Chapter13 - Chapter 13: Ar tificial I ntelligence...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Chapter 13: Ar tificial I ntelligence Computations that mak e it possible for a machine to per ceive, r eason, and act in a manner consistent w ith human behavior for m the field k now n as a r ti fi ci a l i ntel l i gence. T he T ur ing T est A human questioner inputs questions and guesses w hich r espondent is human, based on the answ er s given. I f the questioner cannot tell w hich r espondent is human, the softw ar e passes the test. Chapter 13 Ar t ificial I nt elligence Page 1 Computer Reasoning T o simulate logical r easoning, heur i sti c functi ons ar e often used. istic is an ar tificial measur e of how close the A heur computer 's cur r ent status is to its pr oblem-solving goal. F ( confi g) = g) (# of complet able r ow s, columns, and diagonals for X-player ) (# of completable r ow s, columns, and diagonals for Oplayer ) - 1 X O O X X O O if confi g is a non-w inning configur at ion if confi g is an X-w in if confi g is an O-w in - O O X - - - X X X X O O O X 2-1=1 X X O O X 3-1=2 X O X O O X 2-1=1 O X O O X O 3-1=2 X O O O X X - X X 3-2=1 X O X X 2-2=0 X O X X O 3-2=1 X OXO O O O O O O O X - X X O X 2-2=0 X X X 2-2=0 X X O X O 3-2=1 X X OXO X O O O O O O O X - O X 2-1=1 X X X 3-1=2 X O X X O 3-1=2 X X Chapter 13 Ar t ificial I nt elligence Page 2 OXO XO O O O O O O O O O O X X X X X X X X - O O O X X X2-1=1 X X 2-1=1 X O X X 2-1=1 X O OXO O O O O X X X I n the tic-tac-toe example above, w hen the computer is r eady to mak e an X-move, it uses the heur istic max{F ( confi g), w her e confi g can be the r esult of any O- Exper t Systems By pr ogr amming a computer w ith the assistance of exper ts in a par ticular field, an exper t system can be developed to per for m ver y specialized task s. U ser s in need of exper t assistance complete on-line questionnair es and the exper t system analyzes their r esponses and assigns pr obabilities to var ious diagnoses. Chapter 13 Ar t ificial I nt elligence Page 3 N eur al N etw or k s T o simulate lear ning, cer tain multipr ocessor systems, called neur a l networ ks, have been built to "lear n" to r ecognize par ticular patter ns as cor r ect or incor r ect, based upon a tr below , a neur al ocess. k is used to teach I n the example ial-and-er r or pr netw or a computer ized system how to back a tr uck up to a loading dock . T he physical char acter istics of the tr uck ar e pr ogr ammed, w ith the r elationship betw een the steer ing w heel, the tir es, the cab, and the tr ailer for mally calculated. Star ting at some initial position, the tr uck is back ed up one meter at a time, w ith pr ogr ammed steer ing; the er r or in the r esult is measur ed and factor ed into the next attempt, until the er r or is Chapter 13 Ar t ificial I nt elligence Page 4 N atur al L anguage Pr ocessing Wr itten Compr ehension: H ow can a computer be pr ogr ammed to gr asp the syntax and semantics of a natur al language? John saw the boy in the par k w ith the telescope. Questi on: Whose tel escope Answer : J ohn's i s i t? John saw the boy in the par k w ith the puppy. Questi on: Whose puppy i s iAnswer : T he boy's t? John saw the boy in the par k w ith the statue. Questi on: Whose sta tue i s iAnswer : T he pa r k's t? Chapter 13 Ar t ificial I nt elligence Page 5 Speech Recognition 1. T he PC sound car d conver ts analog sound w aves spok en into a micr ophone into a digital for mat. 0011010101000 0101111101010 0100110101010 0101011110010 1011010100110 1010101010101 0101010101001 "K" 0011010101000 "K" CALM 3. A softw ar e "AH " COMMA 2. A softw ar e 0101111101010 "AH " language "M " COMPARE acoustical 0100110101010 "M " model "P" COMPETE model 0101011110010 "P" compar es the "Y" COMPLETE br eak s the 1011010100110 "Y" COMPUTE phonemes to "OO w or d into 1010101010101 w or ds in its " phonemes. "OO" dictionar y. "T " 0101010101001 "T " 4. Once the softw ar e decides on the most lik ely candidate, it displays that w or d. COM PU T Chapter 13 Ar t ificial I nt elligence Page 6 Robotics Robots ar e pr ogr ammable devices capable of manipulating objects and per for ming task s much lik e humans ar e able to do. One of the mor e difficult pr oblems w hen pr ogr amming a r obot is deter mining w hen it is about to collide w ith something, w hen it has collided w ith something, and w hat to do in r esponse to a collision. Collision Avoidance U se scanner s to deter mine the r obot's pr oximity to other objects, r edir ecting the r obot w hen a collision is imminent. Collision D etection U se sensor s at str ategically located places on the r obot to deter mine if a collision occur s. Collision Reaction Go ar ound? Climb over ? Bounce back ? Run aw ay? D r op dead? Chapter 13 Ar t ificial I nt elligence Page 7 Robot Challenges Other common human actions that ar e difficult to pr ogr am include pr opelled locomotion and manual manipulation. Walk ing Gait H ow can a r obot be pr ogr ammed to pr opel itself for w ar d on "legs" and still maintain its balance? Gr asping H ow can a r obot be pr ogr ammed to gr asp par t of a stack of objects, w ithout toppling the r est of the stack ? Chapter 13 Ar t ificial I nt elligence Page 8 AI in Games N on-player char acter s in games need to appear intelligent, even though they ar e contr olled by the game pr ogr am instead of by a game player . D ead Reck oning By having a game-dr iven pr edator char acter r eact to the anti cipated position of its player dr iven pr ey (using the pr ey's cur r ent position and velocity), a chase can appear mor e r ealisti c. F lock ing By pr oviding a gr oup of char acter s w i th simple goals and behavior s, a "mob mentality" can be i mplemented Chapter 13 Ar t ificial I nt elligence Page 9 ...
View Full Document

Ask a homework question - tutors are online