Anda di halaman 1dari 6

CS 385 Artificial Intelligence Exam 1, September 27, 2006

Do 7 of the following problems. Please submit your work by October 4. You may use your book
and any other non-human sources you wish.
1. riti!ue "uring#s criteria for computer software being $intelligent.$ Propose your own
definition.
%lan "uring&s paper 'omputing (achinery and )ntelligence* in 1+,- proposed what is now
referred to as the ."uring "est& to define artificial intelligence. "he test consists of a human
interrogator whose task it is to hold con/ersations with another human0 acting naturally0 and a
computer programmed to act like a natural human. )f such a piece of software e1isted0 "uring
would label it as successful artificial intelligence.
(y belief is that such a program or interface would indeed appear to be intelligent0 and it
would be0 for the meaning of intelligence as applied knowledge and reasoning. 2owe/er0
artificial intelligence should not be defined in standards of reflecting human action0 because
what makes human beings so remarkable is not the actions they perform0 but the thought
process that underlies such actions.
%lthough0 this itself does not in/alidate "uring&s proposal0 but deepens it. 3uch a machine
would need to be able to react as a human would to a wide range of tasks. "here may /ery
well need to be hundreds or thousands of different scenarios to fulfill0 but some might be
brainstorming and proposing new ideas0 showing appropriate emotional reaction to hardships
and boons0 and gi/ing meaningful ad/ice for /arious described scenarios.
......
%lthough "uring&s criteria are useful for determining if a machine can successfully emulate a
human0 it is not necessarily an indication of intelligence in itself. )ntelligence may be re!uired to
emulate a human0 but if an intelligence is different or greater than that of a human it may not be
e1posed in the test. )t may also be true that an intelligence cannot know through e1perience
what a human should know without being human itself.
) propose di/iding intelligence into two categories4 intelligence and sentience. )ntelligence is
ha/ing the knowledge necessary to accomplish something0 and the ability to perform this
action in whole or in part. 3entience is the ability to identify and analy5e intelligence0 and in
turn to identify and analy5e this intelligence. "hus sentience is recursi/e intelligence0 and
although this may eliminate some humans from the status of sentient0 who is to say that this is
wrong6
7. 8sing truth tables0 pro/e or dispro/e the following4
a. 9P :; 99 P; 9 :;;
P :
P: 9P:; P : 9P;
9:;
P : 9P :; <
9 P :;
" " " = = = = = "
" = = " = " " = "
= " = " " = " = "
= = = " " " " " "
b. 9 P :; 9 P >; 9> P; :
P : >
P > P: P> 9P:; > 9>P; 9P:; 9P>;
9P>;
P :
9>P; :
" " " = = " " " " " "
" = " = = = " = " = "
= " " " = " " " " " "
= = " " = " " " " = =
" " = = " " " " " " "
" = = = " = " = " = "
= " = " " " = = = = "
= = = " " " = = = = "
?ot true.
@. on/ert to list notation and unify if possible. 3how the steps /ia trees as in =igure 7.A.
a. a9B0 b9B0 c;0Y; and a9d0 Y0 e;
b. a9B0 b9B0 c;0Y; and a9d0 Y0 C;
7a. unify99a B 9b B c; Y; 0 9a d Y e;;
DE
unify9a0a; unify99B 9b B c; Y; 0 9d Y e;;
DdFBE
unify 9B0 d; unify999b d c; Y; 0 9Y e;;
Db d c;FYE
unify99b d c; Y;; unify99b d c;0 e; fail.
b. unify99a B 9b B c; Y; 0 9a d Y C;;
DE DdFB0 bcdFY0 bcdFYE
unify9a0a; unify99B 9b B c; Y; 0 9Y C;;
DdFBE DbcdFY. bcdFC E
unify 9B0 d; unify999b d c; Y; 0 9Y C;;
DbdcFYE DbdcFC E
unify99b d c; Y;; unify99b d c;0 C;
substitutions DdFB0 bdcFY0 bcdFCE produce a9d0 b9d0c;0 b9d0c;;
4. )n what order will the nodes in =igure 1 be searched by4
a. depth first search % G = H I J 2 ( D ) K ? L
b. breadth first search % G D L = I 2 ) K H J ( ?
,. "race the backtrack function on =igure 10 assuming a goal of J0 indicating at each step the
contents of 30 3J0 ?3J0 and DL. Put children of 3 at the beginning of ?3J.
3 3 ?3 DL
% M%N M%N M N
G MG%N MG D L %N M N
= M= G %N M= I G D L %N M N
H MH = G %N MH = I G D L %N M N
= M = G %N M= I G D L %N M H N
I MI G %N MI G D L %N M = H N
J MJ I G %N MJ I G D L %N M = H N
Ioal
A. a. "ranslate the following a1ioms into predicate calculus
1. (ary is a lawyer
7. %nn is a computer scientist
@. %nn is rich
4. Jawyers are rich
,. omputer scientists are smart
A. Jawyers ha/e nice cars
7. 3mart0 rich people ha/e nice cars
1. lawyer9mary;
7. computerOscientist9ann;
@. rich9ann;
4. lawyer9B; rich9B;
,. computerOscientist9B; smart9B;
A. lawyer9B; niceOcar9B;
7. smart9B; rich9B; niceOcar9B;
b. Perform the unifications and inferences necessary to pro/e %nn has a nice car
. unify 7 with left side of , to conclude
P4 smart9ann;
unify @ and P with left side of 7 to conclude
+4 niceOcar9ann;
c. Draw the corresponding andFor graph.
niceOcar9B;

rich9B; smart9B;
lawyer9B; computerOscientist9B;

lawyer9mary; rich9ann; computerOscientist9ann;
7. a. "ranslate problem A into P>OJOI
lawyer9mary;.
computerOscientist9ann;.
rich9ann;.
rich9B;4-
lawyer9B;.
smart9B;4-
computerOscientist9B;.
niceOcar9B;4- lawyer9B;.
niceOcar9B;4-
smart9B;0
rich9B;.
b. Qhat will be the P>OJOI !ueries and resulting answers for the following4
1; Does %nn ha/e a nice car6
7; Qho is rich6
6- niceOcar9ann;.
yes
6- rich9B;.
B < ann R
B < mary
c. Qhat are the ad/antagesFdisad/antages of this representation o/er that in problem A6
Disad/antages4
1. You ha/e to learn P>OJOI.
7. One can get mi1ed up as to which side the sufficient and necessary clauses should go and
the meanings of the symbols.
@. De/eloping a more comple1 problem would yield more errors because of the probable
detail of the problem where as words could e1plain the problem better
%d/antages4
1. it is formal and automated
P. Qrite @ of the following predicates in P>OJOI4
a. stepmother9B0 Y;.
stepmother9C0 Y; 4- remarried 9B0 C;0 birthparent 9B0 Y;. FF5
b. cousin9B0 Y;.
cousin9B0Y; 4- 9parent9Q0B; S siblings9Q0C;; S parent 9C0 Y;. FF5
c. subset9J0 JJ; which returns true if J is a subset of JJ
d. flatten9J0 JJ; which returns the elements in list J with possible sublists in JJ with no
sublists. =or e1ample0 if J is M10 M70@N0 MM4NN0 ,; JJ is M10 70 @0 40 ,N.
+. %dd rules that allow noun infiniti/es to the grammar rules in L1ample @.@.A. %n infinitive is to
plus a verb phrase "he noun infinitive can be a subTect 9"o eat is fun.; or a direct obTect 9)
like to eat.;. Parse a sample sentence using these new rules using both data dri/en and goal
dri/en search
1. sentence np /p
7. np n
@. np art n
4. np adT n
,. np art adT n
A. /p /
7. /p / np
P. /p / ad/
+. /p / np ad/
1-. art a
11. art the
17. n man
1@. n dog
14. / likes
1,. / bites
1A. adT brown
17. ad/ slowly
iteration working memory conflict set fire rule
1 sentence 1 1
7 np /p 7 @ 4 , A 7 P + @
@ art n /p A 7 P + 1- 11 17 1@ 11
4 the n /p A 7 P + 1- 17 1@ 17
, the man /p A 7 P + 1- 1@ +
A the man / np ad/ 7 @ 4 , 14 1, 17 1,
7 the man bites np ad/ 7 @ 4 , 17 ,
P the man bites art adT n ad/ 1- 11 17 1@ 1A 17 11
+ the man bites the adT n ad/ 17 1@ 1A 17 1A
1- the man bites the brown n ad/ 17 1@17 1@
11 the man bites the brown dog ad/ 17 17
17 the man bites the brown dog slowly
Figure 1: %
G D L
= I 2 ) K
H J ( ?

Anda mungkin juga menyukai