You are on page 1of 1

S. Russel and P. Norvig. Artificial Intelligence: a Modern Approach. Chapter 1 1.

1 Define in your own words: (a) intelligence, (b) artificial intelligence, (c) agent. 1.2 Read Turing's original paper on A1 (Turing, 19 !) "ttp:##$ind.o%ford&ournals.org#cgi#reprint#'()#*+,#-++. (n t"e
paper, "e discusses se.eral potential ob&ections to "is proposed enterprise and "is test for intelligence. /"ic" ob&ections still carry so$e weig"t0 Are "is refutations .alid0 1an you t"in2 of new ob&ections arising fro$ de.elop$ents since "e wrote t"e paper0 (n t"e paper, "e predicts t"at, by t"e year *!!!, a co$puter will "a.e a +!3 c"ance of passing a fi.e4 $inute Turing Test wit" an uns2illed interrogator. /"at c"ance do you t"in2 a co$puter would "a.e today0 (n anot"er ! years0 1.3 5.ery year t"e 'oebner pri6e "ttp:##www.loebner.net#7ri6ef#loebner4pri6e."t$l is awarded to t"e progra$ t"at co$es closest to passing a .ersion of t"e Turing test. Researc" and report on t"e latest winner of t"e 'oebner pri6e. /"at tec"ni8ues does it use0 9ow does it ad.ance t"e state of t"e art in A(0

1.4 T"ere are well42nown classes of proble$s t"at are intractably difficult for co$puters, and ot"er classes t"at are
pro.ably undecidable. Does t"is $ean t"at A1 is i$possible0 1.5 :uppose we e%tend 5.ans's A;A'<=> progra$ so t"at it can score *!! on a standard (? test. /ould we t"en "a.e a progra$ $ore intelligent t"an a "u$an0 5%plain.

1.6 9ow could introspection4reporting on one's inner t"oug"ts4be inaccurate0 1ould ( be wrong about w"at ('$ t"in2ing0
Discuss.

1.7 5%a$ine t"e A1 literature to disco.er w"et"er t"e following tas2s can currently be sol.ed by co$puters:
a. 7laying a decent ga$e of table tennis (ping4pong). b. Dri.ing in t"e center of 1airo. c. @uying a wee2's wort" of groceries at t"e $ar2et. d. @uying a wee2's wort" of groceries on t"e web. e. 7laying a decent ga$e of bridge at a co$petiti.e le.el. f. Disco.ering and pro.ing new $at"e$atical t"eore$s. g. /riting an intentionally funny story. h. =i.ing co$petent legal ad.ice in a speciali6ed area of law. i. Translating spo2en 5nglis" into spo2en :wedis" in real ti$e. j. 7erfor$ing a co$ple% surgical operation. Aor t"e currently infeasible tas2s, try to find out w"at t"e difficulties are and predict w"en, if e.er, t"ey will be o.erco$e.

1.8 :o$e aut"ors "a.e clai$ed t"at perception and $otor s2ills are t"e $ost i$portant part of intelligence, and t"at
'B"ig"er le.elB capacities are necessarily parasitic4si$ple add4ons to t"ese underlying facilities. 1ertainly, $ost of e.olution and a large part of t"e brain "a.e been de.oted to perception and $otor slulls, w"ereas A1 "as found tas2s suc" as ga$e playing and logical inference to be easier, in $any ways, t"an percei.ing and acting in t"e real world. Do you t"in2 t"at A('s traditional focus on "ig"er4le.el cogniti.e abilities is $isplaced0

1.9 /"y would e.olution tend to result in syste$s t"at act rationally0 /"at goals are suc" syste$s designed to ac"ie.e0 1.10 Are refle% actions (suc" as $o.ing your "and away fro$ a "ot sto.e) rational0 Are t"ey intelligent0 1.11 B:urely co$puters cannot be intelligent4t"ey can do only w"at t"eir progra$$ers tell t"e$.B (s t"e latter state$ent
true, and does it i$ply t"e for$er0

1.12 B:urely ani$als cannot be intelligent4t"ey can do only w"at t"eir genes tell t"e$.B (s t"e latter state$ent true, and
does it i$ply t"e for$er0

1.13 B:urely ani$als, "u$ans, and co$puters cannot be intelligent4t"ey can do only w"at t"eir constituent ato$s are
told to do by t"e laws of p"ysics.B (s t"e latter state$ent true, and does it i$ply t"e for$er0

You might also like