI definition


Whatever the definition is, Collection of
problems and methodologies studied by AI
researchers is an important clue for
investigating AI problems


AI history


History is a mirror. AI researchers are getting
more intelligent


AI research


Integration of multi
-
disciplines.


Bring AI into practice and reality