Belief-optimal Reasoning for Cyber-physical Systems - Computer ...

deadmancrossingraceΤεχνίτη Νοημοσύνη και Ρομποτική

13 Νοε 2013 (πριν από 3 χρόνια και 9 μήνες)

139 εμφανίσεις

CS
B551: E
LEMENTS

OF

A
RTIFICIAL

I
NTELLIGENCE

Instructor: Kris Hauser

http://cs.indiana.edu/~hauserk

1

B
ASICS


Class web site


http://cs.indiana.edu/classes/b551


Textbook


S. Russell and P.
Norvig


Artificial Intelligence: a Modern Approach


3
rd

edition


2
nd

edition can be used, but is not preferable


2

B
ASICS


Instructor


Kris Hauser (
hauserk@indiana.edu
)


AIs


Kai Song (
kaisong@indiana.edu
)


3

O
FFICE

H
OURS


Kris Hauser


Tu

10
-
11,W 12
-
1 in Info E 257 (connector building)


Kai Song


TBA

4

A
GENDA


Intro to AI


Overview of class policies

5

W
HAT

IS

AI?


AI is the reproduction of
human reasoning and
intelligent behavior

by computational methods

6

W
HAT

IS

AI?


AI is
an attempt of

reproduction of
human
reasoning and intelligent behavior

by
computational methods

7

W
HAT

IS

AI?


Discipline that systematizes and automates
reasoning processes to create machines that:

8

Think like humans

Think rationally

Act like humans

Act rationally


The goal of AI is: to build machines that
operate in the same way that humans think


How do humans think?


Build machines according to theory, test how
behavior matches mind’s behavior


Cognitive Science


Manipulation of symbolic knowledge


How does hardware affect reasoning? Discrete
machines, analog minds

9

Think like humans

Think rationally

Act like humans

Act rationally


The goal of AI is: to build machines that perform tasks
that seem to require intelligence when performed by
humans


Take a task at which people are better, e.g.:


Prove a theorem


Play chess


Plan a surgical operation


Diagnose a disease


Navigate in a building


and build a computer system that does it
automatically


But do we want to duplicate human imperfections?


10

Think like humans

Think rationally

Act like humans

Act rationally


The goal of AI is: to build machines that make
the “best” decisions given current knowledge
and resources


“Best” depending on some
utility

function


Influences from economics, control theory


How do self
-
consciousness, hopes, fears,
compulsions, etc. impact intelligence?


Where do utilities come from?


11

Think like humans

Think rationally

Act like humans

Act rationally

W
HAT

IS

I
NTELLIGENCE
?

“If there were machines which bore a resemblance to our
bodies and imitated our actions as closely as possible for
all practical purposes, we should still have two very
certain means of recognizing that they were not real men.
The first is that they could never use words, or put
together signs, as we do in order to declare our thoughts to
others… Secondly, even though some machines might do
some things as well as we do them, or perhaps even better,
they would inevitably fail in others, which would reveal
that they are acting not from understanding, …”

Discourse on the Method, by Descartes (1598
-
1650)


12

W
HAT

IS

I
NTELLIGENCE
?


Turing Test (c. 1950)

13

W
HAT

IS

INTELLIGENCE
?

A
N

A
PPLICATION

OF

THE

T
URING

T
EST


CAPTCHA: Completely Automatic Public Turing
tests to tell Computers and Humans Apart


15

C
HINESE

R
OOM

(J
OHN

S
EARLE
)

16

C
AN

M
ACHINES

A
CT
/T
HINK

I
NTELLIGENTLY
?


Yes, if intelligence is narrowly defined as
information processing


AI has made impressive achievements showing
that tasks initially assumed to require
intelligence can be automated


Each success of AI seems to push further the
limits of what we consider “intelligence”

17

S
OME

A
CHIEVEMENTS


Computers have won over world
champions in several games,
including Checkers, Othello, and
Chess, but still do not do well in
Go


AI techniques are used in many
systems: formal calculus, video
games, route planning, logistics
planning, pharmaceutical drug
design, medical diagnosis,
hardware and software trouble
-
shooting, speech

recognition
,
traffic monitoring,

facial
recognition,

medical
image
analysis, part
inspection
, etc...


DARPA Grand Challenge:

robotic car autonomously
traversed 132 miles of
desert


IBM’s Watson competes with
Jeopardy champs


Some industries (automobile,
electronics) are highly robotized,

while other robots perform brain

and heart surgery, are rolling

on Mars, fly autonomously, …,

but home robots still remain

a thing of the future

18

18

C
AN

M
ACHINES

A
CT
/T
HINK

I
NTELLIGENTLY
?


Yes, if intelligence is narrowly defined as
information processing


AI has made impressive achievements showing
that tasks initially assumed to require
intelligence can be automated


Maybe yes, maybe not, if intelligence cannot be
separated from consciousness


Is the machine
experiencing

thought?


Strong vs. Weak AI

19

B
IG

O
PEN

Q
UESTIONS


Is intelligent behavior just information
processing?

(Physical symbol system hypothesis)


If so, can the human brain solve problems
that are inherently intractable for
computers? Will a general theory of
intelligence emerge from neuroscience?


In a human being, where is the interface
between “intelligence” and the rest of
“human nature”


Self
-
consciousness, emotions, compulsions


What is the role of the body?

(Mind
-
body problem)

20

21


AI contributes to building an information
processing model of human beings, just as
Biochemistry contributes to building a model
of human beings based on bio
-
molecular
interactions


Both try to explain how a human being
operates


Both also explore ways to avoid human
imperfections
(in Biochemistry, by engineering new
proteins and drug molecules; in AI, by designing
rational reasoning methods)


Both try to produce new useful technologies


Neither explains (yet?) the true meaning of
being human

M
AIN

A
REAS

OF

AI


Knowledge representation
(including formal logic)


Search, especially heuristic
search (puzzles, games)


Planning


Reasoning under
uncertainty, including
probabilistic reasoning


Learning


Robotics and perception


Natural language processing

22

Search

Knowledge

rep.

Planning

Reasoning

Learning

Agent

Robotics

Perception

Natural

language

...

Expert

Systems

Constraint

satisfaction

B
ITS

OF

H
ISTORY


1956:

The name “Artificial Intelligence” is coined


60’s:
Search and games, formal logic and theorem
proving


70’s:

Robotics, perception, knowledge
representation, expert systems


80’s:

More expert systems, AI becomes an
industry


90’s:

Rational agents, probabilistic reasoning,
machine learning


00’s:

Systems integrating many AI methods,
machine learning, natural language processing,
reasoning under uncertainty, robotics again

23

AI R
EFERENCES


Conferences


IJCAI, ECAI, AAAI, NIPS


Journals


AI, Comp. I, IEEE Trans. Pattern Anal. Mach. Intel.,
IEEE Int. Sys., JAIR


Societies


AAAI, SIGART, AISB


AI Magazine (Editor: IU’s David Leake)

24

C
AREERS

IN

AI


‘Pure’ AI


Academia, industry labs


Applied AI


Almost any area of CS!


NLP, vision, robotics


Economics


Cognitive Science

25

S
YLLABUS


Introduction to AI


Philosophy, history, agent frameworks


Search


Uninformed search, heuristic search,
heuristics, game
playing


Reasoning under uncertainty


Probability, planning under uncertainty, Bayesian
networks, probabilistic inference, temporal sequences


Machine learning


Neural nets, decision tree learning, support vector
machines, etc.


Applications


Constraint satisfaction,
motion planning, computer
vision


26

27

B335

Robotics

I400

B552

B553

Knowledge representation and
learning

B657

Computer Vision

Biologically
-
inspired computing

B659

I486

Q360

B651

Natural Language Processing

E626

Game theory

Q570

Topics in AI

B551

S626

S675

B555

B556

B553

Algorithms for Optimization
and Learning

C
LASS

P
OLICIES

28

P
REREQUISITES


C211


I recommend:


Two semesters programming


Basic knowledge of data structures


Basic knowledge of algorithmic complexity

29

P
ROGRAMMING

A
SSIGNMENTS


Projects will be written in
Python


Easy to learn


2 weeks for each assignment

30

G
RADING


75% Homework


6 assignments, lowest score will be dropped


25% Final

31

H
OMEWORK

P
OLICY


Due at end of class on due date


Typically Thursdays


No
“slip days”


Extensions only granted in rare cases


Require advance notice except emergencies

32

F
INAL

P
ROJECT


Encouraged if you are intending to do
research or coursework in AI, pursue
higher degree


Individual or small groups (up to 3)


Counts
as two homework assignments


Content


Software, new research, or technical report


Mid
-
semester project proposal


End
-
of
-
year report and in
-
class presentation

E
NROLLMENT


Add/drop deadline w/o penalty: Aug 27


Waitlist deadline: Aug 25

34

T
AKEAWAYS


AI has many interpretations


Act vs. think, human
-
like vs. rational


Concept has evolved


“Intelligence” has many interpretations


Turing test


Chinese room


AI success stories from each perspective

35

H
OMEWORK


Register


Textbook


Survey


http://cs.indiana.edu/classes/b551


Readings:


R&N Ch. 1, 26 (introduction and historical
perspectives)


R&N 3.1
-
3


36