course july lecture02

7/30/2019 Course July Lecture02

1/26

CS 416

Artificial Intelligence

Lectu re 2

Introduct ion


2/26

CS at UVa

$11M in research grants each year

Top 5% of research is funded by NSF

Faculty trips to NSF set national funding priorities

Free MSFT Visual Studio for all students

75% facul ty g row th in past six years

Undergrad research awards from CRA

Highest s tart ing salary ( in SEAS) for ugrads


3/26

Textbook

This is a great book

2nd edition released three years ago

Most widely used in U.S. universities

Its so good. Im going to make you read it!

Homework Read chapters 1 and 2


4/26

Survey Results

Languages

Supermajority prefers C++

Three people indicated theyll need C++ help

LISP? Math

Many w/o stat

7 w/o diffyq

14 w/o linear algebra


5/26

5 people w/o GUI experience

4 people w/o MSFT Windows

14 people dont play so many video games

Where have you done the most programming? 216 17

Graphics 15

201/202 6

OS 2


6/26

AI apps

Chess, google, spam filter, finance, chatterbot, games,

vacuum

12% of CPU for AI tasks in games?

More about magic tricks than AI?

iRoomba - Rodney Brooks (MIT) company


7/26

Languages

Is AI special in its PL needs? AI research used to be more symbolic

A language had to make it easy to create symbols and tomanipulate them

Some symbols would operate on other symbols

LISP supported programs as data and dynamic typing

Modern AI is more quantitative

No language has emerged with an advantage

Our language choice cannot distract from learning AI


8/26

Languages

C++ - Common industry language

C gets a little closer to real-time OS

Perl the duct tape of the Internetmakes the easy things easy and

the hard things impossible theres more than one way to do it

Pythontheres only one way to do it

Scheme easy to learn but difficult to extend

Common Lispthe programmable programming language nontrivia

to learn but a decidedly different experience from programming in

imperative languages


9/26

What is expected of you

Youll have to do math

Neural network update function

Multidimensional function

minimization ProbabilityBayes Rule

We will teach necessary parts ofstatistics and linear algebra

Tcx ji

cx

jiw

Pw

,

,

2

)(

)()|()|(

XP

YPYXPXYP

Calculus expected.

Probability and Linea

Algebra beneficial.


10/26

What is expected of you

You have to program

The programming assignments are non-trivial

C++

Requires integration with existing code libraries

Input/output handling (images, for example)

We do not teach programming in this course

CS 216 expected.

Additional programmin

experience beneficial


11/26

AI Systems

Thermostat

Tic-Tac-Toe

Your car

Chess

Google

Babblefish

This thingAsimo


12/26

Examples

Chess: Deep Junior (IBM) tied Kasparov in 2003 match

ATRs DB Android

Hondas Asimo

Ritsumeikan University

RHex Hexapod
http://kesisleme.eecs.umich.edu/filedisplay.php?which=f&action=movie&id=8http://kesisleme.eecs.umich.edu/filedisplay.php?which=f&action=movie&id=9http://kesisleme.eecs.umich.edu/filedisplay.php?which=f&action=movie&id=66


13/26

AI Techniques

Rule-based

Fuzzy Logic

Neural Networks

Genetic Algorithms

Exhaustive search

Expert Systems

Logic


14/26

How to Categorize These Systems

Systems that think l ike humans

Systems that act l ike humans

Systems that think rat ional ly

Systems that act rat ionally


15/26

Systems that think/act like humans

Its hard to study things you cant observe

How can I know how you think?

Observation is difficult (changing with fMRI). For the most part, you

are a black box

Cognitive Science

How can I know how you act?

Observation is possible, but hard to control all aspects of

experimental conditions.

Turing Test


16/26

Alan TuringBuilding a Brain

World War II mot ivated computer advances

Code breaking (1943, Colossus) Used to decipher

telegrams encrypted using Germanys encryption machine

Electronic Numerical Integrator and Computer (ENIAC, 1946)

Tur ing great ly invo lved with B r i t ish ef for ts to bu i ld

com puters and crack codes (Bletchley Park)

Arrested for being a homosexual in 1952 and denied security clearance

Committed suicide in 1954


17/26

Systems that think/act rationally

Rely on log ic i tsel f rather than human tomeasu re co rrectness

Thinking rationally (logically)

Socrates is a human; All humans are mortal; Socrates is morta

Logic formulas for synthesizing outcomes

Acting rationally (logically)

Even if method is illogical, the observed behavior must berational


18/26

Perspective of this Course

We wil l invest igate the general pr in cip les o frat ional agents

Not restricted to human actions and human environments

Not restricted to human thought

Not confined to only using laws of logic

Anything goes so long as it produces rational behavior


19/26

What is AI?

The use of computers to solve problems thatpreviously could only be solved by applying human

intelligence. thus something can fit this definition

today, but, once we see how the program works and

understand the problem, we will not think of it as AI

anymore (David Parnas)


20/26

Foundations - Philosophy

Aristotle (384 B.C.E.) Author of logical syllogisms

da Vinci (1452)designed, but didnt build, first mechanical

calculator

Descartes (1596) can human free will be captured by a

machine? Is animal behavior more mechanistic?

Necessary connection between logic and action is

discovered


21/26

Foundations - Mathematics

Leveraging uncertainty (Cardano 1501)

Boolean logic (Boole, 1847)

Analysis of limits to what can be computed

Intractability (1965) time required to solve problemscales exponentially with the size of problem instance

NP-complete (1971) Formal classification of problems as

intractable


22/26

Foundations - Economics

Game Theory study of rational behavior in small games

Operations Research study of rational behavior in

complex systems

Herbert Simon (1916 2001) AI researcher who received

Nobel Prize in Economics for showing people accomplish

satisficing solutions, those that are good enough


23/26

Foundations - Neuroscience

How do bra ins work? Early studies (1824) relied on injured and abnormal people to understand what

parts of brain do

More recent studies use accurate sensors to correlate brain activity to humanthought

By monitoring individual neurons, monkeys can now control a computermouse using thought alone

Melody Moore at GaStatelocked-in syndrome

(Gordon) Moores law states computers will have as many gates as humanshave neurons in 2020

How close are we to having a mechanical brain? Parallel computation, remapping, interconnections, binary vs. gradient


24/26

Foundations - Psychology

Helmholtz and Wundt (1821) started to make psychology ascience by carefully controlling experiments

The brain processes information (1842)

Sense Think Act

Cognitive science started at a MIT workshop in 1956 with

the publication of three very influential papers


25/26

Foundations Control Theory

Machines can modify their behavior in response to theenvironment (sense / action loop)

Water-flow regulator (250 B.C.E), steam engine governor,

thermostat

The theory of stable feedback systems (1894)

Build systems that transition from initial

state to goal state with minimum energy

In 1950, control theory could only describe

linear systems and AI largely rose as aresponse to this shortcoming


26/26

Foundations - Linguistics

Speech demons trates so much of humanintel l igence

Analysis of human language reveals thought taking place in

ways not understood in other settings

Children can create sentences they have never heard

before

Language and thought are believed to be tightly

intertwined

course july lecture02

Documents