%!
%%BoundingBox: (atend)
%%Pages: (atend)
%%DocumentFonts: (atend)
%%EndComments
%
% FrameMaker PostScript Prolog 3.0, for use with FrameMaker 3.0
% Copyright (c) 1986,87,89,90,91 by Frame Technology Corporation.
% All rights reserved.
%
% Known Problems:
% Due to bugs in Transcript, the 'PS-Adobe-' is omitted from line 1
/FMversion (3.0) def
% Set up Color vs. Black-and-White
/FMPrintInColor systemdict /colorimage known
systemdict /currentcolortransfer known or def
% Uncomment this line to force b&w on color printer
% /FMPrintInColor false def
/FrameDict 195 dict def
systemdict /errordict known not {/errordict 10 dict def
errordict /rangecheck {stop} put} if
% The readline in 23.0 doesn't recognize cr's as nl's on AppleTalk
FrameDict /tmprangecheck errordict /rangecheck get put
errordict /rangecheck {FrameDict /bug true put} put
FrameDict /bug false put
mark
% Some PS machines read past the CR, so keep the following 3 lines together!
currentfile 5 string readline
00
0000000000
cleartomark
errordict /rangecheck FrameDict /tmprangecheck get put
FrameDict /bug get {
/readline {
/gstring exch def
/gfile exch def
/gindex 0 def
{
gfile read pop
dup 10 eq {exit} if
dup 13 eq {exit} if
gstring exch gindex exch put
/gindex gindex 1 add def
} loop
pop
gstring 0 gindex getinterval true
} def
} if
/FMVERSION {
FMversion ne {
/Times-Roman findfont 18 scalefont setfont
100 100 moveto
(FrameMaker version does not match postscript_prolog!)
dup =
show showpage
} if
} def
/FMLOCAL {
FrameDict begin
0 def
end
} def
/gstring FMLOCAL
/gfile FMLOCAL
/gindex FMLOCAL
/orgxfer FMLOCAL
/orgproc FMLOCAL
/organgle FMLOCAL
/orgfreq FMLOCAL
/yscale FMLOCAL
/xscale FMLOCAL
/manualfeed FMLOCAL
/paperheight FMLOCAL
/paperwidth FMLOCAL
/FMDOCUMENT {
array /FMfonts exch def
/#copies exch def
FrameDict begin
0 ne dup {setmanualfeed} if
/manualfeed exch def
/paperheight exch def
/paperwidth exch def
/yscale exch def
/xscale exch def
currenttransfer cvlit /orgxfer exch def
currentscreen cvlit /orgproc exch def
/organgle exch def /orgfreq exch def
setpapername
manualfeed {true} {papersize} ifelse
{manualpapersize} {false} ifelse
{desperatepapersize} if
end
} def
/pagesave FMLOCAL
/orgmatrix FMLOCAL
/landscape FMLOCAL
/FMBEGINPAGE {
FrameDict begin
/pagesave save def
3.86 setmiterlimit
/landscape exch 0 ne def
landscape {
90 rotate 0 exch neg translate pop
}
{pop pop}
ifelse
xscale yscale scale
/orgmatrix matrix def
gsave
} def
/FMENDPAGE {
grestore
pagesave restore
end
showpage
} def
/FMFONTDEFINE {
FrameDict begin
findfont
ReEncode
1 index exch
definefont
FMfonts 3 1 roll
put
end
} def
/FMFILLS {
FrameDict begin
array /fillvals exch def
end
} def
/FMFILL {
FrameDict begin
fillvals 3 1 roll put
end
} def
/FMNORMALIZEGRAPHICS {
newpath
0.0 0.0 moveto
1 setlinewidth
0 setlinecap
0 0 0 sethsbcolor
0 setgray
} bind def
/fx FMLOCAL
/fy FMLOCAL
/fh FMLOCAL
/fw FMLOCAL
/llx FMLOCAL
/lly FMLOCAL
/urx FMLOCAL
/ury FMLOCAL
/FMBEGINEPSF {
end
/FMEPSF save def
/showpage {} def
FMNORMALIZEGRAPHICS
[/fy /fx /fh /fw /ury /urx /lly /llx] {exch def} forall
fx fy translate
rotate
fw urx llx sub div fh ury lly sub div scale
llx neg lly neg translate
} bind def
/FMENDEPSF {
FMEPSF restore
FrameDict begin
} bind def
FrameDict begin
/setmanualfeed {
%%BeginFeature *ManualFeed True
statusdict /manualfeed true put
%%EndFeature
} def
/max {2 copy lt {exch} if pop} bind def
/min {2 copy gt {exch} if pop} bind def
/inch {72 mul} def
/pagedimen {
paperheight sub abs 16 lt exch
paperwidth sub abs 16 lt and
{/papername exch def} {pop} ifelse
} def
/papersizedict FMLOCAL
/setpapername {
/papersizedict 14 dict def
papersizedict begin
/papername /unknown def
/Letter 8.5 inch 11.0 inch pagedimen
/LetterSmall 7.68 inch 10.16 inch pagedimen
/Tabloid 11.0 inch 17.0 inch pagedimen
/Ledger 17.0 inch 11.0 inch pagedimen
/Legal 8.5 inch 14.0 inch pagedimen
/Statement 5.5 inch 8.5 inch pagedimen
/Executive 7.5 inch 10.0 inch pagedimen
/A3 11.69 inch 16.5 inch pagedimen
/A4 8.26 inch 11.69 inch pagedimen
/A4Small 7.47 inch 10.85 inch pagedimen
/B4 10.125 inch 14.33 inch pagedimen
/B5 7.16 inch 10.125 inch pagedimen
end
} def
/papersize {
papersizedict begin
/Letter {lettertray letter} def
/LetterSmall {lettertray lettersmall} def
/Tabloid {11x17tray 11x17} def
/Ledger {ledgertray ledger} def
/Legal {legaltray legal} def
/Statement {statementtray statement} def
/Executive {executivetray executive} def
/A3 {a3tray a3} def
/A4 {a4tray a4} def
/A4Small {a4tray a4small} def
/B4 {b4tray b4} def
/B5 {b5tray b5} def
/unknown {unknown} def
papersizedict dup papername known {papername} {/unknown} ifelse get
end
/FMdicttop countdictstack 1 add def
statusdict begin stopped end
countdictstack -1 FMdicttop {pop end} for
} def
/manualpapersize {
papersizedict begin
/Letter {letter} def
/LetterSmall {lettersmall} def
/Tabloid {11x17} def
/Ledger {ledger} def
/Legal {legal} def
/Statement {statement} def
/Executive {executive} def
/A3 {a3} def
/A4 {a4} def
/A4Small {a4small} def
/B4 {b4} def
/B5 {b5} def
/unknown {unknown} def
papersizedict dup papername known {papername} {/unknown} ifelse get
end
stopped
} def
/desperatepapersize {
statusdict /setpageparams known
{
paperwidth paperheight 0 1
statusdict begin
{setpageparams} stopped pop
end
} if
} def
/savematrix {
orgmatrix currentmatrix pop
} bind def
/restorematrix {
orgmatrix setmatrix
} bind def
/dmatrix matrix def
/dpi 72 0 dmatrix defaultmatrix dtransform
dup mul exch dup mul add sqrt def
/freq dpi 18.75 div 8 div round dup 0 eq {pop 1} if 8 mul dpi exch div def
/sangle 1 0 dmatrix defaultmatrix dtransform exch atan def
/DiacriticEncoding [
/.notdef /.notdef /.notdef /.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef /.notdef /.notdef /.notdef
/.notdef /.notdef /.notdef /.notdef /space /exclam /quotedbl
/numbersign /dollar /percent /ampersand /quotesingle /parenleft
/parenright /asterisk /plus /comma /hyphen /period /slash /zero /one
/two /three /four /five /six /seven /eight /nine /colon /semicolon
/less /equal /greater /question /at /A /B /C /D /E /F /G /H /I /J /K
/L /M /N /O /P /Q /R /S /T /U /V /W /X /Y /Z /bracketleft /backslash
/bracketright /asciicircum /underscore /grave /a /b /c /d /e /f /g /h
/i /j /k /l /m /n /o /p /q /r /s /t /u /v /w /x /y /z /braceleft /bar
/braceright /asciitilde /.notdef /Adieresis /Aring /Ccedilla /Eacute
/Ntilde /Odieresis /Udieresis /aacute /agrave /acircumflex /adieresis
/atilde /aring /ccedilla /eacute /egrave /ecircumflex /edieresis
/iacute /igrave /icircumflex /idieresis /ntilde /oacute /ograve
/ocircumflex /odieresis /otilde /uacute /ugrave /ucircumflex
/udieresis /dagger /.notdef /cent /sterling /section /bullet
/paragraph /germandbls /registered /copyright /trademark /acute
/dieresis /.notdef /AE /Oslash /.notdef /.notdef /.notdef /.notdef
/yen /.notdef /.notdef /.notdef /.notdef /.notdef /.notdef
/ordfeminine /ordmasculine /.notdef /ae /oslash /questiondown
/exclamdown /logicalnot /.notdef /florin /.notdef /.notdef
/guillemotleft /guillemotright /ellipsis /.notdef /Agrave /Atilde
/Otilde /OE /oe /endash /emdash /quotedblleft /quotedblright
/quoteleft /quoteright /.notdef /.notdef /ydieresis /Ydieresis
/fraction /currency /guilsinglleft /guilsinglright /fi /fl /daggerdbl
/periodcentered /quotesinglbase /quotedblbase /perthousand
/Acircumflex /Ecircumflex /Aacute /Edieresis /Egrave /Iacute
/Icircumflex /Idieresis /Igrave /Oacute /Ocircumflex /.notdef /Ograve
/Uacute /Ucircumflex /Ugrave /dotlessi /circumflex /tilde /macron
/breve /dotaccent /ring /cedilla /hungarumlaut /ogonek /caron
] def
/ReEncode {
dup
length
dict begin
{
1 index /FID ne
{def}
{pop pop} ifelse
} forall
0 eq {/Encoding DiacriticEncoding def} if
currentdict
end
} bind def
/graymode true def
/bwidth FMLOCAL
/bpside FMLOCAL
/bstring FMLOCAL
/onbits FMLOCAL
/offbits FMLOCAL
/xindex FMLOCAL
/yindex FMLOCAL
/x FMLOCAL
/y FMLOCAL
/setpattern {
/bwidth exch def
/bpside exch def
/bstring exch def
/onbits 0 def /offbits 0 def
freq sangle landscape {90 add} if
{/y exch def
/x exch def
/xindex x 1 add 2 div bpside mul cvi def
/yindex y 1 add 2 div bpside mul cvi def
bstring yindex bwidth mul xindex 8 idiv add get
1 7 xindex 8 mod sub bitshift and 0 ne
{/onbits onbits 1 add def 1}
{/offbits offbits 1 add def 0}
ifelse
}
setscreen
{} settransfer
offbits offbits onbits add div FMsetgray
/graymode false def
} bind def
/grayness {
FMsetgray
graymode not {
/graymode true def
orgxfer cvx settransfer
orgfreq organgle orgproc cvx setscreen
} if
} bind def
/HUE FMLOCAL
/SAT FMLOCAL
/BRIGHT FMLOCAL
/Colors FMLOCAL
FMPrintInColor
{
/HUE 0 def
/SAT 0 def
/BRIGHT 0 def
% array of arrays Hue and Sat values for the separations [HUE BRIGHT]
/Colors
[[0 0 ] % black
[0 0 ] % white
[0.00 1.0] % red
[0.37 1.0] % green
[0.60 1.0] % blue
[0.50 1.0] % cyan
[0.83 1.0] % magenta
[0.16 1.0] % comment / yellow
] def
/BEGINBITMAPCOLOR {
BITMAPCOLOR} def
/BEGINBITMAPCOLORc {
BITMAPCOLORc} def
/BEGINBITMAPTRUECOLOR {
BITMAPTRUECOLOR } def
/BEGINBITMAPTRUECOLORc {
BITMAPTRUECOLORc } def
/K {
Colors exch get dup
0 get /HUE exch store
1 get /BRIGHT exch store
HUE 0 eq BRIGHT 0 eq and
{1.0 SAT sub setgray}
{HUE SAT BRIGHT sethsbcolor}
ifelse
} def
/FMsetgray {
/SAT exch 1.0 exch sub store
HUE 0 eq BRIGHT 0 eq and
{1.0 SAT sub setgray}
{HUE SAT BRIGHT sethsbcolor}
ifelse
} bind def
}
{
/BEGINBITMAPCOLOR {
BITMAPGRAY} def
/BEGINBITMAPCOLORc {
BITMAPGRAYc} def
/BEGINBITMAPTRUECOLOR {
BITMAPTRUEGRAY } def
/BEGINBITMAPTRUECOLORc {
BITMAPTRUEGRAYc } def
/FMsetgray {setgray} bind def
/K {
pop
} def
}
ifelse
/normalize {
transform round exch round exch itransform
} bind def
/dnormalize {
dtransform round exch round exch idtransform
} bind def
/lnormalize {
0 dtransform exch cvi 2 idiv 2 mul 1 add exch idtransform pop
} bind def
/H {
lnormalize setlinewidth
} bind def
/Z {
setlinecap
} bind def
/fillvals FMLOCAL
/X {
fillvals exch get
dup type /stringtype eq
{8 1 setpattern}
{grayness}
ifelse
} bind def
/V {
gsave eofill grestore
} bind def
/N {
stroke
} bind def
/M {newpath moveto} bind def
/E {lineto} bind def
/D {curveto} bind def
/O {closepath} bind def
/n FMLOCAL
/L {
/n exch def
newpath
normalize
moveto
2 1 n {pop normalize lineto} for
} bind def
/Y {
L
closepath
} bind def
/x1 FMLOCAL
/x2 FMLOCAL
/y1 FMLOCAL
/y2 FMLOCAL
/rad FMLOCAL
/R {
/y2 exch def
/x2 exch def
/y1 exch def
/x1 exch def
x1 y1
x2 y1
x2 y2
x1 y2
4 Y
} bind def
/RR {
/rad exch def
normalize
/y2 exch def
/x2 exch def
normalize
/y1 exch def
/x1 exch def
newpath
x1 y1 rad add moveto
x1 y2 x2 y2 rad arcto
x2 y2 x2 y1 rad arcto
x2 y1 x1 y1 rad arcto
x1 y1 x1 y2 rad arcto
closepath
16 {pop} repeat
} bind def
/C {
grestore
gsave
R
clip
} bind def
/FMpointsize FMLOCAL
/F {
FMfonts exch get
FMpointsize scalefont
setfont
} bind def
/Q {
/FMpointsize exch def
F
} bind def
/T {
moveto show
} bind def
/RF {
rotate
0 ne {-1 1 scale} if
} bind def
/TF {
gsave
moveto
RF
show
grestore
} bind def
/P {
moveto
0 32 3 2 roll widthshow
} bind def
/PF {
gsave
moveto
RF
0 32 3 2 roll widthshow
grestore
} bind def
/S {
moveto
0 exch ashow
} bind def
/SF {
gsave
moveto
RF
0 exch ashow
grestore
} bind def
/B {
moveto
0 32 4 2 roll 0 exch awidthshow
} bind def
/BF {
gsave
moveto
RF
0 32 4 2 roll 0 exch awidthshow
grestore
} bind def
/G {
gsave
newpath
normalize translate 0.0 0.0 moveto
dnormalize scale
0.0 0.0 1.0 5 3 roll arc
closepath fill
grestore
} bind def
/A {
gsave
savematrix
newpath
2 index 2 div add exch 3 index 2 div sub exch
normalize 2 index 2 div sub exch 3 index 2 div add exch
translate
scale
0.0 0.0 1.0 5 3 roll arc
restorematrix
stroke
grestore
} bind def
/x FMLOCAL
/y FMLOCAL
/w FMLOCAL
/h FMLOCAL
/xx FMLOCAL
/yy FMLOCAL
/ww FMLOCAL
/hh FMLOCAL
/FMsaveobject FMLOCAL
/FMoptop FMLOCAL
/FMdicttop FMLOCAL
/BEGINPRINTCODE {
/FMdicttop countdictstack 1 add def
/FMoptop count 4 sub def
/FMsaveobject save def
userdict begin
/showpage {} def
FMNORMALIZEGRAPHICS
3 index neg 3 index neg translate
} bind def
/ENDPRINTCODE {
count -1 FMoptop {pop pop} for
countdictstack -1 FMdicttop {pop end} for
FMsaveobject restore
} bind def
/gn {
0
{ 46 mul
cf read pop
32 sub
dup 46 lt {exit} if
46 sub add
} loop
add
} bind def
/str FMLOCAL
/cfs {
/str sl string def
0 1 sl 1 sub {str exch val put} for
str def
} bind def
/ic [
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0223
0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0223
0
{0 hx} {1 hx} {2 hx} {3 hx} {4 hx} {5 hx} {6 hx} {7 hx} {8 hx} {9 hx}
{10 hx} {11 hx} {12 hx} {13 hx} {14 hx} {15 hx} {16 hx} {17 hx} {18 hx}
{19 hx} {gn hx} {0} {1} {2} {3} {4} {5} {6} {7} {8} {9} {10} {11} {12}
{13} {14} {15} {16} {17} {18} {19} {gn} {0 wh} {1 wh} {2 wh} {3 wh}
{4 wh} {5 wh} {6 wh} {7 wh} {8 wh} {9 wh} {10 wh} {11 wh} {12 wh}
{13 wh} {14 wh} {gn wh} {0 bl} {1 bl} {2 bl} {3 bl} {4 bl} {5 bl} {6 bl}
{7 bl} {8 bl} {9 bl} {10 bl} {11 bl} {12 bl} {13 bl} {14 bl} {gn bl}
{0 fl} {1 fl} {2 fl} {3 fl} {4 fl} {5 fl} {6 fl} {7 fl} {8 fl} {9 fl}
{10 fl} {11 fl} {12 fl} {13 fl} {14 fl} {gn fl}
] def
/sl FMLOCAL
/val FMLOCAL
/ws FMLOCAL
/im FMLOCAL
/bs FMLOCAL
/cs FMLOCAL
/len FMLOCAL
/pos FMLOCAL
/ms {
/sl exch def
/val 255 def
/ws cfs
/im cfs
/val 0 def
/bs cfs
/cs cfs
} bind def
400 ms
/ip {
is
0
cf cs readline pop
{ ic exch get exec
add
} forall
pop
} bind def
/wh {
/len exch def
/pos exch def
ws 0 len getinterval im pos len getinterval copy pop
pos len
} bind def
/bl {
/len exch def
/pos exch def
bs 0 len getinterval im pos len getinterval copy pop
pos len
} bind def
/s1 1 string def
/fl {
/len exch def
/pos exch def
/val cf s1 readhexstring pop 0 get def
pos 1 pos len add 1 sub {im exch val put} for
pos len
} bind def
/hx {
3 copy getinterval
cf exch readhexstring pop pop
} bind def
/h FMLOCAL
/w FMLOCAL
/d FMLOCAL
/lb FMLOCAL
/bitmapsave FMLOCAL
/is FMLOCAL
/cf FMLOCAL
/wbytes {
dup
8 eq {pop} {1 eq {7 add 8 idiv} {3 add 4 idiv} ifelse} ifelse
} bind def
/BEGINBITMAPBWc {
1 {} COMMONBITMAPc
} bind def
/BEGINBITMAPGRAYc {
8 {} COMMONBITMAPc
} bind def
/BEGINBITMAP2BITc {
2 {} COMMONBITMAPc
} bind def
/COMMONBITMAPc {
/r exch def
/d exch def
gsave
translate rotate scale /h exch def /w exch def
/lb w d wbytes def
sl lb lt {lb ms} if
/bitmapsave save def
r
/is im 0 lb getinterval def
ws 0 lb getinterval is copy pop
/cf currentfile def
w h d [w 0 0 h neg 0 h]
{ip} image
bitmapsave restore
grestore
} bind def
/BEGINBITMAPBW {
1 {} COMMONBITMAP
} bind def
/BEGINBITMAPGRAY {
8 {} COMMONBITMAP
} bind def
/BEGINBITMAP2BIT {
2 {} COMMONBITMAP
} bind def
/COMMONBITMAP {
/r exch def
/d exch def
gsave
translate rotate scale /h exch def /w exch def
/bitmapsave save def
r
/is w d wbytes string def
/cf currentfile def
w h d [w 0 0 h neg 0 h]
{cf is readhexstring pop} image
bitmapsave restore
grestore
} bind def
/proc1 FMLOCAL
/proc2 FMLOCAL
/newproc FMLOCAL
/Fmcc {
/proc2 exch cvlit def
/proc1 exch cvlit def
/newproc proc1 length proc2 length add array def
newproc 0 proc1 putinterval
newproc proc1 length proc2 putinterval
newproc cvx
} bind def
/ngrayt 256 array def
/nredt 256 array def
/nbluet 256 array def
/ngreent 256 array def
/gryt FMLOCAL
/blut FMLOCAL
/grnt FMLOCAL
/redt FMLOCAL
/indx FMLOCAL
/cynu FMLOCAL
/magu FMLOCAL
/yelu FMLOCAL
/k FMLOCAL
/u FMLOCAL
/colorsetup {
currentcolortransfer
/gryt exch def
/blut exch def
/grnt exch def
/redt exch def
0 1 255 {
/indx exch def
/cynu 1 red indx get 255 div sub def
/magu 1 green indx get 255 div sub def
/yelu 1 blue indx get 255 div sub def
/k cynu magu min yelu min def
/u k currentundercolorremoval exec def
nredt indx 1 0 cynu u sub max sub redt exec put
ngreent indx 1 0 magu u sub max sub grnt exec put
nbluet indx 1 0 yelu u sub max sub blut exec put
ngrayt indx 1 k currentblackgeneration exec sub gryt exec put
} for
{255 mul cvi nredt exch get}
{255 mul cvi ngreent exch get}
{255 mul cvi nbluet exch get}
{255 mul cvi ngrayt exch get}
setcolortransfer
{pop 0} setundercolorremoval
{} setblackgeneration
} bind def
/tran FMLOCAL
/fakecolorsetup {
/tran 256 string def
0 1 255 {/indx exch def
tran indx
red indx get 77 mul
green indx get 151 mul
blue indx get 28 mul
add add 256 idiv put} for
currenttransfer
{255 mul cvi tran exch get 255.0 div}
exch Fmcc settransfer
} bind def
/BITMAPCOLOR {
/d 8 def
gsave
translate rotate scale /h exch def /w exch def
/bitmapsave save def
colorsetup
/is w d wbytes string def
/cf currentfile def
w h d [w 0 0 h neg 0 h]
{cf is readhexstring pop} {is} {is} true 3 colorimage
bitmapsave restore
grestore
} bind def
/BITMAPCOLORc {
/d 8 def
gsave
translate rotate scale /h exch def /w exch def
/lb w d wbytes def
sl lb lt {lb ms} if
/bitmapsave save def
colorsetup
/is im 0 lb getinterval def
ws 0 lb getinterval is copy pop
/cf currentfile def
w h d [w 0 0 h neg 0 h]
{ip} {is} {is} true 3 colorimage
bitmapsave restore
grestore
} bind def
/BITMAPTRUECOLORc {
gsave
translate rotate scale /h exch def /w exch def
/bitmapsave save def
/is w string def
ws 0 w getinterval is copy pop
/cf currentfile def
w h 8 [w 0 0 h neg 0 h]
{ip} {gip} {bip} true 3 colorimage
bitmapsave restore
grestore
} bind def
/BITMAPTRUECOLOR {
gsave
translate rotate scale /h exch def /w exch def
/bitmapsave save def
/is w string def
/gis w string def
/bis w string def
/cf currentfile def
w h 8 [w 0 0 h neg 0 h]
{ cf is readhexstring pop }
{ cf gis readhexstring pop }
{ cf bis readhexstring pop }
true 3 colorimage
bitmapsave restore
grestore
} bind def
/BITMAPTRUEGRAYc {
gsave
translate rotate scale /h exch def /w exch def
/bitmapsave save def
/is w string def
ws 0 w getinterval is copy pop
/cf currentfile def
w h 8 [w 0 0 h neg 0 h]
{ip gip bip w gray} image
bitmapsave restore
grestore
} bind def
/ww FMLOCAL
/r FMLOCAL
/g FMLOCAL
/b FMLOCAL
/i FMLOCAL
/gray {
/ww exch def
/b exch def
/g exch def
/r exch def
0 1 ww 1 sub { /i exch def r i get .299 mul g i get .587 mul
b i get .114 mul add add r i 3 -1 roll floor cvi put } for
r
} bind def
/BITMAPTRUEGRAY {
gsave
translate rotate scale /h exch def /w exch def
/bitmapsave save def
/is w string def
/gis w string def
/bis w string def
/cf currentfile def
w h 8 [w 0 0 h neg 0 h]
{ cf is readhexstring pop
cf gis readhexstring pop
cf bis readhexstring pop w gray} image
bitmapsave restore
grestore
} bind def
/BITMAPGRAY {
8 {fakecolorsetup} COMMONBITMAP
} bind def
/BITMAPGRAYc {
8 {fakecolorsetup} COMMONBITMAPc
} bind def
/ENDBITMAP {
} bind def
end
/ALDsave FMLOCAL
/ALDmatrix matrix def ALDmatrix currentmatrix pop
/StartALD {
/ALDsave save def
savematrix
ALDmatrix setmatrix
} bind def
/InALD {
restorematrix
} bind def
/DoneALD {
ALDsave restore
} bind def
%%EndProlog
%%BeginSetup
(3.0) FMVERSION
0.98 0.98 595.3 839.08 0 1 14 FMDOCUMENT
0 0 /Helvetica-Narrow-Bold FMFONTDEFINE
1 0 /Times-Roman FMFONTDEFINE
2 0 /Helvetica-Bold FMFONTDEFINE
3 0 /Times-Italic FMFONTDEFINE
4 0 /Times-Bold FMFONTDEFINE
5 0 /AvantGarde-Demi FMFONTDEFINE
6 0 /Courier-Bold FMFONTDEFINE
32 FMFILLS
0 0 FMFILL
1 .1 FMFILL
2 .3 FMFILL
3 .5 FMFILL
4 .7 FMFILL
5 .9 FMFILL
6 .97 FMFILL
7 1 FMFILL
8 <0f1e3c78f0e1c387> FMFILL
9 <0f87c3e1f0783c1e> FMFILL
10 FMFILL
11 FMFILL
12 <8142241818244281> FMFILL
13 <03060c183060c081> FMFILL
14 <8040201008040201> FMFILL
16 1 FMFILL
17 .9 FMFILL
18 .7 FMFILL
19 .5 FMFILL
20 .3 FMFILL
21 .1 FMFILL
22 0.03 FMFILL
23 0 FMFILL
24 FMFILL
25 FMFILL
26 <3333333333333333> FMFILL
27 <0000ffff0000ffff> FMFILL
28 <7ebddbe7e7dbbd7e> FMFILL
29 FMFILL
30 <7fbfdfeff7fbfdfe> FMFILL
%%EndSetup
%%Page: "0" 1
%%BeginPaperSize: A4
%%EndPaperSize
595.3 839.08 0 FMBEGINPAGE
232.07 655.74 151.92 655.74 2 L
0 X
0 K
V
1.17 H
0 Z
N
0 16 Q
(REPRESENT) 151.92 657.43 T
240.58 655.74 231.11 655.74 2 L
V
N
(A) 231.11 657.43 T
422.48 655.74 239.62 655.74 2 L
V
N
(TIONS AS CONTROL SUB-ST) 239.62 657.43 T
430.99 655.74 421.52 655.74 2 L
V
N
(A) 421.52 657.43 T
455.53 655.74 430.03 655.74 2 L
V
N
(TES) 430.03 657.43 T
1 12 Q
(Aaron Sloman) 268.92 636.1 T
(School of Computer Science) 234.94 618.1 T
(The University of Birmingham) 229.27 600.1 T
2 F
(\050DRAFT VERSION mARCH 1994\051) 211.44 570.1 T
126.48 521.28 69.73 521.28 2 L
V
1.02 H
N
2 14 Q
(Abstract) 69.73 522.77 T
1 12 Q
0.31 (Since \336rst presenting a paper criticising excessive reliance on logical representations in AI at the) 69.73 503.1 P
2.86 (second IJCAI at Imperial College London in 1971, I have been trying to understand what) 69.73 489.1 P
1.92 (representations are and why human beings seem to need so many dif) 69.73 475.1 P
1.92 (ferent kinds, tailored to) 420.05 475.1 P
0.4 (dif) 69.73 461.1 P
0.4 (ferent purposes. This position paper presents the beginnings of a general answer starting from) 82.84 461.1 P
1.47 (the notion that an intelligent agent is essentially a control system with multiple control states,) 69.73 447.1 P
1.87 (many of which contain information \050both factual and non-factual\051, albeit not necessarily in a) 69.73 433.1 P
1.71 (propositional form. The paper attempts to give a general characterisation of the notion of the) 69.73 419.1 P
0.78 (syntax of an information store, in terms of types of variation the relevant mechanisms can cope) 69.73 405.1 P
0.07 (with. Dif) 69.73 391.1 P
0.07 (ferent kinds of syntax can support dif) 112.89 391.1 P
0.07 (ferent kinds of semantics, and serve dif) 292.3 391.1 P
0.07 (ferent kinds) 480.7 391.1 P
-0.2 (of purposes. Similarly concepts of semantics, pragmatics and inference are generalised to apply to) 69.73 377.1 P
1.48 (information-bearing sub-states in control systems. A number of common but incorrect notions) 69.73 363.1 P
0.93 (about representation are criticised \050such as that pictures are in some way isomorphic with what) 69.73 349.1 P
3.24 (they represent\051, and a \336rst attempt is made to characterise dimensions in which forms of) 69.73 335.1 P
(representations can dif) 69.73 321.1 T
(fer) 178.09 321.1 T
(, including the explicit/implicit dimension.) 190.92 321.1 T
FMENDPAGE
%%EndPage: "0" 2
%%Page: "1" 2
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(1) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
223.71 769.87 143.56 769.87 2 L
0 X
V
1.17 H
0 Z
N
0 16 Q
(REPRESENT) 143.56 771.57 T
232.22 769.87 222.75 769.87 2 L
V
N
(A) 222.75 771.57 T
414.12 769.87 231.26 769.87 2 L
V
N
(TIONS AS CONTROL SUB-ST) 231.26 771.57 T
422.62 769.87 413.16 769.87 2 L
V
N
(A) 413.16 771.57 T
447.17 769.87 421.66 769.87 2 L
V
N
(TES) 421.66 771.57 T
1 12 Q
(Aaron Sloman) 260.55 750.23 T
(School of Computer Science) 226.58 732.23 T
(The University of Birmingham) 220.91 714.23 T
153.48 671.42 50.88 671.42 2 L
V
1.02 H
N
2 14 Q
(\0501\051 Introduction) 50.88 672.9 T
1 12 Q
-0.22 (This is a \322position paper\323 introducing one key idea: The study of the nature and role of representations) 50.88 653.23 P
0.23 (in intelligent systems can make most progress if based on the assumption that a mind \050or a brain\051 is a) 50.88 639.23 P
2.6 (sophisticated self-modifying control system with multiple control states, many of which contain) 50.88 625.23 P
-0.22 (information, of dif) 50.88 611.23 P
-0.22 (ferent kinds and in dif) 139.17 611.23 P
-0.22 (ferent forms \050Sloman 1993b\051. A representation, then, is a sub-) 244.02 611.23 P
0.32 (state in a control system. Dif) 50.88 597.23 P
0.32 (ferent sorts of representations will be found in dif) 190.15 597.23 P
0.32 (ferent control systems,) 430.3 597.23 P
(and even within dif) 50.88 583.23 T
(ferent sub-mechanisms in the same control system.) 143.6 583.23 T
0.3 0.34 (This idea, illustrating the \322design-based\323 approach to the study of mind \050Sloman 1993c\051, a) 72.14 563.23 B
-0.19 (development of Dennett\325) 50.88 549.23 P
-0.19 (s \0501978\051 notion of the \322design stance\323, is at odds with many common ideas of) 169.75 549.23 P
0.3 0.53 (representation, which are usually abstracted from an inadequate survey of types of) 50.88 535.23 B
3 F
0.3 0.53 (external) 496.81 535.23 B
1 F
0.3 0.47 (representations \050e.g. sentences and pictures\051. In particular) 50.88 521.23 B
0.3 0.47 (, our analysis of representations as) 356.62 521.23 B
0.3 0.53 (information-bearing control states undermines the idea that there are basically two kinds of) 50.88 507.23 B
0.3 0.32 (representations \050a\051 verbal/symbolic and \050b\051 pictorial/analogical/iconic, and such notions as that) 50.88 493.23 B
-0.26 (representations can be unambiguously classi\336ed as declarative or procedural. By looking at the variety) 50.88 479.23 P
0.3 0.04 (of types of information-rich sub-states in control systems, we\325ll f) 50.88 465.23 B
0.3 0.04 (ind a much richer variety than such) 367.46 465.23 B
0.3 0.08 (simple theories allow) 50.88 451.23 B
0.3 0.08 (. \050A recent survey) 154.98 451.23 B
0.3 0.08 (, Narayanan 1993, includes both papers that assume some of) 242.47 451.23 B
0.3 0.35 (these distinctions, and others that criticise them.\051 Further) 50.88 437.23 B
0.3 0.35 (, the common assumption that mental) 346.47 437.23 B
0.3 0.38 (representations are all consciously accessible is challenged by our analysis, since many of the) 50.88 423.23 B
0.3 0.23 (information-bearing control states discussed below have nothing to do with consciousness. More) 50.88 409.23 B
0.3 0.26 (surprisingly) 50.88 395.23 B
0.3 0.26 (, the analysis of representations as sub-states in control systems also undermines the) 110.55 395.23 B
0.28 (notion that symbols or representations are necessarily physical objects \050as suggested by the \322physical) 50.88 381.23 P
0.3 0.72 (symbol system\323 hypothesis of Newell and Simon 1981\051, or even that every part of every) 50.88 367.23 B
0.3 0.21 (representation must have a distinct underlying physical object as its implementation. Finally) 50.88 353.23 B
0.3 0.21 (, this) 515.18 353.23 B
(analysis leads to generalisations of the notions of syntax, semantics, pragmatics, and inference.) 50.88 339.23 T
0.3 0.01 (The \322control system\323 viewpoint adopted here is also at odds with a standard notion of a \322control) 72.14 319.23 B
-0.15 (system\323 that is limited to the kinds of systems normally studied by physicists and control engineers, in) 50.88 305.23 P
0.3 0.14 (which the behaviour can be completely described by f) 50.88 291.23 B
0.3 0.14 (ixed systems of partial dif) 318.88 291.23 B
0.3 0.14 (ferential equations) 448.15 291.23 B
0.3 0 (linking a f) 50.88 277.23 B
0.3 0 (ixed set of numerical variables. These conventional control systems all have a f) 100.14 277.23 B
0.3 0 (ixed degree) 483.92 277.23 B
0.3 0.24 (of complexity) 50.88 263.23 B
0.3 0.24 (, corresponding to a f) 120.49 263.23 B
0.3 0.24 (ixed set of dimensions or quantities that can vary) 228.59 263.23 B
0.3 0.24 (, and a f) 477.44 263.23 B
0.3 0.24 (ixed) 518.48 263.23 B
0.3 0.52 (collection of relations between these variables, usually expressed as equations or numerical) 50.88 249.23 B
(inequalities.) 50.88 235.23 T
0.05 (By contrast, control systems that exhibit intelligence, such as humans and other hominids, appear) 72.14 215.23 P
0.3 0.07 (to have architectures that are not only much richer) 50.88 201.23 B
0.3 0.07 (, with far more functional dif) 297.43 201.23 B
0.3 0.07 (ferentiation between) 440 201.23 B
0.3 0.24 (components than in standard control systems, but also do not have a f) 50.88 187.23 B
0.3 0.24 (ixed architecture, since the) 403.89 187.23 B
0.3 0.29 (number and variety of components and connections between components can change over time.) 50.88 173.23 B
0.3 0.29 (Furthermore, within such systems, many of the control states exhibit changes that are more like) 50.88 159.23 B
0.19 (changing structures \050e.g. trees and networks whose components and links change over time\051 than like) 50.88 145.23 P
0.3 0.23 (changing values of numerical variables \050e.g. voltage, pressure, velocity) 50.88 131.23 B
0.3 0.23 (, etc.\051 The lack of a static) 410.48 131.23 B
0.3 0.01 (architecture, and the use of information states that change in structure rather than in values of a f) 50.88 117.23 B
0.3 0.01 (ixed) 519.17 117.23 B
-0.16 (set of variables together imply that standard mathematics of control engineers will need to be enriched) 50.88 103.23 P
FMENDPAGE
%%EndPage: "1" 3
%%Page: "2" 3
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(2) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
(in order to cope with intelligent systems.) 50.88 774.23 T
0.3 0.33 (I am not suggesting that there is a sharp dividing line between intelligent and unintelligent) 72.14 754.23 B
0.3 0.08 (systems, nor that we know in any detail how biological control systems actually work. Instead I am) 50.88 740.23 B
0.3 0.19 (of) 50.88 726.23 B
0.3 0.19 (fering a theoretical framework within which dif) 61.03 726.23 B
0.3 0.19 (ferent sorts of systems can be distinguished and) 300 726.23 B
0.3 0.17 (classif) 50.88 712.23 B
0.3 0.17 (ied, and which may turn out later to be useful for describing and analysing both natural and) 82.02 712.23 B
0.3 0.06 (artif) 50.88 698.23 B
0.3 0.06 (icial behaving systems, in termsf of their architecture, their functional subdivisions, the types of) 70.49 698.23 B
(information-bearing states and the types of causal interactions underlying their observable behaviour) 50.88 684.23 T
(.) 533.53 684.23 T
419.43 651.42 50.88 651.42 2 L
V
1.02 H
0 Z
N
2 14 Q
(\0502\051 The importance of \322abstract\323, or \322virtual\323, machines) 50.88 652.9 T
1 12 Q
1.51 (These ideas do not \336t in well with the notion of a \322physical symbol system\323 because most of the) 50.88 633.23 P
2.12 (representing structures that are of interest turn out not to be physical: they are not detectable or) 50.88 619.23 P
0.06 (measurable by physical instruments, and they do not obey the laws of physics neither do they disobey) 50.88 605.23 P
2.74 (them) 50.88 591.23 P
1 14 Q
3.2 (\321) 80.61 591.23 P
1 12 Q
2.74 ( the laws are irrelevant to them. The use of position of a screw to represent required) 94.6 591.23 P
0.59 (temperature in a thermostat, or the state of a rotary \322governor\323 on a steam engine to represent speed) 50.88 577.23 P
0.67 (would be clear examples of physical symbols. By contrast many of the representing structures in AI) 50.88 563.23 P
2.71 (systems are not physical, but are components of what computer scientists normally describe as) 50.88 549.23 P
0.53 (\322virtual machines\323 or \322abstract machines\323 in the sense in which a programming language such as C,) 50.88 535.23 P
0.59 (Lisp or Prolog de\336nes a virtual machine with a speci\336c ontology \050permitted abstract data-structures\051) 50.88 521.23 P
(and behavioural capabilities \050i.e. operations on those data-structures\051.) 50.88 507.23 T
0.3 0.08 (T) 72.14 487.23 B
0.3 0.08 (ypical software data-structures such as lists, networks, or arrays are not physical objects: their) 78.7 487.23 B
-0.26 (laws of behaviour are not those of physical objects: for instance it is commonplace for list) 50.88 473.23 P
4 F
-0.26 (A) 481.08 473.23 P
1 F
-0.26 ( to contain) 489.74 473.23 P
4 F
0.3 0.01 (B) 50.88 459.23 B
1 F
0.3 0.01 ( as an element whilst) 58.89 459.23 B
4 F
0.3 0.01 (B) 164.18 459.23 B
1 F
0.3 0.01 ( contains) 172.19 459.23 B
4 F
0.3 0.01 (A) 218.85 459.23 B
1 F
0.3 0.01 ( as an element, whereas such mutual containment is not possible) 227.51 459.23 B
0.3 0.09 (for physical objects. If) 50.88 445.23 B
4 F
0.3 0.09 (cons) 165.13 445.23 B
1 F
0.3 0.09 ( is the standard list constructor that creates a new list given an arbitrary) 188.13 445.23 B
0.3 0.24 (item) 50.88 431.23 B
4 F
0.3 0.24 (X) 76.7 431.23 B
1 F
0.3 0.24 ( and an existing list) 85.6 431.23 B
4 F
0.3 0.24 (L) 188.67 431.23 B
1 F
0.3 0.24 (, and) 196.91 431.23 B
4 F
0.3 0.24 (head) 225.26 431.23 B
1 F
0.3 0.24 ( and) 250.88 431.23 B
4 F
0.3 0.24 (tail) 276 431.23 B
1 F
0.3 0.24 ( are the standard functions for accessing the f) 293.62 431.23 B
0.3 0.24 (irst) 523.8 431.23 B
0.3 0.34 (element of a list and the rest of the list, then examples of such non-physical laws would be the) 50.88 417.23 B
(following, which are not laws of physics, but much more like laws of mathematics:) 50.88 403.23 T
4 F
(head\050cons\050X, L\051\051 = X) 241.65 381.23 T
(tail\050cons\050X, L\051\051 = L) 245.98 358.23 T
1 F
0.3 0.19 (Computer programs typically manipulate states in) 72.14 337.23 B
3 F
0.3 0.19 (virtual) 325.53 337.23 B
1 F
0.3 0.19 ( machines, e.g. machines containing) 358.81 337.23 B
0.3 0.13 (numbers, strings, arrays, lists, tables, procedures, etc. rather than states in a physical machine, like) 50.88 323.23 B
0.3 0.3 (voltage, current or physical location. Even machine code instructions manipulate bit patterns in) 50.88 309.23 B
0.3 0.19 (abstract address spaces, rather than physical objects. Of course, the virtual machine processes are) 50.88 295.23 B
0.3 0.01 (implemented using physical processes. However) 50.88 281.23 B
0.3 0.01 (, there need not be any one-to-one mapping between) 285.52 281.23 B
0.13 (virtual machine states or structures and those of the underlying physical machine. For example a very) 50.88 267.23 P
0.3 0.19 (lar) 50.88 253.23 B
0.3 0.19 (ge sparse array in a computer can contain far more cells than there are atoms in the underlying) 63.9 253.23 B
0.3 0.16 (physical machine, or even in the universe. \050Sparse arrays use a technique in which array locations) 50.88 239.23 B
0.3 0.06 (containing a \322default\323 value are not explicitly recorded, whereas those with a non-default value are.) 50.88 225.23 B
0.14 (This can save a lot of space if the vast majority of locations contain the default value. Whether a very) 50.88 211.23 P
0.07 (lar) 50.88 197.23 P
0.07 (ge array uses this technique or explicit storage of the contents of all locations may be impossible to) 63.31 197.23 P
0.3 0.33 (tell simply by accessing the array: the two implementation techniques can provide functionally) 50.88 183.23 B
0.3 0.3 (equivalent abstract data-structures. However) 50.88 169.23 B
0.3 0.3 (, if the array contains more locations than could be) 278.49 169.23 B
0.3 0.22 (accommodated in the physical universe it must use some mechanism other than explicit storage.\051) 50.88 155.23 B
0.3 0.03 (Similarly) 50.88 141.23 B
0.3 0.03 (, a logic-based database containing a theorem prover could contain inf) 95.01 141.23 B
0.3 0.03 (initely many items of) 436.1 141.23 B
(information, without using in\336nitely many physical components!) 50.88 127.23 T
0.19 (V) 72.14 107.23 P
0.19 (irtual or abstract machines have many properties that make them suitable for building behaving) 80.08 107.23 P
0.3 0.22 (systems that need to be able to take in and process complex and rapidly changing information. A) 50.88 93.23 B
FMENDPAGE
%%EndPage: "2" 4
%%Page: "3" 4
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(3) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.3 0.32 (network in an abstract machine, containing many nodes with many links, can be constructed or) 50.88 774.23 B
-0.16 (reor) 50.88 760.23 P
-0.16 (ganised far more rapidly than any physical network of comparable complexity) 69.97 760.23 P
-0.16 (, which is one of the) 443.36 760.23 P
0.04 (reasons why virtual machines are so important for intelligent systems. For example, by using) 50.88 746.23 P
3 F
0.04 (pointers) 500.54 746.23 P
1 F
-0.27 (to complex structures it is very easy to have the ef) 50.88 732.23 P
-0.27 (fect of many copies of the same structure in dif) 288.41 732.23 P
-0.27 (ferent) 511.88 732.23 P
-0.22 (locations, and then to produce the ef) 50.88 718.23 P
-0.22 (fect of changing all copies simply by changing the single structure) 223.57 718.23 P
(they all point to.) 50.88 704.23 T
0.3 0.22 (Nevertheless all the virtual machines to be found in actual behaving systems are \050ultimately\051) 72.14 684.23 B
3 F
0.3 0.33 (implemented) 50.88 670.23 B
1 F
0.3 0.33 ( in physical machines, be they brains or computers. A virtual machine need not be) 115.76 670.23 B
-0.02 (directly implemented in a physical machine: it could use a simpler abstract machine. For example, the) 50.88 656.23 P
0.3 0.2 (virtual machine corresponding to a high level programming language is typically implemented in) 50.88 642.23 B
0.25 (terms of the lower level abstract machine corresponding to the \322instruction set\323 of the host computer) 50.88 628.23 P
0.25 (,) 536.85 628.23 P
0.1 (such as a V) 50.88 614.23 P
0.1 (AX, or HP-P) 104.59 614.23 P
0.1 (A, or SP) 165.99 614.23 P
0.1 (ARC, all of which are themselves abstract machines capable of being) 206.07 614.23 P
(implemented in physical machines in dif) 50.88 600.23 T
(ferent ways.) 245.54 600.23 T
0.3 0.28 (This relationship of \322implementation\323 that can hold between machines at dif) 72.14 580.23 B
0.3 0.28 (ferent levels of) 463.1 580.23 B
0.3 0.19 (abstraction, and between an abstract and a physical machine, could turn out to be one of the most) 50.88 566.23 B
-0.11 (important contributions of computer science to the study of mind: far more important than the concept) 50.88 552.23 P
0.3 0.02 (of an algorithm, for example, even though the latter has received more attention \050e.g. Penrose 1989\051.) 50.88 538.23 B
-0.08 (For now I shall assume that the reader is familiar with the typical hierarchies of implementation levels) 50.88 524.23 P
(that can be found in computer) 50.88 510.23 T
(-based systems.) 194.2 510.23 T
0.17 (Even if biological systems are very dif) 72.14 490.23 P
0.17 (ferent and far more complex and subtle, it can be useful to) 258.16 490.23 P
0.3 0.11 (think of implementation hierarchies in computers as providing a relatively simple, well-understood) 50.88 476.23 B
0.3 0.09 (illustration of the concept of a behaving system that needs to be understood as having its behaviour) 50.88 462.23 B
0.3 0.17 (controlled in part by an abstract machine. This is especially true if we consider internal as well as) 50.88 448.23 B
0.3 0.05 (external behaviour: for the internal behaviour that is of interest in understanding problem-solving or) 50.88 434.23 B
0.3 0.39 (goal-directed actions in humans is generally the behaviour of abstract, not physical machines.) 50.88 420.23 B
(Processes of inference, reasoning, interpretation are not physical processes.) 50.88 406.23 T
-0.15 (I am not saying that it is) 72.14 386.23 P
3 F
-0.15 (only) 190.01 386.23 P
1 F
-0.15 ( these complex abstract machines that have information-bearing sub-) 210.66 386.23 P
0.3 0.12 (states. Even a simple thermostat controlling a room heater has a sub-state that can be thought of as) 50.88 372.23 B
0.3 0.22 (representing the current ambient temperature and another sub-state corresponding to the required) 50.88 358.23 B
0.3 0.14 (temperature. These sub-states can vary independently of each other) 50.88 344.23 B
0.3 0.14 (, and they have dif) 384.49 344.23 B
0.3 0.14 (ferent causal) 476.98 344.23 B
0.3 0.01 (roles in the system, but they are part of an integrated system that behaves as a whole. Thus, although) 50.88 330.23 B
-0.09 (the control systems that are of most interest in the study of mind are far more complex than this, it can) 50.88 316.23 P
0.22 (still be illuminating for a survey of control systems to include the simpler systems, including systems) 50.88 302.23 P
0.3 0.15 (that can easily be represented by a f) 50.88 288.23 B
0.3 0.15 (ixed set of physical measurements and a f) 229.33 288.23 B
0.3 0.15 (ixed set of equations) 437.05 288.23 B
0.3 0.05 (linking them. Limiting cases can be part of a concept even though most instances are very dif) 50.88 274.23 B
0.3 0.05 (ferent,) 508.56 274.23 B
0.3 0.05 (just as a circle is an interesting limiting case of the concept of an ellipse, despite not having a major) 50.88 260.23 B
0.3 0.26 (and a minor axis, and zero is a limiting case of the concept of a number that can be the result of) 50.88 246.23 B
(counting or measuring.) 50.88 232.23 T
323.04 199.42 50.88 199.42 2 L
V
1.02 H
0 Z
N
2 14 Q
(\0503\051 Causal relations in abstract machines) 50.88 200.9 T
1 12 Q
1.19 (A full discussion of the causal relations between processes in virtual and physical machines would) 50.88 181.23 P
-0.26 (require a lengthy paper on its own: I am going to assume the common sense view that computer events) 50.88 167.23 P
2.29 (best described in software or program terms can cause events both in the virtual machines they) 50.88 153.23 P
2.47 (manipulate \050e.g. adding a new subnet to a network, or re-formatting a page of text\051 and in the) 50.88 139.23 P
1.23 (underlying physical processes \050e.g. getting many transistors to change their state\051. More obviously) 50.88 125.23 P
1.23 (,) 536.85 125.23 P
-0.29 (physical events, such as arrival of an electrical signal at an interface can cause software events, such as) 50.88 111.23 P
FMENDPAGE
%%EndPage: "3" 5
%%Page: "4" 5
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(4) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0 (transfer of control to a new sub-program or creation of a new data-structure such as a character string.) 50.88 774.23 P
(In other words, causal in\337uences can cross implementation boundaries.) 50.88 760.23 T
0.3 0.17 (Philosophical objections to this assumption tend to use ar) 72.14 740.23 B
0.3 0.17 (guments claiming that only physical) 359.12 740.23 B
0.3 0.12 (events can be real causes or ef) 50.88 726.23 B
0.3 0.12 (fects. There are many problems with such objections, partly because) 201.3 726.23 B
0.3 0.15 (they unjustif) 50.88 712.23 B
0.3 0.15 (iably assume that there is a unique and well-def) 112.76 712.23 B
0.3 0.15 (ined) 350.15 712.23 B
3 F
0.3 0.15 (physical) 374.85 712.23 B
1 F
0.3 0.15 ( layer of reality) 416.02 712.23 B
0.3 0.15 (, and also) 492.28 712.23 B
0.3 0 (because they imply rejection of most of our patently useful ordinary ways of talking about causation,) 50.88 698.23 B
0.3 0.09 (e.g. when we describe social events as causing political events, or bad news as causing distress and) 50.88 684.23 B
(distress as causing a \337ood of tears.) 50.88 670.23 T
0.11 (The claim then is that sub-states of a control system contain information, and can be described as) 72.14 650.23 P
0.3 0.02 (being \322representations\323 in a new technical theory-based sense of the word that is intended ultimately) 50.88 636.23 B
0.3 0.32 (to replace and explicate the ordinary notion of a representation. More precisely: each state of a) 50.88 622.23 B
-0.23 (component of a control system is a representation-instance, and the range of states that are possible for) 50.88 608.23 P
0.3 0.12 (that component def) 50.88 594.23 B
0.3 0.12 (ines a representational system, or what Donald Peterson \0501994\051 calls a \322form of) 145.57 594.23 B
0.3 0.35 (representation\323, more commonly referred to as a \322notation\323 or \322formalism\323. A thermostat uses) 50.88 580.23 B
0.3 0.01 (particularly simple formalisms, because the sub-states of a thermostat that are relevant to its function) 50.88 566.23 B
0.3 0.42 (vary only in particularly simple ways, whereas some other control systems, such as computer) 50.88 552.23 B
(operating systems, use far more complex formalisms in their information-bearing sub-states.) 50.88 538.23 T
211.86 505.42 50.88 505.42 2 L
V
1.02 H
0 Z
N
2 14 Q
(\0504\051 Preliminary overview) 50.88 506.9 T
1 12 Q
0.78 (My ideas on all this are still in a half-baked form, and this paper does not so much report results as) 50.88 487.23 P
0.18 (invite contributions to an ongoing investigation. The basic idea that representations are control states,) 50.88 473.23 P
(or aspects of control states. can be elaborated as follows.) 50.88 459.23 T
4 14 Q
(\245) 50.88 442.23 T
1 12 Q
0.3 0.07 (Many complex behaving systems have diverse independently variable interacting sub-states, with) 62.22 442.23 B
-0.06 (dif) 62.22 428.23 P
-0.06 (ferent functions \050dif) 75.32 428.23 P
-0.06 (ferent causal roles\051. Sometimes dif) 170.91 428.23 P
-0.06 (ferent sub-states are in physically separate) 337.33 428.23 P
0.3 0.25 (sub-mechanisms, and sometimes they are superimposed in one mechanism \050like superimposed) 62.22 414.23 B
(wave-forms\051.) 62.22 400.23 T
4 14 Q
(\245) 50.88 384.23 T
1 12 Q
0.3 0.08 (The functional dif) 62.22 384.23 B
0.3 0.08 (ferentiation between dif) 150.55 384.23 B
0.3 0.08 (ferent sub-states is typically far greater for systems that) 267.31 384.23 B
(we would regard as intelligent than for non-intelligent systems \050examples of both are given below\051.) 62.22 370.23 T
4 14 Q
(\245) 50.88 354.23 T
1 12 Q
0.3 0.44 (Some of the sub-states \322contain\323 information, used for dif) 62.22 354.23 B
0.3 0.44 (ferent purposes: sensory buf) 368.38 354.23 B
0.3 0.44 (fers,) 517.09 354.23 B
0.3 0.64 (perceptual summaries, long term memories, plans, feedback control signals. I call these) 62.22 340.23 B
0.3 0.04 (\322information\323 states. At this stage there is no presumption about) 62.22 326.23 B
3 F
0.3 0.04 (how) 379.02 326.23 B
1 F
0.3 0.04 ( the information is embodied) 399.11 326.23 B
0.3 0.44 (in the state: dif) 62.22 312.23 B
0.3 0.44 (ferent embodiments are to be found in dif) 142.01 312.23 B
0.3 0.44 (ferent sorts of plants, animals and) 361.49 312.23 B
0.3 0.03 (machines. In a simple thermostat, information about temperature might be embodied in the degree) 62.22 298.23 B
0.3 0.28 (of curvature of a bi-metallic strip, whereas quite dif) 62.22 284.23 B
0.3 0.28 (ferent embodiments can be found in more) 326.93 284.23 B
0.3 0.2 (sophisticated controllers and in biological systems, including chemical states, neural states and) 62.22 270.23 B
(software states.) 62.22 256.23 T
4 14 Q
(\245) 50.88 240.23 T
1 12 Q
0.3 0.12 (The notion of \322information\323 used here will not be def) 62.22 240.23 B
0.3 0.12 (ined in advance. This is one of those many) 327.57 240.23 B
0.3 0.14 (concepts that can only be made completely clear in the context of a fully developed explanatory) 62.22 226.23 B
0.3 0.22 (theory \050in deep science def) 62.22 212.23 B
0.3 0.22 (initions come last, not f) 197.84 212.23 B
0.3 0.22 (irst\051. While we lack a comprehensive theory) 316.76 212.23 B
0 (characterising control architectures, the kinds of sub-states that can occur within such architectures,) 62.22 198.23 P
0.3 0.16 (and the kinds of causal roles such substates can fulf) 62.22 184.23 B
0.3 0.16 (il, we cannot yet give a precise def) 320.14 184.23 B
0.3 0.16 (inition of) 493.81 184.23 B
0.3 0.06 (\322information-bearing sub-state\323. However) 62.22 170.23 B
0.3 0.06 (, it is easy to give examples of machines and or) 267.61 170.23 B
0.3 0.06 (ganisms) 500.17 170.23 B
0.24 (with sub-states that contain information about the current environment, information about previous) 62.22 156.23 P
0.3 0.17 (events, information about states to be achieved, information about how to achieve or prevent or) 62.22 142.23 B
-0.07 (maintain states of af) 62.22 128.23 P
-0.07 (fairs, and information about what to do next. In each case the sub-state contains) 158.72 128.23 P
0.3 0.17 (information) 62.22 114.23 B
3 F
0.3 0.17 ( about) 120.66 114.23 B
1 F
0.3 0.17 ( something, information that can inf) 152.27 114.23 B
0.3 0.17 (luence processes within the same or other) 331.61 114.23 B
(sub-components of the architecture.) 62.22 100.23 T
FMENDPAGE
%%EndPage: "4" 6
%%Page: "5" 6
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(5) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
4 14 Q
0 X
(\245) 50.88 774.23 T
1 12 Q
0.3 0.15 (A full theory of representations would account for the dif) 62.22 774.23 B
0.3 0.15 (ferent kinds of information that can be) 347.97 774.23 B
0.3 0.07 (contained in control states, and the dif) 62.22 760.23 B
0.3 0.07 (ferent roles they can play) 249.47 760.23 B
0.3 0.07 (. This can be broken down further) 372.94 760.23 B
0.3 0.03 (into: the dif) 62.22 746.23 B
0.3 0.03 (ferent forms in which the information can be expressed \050syntax\051, the dif) 119 746.23 B
0.3 0.03 (ferent kinds of) 468.85 746.23 B
-0.3 (things that can be referred to or expressed \050semantics\051, the dif) 62.22 732.23 P
-0.3 (ferent operations that can be applied to) 355.12 732.23 P
-0.08 (the information \050generalised inference\051 and the dif) 62.22 718.23 P
-0.08 (ferent uses to which the information states can be) 303.33 718.23 P
(put \050pragmatics\051. \050These notions are all elaborated below) 62.22 704.23 T
(.\051) 335.22 704.23 T
4 14 Q
(\245) 50.88 688.23 T
1 12 Q
0.3 0.14 (Which concepts are applicable to a particular system will depend on its architecture: in a simple) 62.22 688.23 B
0.16 (architecture such as that of a thermostat a sub-state \050e.g curvature of a bi-metallic strip\051 can be said) 62.22 674.23 P
0.3 0.45 (to contain information about the ambient temperature, but there is no dif) 62.22 660.23 B
0.3 0.45 (ferentiation of the) 445.23 660.23 B
-0.05 (environment into things with properties and relationships, and in particular no representation of any) 62.22 646.23 P
3 F
-0.19 (thing) 62.22 632.23 P
1 F
-0.19 ( as having a temperature. Moreover) 86.87 632.23 P
-0.19 (, the use to which the information can be put is very limited) 255.97 632.23 P
0.3 0.17 (and unvarying. By contrast, in a robot with a visual system whose architecture is similar to that) 62.22 618.23 B
0.3 0.06 (sketched below) 62.22 604.23 B
0.3 0.06 (, there is information about the environment at various levels of abstraction, based) 136.86 604.23 B
0.3 0.15 (on considerable amounts of interpretation in the light of prior information, and several potential) 62.22 590.23 B
0.02 (uses to which the same information can be put. For instance, the same item of information could on) 62.22 576.23 P
-0.26 (some occasions be used to check whether a goal has been achieved, to trigger a new goal, to con\336rm) 62.22 562.23 P
0.3 0.12 (a hypothesis, to interpret new sensory data, or to derive new factual information via a process of) 62.22 548.23 B
(reasoning.) 62.22 534.23 T
4 14 Q
(\245) 50.88 518.23 T
1 12 Q
0.3 0.06 (W) 62.22 518.23 B
0.3 0.06 (e must distinguish the representational needs of a designer concerned with building, modifying,) 72.63 518.23 B
0.3 0.17 (maintaining or describing a system, and the needs of the system itself. Computer scientists, and) 62.22 504.23 B
0.3 0.09 (some AI theorists concentrate more on notations for use by researchers and designers. I am more) 62.22 490.23 B
0.3 0.16 (concerned with run-time representation in control states of working systems, not design-time or) 62.22 476.23 B
0.1 (compile-time formalisms, nor external description. \050Of course, where the working system is itself a) 62.22 462.23 P
0.3 0.43 (designer or engineer) 62.22 448.23 B
0.3 0.43 (, or attempts to understand itself, both sorts of representation will play) 168.75 448.23 B
(important roles.\051) 62.22 434.23 T
4 14 Q
(\245) 50.88 418.23 T
1 12 Q
0.3 0 (I believe that when we fully understand design principles for sophisticated control systems, a great) 62.22 418.23 B
0.3 0.32 (many philosophical problems about the nature of mind, language, representation, etc. will be) 62.22 404.23 B
0.12 (transformed: though not necessarily solved, for in many cases the transformation will consist in the) 62.22 390.23 P
-0.07 (demonstration that the original questions were based on confused or false assumptions, made use of) 62.22 376.23 P
0.3 0.1 (muddled concepts, etc. The transformed questions will then be capable of being answered on the) 62.22 362.23 B
0.3 0.09 (basis of either theoretical \050e.g. mathematical or logical\051 analysis or empirical investigation. \050This) 62.22 348.23 B
0.29 (does not imply that the original questions were of no value: muddled ideas can drive advances that) 62.22 334.23 P
(lead to their own replacement.\051) 62.22 320.23 T
4 14 Q
(\245) 50.88 304.23 T
1 12 Q
0.3 0.01 (In particular) 62.22 304.23 B
0.3 0.01 (, I expect that we shall discover that many current disputes about the relative merits of) 121.07 304.23 B
0.3 0 (\322symbolic\323 and \322connectionist\323 models are as muddled as old investigations attempting to f) 62.22 290.23 B
0.3 0 (ind out) 505.88 290.23 B
(which combinations of earth, air) 62.22 276.23 T
(, \336re and water constituted dif) 217.29 276.23 T
(ferent known substances.) 360.97 276.23 T
4 14 Q
(\245) 50.88 260.23 T
1 12 Q
0.3 0.32 (When we have a good overview of design options and the dif) 62.22 260.23 B
0.3 0.32 (ferent functional roles that sub-) 377.9 260.23 B
0.3 0.08 (mechanisms and control states can have, this will provide a new basis for a conceptual taxonomy) 62.22 246.23 B
0.27 (for talking about types of representations, which will be much deeper and richer than current naive) 62.22 232.23 P
(distinctions such as declarative/procedural, verbal/pictorial, symbolic/connectionist.) 62.22 218.23 T
0.3 0.37 (From this standpoint the questions that lead to deeper understanding are) 72.14 198.23 B
3 F
0.3 0.37 (design) 450.65 198.23 B
1 F
0.3 0.37 ( questions:) 484.2 198.23 B
0.3 0 (questions about how to design intelligent, sentient, autonomous agents, with their own desires, goals,) 50.88 184.23 B
-0.07 (and so on. Design questions need to be addressed in the context of) 50.88 170.23 P
3 F
-0.07 (r) 371.07 170.23 P
-0.07 (equir) 375.29 170.23 P
-0.07 (ements) 400.16 170.23 P
1 F
-0.07 ( \050Sloman 1993c\051. And) 433.47 170.23 P
-0.03 (at present we understand little about the requirements driving the design of intelligent agents, whether) 50.88 156.23 P
(in the laboratory) 50.88 142.23 T
(, or in biological evolutionary processes.) 129.37 142.23 T
FMENDPAGE
%%EndPage: "5" 7
%%Page: "6" 7
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(6) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
81.97 771.42 50.88 771.42 2 L
0 X
V
1.02 H
0 Z
N
2 14 Q
(\0505\051 A) 50.88 772.9 T
264.96 771.42 81.45 771.42 2 L
V
N
(void simplistic dichotomies) 81.45 772.9 T
1 12 Q
1.06 (People who study representations tend to examine only a small range of cases, and as a result they) 50.88 753.23 P
3.98 (produce shallow or simplistic distinctions: verbal/visual, analog/digital, symbolic/sub-symbolic,) 50.88 739.23 P
1.13 (procedural/declarative, explicit/implicit etc. An examination of the history of mathematics, science,) 50.88 725.23 P
0.67 (engineering, and culture reveals a much greater diversity of types, including many ad-hoc notations,) 50.88 711.23 P
-0.28 (such as the standard arabic notation for numbers in which concatenation representations multiplication) 50.88 697.23 P
2.45 (by 10 and addition, a type of rule that has no application outside of arithmetic. From a design) 50.88 683.23 P
1.41 (standpoint we need to understand why diverse notations are to be expected: the answer lies in the) 50.88 669.23 P
1.99 (variety of functions ful\336lled by dif) 50.88 655.23 P
1.99 (ferent information-bearing sub-states in a sophisticated control) 226.82 655.23 P
(system \050see also Sloman 1985a\051.) 50.88 641.23 T
0.3 0.07 (There are many dif) 72.14 621.23 B
0.3 0.07 (ferent trade-of) 165.71 621.23 B
0.3 0.07 (fs in designing and selecting types of information-rich control) 235.81 621.23 B
0.3 0.03 (states \050i.e. representations\051 for \322run-time\323 use, and these lead to a wide variety of design options, far) 50.88 607.23 B
0.3 0.12 (more than the common categorisations ref) 50.88 593.23 B
0.3 0.12 (lect. For example, so-called verbal or symbolic notations) 258.5 593.23 B
0.3 0.13 (often use \322pictorial\323 relationships, such as ordering, to represent similarly structured relationships.) 50.88 579.23 B
0.3 0.04 (\322She shot him and drove away\323 usually implies a dif) 50.88 565.23 B
0.3 0.04 (ferent temporal ordering from \322She drove away) 308.01 565.23 B
-0.09 (and shot him\323, and to that extent such verbal forms include a pictorial or iconic function, undermining) 50.88 551.23 P
(the common distinction between the two.) 50.88 537.23 T
0.3 0.17 (Moreover) 72.14 517.23 B
0.3 0.17 (, many programming languages that are normally described as \322procedural\323 make it) 120.35 517.23 B
0.3 0.03 (possible to use data-structures as information stores that are interpreted, just like \322facts\323 in allegedly) 50.88 503.23 B
-0.17 (declarative languages. At the same time, supposedly declarative languages like Prolog usually provide) 50.88 489.23 P
0.3 0.28 (means for expressing not only what is to be done but also in what order various steps should be) 50.88 475.23 B
0.3 0.05 (followed, and to that extent it is as procedural as any other language, except that the mappings from) 50.88 461.23 B
0.3 0.04 (syntactic ordering to process ordering is complicated by backtracking. The terminological confusion) 50.88 447.23 B
0.3 0.12 (is not helped by those who apply the word \322declarative\323 to functional programming languages like) 50.88 433.23 B
(Scheme.) 50.88 419.23 T
0.3 0.15 (A f) 72.14 399.23 B
0.3 0.15 (irst step towards getting a deeper understanding of the role of representations in intelligent) 87.88 399.23 B
0.3 0.15 (\050and non-intelligent\051 behaving systems is to understand the requirements they may need to satisfy) 50.88 385.23 B
0.3 0.15 (.) 536.85 385.23 B
0.3 0.32 (Some requirements concern their syntactic richness, others concern their manipulability) 50.88 371.23 B
0.3 0.32 (, others) 501.98 371.23 B
0.3 0.08 (concern speed with which they can be created, changed or used, and so on. T) 50.88 357.23 B
0.3 0.08 (o illustrate the variety) 429.82 357.23 B
0.3 0.08 (,) 536.85 357.23 B
-0.29 (below are some examples of systems that would not normally be described as intelligent and some that) 50.88 343.23 P
(would be. \050I am not suggesting that this is a clear distinction based on generally agreed criteria.\051) 50.88 329.23 T
407 296.42 50.88 296.42 2 L
V
N
2 14 Q
(\0506\051 Examples of information states in control systems) 50.88 297.9 T
3 12 Q
(The following would normally be described as components of non-intelligent systems:-) 50.88 272.23 T
1 F
(1.) 50.88 252.23 T
(The state of a temperature sensor in a thermostat.) 73.55 252.23 T
(2.) 50.88 236.23 T
(The state of the temperature control in a thermostat.) 76.39 236.23 T
(3.) 50.88 220.23 T
(The state of the accelerator) 76.39 220.23 T
(, choke, steering wheel or brakes, in a car) 205.79 220.23 T
(.) 403.95 220.23 T
(4.) 50.88 204.23 T
-0.32 (The /etc/rc and /etc/rc.local start-up scripts and many other system con\336guration \336les in a typical) 76.39 204.23 P
(Unix operating system.) 76.39 190.23 T
(5.) 50.88 174.23 T
(The internal, constantly changing, tables and processes in a running operating system.) 76.39 174.23 T
(6.) 50.88 158.23 T
(Information about the state of network links in a networked communication system.) 76.39 158.23 T
(7.) 50.88 142.23 T
(A staf) 76.39 142.23 T
(f payroll database.) 105.15 142.23 T
3 F
(The following would normally be described as involving intelligence:-) 50.88 116.23 T
1 F
(1.) 50.88 96.23 T
(Current visual \050or other\051 percepts in a person or other animal.) 73.55 96.23 T
FMENDPAGE
%%EndPage: "6" 8
%%Page: "7" 8
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(7) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
(2.) 50.88 774.23 T
(Particular memories of previous situations, things, etc. in a person or robot.) 76.39 774.23 T
(3.) 50.88 758.23 T
(General beliefs \050about the environment, oneself, etc.\051, e.g. that unsupported objects fall, or that) 76.39 758.23 T
(people who smoke tend to smell unpleasant.) 76.39 744.23 T
(4.) 50.88 728.23 T
(Motives, desires, preferences, attitudes \050af) 76.39 728.23 T
(fective states at various levels of abstraction, and) 279.01 728.23 T
(dif) 76.39 714.23 T
(ferent degrees of generality and persistence\051.) 89.5 714.23 T
(5.) 50.88 698.23 T
(Personality \050a collection of very high level dispositions\051.) 76.39 698.23 T
(6.) 50.88 682.23 T
(Intentions, plans \050long term, short term, current, future\051.) 76.39 682.23 T
(7.) 50.88 666.23 T
(Information underlying motor skills \050e.g. violin playing\051.) 76.39 666.23 T
(8.) 50.88 650.23 T
(Information underlying linguistic skills \050phonological, syntactic, lexical information, etc.\051.) 76.39 650.23 T
(9.) 50.88 634.23 T
(Current information in various motor) 76.39 634.23 T
(-control feedback systems, e.g. posture control, hand-) 254.71 634.23 T
(movements.) 76.39 620.23 T
(10.) 50.88 604.23 T
-0.08 (Internal self-monitoring states \050internal records of bodily states, mental states, dispositions, etc.\051.) 76.39 604.23 P
0.44 (These are merely illustrative of the variety of types of information states that we need to understand.) 50.88 584.23 P
0.13 (Some are short term, some enduring long term states, some mostly passive \050changed by other things\051,) 50.88 570.23 P
2.03 (some active \050i.e.) 50.88 556.23 P
3 F
2.03 (sour) 138.56 556.23 P
2.03 (ces) 159.45 556.23 P
1 F
2.03 ( of change\051, others mediators or modi\336ers of processes that they do not) 174.76 556.23 P
3.21 (themselves initiate, some directly manipulable by other mechanisms and others only indirectly) 50.88 542.23 P
2.76 (modi\336able \050e.g. by training\051, some consciously accessible, some not. In human beings the vast) 50.88 528.23 P
(majority of information states are not consciously accessible.) 50.88 514.23 T
0.3 0.1 (I am not assuming that there is a sharp and well-understood distinction between intelligent and) 72.14 494.23 B
0.3 0.16 (non-intelligent systems, and I shall say nothing about how that distinction might be clarif) 50.88 480.23 B
0.3 0.16 (ied, for I) 496.21 480.23 B
0.3 0.38 (think it will eventually be replaced by a whole family of dif) 50.88 466.23 B
0.3 0.38 (ferent distinctions concerned with) 363.44 466.23 B
0.3 0.06 (presence of absence of dif) 50.88 452.23 B
0.3 0.06 (ferent collections of capabilities: taxonomies are usually more useful than) 178.63 452.23 B
(dichotomies.) 50.88 438.23 T
295 405.42 50.88 405.42 2 L
V
1.02 H
0 Z
N
2 14 Q
(\0507\051 What is a form of representation?) 50.88 406.9 T
1 12 Q
-0.08 (Information of all those kinds can be embodied in states of complex systems using dif) 50.88 387.23 P
-0.08 (ferent notations,) 461.98 387.23 P
2.5 (or \322forms of representation\323. Confusion can follow from the common practice of referring to a) 50.88 373.23 P
2.95 (notation or formalism, such as predicate calculus, or algebra, as a \322representation\323, as well as) 50.88 359.23 P
1.54 (individual instances. I shall try to avoid this ambiguity by restricting the word \322representation\323 to) 50.88 345.23 P
3 F
0.33 (instances) 50.88 331.23 P
1 F
0.33 (, and will use the expression \322formalism\323, or \322notation\323 or \322form of representation\323 to refer) 95.52 331.23 P
1.52 (to the general forms that individual representations are instances of. In this terminology) 50.88 317.23 P
1.52 (, predicate) 488.38 317.23 P
1.5 (calculus would be a formalism or notation but not a representation, whereas a particular predicate) 50.88 303.23 P
0.67 (calculus expression would be a representation. The word \322symbol\323 is also used to refer to particular) 50.88 289.23 P
0.05 (instances, usually those that lack meaningful internal structure, while \322symbolism\323 is sometimes used) 50.88 275.23 P
1.73 (as a synonym for \322notation\323 or \322formalism\323. The word \322language\323 is fairly close to what is here) 50.88 261.23 P
0.9 (described as a notation or formalism, though it is often thought of as restricted to external forms of) 50.88 247.23 P
0.61 (communication between complete intelligent agents, whereas what I am talking about could be used) 50.88 233.23 P
0.82 (internally for communication between parts of a behaving system, and not necessarily an intelligent) 50.88 219.23 P
(system.) 50.88 205.23 T
0.3 0.26 (What I am here calling a \322formalism\323, \322language\323, or \322notation\323 will normally have syntax,) 72.14 185.23 B
0.3 0.15 (semantics, pragmatics, and inference rules which determine the consequences of transforming one) 50.88 171.23 B
-0.12 (state into another) 50.88 157.23 P
-0.12 (. Not all notations need have all these features: e.g. a notation used entirely in control) 132.59 157.23 P
0.3 0.31 (signals need not include any inference capabilities. In some simple cases there may be no clear) 50.88 143.23 B
(distinction between pragmatics and semantics.) 50.88 129.23 T
0.3 0.15 (W) 72.14 109.23 B
0.3 0.15 (e must take care to interpret these ideas in a suf) 82.65 109.23 B
0.3 0.15 (f) 320.07 109.23 B
0.3 0.15 (iciently general way) 323.55 109.23 B
0.3 0.15 (. Features of \322external\323) 423.61 109.23 B
0.3 0.33 (notations can mislead us into adopting over) 50.88 95.23 B
0.3 0.33 (-simplif) 275.26 95.23 B
0.3 0.33 (ied theories of representation. For example,) 315.23 95.23 B
FMENDPAGE
%%EndPage: "7" 9
%%Page: "8" 9
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(8) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.3 0.08 (notations in a computer or brain may use a topology that cannot be directly represented on paper or) 50.88 774.23 B
0.3 0.27 (any other two dimensional surface. W) 50.88 760.23 B
0.3 0.27 (e can aim for full generality by considering how to design) 243.66 760.23 B
0.3 0.08 (behaving systems. From this \322design\323 perspective, a notation can be thought of as:) 50.88 746.23 B
539.93 744.93 460.74 744.93 2 L
V
0.59 H
0 Z
N
0.3 0.08 (a set of possible) 460.74 746.23 B
300.71 730.93 50.88 730.93 2 L
V
N
(information states that can occur in control systems.) 50.88 732.23 T
-0.26 (It is possible for dif) 72.14 712.23 P
-0.26 (ferent notations to be used by dif) 164.82 712.23 P
-0.26 (ferent sub-states of the same complex system:) 320.92 712.23 P
0.01 (I shall try to show that intelligent systems typically require many dif) 50.88 698.23 P
0.01 (ferent kinds of notations or forms) 378.91 698.23 P
(of representation, corresponding to dif) 50.88 684.23 T
(ferent functional roles played by dif) 234.86 684.23 T
(ferent sub-states.) 406.84 684.23 T
0.3 0.57 (The relationship between representation and underlying physical structure can be quite) 72.14 664.23 B
-0.1 (unobvious. Often a representation or symbol will be implemented using a perceivable physical pattern) 50.88 650.23 P
0.3 0.21 (or structure, but the physical object does not uniquely determine the representation or symbol. In) 50.88 636.23 B
0.3 0.16 (dif) 50.88 622.23 B
0.3 0.16 (ferent contexts the same physical pattern can be an instance of dif) 64.46 622.23 B
0.3 0.16 (ferent notations, like the joke) 393.64 622.23 B
(utterances that can be interpreted as saying entirely dif) 50.88 608.23 T
(ferent things in dif) 312.45 608.23 T
(ferent languages.) 401.18 608.23 T
0.3 0.24 (Below I shall try to show how the familiar concepts of grammar) 72.14 588.23 B
0.3 0.24 (, pragmatics, semantics, and) 397.25 588.23 B
0.3 0.1 (inference can be generalised to correspond to this general notion of a representation, though not all) 50.88 574.23 B
0.3 0.19 (examples will exhibit the full richness of these concepts. It is common in connection with natural) 50.88 560.23 B
0.22 (languages to contrast syntax \050grammar\051, semantics \050meaning\051 and pragmatics. There are considerable) 50.88 546.23 P
-0.12 (dif) 50.88 532.23 P
-0.12 (\336culties in making these notions very precise in their full generality) 63.99 532.23 P
-0.12 (. However) 386.47 532.23 P
-0.12 (, attempting precision) 435.82 532.23 P
0.3 0.12 (at this stage would be premature: we f) 50.88 518.23 B
0.3 0.12 (irst need a full theory of information-bearing substates and a) 239.73 518.23 B
0.03 (survey of their possible structures and uses. W) 50.88 504.23 P
0.03 (ithin such a theoretical framework it should be possible) 273.79 504.23 P
0.3 0.04 (to give more precise and systematic taxonomies and def) 50.88 490.23 B
0.3 0.04 (initions than we can of) 323.2 490.23 B
0.3 0.04 (fer now) 434.34 490.23 B
0.3 0.04 (. Nevertheless) 471.1 490.23 B
(we can already of) 50.88 476.23 T
(fer an approximate \336rst-draft analysis.) 135.57 476.23 T
410.09 443.42 50.88 443.42 2 L
V
1.02 H
N
2 14 Q
(\0508\051 The syntax \050or grammatical form\051 of a control state) 50.88 444.9 T
1 12 Q
-0.24 (A notation in this general sense corresponds to a \322set of possible sub-states\323 of a behaving system, just) 50.88 425.23 P
-0.24 (as a conventional grammar determines a set of possible sentences. In some cases, like the temperature-) 50.88 411.23 P
1.5 (representing sub-state of a thermostat, the set of possible states has a very simple structure, e.g. a) 50.88 397.23 P
0.15 (linear continuum. In other cases it is far more complex, like the set of possible states of the \322start-up\323) 50.88 383.23 P
0.53 (\336les of an operating system, the set of programs permitted by a programming language, or the set of) 50.88 369.23 P
0.86 (sentences in a natural language. In this context it is useful to generalise the notion of \322grammar\323 to) 50.88 355.23 P
1.25 (refer to the set of structures of such states. This is a natural generalisation of the normal linguistic) 50.88 341.23 P
2.21 (notion of grammar as determining structures of possible sentences, each of which has a speci\336c) 50.88 327.23 P
(grammatical form, or syntax.) 50.88 313.23 T
0.3 0.19 (One dif) 72.14 293.23 B
0.3 0.19 (ference between our standpoint and that of linguistics is that linguists tend to think of) 109.86 293.23 B
0.3 0.07 (sentences as static entities whereas inside a behaving system representational states need to change,) 50.88 279.23 B
0.13 (for instance during reasoning, learning, planning, perceiving, as well as during control processes with) 50.88 265.23 P
0.3 0.25 (feedback. So actual variability of representations is important, as opposed to the mere variety of) 50.88 251.23 B
0.3 0.14 (possibilities for static structures. The) 50.88 237.23 B
3 F
0.3 0.14 (kind) 238.11 237.23 B
1 F
0.3 0.14 ( of variability that is permitted depends on the grammar) 259.34 237.23 B
0.3 0.14 (.) 536.85 237.23 B
0.18 (\050For now we leave open how to de\336ne the boundary between structure and content of a state. It is not) 50.88 223.23 P
0.3 0.04 (very important for us whether grammar of a language incorporates all the individual lexical items or) 50.88 209.23 B
0.3 0.09 (only classes of such items. In the latter case, the actual words in a sentence would not be part of its) 50.88 195.23 B
0.15 (syntax, only their lexical classes would be. Similarly) 50.88 181.23 P
0.15 (, it is not very important, for present purposes, to) 303.99 181.23 P
0.3 0.02 (say whether two states of a thermostat corresponding to dif) 50.88 167.23 B
0.3 0.02 (ferent temperatures have dif) 337.94 167.23 B
0.3 0.02 (ferent syntax,) 474.03 167.23 B
(or the same syntax but dif) 50.88 153.23 T
(ferent content.\051) 174.91 153.23 T
0.3 0.09 (Although I use the words \322grammar\323 and \322syntax\323 almost interchangeably I shall endeavour to) 72.14 133.23 B
-0.09 (restrict \322grammar\323 to refer to the characterisation of a whole class of possible structures, and \322syntax\323) 50.88 119.23 P
0.06 (to refer to the structure of an instance. In this sense a) 50.88 105.23 P
3 F
0.06 (language) 307.76 105.23 P
1 F
0.06 ( has a grammar) 352.4 105.23 P
0.06 (, whereas a) 425.71 105.23 P
3 F
0.06 (sentence) 482.51 105.23 P
1 F
0.06 (, or) 523.8 105.23 P
3 F
0.3 0.16 (phrase) 50.88 91.23 B
1 F
0.3 0.16 ( or) 84.46 91.23 B
3 F
0.3 0.16 (clause) 101.67 91.23 B
1 F
0.3 0.16 ( has a syntax. The syntax of an instance and its grammatical form or grammatical) 133.26 91.23 B
FMENDPAGE
%%EndPage: "8" 10
%%Page: "9" 10
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(9) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.3 0.29 (category are the same thing. The grammar of a language is the same thing as the set of possible) 50.88 774.23 B
(syntactic forms that can occur in the language.) 50.88 760.23 T
-0.19 (Syntax is not a physical property of a representing structure, but depends on the capabilities of the) 72.14 740.23 P
-0.23 (system that uses that structure and other structures. As previously remarked, physically identical states) 50.88 726.23 P
0.3 0.33 (need not implement representations with the same syntax. A particular string of characters in a) 50.88 712.23 B
0.3 0.02 (language, e.g. \322) 50.88 698.23 B
4 F
0.3 0.02 (x+y) 126.74 698.23 B
1 F
0.3 0.02 (\323 does not uniquely determine the underlying grammar of the language: the same) 145.63 698.23 B
-0.02 (string can occur in dif) 50.88 684.23 P
-0.02 (ferent languages with dif) 155.85 684.23 P
-0.02 (ferent grammars. In many programming languages that) 275.15 684.23 P
-0.23 (would be a string of three symbols, e.g. two variables separated by an in\336x operator) 50.88 670.23 P
-0.23 (, whereas in space-) 449.28 670.23 P
0.3 0.12 (delimited languages like LISP it could be a single symbol composed of three characters. Similarly) 50.88 656.23 B
0.3 0.12 (,) 536.85 656.23 B
0.3 0.13 (you cannot infer the underlying notation, or representational formalism, by inspecting the physical) 50.88 642.23 B
0.3 0.59 (form of a particular representational state. What determines the syntax \050or grammar\051 of an) 50.88 628.23 B
-0.29 (information-rich sub-state is what kinds of variation in that state are possible and signi\336cant in relation) 50.88 614.23 P
0.3 0.16 (to controlling internal or external behaviour) 50.88 600.23 B
0.3 0.16 (. \050This is one reason why introspection concerning the) 268.97 600.23 B
0.3 0.18 (nature of thought processes or mental images can be totally misleading: inspection does not show) 50.88 586.23 B
(what variation is possible, nor what the ef) 50.88 572.23 T
(fects are.\051) 251.19 572.23 T
0.3 0.58 (T) 72.14 552.23 B
0.3 0.58 (ypically what the syntax of some information state is will be determined by whatever) 79.2 552.23 B
0.19 (mechanism actually uses those states and the variety of possibilities that it can cope with. The system) 50.88 538.23 P
0.3 0.02 (need not use any) 50.88 524.23 B
3 F
0.3 0.02 (explicit) 135.75 524.23 B
1 F
0.3 0.02 ( formalisation of the grammar) 171.25 524.23 B
0.3 0.02 (. \050When there is an explicit grammar) 316.38 524.23 B
0.3 0.02 (, e.g. in a) 494.75 524.23 B
0.3 0.1 (grammar) 50.88 510.23 B
0.3 0.1 (-driven parsing program, the syntactic specif) 94.65 510.23 B
0.3 0.1 (ication will generally use a dif) 314.61 510.23 B
0.3 0.1 (ferent notation,) 464.74 510.23 B
0.3 0 (with a dif) 50.88 496.23 B
0.3 0 (ferent syntax from that which it def) 97.24 496.23 B
0.3 0 (ines: e.g. a formalism other than English can be used by) 268.3 496.23 B
(linguists to specify the grammar of English.\051) 50.88 482.23 T
0.3 0.05 (A further complication is that any particular information-bearing sub-state will generally have a) 72.14 462.23 B
0.3 0.08 (syntax at more than one level of abstraction. English sentences have a syntax that is independent of) 50.88 448.23 B
-0.05 (whether they are written or spoken. However) 50.88 434.23 P
-0.05 (, at a lower level, spoken English has a structure in terms) 266.92 434.23 P
0.3 0.11 (of phonemes \050or possibly other units of sound-structure\051 whereas written English has a structure in) 50.88 420.23 B
0.3 0.21 (terms of sequences of letters and spaces and punctuation marks. The lower level syntax is rich in) 50.88 406.23 B
0.3 0.23 (generative power) 50.88 392.23 B
0.3 0.23 (, in that there are far more potential words composed of sequences of letters, or) 137.34 392.23 B
-0.26 (sequences of phonemes, than actually exist in the language. This is not a trivial detail: it is part of what) 50.88 378.23 P
0.3 0.12 (accounts for the ability of a language to evolve. At a still lower level it is ar) 50.88 364.23 B
0.3 0.12 (guable that there is yet) 427.3 364.23 B
0.3 0.02 (another form of syntactic structure corresponding to the fact that certain stroke patterns are available) 50.88 350.23 B
0.08 (for forming letters, and in principle additional letters could be constructed as needed. At that level the) 50.88 336.23 P
(syntax is dif) 50.88 322.23 T
(ferent for dif) 109.3 322.23 T
(ferent font styles.) 170.36 322.23 T
0.28 (W) 72.14 302.23 P
0.28 (ithin a computer there are similar levels of structure. For example, lists trees and networks can) 82.98 302.23 P
0.3 0.09 (be constructed out of lower level units that are made of sequences of \322bits\323. In a system with error) 50.88 288.23 B
0.3 0.09 (-) 535.86 288.23 B
-0.04 (correcting memory the bit patterns themselves may be made out of more complex sequences of bits in) 50.88 274.23 P
(terms of which they are encoded.) 50.88 260.23 T
-0.25 (I hope that all this shows that linguistic grammars and the sentences they generate are only special) 72.14 240.23 P
0.3 0.5 (cases of a more general concept. There are very many distinct notations, or representational) 50.88 226.23 B
0.3 0.06 (formalisms, each with its own sets of instances, its own properties, potential uses, etc. For example,) 50.88 212.23 B
0.3 0.1 (conventional grammatical notations cannot capture the syntax of forms of representation that allow) 50.88 198.23 B
3 F
0.07 (continuous) 50.88 184.23 P
1 F
0.07 ( variability) 103.52 184.23 P
0.07 (, or structures with mutual containment \050A contains B and B contains A\051, as can) 155.11 184.23 P
(happen with lists in a computer) 50.88 170.23 T
(.) 200.46 170.23 T
0.3 0.36 (I have previously \050Sloman 1971, Sloman 1975\051 analysed some of the syntactic dif) 72.14 150.23 B
0.3 0.36 (ferences) 497.38 150.23 B
0.3 0.47 (between applicative or logical representations \050where the key syntactic relationship involves) 50.88 136.23 B
0.3 0.55 (\322application\323 of a function sign to ar) 50.88 122.23 B
0.3 0.55 (gument expressions\051 and what I then called analogical) 249.14 122.23 B
0.3 0.36 (representations, in which they key syntactic relationship is the holding of a relation within the) 50.88 108.23 B
0.3 0.7 (representing medium. Applicative syntax allows a very rich variety of forms of semantic) 50.88 94.23 B
FMENDPAGE
%%EndPage: "9" 11
%%Page: "10" 11
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(10) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.3 0.52 (interpretation, since arbitrary procedures can be associated with function names to compute) 50.88 774.23 B
0.3 0.1 (denotation of complete applicative expressions on the basis of denotations of their components, for) 50.88 760.23 B
0.3 0.04 (instance computing the truth values of the sentences \322Mary is richer than T) 50.88 746.23 B
0.3 0.04 (om\323, \322Mary is taller than) 417.35 746.23 B
0.3 0 (T) 50.88 732.23 B
0.3 0 (om\323, \322Mary is cleverer than T) 57.37 732.23 B
0.3 0 (om\323. In the simplest analogical representations syntactic relationships) 202.59 732.23 B
0.3 0.12 (within the representing medium represent relations in the domain depicted, but this does not imply) 50.88 718.23 B
0.3 0.13 (that there is any isomorphism, as 2-D relationships can depict 3-D relationships. In some cases the) 50.88 704.23 B
0.07 (mapping from relations in the representation to represented relationships is deterministic \050e.g. in most) 50.88 690.23 P
0.3 0.01 (conventional maps\051. In other cases the representing relations are locally ambiguous and therefore the) 50.88 676.23 B
-0.08 (mapping is highly context sensitive and \336nding a consistent interpretation requires problem solving or) 50.88 662.23 P
0.3 0.06 (search. \050A lot of AI vision work has been based on this fact.\051 Nothing said so far should be taken to) 50.88 648.23 B
0.3 0.04 (imply that there are only applicative and analogical forms of representations. Many others exist, and) 50.88 634.23 B
0.3 0.08 (moreover) 50.88 620.23 B
0.3 0.08 (, these two can be combined, for instance when the order of sentences depicts the order of) 96.99 620.23 B
(events.) 50.88 606.23 T
0.3 0.06 (Sometimes properties of the representing medium constrain what can be represented because of) 72.14 586.23 B
0.23 (the relatively direct semantic relationships used. Thus what can be depicted in a 2-D surface using an) 50.88 572.23 P
0.3 0.26 (analogical representation will be limited by the variety of relationships available in that surface,) 50.88 558.23 B
0.23 (relationships such as neighbourhood, direction, distance, containment, connectivity) 50.88 544.23 P
0.23 (, and so on. A still) 451.09 544.23 P
0.3 0.03 (wider variety of relationships is available within the class of 3-D structures, but in general it is more) 50.88 530.23 B
0.3 0.16 (dif) 50.88 516.23 B
0.3 0.16 (f) 64.48 516.23 B
0.3 0.16 (icult for us to create and change them, and to grasp their structure \050for instance one part of the) 67.97 516.23 B
0.05 (structure can obscure another\051. The use of datastructures containing pointers to other datastructures in) 50.88 502.23 P
0.3 0.08 (a computer provides some of the benef) 50.88 488.23 B
0.3 0.08 (its of pictorial syntax since pointers can be allocated classes,) 241.56 488.23 B
0.09 (giving analogues of closeness, direction and other spatial properties, without the constraints of 2-D or) 50.88 474.23 P
0.07 (3-D spaces, though at the price of loss of continuity \050since datastructures are essentially discrete\051, and) 50.88 460.23 P
-0.24 (the need for specialised procedures for traversing and manipulating links. At this stage it is not clear to) 50.88 446.23 P
0.3 0.01 (what extent human brains use this kind of virtual machine. They are particularly useful for processes) 50.88 432.23 B
0.3 0.05 (that require temporary construction of trees or networks. I have also previously suggested that some) 50.88 418.23 B
0.3 0.17 (aspects of the development of children\325) 50.88 404.23 B
0.3 0.17 (s counting skills could be explained by the construction of) 247.84 404.23 B
(networks of datastructures. \050Chapter 8, in Sloman 1978\051.) 50.88 390.23 T
263.92 357.42 50.88 357.42 2 L
V
1.02 H
0 Z
N
2 14 Q
(\0509\051 Pragmatics of a control state) 50.88 358.9 T
1 12 Q
0.02 (The concept of pragmatics used in connection with natural language is somewhat ambiguous, as there) 50.88 339.23 P
0.31 (are both narrow technical notions and broader more intuitive notions \050see Gazdar 1979\051. Some of the) 50.88 325.23 P
0.52 (technical linguistic concepts of pragmatics are closely bound up with the notion of linguistic context) 50.88 311.23 P
5.44 (within which speaker and hearer share presuppositions that can provide a framework for) 50.88 297.23 P
0.64 (disambiguation of referring expressions, for instance pronouns such as \322it\323 or \322she\323. Although these) 50.88 283.23 P
3.3 (problems of \322indexicals\323 are normally associated with communication in an external language) 50.88 269.23 P
2.61 (between intelligent agents, analogous situations can occur within components of a machine, for) 50.88 255.23 P
1.61 (example when the instruction address register in a computer is to be interpreted as) 50.88 241.23 P
3 F
1.61 (r) 470.45 241.23 P
1.61 (elative) 474.67 241.23 P
1 F
1.61 ( to the) 506.65 241.23 P
(beginning of the currently executing procedure, whose address is in a dif) 50.88 227.23 T
(ferent register) 399.73 227.23 T
(.) 466.01 227.23 T
0.3 0.37 (W) 72.14 207.23 B
0.3 0.37 (e can generalize further by adopting a notion of pragmatics which is concerned with the) 82.87 207.23 B
3 F
0.3 0.32 (purposes) 50.88 193.23 B
1 F
0.3 0.32 ( for which utterances are used, or their functional roles within the lar) 96.77 193.23 B
0.3 0.32 (ger system. Thus) 452.8 193.23 B
0.3 0.08 (questions, assertions, requests, commands and promises all have dif) 50.88 179.23 B
0.3 0.08 (ferent communicative purposes,) 383.53 179.23 B
0.3 0.07 (and the general notion of pragmatics of a notation can be def) 50.88 165.23 B
0.3 0.07 (ined as the set of possible purposes for) 349.38 165.23 B
-0.26 (which it is used, together with the rules that link forms of expression to purposes. These rules are often) 50.88 151.23 P
0.3 0.1 (very loose in natural languages, since, for example, utterances in the form of assertions \050\322Someone) 50.88 137.23 B
0.3 0.38 (left the door open\323\051 and questions \050\322Can you reach the door handle?\323\051 are often intended, and) 50.88 123.23 B
0.3 0.16 (interpreted, as requests or commands \050\322Please shut the door\323\051. As such examples show) 50.88 109.23 B
0.3 0.16 (, pragmatic) 484.01 109.23 B
0.3 0 (roles of linguistic expressions depend not only on their syntactic form but also on context, the source) 50.88 95.23 B
FMENDPAGE
%%EndPage: "10" 12
%%Page: "11" 12
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(1) 288.99 65.15 T
(1) 294.54 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.3 0.32 (of the utterance, and relationships between speaker and hearers. The broad study of pragmatics) 50.88 774.23 B
0.3 0.17 (includes analysing the way in which syntactic forms, context, the capabilities and relationships of) 50.88 760.23 B
0.3 0.21 (utterers and receivers help to determine which uses are possible and under which conditions they) 50.88 746.23 B
(occur) 50.88 732.23 T
(.) 76.86 732.23 T
0.3 0.17 (W) 72.14 712.23 B
0.3 0.17 (e can further generalise this notion of pragmatics to cover the dif) 82.67 712.23 B
0.3 0.17 (ferent functional roles that) 407.59 712.23 B
0.3 0.48 (information-bearing sub-states of a complex control system can have, without assuming that) 50.88 698.23 B
0.3 0.11 (autonomous intelligent language-using agents are involved. This general notion includes ef) 50.88 684.23 B
0.3 0.11 (fects on) 501.13 684.23 B
0.3 0.04 (other sub-states, and the ways in which dif) 50.88 670.23 B
0.3 0.04 (ferences in syntactic forms and other factors contribute to) 258.87 670.23 B
-0.18 (those ef) 50.88 656.23 P
-0.18 (fects. The notion of \322purpose\323 or \322function\323 that I am using does not require a human designer) 88.11 656.23 P
-0.2 (or any conscious intention. It assumes only that we are talking about an integrated system in which the) 50.88 642.23 P
0.3 0.16 (various components somehow co-operate in a systematic way to produce behaviour) 50.88 628.23 B
0.3 0.16 (. The function,) 466.45 628.23 B
0.3 0.19 (purpose, or role of a component is then def) 50.88 614.23 B
0.3 0.19 (ined in terms of how it contributes towards the overall) 266.86 614.23 B
-0.04 (ef) 50.88 600.23 P
-0.04 (fect, whether directly or indirectly) 59.98 600.23 P
-0.04 (. This seems to be how the concept of anatomical or physiological) 223.23 600.23 P
0.3 0.09 (function is used biology) 50.88 586.23 B
0.3 0.09 (, though in some cases biologists make the \050in my view unnecessary\051 claim) 168.82 586.23 B
0.3 0.02 (that their usage is justif) 50.88 572.23 B
0.3 0.02 (ied by evolutionary selection of a design. Although I shall not ar) 163.78 572.23 B
0.3 0.02 (gue over this) 477.1 572.23 B
-0.25 (here, I regard that justi\336cation as spurious since what happens now is suf) 50.88 558.23 P
-0.25 (\336cient justi\336cation for talking) 397.69 558.23 P
0.3 0.1 (about the function of the heart or liver) 50.88 544.23 B
0.3 0.1 (, no matter how or) 239.38 544.23 B
0.3 0.1 (ganisms came to have hearts and livers. In) 330.45 544.23 B
0.3 0.22 (exactly that sense we can talk about the function of the sun in relation to life on earth or weather) 50.88 530.23 B
(patterns.) 50.88 516.23 T
(Ef) 50.88 496.23 T
(fects that are relevant to the pragmatics of a control subsystem include such functions as:) 61.98 496.23 T
4 14 Q
(\245) 50.88 479.23 T
1 12 Q
( initiating a new process,) 62.22 479.23 T
4 14 Q
(\245) 50.88 463.23 T
1 12 Q
(terminating an existing process,) 62.22 463.23 T
4 14 Q
(\245) 50.88 447.23 T
1 12 Q
(suspending or interrupting processes,) 62.22 447.23 T
4 14 Q
(\245) 50.88 431.23 T
1 12 Q
(modifying processes,) 62.22 431.23 T
4 14 Q
(\245) 50.88 415.23 T
1 12 Q
(altering the information stores that control processes,) 62.22 415.23 T
3.35 (and many more. W) 50.88 395.23 P
3.35 (ithin traditional control theory) 152.38 395.23 P
3.35 (, notions such as positive feedback, negative) 306.57 395.23 P
-0.28 (feedback, ampli\336cation and damping are all concerned with this generalised conception of pragmatics.) 50.88 381.23 P
-0.05 (If we allow a richer variety of control states, and a richer variety of causal interactions then our notion) 50.88 367.23 P
-0.07 (of pragmatics will be correspondingly richer) 50.88 353.23 P
-0.07 (. For example, in a typical control mechanism made up of) 263.39 353.23 P
0.02 (a \336xed set of measurable states each of which admits continuous one-dimensional variation, where all) 50.88 339.23 P
0.12 (the interactions can be expressed in a set of partial dif) 50.88 325.23 P
0.12 (ferential equations, the notion of one subsystem) 309.63 325.23 P
0.82 (asking another a question makes little sense. However) 50.88 311.23 P
0.82 (, where sub-systems include factual databases) 315.94 311.23 P
0.31 (that can be interrogated by) 50.88 297.23 P
0.31 (, or given new information by) 179.22 297.23 P
0.31 (, other sub-systems we can \336nd analogues of) 323.2 297.23 P
-0.2 (assertions and questions. Simpli\336ed analogues of commands are available in many mechanisms where) 50.88 283.23 P
0.51 (a signal sent by one component can initiate behaviour in another) 50.88 269.23 P
0.51 (. More complex sorts of commands,) 364.42 269.23 P
4.08 (closer to linguistic commands, are possible where the control signals have internal structure) 50.88 255.23 P
(determining semantic content and can vary in complexity) 50.88 241.23 T
(, as sentences can.) 325.57 241.23 T
0.3 0.35 (On the basis of this simple introduction, readers familiar with modern computing systems,) 72.14 221.23 B
-0.1 (especially networked systems, will easily think of a wide variety of functional roles that could be used) 50.88 207.23 P
0.3 0.33 (to provide a taxonomy of types of pragmatic roles for sub-states in complex behaving systems,) 50.88 193.23 B
0.3 0.23 (including analogues of the problems of disambiguating ambiguous signals on the basis of shared) 50.88 179.23 B
0.3 0.05 (presuppositions, the use of \322indexical\323 expressions similar to \322here\323, \322now\323, \322it\323 and similar things.) 50.88 165.23 B
0.3 0.09 (One of the characteristics of computing systems, unlike previously known mechanisms, is that they) 50.88 151.23 B
-0.12 (were originally designed to take over some human mental tasks, and therefore it is not at all surprising) 50.88 137.23 P
0.3 0.25 (that they should provide much richer illustrations \050and potential explanatory models\051 of familiar) 50.88 123.23 B
(concepts of representations than other sorts of mechanisms do.) 50.88 109.23 T
FMENDPAGE
%%EndPage: "11" 13
%%Page: "12" 13
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(12) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
266.26 771.42 50.88 771.42 2 L
0 X
V
1.02 H
0 Z
N
2 14 Q
(\05010\051 Semantics of a control state) 50.88 772.9 T
1 12 Q
-0.04 (Analysing the notion of meaning, or semantics, is a notoriously dif) 50.88 753.23 P
-0.04 (\336cult philosophical problem, partly) 370.73 753.23 P
0.43 (because there are several dif) 50.88 739.23 P
0.43 (ferent concepts with quite complex relations between them. For instance) 188.24 739.23 P
1.1 (Frege \050following Mill and others\051 attempted to make a distinction between sense \050connotation\051 and) 50.88 725.23 P
0.57 (reference \050denotation\051, but it turned out a far more slippery task than he at \336rst thought \050as he found) 50.88 711.23 P
0.74 (out when he later began to analyse various tricky special cases, such as the use of the pronoun \322I\323\051.) 50.88 697.23 P
0.83 (This is no place to attempt a full analysis of possible notions of meaning and reference. However) 50.88 683.23 P
0.83 (, I) 529.03 683.23 P
-0.1 (claim that there are no well-de\336ned concepts that correspond to our normal use of these words. Rather) 50.88 669.23 P
2.47 (there are many dif) 50.88 655.23 P
2.47 (ferent features to be found in the contexts in which we ordinarily talk about) 145.66 655.23 P
0.06 (meaning, and dif) 50.88 641.23 P
0.06 (ferent subsets of those features can occur in connection with control states of various) 131.72 641.23 P
2.31 (kinds of machines. There is no \322correct\323 set of features that uniquely justi\336es the attribution of) 50.88 627.23 P
2.31 (meaning. Rather there are dif) 50.88 613.23 P
2.31 (ferent subsets with interestingly dif) 200.13 613.23 P
2.31 (ferent properties, and instead of) 378.39 613.23 P
0.17 (ar) 50.88 599.23 P
0.17 (guing about which features correspond to) 59.98 599.23 P
3 F
0.17 (the) 262.86 599.23 P
1 F
0.17 ( concept of \322having a meaning\323 we should explore the) 277.51 599.23 P
(similarities and dif) 50.88 585.23 T
(ferences implied by dif) 140.61 585.23 T
(ferent subsets.) 251.31 585.23 T
0.15 (I have previously tried to show \050Sloman 1985b\051 that a signi\336cant subset of features characteristic) 72.14 565.23 P
0.04 (of uses of symbols with meaning can be found even in the way in which a computer \322understands\323 its) 50.88 551.23 P
0.3 0.08 (machine language \050i.e. without the need for any AI programming\051. For example, in typical machine) 50.88 537.23 B
0.3 0.04 (languages certain bit patterns are used \050by the machine\051 to refer to locations in the machine\325) 50.88 523.23 B
0.3 0.04 (s virtual) 500.22 523.23 B
0.02 (address space, others are used to specify instructions to be performed and others are used as if to refer) 50.88 509.23 P
0.3 0.12 (to numbers, for example when they are incremented during repeated operations. Instead of ar) 50.88 495.23 B
0.3 0.12 (guing) 512.05 495.23 B
-0.03 (over whether these are) 50.88 481.23 P
3 F
-0.03 (r) 161.99 481.23 P
-0.03 (eal) 166.21 481.23 P
1 F
-0.03 ( instances of meaning, or reference, we should accept that they are in some) 180.86 481.23 P
0.3 0.06 (ways like and in some ways unlike other instances, and we should try to analyse the implications of) 50.88 467.23 B
0.3 0.05 (these similarities and dif) 50.88 453.23 B
0.3 0.05 (ferences. \050The example illustrates that the same structure can have dif) 170.49 453.23 B
0.3 0.05 (ferent) 511.63 453.23 B
(semantic roles in dif) 50.88 439.23 T
(ferent pragmatic contexts in a control system.\051) 148.26 439.23 T
0.3 0.39 (One of the characteristics of typically human uses of symbols is that they occur within an) 72.14 419.23 B
0.3 0.02 (architecture that supports not only states and processes involving notions of truth and falsity) 50.88 405.23 B
0.3 0.02 (, such as) 498.11 405.23 B
0.27 (believing, perceiving, inferring, predicting and planning but also motivational or af) 50.88 391.23 P
0.27 (fective states such) 452.06 391.23 P
0.3 0.69 (as wanting, hoping, fearing, enjoying and disliking. Fully human-like uses of symbols or) 50.88 377.23 B
-0.01 (representations with meanings will not be possible in machines or other or) 50.88 363.23 P
-0.01 (ganisms whose architecture) 407.31 363.23 P
-0.25 (is not rich enough to provide the required context for those uses. In other words,) 50.88 349.23 P
539.85 347.93 435.17 347.93 2 L
V
0.59 H
N
-0.25 (human-like semantics) 435.17 349.23 P
452.56 333.93 50.88 333.93 2 L
V
N
0.3 0.09 (presupposes human-like pragmatics, and that requires a human-like functional dif) 50.88 335.23 B
539.94 333.93 452.34 333.93 2 L
V
N
0.3 0.09 (ferentiation in the) 452.34 335.23 B
181.5 319.93 50.88 319.93 2 L
V
N
-0.27 (control-system architecture) 50.88 321.23 P
-0.27 (. \050The actual physical architecture could be completely dif) 181.5 321.23 P
-0.27 (ferent, however) 458.54 321.23 P
-0.27 (.\051) 532.86 321.23 P
0.3 0.05 (At this stage it is too early to say exactly what the architectural requirements for all those states) 72.14 301.23 B
0.3 0.29 (are, though this is the topic of work currently being done at the University of Birmingham \050e.g.) 50.88 287.23 B
0.3 0.09 (Sloman 1993b, Beaudoin and Sloman 1993\051. In any case, there are many dif) 50.88 273.23 B
0.3 0.09 (ferent types of systems) 426.74 273.23 B
0.28 (that we need to study as part of the general study of \322design space\323, the space of possible designs for) 50.88 259.23 P
(interesting behaving systems.) 50.88 245.23 T
0.3 0.17 (W) 72.14 225.23 B
0.3 0.17 (e can make some negative points against over) 82.67 225.23 B
0.3 0.17 (-simple theories of semantics. For example, it) 311.36 225.23 B
0.3 0.76 (would a mistake to require all semantic relations to involve causal connections between) 50.88 211.23 B
0.3 0.33 (representations and what they represent. This would rule out representations of non-existent or) 50.88 197.23 B
0.3 0.42 (impossible states of af) 50.88 183.23 B
0.3 0.42 (fairs: any planning system needs to be able to create representations of) 168.12 183.23 B
0.3 0.14 (alternative possible situations most of which will never actually exist. Non-existent objects cannot) 50.88 169.23 B
0.3 0.06 (enter into causal relations. Another mistake would require all representations to share structure with) 50.88 155.23 B
0.3 0.05 (what they represent: many counter) 50.88 141.23 B
0.3 0.05 (-examples are to be found in natural languages, including the fact) 219.41 141.23 B
0.15 (that the very same thing can be referred to either by a very simple pronoun, or by a variety of phrases) 50.88 127.23 P
0.3 0.01 (of dif) 50.88 113.23 B
0.3 0.01 (fering structure. A less obvious mistake is to require) 77.34 113.23 B
3 F
0.3 0.01 (every) 334.46 113.23 B
1 F
0.3 0.01 ( case of meaning or representation to) 360.48 113.23 B
0.3 0.2 (accord with a general system or convention: for that would rule out the possibility of temporarily) 50.88 99.23 B
FMENDPAGE
%%EndPage: "12" 14
%%Page: "13" 14
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(13) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.3 0.1 (using an object arbitrarily to represent another for a short time, as sometimes happens, for example) 50.88 774.23 B
0 (when choosing a pepper pot, a match box and a mark on a table to represent the con\336guration of three) 50.88 760.23 P
0.3 0.19 (vehicles at the time of a road accident. These selections need not involve any regular association,) 50.88 746.23 B
0.3 0.18 (systematic relationship or social conventions, and the semantic relations might last only for a few) 50.88 732.23 B
0.3 0.5 (seconds until a dif) 50.88 718.23 B
0.3 0.5 (ferent event is discussed. If this kind of arbitrary and temporary semantic) 148.63 718.23 B
0.28 (relationship can be useful in external representations we should expect similar phenomena in internal) 50.88 704.23 P
0.3 0.29 (representations, especially those involving current percepts which can involve unknown objects) 50.88 690.23 B
(encountered for only a short time. Any general theory of semantics should allow this.) 50.88 676.23 T
0.3 0.03 (A more obvious semantic requirement is that where representation is context-free the variability) 72.14 656.23 B
0.3 0.01 (of a sub-state should at least match quantitatively the variety of things that it is required to represent.) 50.88 642.23 B
0.3 0.08 (A set of eight binary switches is capable of only 256 possible states, and so would be inadequate to) 50.88 628.23 B
0.05 (represent states of a chess board, since far more con\336gurations are possible in chess. However) 50.88 614.23 P
0.05 (, the set) 502.77 614.23 P
0.3 0.34 (of possible sub-states of a collection of switches grows very rapidly as the number of switches) 50.88 600.23 B
0.3 0.01 (increases, so that a relatively small collection of switches can cover a huge variety of possible states.) 50.88 586.23 B
0.3 0.1 (This is one of the characteristics of computing mechanisms which accounts for their generality and) 50.88 572.23 B
(power) 50.88 558.23 T
(.) 80.19 558.23 T
0.18 (Where the representing medium is the state of a virtual machine, and the syntax allows inde\336nite) 72.14 538.23 P
0.3 0.08 (structural extension, as in the case of natural languages, predicate calculus, and many programming) 50.88 524.23 B
0.3 0.41 (languages, the semantics may allow an inf) 50.88 510.23 B
0.3 0.41 (inite variety of dif) 271.77 510.23 B
0.3 0.41 (ferent situations to be dif) 367.64 510.23 B
0.3 0.41 (ferently) 499.67 510.23 B
-0.2 (represented. This implies that the semantic rules mapping representing structures to things represented) 50.88 496.23 P
0.3 0.56 (cannot be full explicit: they must themselves include generative power) 50.88 482.23 B
0.3 0.56 (. This is obviously a) 430.58 482.23 B
0.3 0.15 (requirement for many forms of creativity) 50.88 468.23 B
0.3 0.15 (, in mathematics, science, art and everyday life, when we) 254.53 468.23 B
(cope ef) 50.88 454.23 T
(fectively with novel situations.) 85.62 454.23 T
0.17 (Of course, the physical mechanisms in the system may impose a limit on that variety) 72.14 434.23 P
0.17 (, as memory) 480.54 434.23 P
0.3 0.22 (limits in a computer can limit the size of programs or datastructures. But the potential variability) 50.88 420.23 B
0.06 (inherent in the virtual machine may be far greater than the actual implementation allows. In the world) 50.88 406.23 P
0.3 0.25 (of computing this manifests itself in programming languages that allow programs and data to be) 50.88 392.23 B
0.3 0.36 (created that current computers cannot accommodate, driving the development of new forms of) 50.88 378.23 B
0.3 0.25 (hardware, including memory management systems with lar) 50.88 364.23 B
0.3 0.25 (ger address spaces. When a computer) 350.05 364.23 B
0.3 0.15 (becomes available with a new bigger memory system it may be possible for the same programs as) 50.88 350.23 B
0.3 0.19 (before to run in it, but tackling lar) 50.88 336.23 B
0.3 0.19 (ger) 222.33 336.23 B
0.3 0.19 (, more complex, problems. Similarly) 237.73 336.23 B
0.3 0.19 (, it is possible that once) 421.01 336.23 B
-0.29 (animal brains started to use virtual machines that were not inherently limited by the physical structures) 50.88 322.23 P
0.3 0.06 (available, this could increase the evolutionary pressure towards bigger brains to enable those virtual) 50.88 308.23 B
0.2 (machines to be used more extensively) 50.88 294.23 P
0.2 (. \050This is one way of looking at the old distinction in linguistics) 233.29 294.23 P
0.3 0.03 (between \322competence\323 and \322performance\323, where the latter is limited by implementation details, but) 50.88 280.23 B
(not the former) 50.88 266.23 T
(.\051) 118.83 266.23 T
0.3 0.11 (Additional considerations from the design standpoint, include: how quickly required structures) 72.14 246.23 B
-0.2 (can be created, how quickly they can be changed as needed, how quickly the required substructure can) 50.88 232.23 P
-0.19 (be found in a very lar) 50.88 218.23 P
-0.19 (ge database of structures, and how well a collection of permitted structures \336t the) 152.62 218.23 P
0.3 0.26 (purposes for which they are to be used. \050Compare the criteria for adequacy of representations in) 50.88 204.23 B
0.3 0.16 (McCarthy and Hayes 1969.\051 Experienced software designers learn a vast stock of representational) 50.88 190.23 B
0.3 0.29 (schemata and are able to select appropriate ones for new design tasks with far greater ease than) 50.88 176.23 B
0.3 0.51 (equally clever individuals who merely know the programming language concerned. Explicit) 50.88 162.23 B
0.3 0.56 (codif) 50.88 148.23 B
0.3 0.56 (ication of the knowledge of such experienced designers would be of enormous help in) 77.65 148.23 B
0.3 0.21 (formulating a general account of semantic capabilities of dif) 50.88 134.23 B
0.3 0.21 (ferent forms of representation, but is) 355.23 134.23 B
(likely to be a dif) 50.88 120.23 T
(\336cult task, like all attempts to articulate complex intuitive knowledge.) 129.28 120.23 T
FMENDPAGE
%%EndPage: "13" 15
%%Page: "14" 15
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(14) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
63.32 771.42 50.88 771.42 2 L
0 X
V
1.02 H
0 Z
N
2 14 Q
(\0501) 50.88 772.9 T
340.91 771.42 62.55 771.42 2 L
V
N
(1\051 Inferences \050reasoning\051 in control states) 62.55 772.9 T
1 12 Q
0.62 (Inference \050reasoning\051 is normally de\336ned as derivation of new propositions from old ones subject to) 50.88 753.23 P
0.65 (the constraint that it is impossible for the old ones to be true and the new ones false simultaneously) 50.88 739.23 P
0.65 (.) 536.85 739.23 P
2.3 (W) 50.88 725.23 P
2.3 (e can generalize the concept of \322inference\323 to include all manipulation of information-bearing) 61.24 725.23 P
0.87 (structures to obtain new structures that enable something to be done that could not be done directly) 50.88 711.23 P
4.18 (with the originals. \050A similar though slightly less general proposal, based on the notion of) 50.88 697.23 P
1.28 (\322denotation\323 was made in Sloman 1971.\051 Examples of this generalized notion of inference include) 50.88 683.23 P
(such processes as:) 50.88 669.23 T
4 14 Q
(\245) 50.88 652.23 T
1 12 Q
(creating an interpretation of a sensory array \050see discussion of vision, below\051,) 62.22 652.23 T
4 14 Q
(\245) 50.88 636.23 T
1 12 Q
(deriving a parse-tree from a sentence and a set of syntactic rules,) 62.22 636.23 T
4 14 Q
(\245) 50.88 620.23 T
1 12 Q
(forming a plan,) 62.22 620.23 T
4 14 Q
(\245) 50.88 604.23 T
1 12 Q
(translating \050or compiling\051 a high level plan \050or program\051 into a more detailed lower) 62.22 604.23 T
(-level language,) 461.36 604.23 T
4 14 Q
(\245) 50.88 588.23 T
1 12 Q
-0.15 (searching an abstract space for solution to a problem \050e.g. searching for a plan in a space of possible) 62.22 588.23 P
(plans\051) 62.22 574.23 T
4 14 Q
(\245) 50.88 558.23 T
1 12 Q
0.3 0.04 (using a neural net to transform a vector representing a desired conf) 62.22 558.23 B
0.3 0.04 (iguration of limbs into a vector) 388.24 558.23 B
0.3 0.37 (representing a set of control signals to be sent to muscles or motors in order to produce that) 62.22 544.23 B
(con\336guration.,) 62.22 530.23 T
0.65 (and many more. It should be clear that the variety and power of the inference mechanisms available) 50.88 510.23 P
(will be intimately related to the variability of syntactic forms.) 50.88 496.23 T
0.3 0.13 (This general notion is not restricted to manipulation of propositions, or symbols that represent) 72.14 476.23 B
-0 (propositions, such as are traditionally studied in logic. A very common form of inference in this sense) 50.88 462.23 P
0.3 0.22 (is use of a map to f) 50.88 448.23 B
0.3 0.22 (ind a good route, or to constrain search for some item satisfying geographical) 147.73 448.23 B
0.3 0.07 (constraints \050e.g. f) 50.88 434.23 B
0.3 0.07 (inding a railway station close to a particular town\051, which might be a much slower) 135.39 434.23 B
-0.18 (process if all the available information were in the form of lists of propositions about the locations and) 50.88 420.23 P
0.3 0.15 (features of various towns, roads, rivers, and segments of railways. \050Of course, propositions can be) 50.88 406.23 B
0.3 0.2 (or) 50.88 392.23 B
0.3 0.2 (ganised in such a way as to facilitate searching. In particular) 61.05 392.23 B
0.3 0.2 (, if they are stored in the form of a) 365.48 392.23 B
(virtual 2-D array then sets of propositions can replicate some of the functions of a map.\051) 50.88 378.23 T
0.3 (In this generalized sense almost any transformation of one representing sub-state into another) 72.14 358.23 P
0.3 (, or) 523.57 358.23 P
0.3 0.31 (construction of a new sub-state under the inf) 50.88 344.23 B
0.3 0.31 (luence of others, can be regarded as an example of) 279.65 344.23 B
0.3 0.02 (reasoning or inference, provided that the process fulf) 50.88 330.23 B
0.3 0.02 (ils some useful function within the total system.) 307.05 330.23 B
0.3 0.36 (A full survey of types of inference would require a survey of architectures and the varieties of) 50.88 316.23 B
0.3 0.4 (functions that could be fulf) 50.88 302.23 B
0.3 0.4 (illed by sub-mechanisms. T) 192.57 302.23 B
0.3 0.4 (axonomies of forms of syntax, types of) 335.66 302.23 B
(pragmatics, kinds of semantics and forms of inference would all need to be closely related.) 50.88 288.23 T
0.3 0.44 (External representations used by humans vary enormously in their structure, the kinds of) 72.14 268.23 B
-0.3 (variability they allow) 50.88 254.23 P
-0.3 (, and their functions \050Sloman 1985a\051. Similarly) 152.11 254.23 P
-0.3 (, within a complex system that has) 376.07 254.23 P
0.3 0.16 (multiple independently variable components there can be sub-systems with information states that) 50.88 240.23 B
0.3 0.7 (dif) 50.88 226.23 B
0.3 0.7 (fer widely in syntax, semantics, pragmatics and forms of inference. When the internal) 66.09 226.23 B
-0.28 (representations are based on virtual machines rather than physical structures the variety of possibilities) 50.88 212.23 P
0.3 0.13 (is even greater) 50.88 198.23 B
0.3 0.13 (, since there is a wider variety of types of structures and manipulations of structures) 122.91 198.23 B
(within virtual machines than within physical mechanisms.) 50.88 184.23 T
326.95 151.42 50.88 151.42 2 L
V
N
2 14 Q
(\05012\051 A Partial view of a visual architecture) 50.88 152.9 T
1 12 Q
0.53 (W) 50.88 133.23 P
0.53 (e can illustrate some of this variety by considering an over) 61.24 133.23 P
0.53 (-simpli\336ed view of the architecture of a) 346.8 133.23 P
0.36 (visual sub-system of an intelligent agent. V) 50.88 119.23 P
0.36 (isual systems provide a great deal of scope for combining) 260.51 119.23 P
3.89 (dif) 50.88 105.23 P
3.89 (ferent sorts of notations, as they have many intermediate information states with dif) 63.99 105.23 P
3.89 (ferent) 511.88 105.23 P
-0.01 (representational forms and dif) 50.88 91.23 P
-0.01 (ferent functions and causal links. The diagram below gives a crude \050and) 194.86 91.23 P
FMENDPAGE
%%EndPage: "14" 16
%%Page: "15" 16
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(15) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.67 (partly conjectural\051 indication of this variety) 50.88 774.23 P
0.67 (. Some intermediate representations will be very close in) 262.32 774.23 P
5.7 (form to the original optic array) 50.88 382.23 P
5.7 (, preserving neighbourhood relationships and other spatial) 226.83 382.23 P
0.85 (relationships, whereas others lose spatial structure \050e.g. histograms showing relative global amounts) 50.88 368.23 P
0.15 (of dif) 50.88 354.23 P
0.15 (ferent features\051 and some will include relatively abstract information \050e.g. descriptions of causal) 77.12 354.23 P
0.79 (and functional properties of perceived items\051. Some forms of \322abstract maps\323 may need to combine) 50.88 340.23 P
0.04 (the dif) 50.88 326.23 P
0.04 (ferent types of information, e.g. for indexing purposes and controlling visual search. One of the) 81.68 326.23 P
2.27 (many interesting questions is to what extent visual systems use virtual structures with unlimited) 50.88 312.23 P
-0.14 (complexity) 50.88 298.23 P
-0.14 (, and to what extent they are limited by \336xed size structures, like retinal maps. There seems) 104.07 298.23 P
-0.02 (to be no limit in principle to the structural complexity of scenes that can be perceived \050including more) 50.88 284.23 P
1.25 (and more complex diagrams for example\051. But there may be limits to what can be done with such) 50.88 270.23 P
0.06 (scenes, for instance the number of items that can be directly operated on simultaneously by high level) 50.88 256.23 P
(processes.) 50.88 242.23 T
0.02 (If all this is correct, understanding how vision works must be an important step towards a general) 72.14 222.23 P
0.3 0.08 (understanding of how to build systems making use of multiple forms of representation. \050For a more) 50.88 208.23 B
(detailed analysis or requirements for visual perception see Sloman 1989\051.) 50.88 194.23 T
0.3 0.06 (The internal information structures within a perceptual system depend not only on the nature of) 72.14 174.23 B
-0.09 (the environment but also on the agent\325) 50.88 160.23 P
-0.09 (s general needs and current purposes, and conceptual apparatus.) 234.2 160.23 P
-0.14 (Although house-\337ies and humans share the same physical environment it is likely that their perceptual) 50.88 146.23 P
-0.2 (systems store very dif) 50.88 132.23 P
-0.2 (ferent kinds of information and use it for very dif) 154.99 132.23 P
-0.2 (ferent purposes. There will also) 389.11 132.23 P
0.3 0.02 (be diversity among humans, depending on cultural background and individual learning and interests:) 50.88 118.23 B
0.3 0.1 (for example, within two people the process of looking at the same Chinese text could produce very) 50.88 104.23 B
0.3 0.03 (dif) 50.88 90.23 B
0.3 0.03 (ferent intermediate visual databases if only one of them is f) 64.08 90.23 B
0.3 0.03 (luent reader of Chinese. Even within a) 352.73 90.23 B
50.88 85.75 539.85 782.23 C
50.88 401.73 539.85 770.23 C
50.88 401.73 546.88 770.23 R
7 X
0 K
V
90 450 65.19 11.34 278.02 746.04 G
0.5 H
2 Z
0 X
90 450 65.19 11.34 278.02 746.04 A
5 12 Q
(SCENES) 252.51 742.6 T
7 X
90 450 12.4 11.6 260.02 703.57 G
0 X
90 450 12.4 11.6 260.02 703.57 A
7 X
90 450 12.4 11.6 292.82 703.57 G
0 X
90 450 12.4 11.6 292.82 703.57 A
0 F
(Images) 260.27 656.21 T
(V) 212.21 570.18 T
(isible surface descriptions.) 218.61 570.18 T
(Object or scene centred descriptions) 186.59 525.84 T
138.43 503.63 408.02 541.23 18 RR
N
174.42 558.38 383.22 594.38 18 RR
N
239.22 651.18 315.22 673.57 11.2 RR
N
155.5 444.25 412.29 487.98 18 RR
7 X
V
0 X
N
176.93 450.71 398.12 481.53 R
7 X
V
0 X
(Information relevant to planning, learning,) 176.93 473.53 T
(inferring, control, monitoring, etc.) 176.93 459.53 T
247.82 736.95 257.62 712.77 2 L
1 H
N
278.22 734.55 287.82 716.15 2 L
N
307.02 736.15 295.02 715.35 2 L
N
273.42 736.15 264.62 713.75 2 L
N
256.83 682.97 252.02 671.98 250.27 683.85 253.55 683.41 4 Y
V
254.42 689.97 253.55 683.41 2 L
N
264.93 684.23 261.62 672.69 258.31 684.23 261.62 684.23 4 Y
V
261.62 689.25 261.62 684.23 2 L
N
274.26 683.89 272.82 671.98 267.73 682.84 270.99 683.37 4 Y
V
269.62 692.13 271 683.36 2 L
N
282.26 682.94 276.01 672.7 275.88 684.7 279.07 683.82 4 Y
V
281.62 693.57 279.08 683.82 2 L
N
292.76 682.16 286.41 671.98 286.4 683.98 289.58 683.07 4 Y
V
291.22 689.25 289.58 683.07 2 L
N
304.51 683.91 303.21 671.98 297.99 682.78 301.25 683.34 4 Y
V
300.02 690.69 301.26 683.34 2 L
N
241.44 637.81 232.01 630.38 236.07 641.67 238.76 639.74 4 Y
V
246.42 650.54 238.76 639.73 2 L
N
263.4 641.63 259.21 630.38 256.8 642.14 260.1 641.88 4 Y
V
260.82 651.17 260.11 641.88 2 L
N
292.44 642.67 289.61 631.01 285.83 642.4 289.14 642.54 4 Y
V
288.82 650.54 289.15 642.53 2 L
N
314.7 642.65 317.61 631.01 308.97 639.34 311.83 640.99 4 Y
V
306.42 650.54 311.84 640.99 2 L
N
231.25 551.86 226.22 540.96 224.71 552.87 227.98 552.36 4 Y
V
228.82 557.9 227.99 552.36 2 L
N
278.01 553.12 273.42 542.03 271.43 553.86 274.72 553.49 4 Y
V
275.22 557.9 274.73 553.49 2 L
N
334.01 553.12 329.42 542.03 327.43 553.86 330.72 553.49 4 Y
V
331.22 557.9 330.72 553.49 2 L
N
272.13 498.9 268.82 487.36 265.51 498.9 268.82 498.9 4 Y
V
268.82 504.56 268.82 498.9 2 L
N
(of shape, motion, causal relations etc.) 184.13 509.84 T
246.98 640.56 253.42 650.69 253.33 638.69 250.16 639.63 4 Y
V
247.82 631.15 250.16 639.62 2 L
N
271.78 640.56 278.21 650.69 278.13 638.69 274.95 639.63 4 Y
V
272.62 631.15 274.96 639.62 2 L
N
294.18 640.56 300.61 650.69 300.53 638.69 297.35 639.63 4 Y
V
295.02 631.15 297.36 639.62 2 L
N
300.56 546.14 302.22 558.03 307.11 547.07 303.83 546.61 4 Y
V
304.62 540.96 303.84 546.6 2 L
N
244.16 548.71 251.01 558.57 250.43 546.58 247.3 547.65 4 Y
V
245.42 542.03 247.3 547.64 2 L
N
293.23 493.81 294.21 505.77 299.71 495.11 296.47 494.46 4 Y
V
297.42 489.52 296.48 494.45 2 L
N
238.62 495.34 245.42 505.22 244.89 493.23 241.75 494.28 4 Y
V
239.82 488.44 241.75 494.28 2 L
N
226.53 499.44 223.22 487.9 219.91 499.44 223.22 499.44 4 Y
V
223.22 505.1 223.22 499.44 2 L
N
319.33 499.44 316.02 487.9 312.71 499.44 316.02 499.44 4 Y
V
316.02 505.1 316.02 499.44 2 L
N
325.16 645.57 315.01 651.98 327.01 651.92 326.08 648.74 4 Y
V
433.66 621.77 443.8 615.36 431.81 615.42 432.73 618.6 4 Y
V
326.09 648.74 432.74 618.59 2 L
0 Z
N
4 F
(Other) 451.36 610.17 T
(modalities) 451.36 593.8 T
1 F
(:) 504 593.8 T
448 496.99 518.59 625.79 R
2 Z
N
376.25 610.39 366.21 616.96 378.2 616.71 377.23 613.55 4 Y
V
436.16 599.53 446.2 592.96 434.2 593.21 435.18 596.37 4 Y
V
377.24 613.55 435.2 596.37 2 L
0 Z
N
417.38 542.15 406.21 537.76 413.36 547.4 415.37 544.78 4 Y
V
435.04 563.77 446.21 568.16 439.07 558.52 437.06 561.14 4 Y
V
415.37 544.77 437.06 561.14 2 L
N
410.36 485.56 400.61 478.57 405.17 489.66 407.77 487.61 4 Y
V
434.86 527.57 444.61 534.56 440.04 523.47 437.45 525.52 4 Y
V
407.78 487.61 437.46 525.51 2 L
N
4 F
(touch) 451.36 578.42 T
(hearing) 451.36 563.03 T
(smell) 451.36 547.65 T
(body feed-) 451.36 532.26 T
246 601.04 236.01 594.38 240.95 605.32 243.47 603.18 4 Y
V
250.42 611.44 243.48 603.18 2 L
2 Z
N
267.55 605.57 263.21 594.39 260.96 606.17 264.26 605.87 4 Y
V
264.82 611.98 264.27 605.86 2 L
N
296.36 606.6 293.61 594.92 289.75 606.28 293.05 606.44 4 Y
V
292.82 611.44 293.06 606.43 2 L
N
317.83 606.3 321.61 594.92 312.36 602.57 315.1 604.44 4 Y
V
310.42 611.44 315.11 604.43 2 L
N
250.56 601.72 257.41 611.57 256.83 599.58 253.7 600.65 4 Y
V
251.82 595.03 253.71 600.64 2 L
N
275.36 601.72 282.21 611.57 281.62 599.58 278.49 600.65 4 Y
V
276.62 595.03 278.5 600.64 2 L
N
297.76 601.72 304.61 611.57 304.02 599.58 300.89 600.65 4 Y
V
299.02 595.03 300.9 600.64 2 L
N
0 F
(Intermediate databases) 202.05 617.73 T
190.05 610.65 366.05 633.05 11.2 RR
0 Z
N
397.38 574.18 386.21 578.56 398 580.77 397.69 577.47 4 Y
V
434.23 577.35 445.4 572.96 433.61 570.76 433.92 574.05 4 Y
V
397.7 577.47 433.93 574.05 2 L
N
4 F
(back) 451.36 517.86 T
1 F
(etc.) 451.36 503.47 T
223.12 553.38 218.4 542.35 216.55 554.21 219.83 553.8 4 Y
V
226.9 610.37 219.84 553.79 2 L
0.5 H
2 Z
N
334.26 595.55 334.61 607.54 340.67 597.18 337.46 596.37 4 Y
V
351.62 539.51 337.47 596.36 2 L
N
202.4 596.48 207.05 607.55 208.97 595.7 205.68 596.09 4 Y
V
192.89 488.49 205.69 596.09 2 L
N
372.57 500.45 371.45 488.5 366.07 499.22 369.32 499.84 4 Y
V
348.79 610.37 369.33 499.82 2 L
N
132.37 419.45 440.64 437.16 R
7 X
V
434.08 426.35 132.37 426.35 2 L
0 X
V
1.02 H
0 Z
N
2 14 Q
(Multiple control sub-states in a visual system) 132.37 427.83 T
68.59 411.64 529.22 765.97 R
0.5 H
2 Z
N
89.85 571.81 164.26 653.3 18 RR
7 X
V
0 X
N
4 12 Q
(Histograms) 93.4 632.38 T
143.27 569.84 143.99 581.82 149.73 571.28 146.5 570.56 4 Y
V
153.7 552.8 152.99 540.82 147.25 551.36 150.47 552.08 4 Y
V
146.5 570.56 150.48 552.07 2 L
1 H
0 Z
N
122.78 564.88 123.99 576.82 129.29 566.06 126.03 565.47 4 Y
V
143.2 487.76 141.98 475.83 136.68 486.59 139.94 487.18 4 Y
V
126.03 565.46 139.94 487.17 2 L
N
169.14 608.82 157.17 607.96 166.87 615.03 168.01 611.93 4 Y
V
184.18 621.28 196.15 622.14 186.45 615.07 185.31 618.18 4 Y
V
168.01 611.92 185.32 618.17 2 L
N
164.46 584.25 157.17 593.79 168.4 589.56 166.43 586.91 4 Y
V
179.7 581.36 186.98 571.82 175.75 576.05 177.73 578.7 4 Y
V
166.43 586.9 177.73 578.7 2 L
N
5 10 Q
(\050various\051) 100.48 606.94 T
50.88 85.75 539.85 782.23 C
-2.27 32.1 609.73 824.1 C
FMENDPAGE
%%EndPage: "15" 17
%%Page: "16" 17
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(16) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.3 0.03 (single agent dif) 50.88 774.23 B
0.3 0.03 (ferent intermediate databases in a visual system may be put to very dif) 125.68 774.23 B
0.3 0.03 (ferent kinds of) 468.9 774.23 B
0.03 (use. For example, compare recognition of word-shapes whilst reading and perception of 3-D structure) 50.88 760.23 P
-0.04 (in the environment. Or contrast the use of optical \337ow to control posture and the use of optical \337ow in) 50.88 746.23 P
(perceiving the shapes of objects. The former requires very rapidly computed global information about) 50.88 732.23 T
0.3 0.51 (f) 50.88 718.23 B
0.3 0.51 (low patterns in order to determine whether forward or backward falling motion needs to be) 54.72 718.23 B
(counteracted by contracting muscles, whereas the latter requires far more intricate detail.) 50.88 704.23 T
0.17 (The need for a variety of dif) 72.14 684.23 P
0.17 (ferent notations, or representational forms, in dif) 208.15 684.23 P
0.17 (ferent sub-databases) 442.11 684.23 P
0.3 0.26 (in a visual system is a special case of a general requirement for sensory systems to use dif) 50.88 670.23 B
0.3 0.26 (ferent) 510.57 670.23 B
-0.19 (representing sub-states for dif) 50.88 656.23 P
-0.19 (ferent purposes, e.g. in auditory) 192.98 656.23 P
-0.19 (, tactile and other forms of sensory input.) 343.66 656.23 P
0.3 0.26 (Dif) 50.88 642.23 B
0.3 0.26 (ferent output stages in motor) 67.44 642.23 B
0.3 0.26 (-control subsystems are also likely to require dif) 214 642.23 B
0.3 0.26 (ferent notations) 460.63 642.23 B
0.09 (depending on whether they are concerned with relatively high level plan information or collections of) 50.88 628.23 P
0.3 0 (detailed and rapidly changing low level control signals to motors or muscles. For internal monitoring) 50.88 614.23 B
0.3 0.01 (of internal states, and control of internal processes, such as reasoning, learning or planning, yet more) 50.88 600.23 B
0.3 0.21 (forms of representation may be needed. For the purpose of acquiring and storing vast amounts of) 50.88 586.23 B
0.3 0.51 (content-addressable information, something like a connectionist network containing a lot of) 50.88 572.23 B
0.3 0 (distributed superimposed information could be very useful whereas using the information to solve an) 50.88 558.23 B
(algebraic problem is more likely to use something more like a conventional symbolic representation.) 50.88 544.23 T
0.3 0.08 (Which notations are good for which purposes are engineering design questions, to be answered) 72.14 524.23 B
0.3 0.11 (not by armchair philosophical discussions about abstract requirements for rationality or thought, or) 50.88 510.23 B
0.3 0.57 (intelligence but by detailed analysis of design problems, along with experiments using test) 50.88 496.23 B
(implementations to \336nd out where they work and where they go wrong.) 50.88 482.23 T
0.3 0.14 (Which notations are actually used by dif) 72.14 462.23 B
0.3 0.14 (ferent or) 273.22 462.23 B
0.3 0.14 (ganisms is a dif) 315.53 462.23 B
0.3 0.14 (ferent sort of question, i.e. an) 393.42 462.23 B
0.3 0.34 (empirical question, to be answered by empirical investigations that need to be informed by the) 50.88 448.23 B
0.2 (engineering design considerations. There may be links between the empirical question and the design) 50.88 434.23 P
0.3 0.33 (question insofar as evolutionary pressures favour good design solutions. However) 50.88 420.23 B
0.3 0.33 (, evolution is) 472.64 420.23 B
-0.17 (constrained by what has previously evolved and a failure to take that into account can lead to incorrect) 50.88 406.23 P
(theories about how humans and other animals work.) 50.88 392.23 T
389.91 359.42 50.88 359.42 2 L
V
1.02 H
0 Z
N
2 14 Q
(\05013\051 Hierarchies of dispositions \050desire-like states\051.) 50.88 360.9 T
1 12 Q
0.04 ( Overall control of behaviour is another domain in which intelligent systems are likely to use multiple) 50.88 341.23 P
0.5 (forms of representation, as shown, approximately) 50.88 327.23 P
0.5 (, in the conjectural diagram below) 290.11 327.23 P
0.5 (, which is mainly) 455.72 327.23 P
1.14 (concerned with motivational and af) 50.88 313.23 P
1.14 (fective states. Dif) 225.09 313.23 P
1.14 (ferent levels of control correspond to dif) 311.42 313.23 P
1.14 (ferent) 511.88 313.23 P
-0.13 (layers of information states. Some of them correspond to dispositions that persist over a long time, but) 50.88 299.23 P
1.98 (are \322activated\323 only under certain conditions, which may occur rarely) 50.88 285.23 P
1.98 (, or never) 402.35 285.23 P
1.98 (. Some long term) 451.28 285.23 P
0.27 (dispositions are very general in their ef) 50.88 271.23 P
0.27 (fects and are hard to change \050e.g. personality) 239.46 271.23 P
0.27 (, attitudes\051. Other) 455.38 271.23 P
-0.21 (control states more episodic and transient \050e.g. desires, beliefs, intentions, moods\051. Many are complex,) 50.88 257.23 P
FMENDPAGE
%%EndPage: "16" 18
%%Page: "17" 18
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(17) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
4.13 (richly-structured, sub-states, e.g. political attitudes, involving several dif) 50.88 774.23 P
4.13 (ferent general beliefs,) 427.36 774.23 P
(preferences, desires and principles.) 50.88 760.23 T
(Control states in such a system dif) 72.14 361.33 T
(fer in various ways, e.g. according to such factors as:) 236.49 361.33 T
4 14 Q
(\245) 50.88 344.33 T
1 12 Q
0.18 (how generally applicable they are \050e.g. a generous temperament will in\337uence behaviour towards a) 62.22 344.33 P
(lar) 62.22 330.33 T
(ger class of individuals than the attitude of loyalty to ones friends\051,) 74.65 330.33 T
4 14 Q
(\245) 50.88 314.33 T
1 12 Q
(how long they persist \050desires, moods and emotional states can be short-lived\051,) 62.22 314.33 T
4 14 Q
(\245) 50.88 298.33 T
1 12 Q
(how easy they are to change, and what changes them,) 62.22 298.33 T
4 14 Q
(\245) 50.88 282.33 T
1 12 Q
0.3 0.21 (how directly they inf) 62.22 282.33 B
0.3 0.21 (luence actual \050internal or external\051 behaviour \050e.g. some ethical principles) 167.18 282.33 B
0.3 0.16 (correspond to states that do not directly motivate any sort of behaviour but control the selection) 62.22 268.33 B
0.3 0.02 (when two specif) 62.22 254.33 B
0.3 0.02 (ic sorts of motivation come into conf) 141.12 254.33 B
0.3 0.02 (lict: e.g. a principle that patriotic duty should) 320.39 254.33 B
(come before family loyalty) 62.22 240.33 T
(, or) 191.67 240.33 T
3 F
(vice versa) 210.66 240.33 T
1 F
(\051) 258.94 240.33 T
4 14 Q
(\245) 50.88 224.33 T
1 12 Q
0.3 0.16 (how much of the system they af) 62.22 224.33 B
0.3 0.16 (fect at any given time \050certain alarm states, and moods, seem to) 221.79 224.33 B
(have very global ef) 62.22 210.33 T
(fects on mental and physical processes, even though they are transient\051,) 154.25 210.33 T
4 14 Q
(\245) 50.88 194.33 T
1 12 Q
0.22 (whether they produce only transient ef) 62.22 194.33 P
0.22 (fects or whether they feed back into long term states \050a form) 247.97 194.33 P
(of learning\051.) 62.22 180.33 T
0.3 0.08 (Our ordinary language concepts are not rich and precise enough for describing the full richness) 72.14 160.33 B
0.3 0.09 (and variety of control states. Only when we know more about possible underlying mechanisms and) 50.88 146.33 B
0.3 0.01 (have a good theory about how dif) 50.88 132.33 B
0.3 0.01 (ferent overall architectures meet dif) 213.85 132.33 B
0.3 0.01 (ferent design requirements, will) 386.22 132.33 B
0.3 0.22 (we be able to generate a good taxonomy of mental states \050as the theory of the structure of matter) 50.88 118.33 B
(generated the periodic table of the elements\051.) 50.88 104.33 T
50.88 85.75 539.85 782.23 C
50.88 377.33 539.85 756.23 C
47.23 377.33 543.5 756.23 R
7 X
0 K
V
3 H
2 Z
0 X
90 450 7.5 6.72 192.59 707.97 A
90 450 7.5 6.72 236.08 666.34 A
90 450 7.5 6.72 383.93 704.4 A
90 450 7.5 6.72 305.98 707.97 A
90 450 7.5 6.72 128.51 507.86 G
90 450 7.5 6.72 128.51 507.86 A
90 450 7.5 6.72 269.08 626.99 A
90 450 7.5 6.72 349.51 507.86 G
90 450 7.5 6.72 349.51 507.86 A
90 450 7.5 6.72 472.51 507.86 G
90 450 7.5 6.72 472.51 507.86 A
90 450 7.5 6.72 189.08 551.17 A
90 450 7.5 6.72 226.51 507.86 G
90 450 7.5 6.72 226.51 507.86 A
90 450 7.5 6.72 228.02 582.14 A
90 450 7.5 6.72 291.51 574.17 A
90 450 7.5 7.2 369.51 575.61 A
191.45 516.2 220 507 190.91 499.67 191.18 507.94 4 Y
2 X
V
140.61 509.58 191.18 507.93 2 L
6 H
N
315.7 515.51 344.54 507.24 315.7 498.97 315.7 507.24 4 Y
V
242.54 507.24 315.7 507.24 2 L
N
437.17 515.26 466.01 506.99 437.17 498.72 437.17 506.99 4 Y
V
363.01 506.99 437.17 506.99 2 L
N
126.98 706.03 58.58 706.03 2 L
0 X
V
1.02 H
0 Z
N
2 14 Q
(Long term) 58.58 707.51 T
129.58 547.59 58.83 547.59 2 L
V
N
(Short term) 58.83 549.07 T
210.23 531.89 221 515.63 203.25 523.71 206.74 527.8 4 Y
V
186.01 545.38 206.75 527.79 2 L
2.5 H
2 Z
N
211.46 562.2 194.57 555.98 205.6 570.2 208.53 566.2 4 Y
V
221.61 575.58 208.54 566.19 2 L
2 H
N
345.01 534.95 355.56 518.55 337.92 526.86 341.46 530.91 4 Y
V
298.48 568.46 341.48 530.9 2 L
2.5 H
N
365.46 535.64 355.16 519.08 355.12 538.58 360.29 537.11 4 Y
V
368.61 567.58 360.3 537.1 2 L
N
3 H
90 450 7.5 6.72 320.15 666.99 A
90 450 7.5 6.72 344.95 615.98 A
90 450 7.5 6.72 383.93 629.58 A
224.79 681.04 231.57 671.14 220.57 675.94 222.68 678.49 4 Y
V
199.26 697.68 222.68 678.49 2 L
0.7 H
N
302.59 673.65 312.64 667.08 300.64 667.33 301.62 670.49 4 Y
V
199.26 701.08 301.63 670.48 2 L
N
250.1 672.25 238.24 670.48 247.36 678.27 248.73 675.26 4 Y
V
305.98 701.08 248.74 675.26 2 L
N
367.65 640.48 372.87 629.67 362.72 636.07 365.19 638.27 4 Y
V
309.11 701.08 365.2 638.27 2 L
N
258.5 644.18 263.04 633.07 253.3 640.09 255.9 642.14 4 Y
V
241.79 660.27 255.91 642.13 2 L
N
244.82 598.1 231.15 588.87 237.85 603.94 241.33 601.02 4 Y
V
259.5 622.87 241.34 601.02 2 L
1.5 H
N
285.41 597.19 288 580.89 277.17 593.34 281.29 595.26 4 Y
V
270.13 619.47 281.29 595.26 2 L
N
339 634.71 340.99 622.87 333.03 631.86 336.02 633.28 4 Y
V
323.28 660.27 336.03 633.28 2 L
0.7 H
N
393.23 644.78 390.59 633.07 386.62 644.4 389.92 644.59 4 Y
V
387.06 694.28 389.94 644.58 2 L
N
381.35 592.82 372.87 578.66 372.56 595.16 376.96 593.99 4 Y
V
383.52 619.47 376.97 593.98 2 L
1.5 H
N
359.55 597.34 365.79 582.06 352.41 591.71 355.98 594.53 4 Y
V
344.54 609.26 355.99 594.52 2 L
N
406.62 390.77 162.48 390.77 2 L
V
1.02 H
0 Z
N
(Hierarchies of \050dispositional\051 control) 162.48 392.25 T
411.86 534.45 533.86 712.96 R
7 X
V
4 12 Q
0 X
(Personality) 411.86 704.96 T
(Attitudes and beliefs) 411.86 662.96 T
(Moods \050global states\051) 411.86 634.96 T
(Emotions) 411.86 620.96 T
(Desires, preferences) 411.86 578.96 T
(inclinations, etc.) 411.86 564.96 T
224.44 649.31 228.55 660.59 231.04 648.85 227.74 649.08 4 Y
2 X
V
224.73 606.1 227.75 649.08 2 L
0.7 H
2 Z
N
318.37 697.34 312.64 707.89 323.09 701.98 320.73 699.66 4 Y
V
383.52 636.47 320.74 699.66 2 L
N
185.38 689.71 190.39 700.61 191.92 688.71 188.65 689.21 4 Y
V
177.73 617.1 188.66 689.2 2 L
N
0 X
(Relatively hard to) 53.99 685.22 T
(change, very slow) 53.99 671.22 T
(learning, effects) 53.99 657.22 T
-0.48 (diffuse and indirect) 53.99 643.22 P
2 14 Q
(body monitors) 242.88 452.51 T
234.77 447.35 348.16 467.75 10.2 RR
0.5 H
N
159.07 492.56 140.61 502.58 161.57 503.86 160.32 498.21 4 Y
V
277.35 472.75 160.32 498.21 2 L
3 H
N
320.63 495.87 341.12 500.45 326.17 485.71 323.4 490.79 4 Y
V
284.44 469.84 323.41 490.78 2 L
N
438.83 729.89 140.26 729.89 2 L
V
1.02 H
0 Z
N
(Sources of motivation and action in an agent) 140.26 731.38 T
312.93 654.2 319.72 664.1 319.21 652.11 316.07 653.16 4 Y
2 X
V
291.39 578.66 316.08 653.16 2 L
0 X
V
0.7 H
2 Z
2 X
N
267 553.16 287.84 565.06 276.48 543.92 271.74 548.54 4 Y
V
241.73 518.1 271.74 548.53 2 L
11 X
V
4 H
2 X
N
460.2 532.45 475.7 518.28 455.05 522.08 457.62 527.26 4 Y
0 X
V
377.61 566.58 457.64 527.26 2 L
3 H
N
340.45 554.09 357.48 566.37 349.54 546.93 345 550.51 4 Y
2 X
V
283.08 471.15 345 550.51 2 L
N
129.41 409.01 435.85 440.43 R
7 X
V
0 12 Q
0 X
2.32 (Arrows represent causes of differing strengths, differing time-) 129.41 432.43 P
(scales \050learning\051, some deterministic some probabilistic) 129.41 418.43 T
507.04 515.51 535.88 507.24 507.04 498.97 507.04 507.24 4 Y
2 X
V
486.27 507.24 507.04 507.24 2 L
6 H
N
83.78 515.51 112.61 507.24 83.78 498.97 83.78 507.24 4 Y
V
57.53 507.24 83.78 507.24 2 L
N
5 10 Q
0 X
(Event str) 53.61 482.58 T
(eam) 93.11 482.58 T
50.61 382.58 538.61 745.58 R
0.5 H
0 Z
N
256.32 605.76 265.73 621.1 265.88 603.1 261.1 604.43 4 Y
2 X
V
252.73 573.1 261.1 604.43 2 L
2 H
2 Z
N
335.73 474.1 446.73 489.1 R
7 X
V
0 12 Q
0 X
(re\337ex action) 335.73 481.1 T
50.88 85.75 539.85 782.23 C
-2.27 32.1 609.73 824.1 C
FMENDPAGE
%%EndPage: "17" 19
%%Page: "18" 19
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(18) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.3 0.06 (Dif) 72.14 774.23 B
0.3 0.06 (ferent levels of control may use dif) 88.1 774.23 B
0.3 0.06 (ferent notations or forms of representation. Some may be) 259.83 774.23 B
0.3 0.72 (\322symbolic\323, some neural or \322connectionist\323. It remains an open question which forms of) 50.88 760.23 B
0.3 0.19 (representation are most suitable for which of these dif) 50.88 746.23 B
0.3 0.19 (ferent control states. A hierarchy of control) 322.1 746.23 B
0.27 (states might be composed of a hierarchy of production systems, where each system includes rules for) 50.88 732.23 P
0.3 0.05 (modifying lower level systems. Alternatively) 50.88 718.23 B
0.3 0.05 (, each node might be a neural net which responds to its) 270.52 718.23 B
0.3 0.31 (inputs by modifying weights in a lower level neural net. Both of these implementations may be) 50.88 704.23 B
0.3 0.19 (capable of meeting the same general design requirements, dif) 50.88 690.23 B
0.3 0.19 (fering only in some of their detailed) 358.63 690.23 B
0.02 (consequences, which might turn out to be relatively unimportant and hard to detect \322from outside\323. In) 50.88 676.23 P
0.3 0.32 (order to decide between these options in explaining human capabilities we need both empirical) 50.88 662.23 B
(evidence and a design-based theory about the trade-of) 50.88 648.23 T
(fs.) 309.44 648.23 T
512.75 615.42 50.88 615.42 2 L
V
1.02 H
0 Z
N
2 14 Q
(\05014\051 People use many kinds of information states in solving problems) 50.88 616.9 T
1 12 Q
3.2 (It seems that when solving problems, people often use \322mixed\323 notations, both internally and) 50.88 597.23 P
1.59 (externally) 50.88 583.23 P
1.59 (, as illustrated below) 98.06 583.23 P
1.59 (. Asking people how many dif) 201.32 583.23 P
1.59 (ferent intersection points are possible) 354.29 583.23 P
1.53 (between a circle and triangle in the same plane normally causes people to start imagining various) 50.88 569.23 P
1.58 (continuous transformations of a circle and triangle relative to each other) 50.88 555.23 P
1.58 (. Somehow) 411.72 555.23 P
1.58 (, by doing this) 466.48 555.23 P
3.84 (many people \050though not all\051 are able to discover that there are seven possible numbers of) 50.88 541.23 P
-0.07 (intersections 0,1,2,3,4,5 and 6. Some of the numbers can be achieved in more than one way) 50.88 527.23 P
-0.07 (, e.g. using) 487.69 527.23 P
-0.19 (tangents or incidence of vertices on the circle. When answering this question many people seem to use) 50.88 247.23 P
2.36 (an internal representation that admits transformations like sliding and rotating, rather than being) 50.88 233.23 P
3.3 (restricted to the kind of applicative syntax that allows only insertion, deletion or substitution.) 50.88 219.23 P
-0.22 (However) 50.88 205.23 P
-0.22 (, it is clear that simply transforming images is not enough: the images need to be linked to the) 94.35 205.23 P
0.17 (terms of the question and the intersection points identi\336ed and counted, showing the need to combine) 50.88 191.23 P
(dif) 50.88 177.23 T
(ferent forms of representation.) 63.99 177.23 T
-0.21 (A more subtle requirement is the ability to reason about the transformations that are possible so as) 72.14 157.23 P
0.3 0.23 (to be sure that all signif) 50.88 143.23 B
0.3 0.23 (icantly dif) 170.54 143.23 B
0.3 0.23 (ferent cases have been covered. This is often a requirement for) 222.08 143.23 B
0.3 0.49 (mathematical proofs using diagrams, and seems to require the ability to relate diagrammatic) 50.88 129.23 B
0.3 0.06 (representations to propositions stating general facts about those representations. Another example is) 50.88 115.23 B
50.88 85.75 539.85 782.23 C
50.88 269.8 539.39 523.23 C
50.88 269.8 539.39 523.23 R
7 X
0 K
V
90 450 38.98 38.98 157.18 400.27 G
0.5 H
2 Z
0 X
90 450 38.98 38.98 157.18 400.27 A
7 X
90 450 38.98 38.98 417.29 398.97 G
0 X
90 450 38.98 38.98 417.29 398.97 A
391.49 457.75 363.14 415.23 416.29 404.6 3 Y
N
157.18 435.7 128.83 393.18 181.98 382.55 3 Y
N
7 X
90 450 38.98 38.98 277.85 401.93 G
0 X
90 450 38.98 38.98 277.85 401.93 A
323.92 401.93 295.57 359.41 348.72 348.78 3 Y
N
161.62 344.5 154.32 344.5 2 L
V
1.17 H
0 Z
N
0 16 Q
(0) 154.32 346.2 T
289.18 346.17 281.89 346.17 2 L
V
N
(1) 281.89 347.86 T
413.19 345.21 405.9 345.21 2 L
V
N
(2) 405.9 346.9 T
59.42 277.29 532.77 514.9 R
0.5 H
2 Z
N
421.97 286.24 149.05 286.24 2 L
V
1.02 H
0 Z
N
2 14 Q
(Reasoning about intersections in a plane) 149.05 287.72 T
91.06 467.1 490.23 510.43 R
7 X
V
5 10 Q
0 X
(In how many ways can a cir) 91.06 502.43 T
(cle and a triangle intersect? I.e. how many dif) 226.92 502.43 T
(fer) 448.77 502.43 T
(ent) 461.09 502.43 T
-0.09 (numbers) 91.06 488.43 P
-0.09 (of intersections ar) 135.68 488.43 P
-0.09 (e ther) 220.36 488.43 P
-0.09 (e? Below ar) 247.97 488.43 P
-0.09 (e thr) 306.78 488.43 P
-0.09 (ee examples with 0 intersections,) 327.99 488.43 P
(1 intersection, and) 91.06 474.43 T
(2 intersections. How many other possible numbers ar) 182.95 474.43 T
(e ther) 436.53 474.43 T
(e?) 464.23 474.43 T
135.23 307.93 446.89 334.6 R
7 X
V
0 X
(Most people claim to \322visualise\323 the arrangements, imagine) 135.23 327.93 T
(various changes, and count the r) 135.23 315.93 T
(esulting intersection points.) 294.46 315.93 T
50.88 85.75 539.85 782.23 C
-2.27 32.1 609.73 824.1 C
FMENDPAGE
%%EndPage: "18" 20
%%Page: "19" 20
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(19) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
(the Chinese proof of Pythagoras\325 theorem., summarised below) 50.88 774.23 T
(.) 350.22 774.23 T
0.3 0.35 (In principle the theorem can be stated entirely in terms of geometrical concepts \050including) 72.14 507.44 B
0.3 0.27 (addition of areas\051. It is also possible in principle to prove the theorem by using only logical and) 50.88 493.44 B
-0.27 (algebraic notations, starting from some set of axioms for geometry) 50.88 479.44 P
-0.27 (, though it is ar) 367.42 479.44 P
-0.27 (guable that insofar as) 438.42 479.44 P
0.3 0.18 (any such axioms are supposed to be non-arbitrary they must be derived from some non-axiomatic) 50.88 465.44 B
0.3 0.25 (knowledge of spatial structure, whose precise nature remains to be investigated. Moreover) 50.88 451.44 B
0.3 0.25 (, most) 508.99 451.44 B
-0.14 (people \336nd purely logical and algebraic proofs hard to understand and seem to need the aid of pictures) 50.88 437.44 P
0.3 0.15 (or diagrams. Often thinking of the right picture transforms the problem so that it is much easier to) 50.88 423.44 B
0.3 0.04 (solve. For example the Chinese proof illustrated transforms the problem of proving the theorem into) 50.88 409.44 B
0.3 0.11 (that of showing that the area of a lar) 50.88 395.44 B
0.3 0.11 (ge square is composed of the area of a smaller square and four) 230.45 395.44 B
0.3 0.02 (triangles, which, together with a little algebra, proves the theorem easily) 50.88 381.44 B
0.3 0.02 (, even though the squares on) 401.27 381.44 B
-0.23 (the shorter sides of the triangles are nowhere depicted. Again the use of this mixture of representations) 50.88 367.44 P
0.3 0.18 (seems to require the ability to see that a) 50.88 353.44 B
3 F
0.3 0.18 (particular) 254.06 353.44 B
1 F
0.3 0.18 ( diagram captures all the essential features of a) 304.47 353.44 B
0.3 0.07 (whole range of cases, including all forms of right-angled triangles. A still deeper problem is how to) 50.88 339.44 B
0.3 0.01 (encode knowledge about shapes and conf) 50.88 325.44 B
0.3 0.01 (igurations in such a way as to enable one to think of a good) 250.79 325.44 B
0.3 0.23 (transformation when presented with a dif) 50.88 311.44 B
0.3 0.23 (f) 258.99 311.44 B
0.3 0.23 (icult problem. There seems to be wide variety between) 262.55 311.44 B
0.3 0.25 (people as regards such capabilities. How many dif) 50.88 297.44 B
0.3 0.25 (ferent forms of representation mathematicians) 306.34 297.44 B
(actually use when thinking about such problems remains an open question.) 50.88 283.44 T
0.3 0.01 (I suspect that this use of images or diagrams in mathematical reasoning is very closely related to) 72.14 263.44 B
-0.19 (everyday human capabilities, for example when a child looking at a mechanical device is able to \322see\323) 50.88 249.44 P
0.3 0.09 (how it works \050turning this knob moves that lever which....\051. How such abilities develop in children,) 50.88 235.44 B
0.3 0.3 (what the architectural requirements are, which mechanisms are involved, and how they change,) 50.88 221.44 B
(remains unexplained.) 50.88 207.44 T
0.3 0.23 (It is worth noting that people who claim to use images to solve problems could be deceiving) 72.14 187.44 B
0.3 0.01 (themselves. Introspection is not a reliable way of telling what is going on: self-perception is no more) 50.88 173.44 B
0.3 0.03 (reliable than object perception! In both, the reality is mostly hidden: internal and external perceptual) 50.88 159.44 B
0.08 (systems evolved to give us information that is useful for practical purposes, rather than to tell us what) 50.88 145.44 P
-0.22 (\322really\323 exists, which is why the researches of scientists often produce results that extend or contradict) 50.88 131.44 P
0.3 0.26 (what we perceive. There is no more reason to believe that introspection can tell us the syntax of) 50.88 117.44 B
-0.1 (notations we use internally than there is to believe that external perception can tell us the true physical) 50.88 103.44 P
50.88 85.75 539.85 782.23 C
50.88 523.43 539.85 770.23 C
493.86 541.78 94.24 541.78 2 L
0 X
0 K
V
1.02 H
0 Z
N
2 14 Q
(The Chinese proof of Pythagoras\325 theorem: a \322hybrid\323 proof) 94.24 543.26 T
372.41 663.34 372.41 606.65 457.45 606.65 3 Y
0.5 H
2 Z
N
514.15 691.69 514.15 748.38 429.11 748.38 3 Y
N
104.03 731.26 104.03 674.56 189.07 674.56 3 Y
N
429.11 748.38 372.41 748.38 372.41 663.34 3 Y
N
457.45 606.65 514.15 606.65 514.15 691.69 3 Y
N
5 10 Q
(a) 358.24 630.5 T
(a) 89.42 700.24 T
(b) 134.09 661.26 T
(a) 395.07 752.22 T
(c) 151.81 712.11 T
(c) 476.84 651.76 T
(c) 416.61 634.04 T
(c) 459.13 709.7 T
(c) 405.98 702.61 T
(b) 359.64 702.61 T
(b) 469.48 752.22 T
(b) 519.08 642.37 T
(b) 409.24 595.07 T
(a) 487.19 598.61 T
(a) 515.54 722.63 T
93.4 568.26 316.62 651.75 R
7 X
V
0 X
(\050a+b\051) 143.05 643.75 T
2 F
(2) 169.83 648.55 T
5 F
( =) 175.39 643.75 T
2 12 Q
(c) 192.58 643.75 T
2 10 Q
(2) 199.25 648.55 T
5 F
( + 4x1/2\050ab\051) 204.81 643.75 T
2 12 Q
(a) 143.34 613.75 T
2 10 Q
(2) 150 618.55 T
2 12 Q
(+2ab+b) 155.56 613.75 T
2 10 Q
(2) 197.56 618.55 T
2 12 Q
( = c) 203.12 613.75 T
2 10 Q
(2) 223.46 618.55 T
2 12 Q
( + 2ab) 229.02 613.75 T
(a) 169.34 583.75 T
2 10 Q
(2) 176.01 588.55 T
2 12 Q
( + b) 181.56 583.75 T
2 10 Q
(2) 202.56 588.55 T
2 12 Q
( = c) 208.12 583.75 T
2 10 Q
(2) 235.13 588.55 T
79.22 572.47 327.26 657.51 R
N
162.18 630.19 179.89 601.84 2 L
N
176.35 630.19 162.18 598.3 2 L
N
238.67 632.71 256.39 604.36 2 L
N
252.85 632.71 238.67 600.82 2 L
N
192.61 674.56 348.52 748.97 R
7 X
V
0 12 Q
0 X
(The \336gure on the right shows) 192.61 740.97 T
-0.69 (clearly that the square of \050a + b\051 is) 192.61 726.97 P
-0.83 (equal to: the sum of the square of) 192.61 712.97 P
-0.87 (c plus four triangles of base b and) 192.61 698.97 P
(height a.) 192.61 684.97 T
54.42 532.82 536.31 766.69 R
N
50.88 85.75 539.85 782.23 C
-2.27 32.1 609.73 824.1 C
FMENDPAGE
%%EndPage: "19" 21
%%Page: "20" 21
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(20) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.3 0.12 (structure of physical objects. Thus we can only form conjectures and try to test them in the normal) 50.88 774.23 B
0.3 0.23 (way that scientif) 50.88 760.23 B
0.3 0.23 (ic conjectures are tested: by exploring their implications, seeing if they have the) 134 760.23 B
0.3 0.24 (consequences they are alleged to have, and seeing whether their consequences f) 50.88 746.23 B
0.3 0.24 (it other facts and) 454.6 746.23 B
0.3 0.07 (theories. W) 50.88 732.23 B
0.3 0.07 (e also need to allow the possibility that there are several layers of implementation, using) 106.24 732.23 B
(quite dif) 50.88 718.23 T
(ferent notations in high level and low level virtual machines.) 90.97 718.23 T
0.3 0.66 (There is no uniquely optimal way of representing any given set of information. Often) 72.14 698.23 B
0.3 0.18 (representational forms have properties that suit them well to particular problems, even though the) 50.88 684.23 B
0.3 0.37 (same information may be better represented in a dif) 50.88 670.23 B
0.3 0.37 (ferent way for other purposes. This can be) 319.93 670.23 B
(illustrated with a slightly modi\336ed version of a well known example.) 50.88 656.23 T
258.51 623.42 50.88 623.42 2 L
V
1.02 H
0 Z
N
2 14 Q
(\05015\051 The magic square example) 50.88 624.9 T
1 12 Q
2.39 (Algebraic and arithmetical notations are extremely useful for solving many problems relating to) 50.88 605.23 P
0.25 (numbers. For example, if you are given the values of six variables,) 50.88 591.23 P
6 14 Q
0.7 (a) 376.64 591.23 P
1 12 Q
0.25 (,) 385.04 591.23 P
6 14 Q
0.7 (b) 391.28 591.23 P
1 12 Q
0.25 (,) 399.68 591.23 P
6 14 Q
0.7 (d) 405.92 591.23 P
1 12 Q
0.25 (,) 414.32 591.23 P
6 14 Q
0.7 (e) 420.57 591.23 P
1 12 Q
0.25 (,) 428.96 591.23 P
6 14 Q
0.7 (g) 435.21 591.23 P
1 12 Q
0.25 (, and) 443.61 591.23 P
6 14 Q
0.7 (h) 470.42 591.23 P
1 12 Q
0.25 ( and the task) 478.81 591.23 P
0.79 (of computing three more,) 50.88 577.23 P
6 14 Q
2.21 (c) 178.95 577.23 P
1 12 Q
0.79 (,) 187.35 577.23 P
6 14 Q
2.21 (f) 194.13 577.23 P
1 12 Q
0.79 ( and) 202.53 577.23 P
6 14 Q
2.21 (i) 227.42 577.23 P
1 12 Q
0.79 (, then the following three equations would make the task very) 235.82 577.23 P
(easy) 50.88 563.23 T
(, using simple algebraic transformations:) 71.41 563.23 T
6 14 Q
(a + b + c = 15) 236.6 544.9 T
(d + e + f = 15) 236.6 527.9 T
(g + h + i = 15) 236.6 510.9 T
1 12 Q
0.9 (Moreover) 50.88 492.23 P
0.9 (, checking subsequently that the values all satis\336ed the following equations would also be) 97.69 492.23 P
(very easy:) 50.88 478.23 T
6 14 Q
(a + d + g = 15) 236.6 459.9 T
(b + e + h = 15) 236.6 442.9 T
(c + f + i = 15) 236.6 425.9 T
(a + e + i = 15) 236.6 408.9 T
(c + e + g = 15) 236.6 391.9 T
1 12 Q
0.3 0.06 (By contrast, a dif) 72.14 373.23 B
0.3 0.06 (ferent task based on the same set of relationships, expressed in the same set of) 156.45 373.23 B
0.3 0.17 (equations, is much more tedious, namely the task of f) 50.88 359.23 B
0.3 0.17 (inding all the possible one-to-one mappings,) 317.95 359.23 B
(satisfying the above equations, from the numbers:) 50.88 345.23 T
6 14 Q
(1,2,3,4,5,6,7,8,9) 224.01 326.9 T
1 12 Q
(onto the letters:) 50.88 308.23 T
6 14 Q
(a,b,c,d,e,f,g,h,i) 224.01 289.9 T
1 12 Q
1.08 (This can be done by exhaustive search through all possible ways of mapping the numbers onto the) 50.88 271.23 P
1.98 (letters, but the search space has factorial 9, i.e. 362880, nodes. This would be very simple for a) 50.88 257.23 P
(computer program, but is not easy for people: it would take a lot of time and be error prone.) 50.88 243.23 T
-0.11 (It turns out very much easier for some people to perform the task if they switch to a non-algebraic) 72.14 223.23 P
-0.22 (notation that reveals important features of the problem that enable the solution to be found in far fewer) 50.88 209.23 P
0.3 0.38 (steps. The crucial point is that the problem of f) 50.88 195.23 B
0.3 0.38 (inding a mapping that satisf) 296.2 195.23 B
0.3 0.38 (ies the equations is) 440.73 195.23 B
0.3 0.34 (isomorphic to the problem of creating a 3 by 3 \322magic square\323, in which all rows columns and) 50.88 181.23 B
-0.23 (diagonals add up to 15, using only the numbers) 50.88 167.23 P
6 14 Q
-0.64 (1) 278.67 167.23 P
1 12 Q
-0.23 ( to) 287.07 167.23 P
6 14 Q
-0.64 (9) 301.93 167.23 P
1 12 Q
-0.23 (. The correspondence between the two tasks can) 310.33 167.23 P
-0.11 (be seen by labelling the locations in the square with the variables) 50.88 153.23 P
6 14 Q
-0.32 (a) 364.63 153.23 P
1 12 Q
-0.11 ( to) 373.03 153.23 P
6 14 Q
-0.32 (i) 388.13 153.23 P
1 12 Q
-0.11 (, as in the diagram below) 396.52 153.23 P
-0.11 (. It is) 515.76 153.23 P
0.3 0.11 (also necessary to check the equations against the collinear triples in the square. That task would be) 50.88 139.23 B
0.3 0.18 (made harder if the letters within each equation were re-ordered, and the f) 50.88 125.23 B
0.3 0.18 (irst three equations were) 417.09 125.23 B
0.3 0.08 (mixed up with the others. This shows that the equations as they stand share some pictorial structure) 50.88 111.23 B
(with the diagram, whose properties are very helpful in \336nding a solution, as I\325ll now explain.) 50.88 97.23 T
FMENDPAGE
%%EndPage: "20" 22
%%Page: "21" 22
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(21) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.13 (The visually evident structure of the square shows that there are three dif) 72.14 774.23 P
0.13 (ferent kinds of locations) 423.21 774.23 P
0.3 0.12 (with dif) 50.88 760.23 B
0.3 0.12 (ferent roles in the solution, four mid-edge locations each occurring in two collinear triples,) 89.61 760.23 B
0.3 0.19 (four corner locations each occurring in three triples, and the centre location, which occurs in four) 50.88 746.23 B
0.3 0.6 (triples. In any solution to the problem there must therefore also be three kinds of numbers) 50.88 732.23 B
0.3 0.22 (corresponding to three kinds of locations: four numbers occurring in two equations, four in three) 50.88 564.23 B
-0.03 (equations, and one in four equations. This is easily checked:) 50.88 550.23 P
6 14 Q
-0.07 (a) 342.42 550.23 P
1 12 Q
-0.03 (, for example occurs in three equations,) 350.82 550.23 P
-0.09 (so it must correspond to a corner location, as shown, and) 50.88 536.23 P
6 14 Q
-0.26 (e) 325.33 536.23 P
1 12 Q
-0.09 ( occurs in four) 333.73 536.23 P
-0.09 (, so it must be mapped to the) 402.58 536.23 P
(centre.) 50.88 522.23 T
0.3 0.01 (If we check for each number how many triples adding up to 15 contain it, we can use the size of) 72.14 502.23 B
0.3 0.25 (the set to assign the number to a particular category) 50.88 488.23 B
0.3 0.25 (, which constrains the mappings onto letters,) 313.34 488.23 B
0.3 0.23 (considerably reducing the search space by eliminating inappropriate mappings, leaving a total of) 50.88 474.23 B
0.3 0.04 (factorial 4 x factorial 4 = 576 mappings to consider) 50.88 460.23 B
0.3 0.04 (. However) 301.07 460.23 B
0.3 0.04 (, consideration of most of those can be) 351.17 460.23 B
0.05 (avoided by propagating partial solutions over the square, using the observation that symmetry implies) 50.88 446.23 P
0.3 0.05 (that the f) 50.88 432.23 B
0.3 0.05 (irst assignment can go to) 93.94 432.23 B
3 F
0.3 0.05 (any) 218.97 432.23 B
1 F
0.3 0.05 ( location of the appropriate sort. This can be used to cut down) 236.44 432.23 B
0.3 0.18 (the search enormously) 50.88 418.23 B
0.3 0.18 (. After a few numbers are in, the rest are f) 162.44 418.23 B
0.3 0.18 (ixed, and remaining solutions are) 373.27 418.23 B
(found by re\337ecting and rotating the original solution.) 50.88 404.23 T
0.05 (T) 72.14 384.23 P
0.05 (o illustrate how the search is cut down, start with the number) 78.63 384.23 P
6 14 Q
0.13 (1) 374.64 384.23 P
1 12 Q
0.05 ( and consider which pairs can be) 383.04 384.23 P
-0.16 (combined with it. Because) 50.88 370.23 P
6 14 Q
-0.45 (15-1=14) 180.48 370.23 P
1 12 Q
-0.16 (, the numbers in each pair must add up to) 239.24 370.23 P
6 14 Q
-0.45 (14) 438.49 370.23 P
1 12 Q
-0.16 (. Since the lar) 455.28 370.23 P
-0.16 (gest) 520.53 370.23 P
0.16 (other number is) 50.88 356.23 P
6 14 Q
0.46 (9) 129.65 356.23 P
1 12 Q
0.16 ( and) 138.04 356.23 P
6 14 Q
0.46 (14-9=5) 161.69 356.23 P
1 12 Q
0.16 (, that rules out the numbers) 212.06 356.23 P
6 14 Q
0.46 (2) 346.63 356.23 P
1 12 Q
0.16 (,) 355.02 356.23 P
6 14 Q
0.46 (3) 361.18 356.23 P
1 12 Q
0.16 ( and) 369.58 356.23 P
6 14 Q
0.46 (4) 393.22 356.23 P
1 12 Q
0.16 ( as too small to be combined) 401.62 356.23 P
0.12 (with) 50.88 342.23 P
6 14 Q
0.35 (1) 75.32 342.23 P
1 12 Q
0.12 (. The only remaining candidate triples are quickly found to be \050) 83.72 342.23 P
6 14 Q
0.35 (1 5 9) 388.86 342.23 P
1 12 Q
0.12 (\051 and \050) 431.53 342.23 P
6 14 Q
0.35 (1 6 8) 463.08 342.23 P
1 12 Q
0.12 (\051. Thus) 505.75 342.23 P
0.3 0.24 (the number) 50.88 328.23 B
6 14 Q
0.84 0.24 (1) 111.43 328.23 B
1 12 Q
0.3 0.24 ( occurs in only two triples and must therefore be a \322mid-edge\323 number) 120.07 328.23 B
0.3 0.24 (. It could be) 478.76 328.23 B
0.3 0.06 (mapped onto locations labelled) 50.88 314.23 B
6 14 Q
0.84 0.06 (b d f) 206.61 314.23 B
1 12 Q
0.3 0.06 ( or) 250.58 314.23 B
6 14 Q
0.84 0.06 (h) 267.42 314.23 B
1 12 Q
0.3 0.06 (, and at this stage, with no other constraints, we could) 275.88 314.23 B
-0.13 (arbitrarily choose) 50.88 300.23 P
6 14 Q
-0.35 (b) 137.89 300.23 P
1 12 Q
-0.13 (. A similar check shows that) 146.28 300.23 P
6 14 Q
-0.35 (5) 284.43 300.23 P
1 12 Q
-0.13 ( can occur in three more triples, with \050) 292.83 300.23 P
6 14 Q
-0.35 (2 8) 475.34 300.23 P
1 12 Q
-0.13 (\051 \050) 500.17 300.23 P
6 14 Q
-0.35 (3 7) 511.03 300.23 P
1 12 Q
-0.13 (\051) 535.86 300.23 P
0.09 (\050) 50.88 286.23 P
6 14 Q
0.24 (4 6) 54.87 286.23 P
1 12 Q
0.09 (\051 Since it occurs in four triples) 80.3 286.23 P
6 14 Q
0.24 (5) 229.79 286.23 P
1 12 Q
0.09 ( must go in the middle, and be mapped onto) 238.19 286.23 P
6 14 Q
0.24 (e) 452.58 286.23 P
1 12 Q
0.09 (. Since the triple) 460.98 286.23 P
-0.2 (\050) 50.88 272.23 P
6 14 Q
-0.57 (1 5 9) 54.87 272.23 P
1 12 Q
-0.2 (\051 starts with a mid-edge location and goes through the middle, the diagram shows that it must) 95.71 272.23 P
-0.03 (end with a mid-edge location, and that means that) 50.88 258.23 P
6 14 Q
-0.08 (9) 292.46 258.23 P
1 12 Q
-0.03 ( must be mapped onto) 300.85 258.23 P
6 14 Q
-0.08 (h) 409.65 258.23 P
1 12 Q
-0.03 (. From this it follows that) 418.05 258.23 P
-0.06 (the other pair of numbers combining with) 50.88 244.23 P
6 14 Q
-0.17 (1) 253.32 244.23 P
1 12 Q
-0.06 (, i.e.) 261.71 244.23 P
6 14 Q
-0.17 (6) 285.24 244.23 P
1 12 Q
-0.06 ( and) 293.64 244.23 P
6 14 Q
-0.17 (8) 316.83 244.23 P
1 12 Q
-0.06 ( must lie in an edge triple, and therefore they) 325.23 244.23 P
0.03 (must be in the adjacent corners, i.e.) 50.88 230.23 P
6 14 Q
0.07 (a) 223.59 230.23 P
1 12 Q
0.03 ( and) 231.98 230.23 P
6 14 Q
0.07 (c) 255.35 230.23 P
1 12 Q
0.03 ( \050at this stage either mapping would do, as the situation is) 263.74 230.23 P
0.3 0.33 (still symmetrical\051. As) 50.88 216.23 B
6 14 Q
0.84 0.33 (a) 165.92 216.23 B
1 12 Q
0.3 0.33 ( and) 174.65 216.23 B
6 14 Q
0.84 0.33 (e) 200.2 216.23 B
1 12 Q
0.3 0.33 ( are now f) 208.92 216.23 B
0.3 0.33 (ixed and add up to) 260.73 216.23 B
6 14 Q
0.84 0.33 (14) 360.06 216.23 B
1 12 Q
0.3 0.33 (, their companion) 377.51 216.23 B
6 14 Q
0.84 0.33 (i) 471.58 216.23 B
1 12 Q
0.3 0.33 ( must be) 480.3 216.23 B
6 14 Q
0.84 0.33 (4) 527.8 216.23 B
1 12 Q
0.3 0.33 (;) 536.52 216.23 B
0.3 0.2 (similarly completing the other triples can be done without any further search. Another solution is) 50.88 202.23 B
-0 (obtained by re\337ecting about the middle vertical axis. All remaining solutions can be found by rotation) 50.88 188.23 P
(of those two solutions through 90 degrees. \050Is it obvious that no other solutions are possible?\051) 50.88 174.23 T
0.3 0.39 (Perceiving the symmetries in the whole structure and detecting the dif) 72.14 154.23 B
0.3 0.39 (ferences in types of) 437.96 154.23 B
0.27 (locations both use human abilities to detect spatial structures and spatial relations. I do not pretend to) 50.88 140.23 P
0.05 (have an any explanation of how this is done, though no doubt it has something to do with the fact that) 50.88 126.23 P
0.3 0.12 (the visible structures in a 3 by 3 array form patterns that are instances of general schemata that the) 50.88 112.23 B
0.3 0.05 (visual system has frequently encountered, so that properties of such patterns are well known. In part) 50.88 98.23 B
50.88 85.75 539.85 782.23 C
50.88 578.09 538.89 728.23 C
50.88 578.09 538.89 728.23 R
7 X
0 K
V
68.77 613.05 164.26 705.18 R
V
6 14 Q
0 X
(a b c) 68.77 695.84 T
(d e f) 68.77 665.84 T
(g h i) 68.77 635.84 T
61.51 627.22 155.93 712.26 R
0.5 H
2 Z
N
90.56 627.22 123.25 712.26 R
N
61.51 659.11 155.93 687.46 R
N
256.39 614.85 525.68 712.26 R
7 X
V
5 12 Q
0 X
(T) 256.39 704.26 T
0 F
(here are three kinds of locations \050mid-edge, corner) 261.43 704.26 T
(, and) 498.09 704.26 T
-0.56 (centre\051 so there must be three kinds of numbers to go into) 256.39 690.26 P
(them:) 256.39 676.26 T
( \0501\051 four numbers in two triples \050mid-edge\051) 256.39 662.26 T
( \0502\051 four numbers in three triples \050corner\051) 256.39 648.26 T
( \0503\051 one number in four triples \050centre\051) 256.39 634.26 T
57.96 581.67 529.22 719.35 R
N
75.73 584.1 515.73 604.1 R
7 X
V
508.86 593.28 75.73 593.28 2 L
0 X
V
1.02 H
0 Z
N
2 14 Q
(Using visually obvious differences between locations in a square) 75.73 594.77 T
50.88 85.75 539.85 782.23 C
-2.27 32.1 609.73 824.1 C
FMENDPAGE
%%EndPage: "21" 23
%%Page: "22" 23
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(22) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.3 0.2 (the very syntax of the representation makes certain properties easier to detect than corresponding) 50.88 774.23 B
0.3 0.29 (properties in the equation format. For example f) 50.88 760.23 B
0.3 0.29 (inding how many equations contain the letter) 296.76 760.23 B
6 14 Q
0.84 0.29 (e) 531.46 760.23 B
1 12 Q
0.3 0.03 (requires a search through the nine equations for occurrences of the letter) 50.88 746.23 B
0.3 0.03 (, whereas f) 402.4 746.23 B
0.3 0.03 (inding how many) 454.92 746.23 B
0.3 0.16 (collinear triples go through the central location does not require such a search as there is only one) 50.88 732.23 B
0.3 0.01 (occurrence of the location, and for someone accustomed to such structures it is easy to cycle through) 50.88 718.23 B
0.3 0.16 (the lines counting them. However) 50.88 704.23 B
0.3 0.16 (, the apparently simplicity of the experience hides very complex) 219.06 704.23 B
(processing.) 50.88 690.23 T
0.3 0.26 (These are special cases of sophisticated general human visual capabilities whose underlying) 72.14 670.23 B
0.17 (mechanisms we do not at present understand, although studies of vision by researchers in psychology) 50.88 656.23 P
0.3 0.07 (and AI have explored many dif) 50.88 642.23 B
0.3 0.07 (ferent mechanisms and forms of representation, all with only limited) 203.62 642.23 B
0.14 (success. Only when we have a good theory of vision and spatial inference can be hope to have a deep) 50.88 628.23 P
0.3 0.08 (account of how people combine dif) 50.88 614.23 B
0.3 0.08 (ferent forms of representation in solving such problems. Up to a) 224.35 614.23 B
0.3 0.17 (point people can do these things both with external representations and with internal ones, though) 50.88 600.23 B
0.3 0.16 (controlled experiments sometimes show that an individual\325) 50.88 586.23 B
0.3 0.16 (s conf) 344.9 586.23 B
0.3 0.16 (idence regarding manipulation of) 374.48 586.23 B
-0.04 (internal images is misplaced. For instance people who claim to be able to visualize printed words on a) 50.88 572.23 P
0.3 0.06 (blank wall cannot read them backwards nearly as fast as they can read them forwards, whereas with) 50.88 558.23 B
0.3 0.11 (real printed text the speed dif) 50.88 544.23 B
0.3 0.11 (ference is considerably reduced. \050See also the papers by Reisber) 195.72 544.23 B
0.3 0.11 (g and) 512.81 544.23 B
(Slezak in Narayanan 1993\051.) 50.88 530.23 T
0.2 (It is worth noting that in the case of the magic square the representational sophistication required) 72.14 510.23 P
0.3 0.24 (for f) 50.88 496.23 B
0.3 0.24 (inding the solution without search is far greater than that required to understand the original) 72.7 496.23 B
-0.06 (problem and test a solution. In particular) 50.88 482.23 P
-0.06 (, the \322short-cut\323 requires use of a richer description language,) 244.93 482.23 P
0.3 0 (including descriptions of dif) 50.88 468.23 B
0.3 0 (ferent sorts of triples \050e.g. \322diagonal\323\051, and dif) 187.21 468.23 B
0.3 0 (ferent sorts of numbers, or) 411.69 468.23 B
0.3 0.05 (dif) 50.88 454.23 B
0.3 0.05 (ferent sorts of locations \050\322mid-edge\323, \322corner\323, \322centre\323\051. This extra capability is not required for) 64.13 454.23 B
0.27 (the task of \336nding the solution using a systematic generate and test algorithm, which would be trivial) 50.88 440.23 P
-0.08 (to program, by comparison, especially in a typical AI programming language. It is far easier to write a) 50.88 426.23 P
0.3 0.37 (program to do the exhaustive search than to write one that can go through the search-avoiding) 50.88 412.23 B
0.3 0.19 (manipulations illustrated above. This is a general phenomenon: f) 50.88 398.23 B
0.3 0.19 (inding a solution without \322brute-) 375.41 398.23 B
0.3 0.69 (force\323 combinatorial searching often requires more sophisticated, and often more varied,) 50.88 384.23 B
0.3 0.36 (representational abilities than are required for f) 50.88 370.23 B
0.3 0.36 (inding) 295.76 370.23 B
3 F
0.3 0.36 (some) 332.22 370.23 B
1 F
0.3 0.36 ( solution or checking any proposed) 358.3 370.23 B
0.3 0.32 (solution. In this sense, intellectual laziness requires intelligence. In fact, a major component of) 50.88 356.23 B
0.3 0.39 (intelligence as we know it could be characterised as \322productive laziness\323, though how this is) 50.88 342.23 B
-0.17 (achieved still remains to be explained. In contrast with all this representational versatility the syntactic) 50.88 328.23 P
0.3 0.08 (manipulations of logical and arithmetical notations are relatively easy to model on a computer) 50.88 314.23 B
0.3 0.08 (. \050For) 512.53 314.23 B
(further discussion of transformations of representations see Peterson 1994\051.) 50.88 300.23 T
0.3 0.33 (What is sometimes for) 72.14 280.23 B
0.3 0.33 (gotten is that for the purpose of designing a system with human-like) 187.93 280.23 B
0.09 (intelligence it is not suf) 50.88 266.23 P
0.09 (\336cient to have mechanisms that can use the new representation of the problem) 163.65 266.23 P
0.3 0.08 (in order to solve it in the transformed form. The system would also need) 50.88 252.23 B
3 F
0.3 0.08 (general) 411.32 252.23 B
1 F
0.3 0.08 ( information about) 448.54 252.23 B
0.3 0.08 (dif) 50.88 238.23 B
0.3 0.08 (ferent types of representations and their uses, in order to enable it to think of switching from one) 64.21 238.23 B
0.3 0.02 (notation to another) 50.88 224.23 B
0.3 0.02 (. How is that general information acquired? How is it stored? How is it used? \050Do) 141.72 224.23 B
0.3 0.05 (any other animals have that sort of \322meta-level\323 representational capability?\051 In my own case, I was) 50.88 210.23 B
0.25 (aware of the magic square problem for many years before I noticed that the distinction between three) 50.88 196.23 P
0.04 (types of location in the square could be used drastically to reduce the search space. Why did it take so) 50.88 182.23 P
0.3 0.22 (long? Why is it not blindingly obvious immediately the problem is posed? The answer may have) 50.88 168.23 B
0.3 0.23 (something to do with the way in which the meta-level representational capabilities are deployed.) 50.88 154.23 B
(Perhaps some individuals deploy them far more ef) 50.88 140.23 T
(fectively than others?) 292.16 140.23 T
FMENDPAGE
%%EndPage: "22" 24
%%Page: "23" 24
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(23) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
479.35 771.42 50.88 771.42 2 L
0 X
V
1.02 H
0 Z
N
2 14 Q
(\05016\051 Ef\336ciency of a representation is relative to a virtual machine) 50.88 772.9 T
1 12 Q
-0.22 (T) 50.88 753.23 P
-0.22 (alking about ef) 57.37 753.23 P
-0.22 (\336ciency requires care. Often the question \322Which solution requires fewer steps?\323 does) 128.65 753.23 P
1.76 (not have a unique answer) 50.88 739.23 P
1.76 (. It depends on which implementation machine is being considered. For) 179.15 739.23 P
1.66 (example, on a machine that is well suited to generating sets of numbers and adding them up, the) 50.88 725.23 P
1 (exhaustive search through combinations of numbers may require fewer \322machine instructions\323 than) 50.88 711.23 P
0.95 (any attempt to simulate, on the same machine, the human visual capabilities that enable spatial and) 50.88 697.23 P
-0.24 (topological features to be detected very quickly and easily \050compare Glasgow 1993, Sloman 1993a\051. It) 50.88 683.23 P
0.53 (may turn out that this human capability depends on massive neural networks, whose simulation on a) 50.88 669.23 P
1.39 (standard computer would be very expensive, and which inherently require more storage space and) 50.88 655.23 P
0.66 (more \322atomic\323 operations than the exhaustive combinatorial search. Thus what appears ef) 50.88 641.23 P
0.66 (\336cient and) 488.89 641.23 P
0.72 (ef) 50.88 627.23 P
0.72 (fortless for us may actually) 59.98 627.23 P
0.72 (, in some deep sense, be very inef) 192.66 627.23 P
0.72 (\336cient, though very quick because of) 359.04 627.23 P
0.15 (the parallelism. This illustrates a general point: biological designs, produced by evolution, often trade) 50.88 613.23 P
3.36 (huge amounts of storage space and processing power for speed and versatility) 50.88 599.23 P
3.36 (. The costs are) 460.48 599.23 P
0.34 (sometimes for) 50.88 585.23 P
0.34 (gotten by those who advocate spatial representations on computers as \322more ef) 119.29 585.23 P
0.34 (\336cient\323.) 501.55 585.23 P
(In summary) 50.88 571.23 T
(, the following questions need to be distinguished:) 107.73 571.23 T
(\050a\051) 57.96 554.23 T
-0.38 (For problem P) 79.22 554.23 P
-0.38 (, does formalism F1 or formalism F2 enable solutions to be found in fewer steps at) 146.44 554.23 P
(the level of the formalism?) 79.22 540.23 T
(\050b\051) 57.96 523.23 T
(For problem P) 79.22 523.23 T
(, using implementation machine M, does formalism F1 or formalism F2 enable) 147.19 523.23 T
(solutions to be found in fewer steps at the level of machine M?) 79.22 509.23 T
(\050c\051) 57.96 492.23 T
-0.28 (For problem P) 79.22 492.23 P
-0.28 (, which combination of formalisms and implementations produces the fastest solu-) 146.62 492.23 P
(tion in real time using available physical mechanisms?) 79.22 478.23 T
0.54 (The answer to \050a\051 may be F1 and the answer to \050b\051 F2, and the best answer to \050c\051 may be something) 50.88 458.23 P
(totally dif) 50.88 444.23 T
(ferent.) 97.64 444.23 T
0.13 (Thus asking in the abstract which formalism is better for a problem can be a silly exercise. A full) 72.14 424.23 P
0.3 0.35 (analysis of this problem would require investigation of the relative ef) 50.88 410.23 B
0.3 0.35 (f) 410.53 410.23 B
0.3 0.35 (iciency of implementing) 414.22 410.23 B
0.19 (dif) 50.88 396.23 P
0.19 (ferent kinds of abstract formalisms in dif) 63.99 396.23 P
0.19 (ferent kinds of implementation machines, which may also) 260.74 396.23 P
0.3 0.06 (be virtual machines concerning which similar questions arise relative to lower level implementation) 50.88 382.23 B
0.3 0.78 (machines, and so on. Until we know a lot more about these issues, debates about which) 50.88 368.23 B
0.3 0.31 (representations are best in the abstract, like debates between symbolists and connectionists, are) 50.88 354.23 B
0.3 0.09 (pointless. \050Pat Hayes made similar comments in his 1974 paper criticising my 1971 paper) 50.88 340.23 B
0.3 0.09 (. See also) 492.76 340.23 B
(Hayes 1993.\051) 50.88 326.23 T
307.44 293.42 50.88 293.42 2 L
V
N
2 14 Q
(\05017\051 Information-bearing control states) 50.88 294.9 T
1 12 Q
1.12 (In the light of discussion so far we can now see that information-bearing control states involve the) 50.88 275.23 P
(following, in varying degrees of sophistication.) 50.88 261.23 T
5 F
(1. An underlying medium and a mechanism that manipulates it.) 50.88 239.23 T
1 F
2.66 (The medium and the mechanism determine the kind of variability supported \050the syntax of the) 50.88 219.23 P
1.58 (information states\051, the kinds of inferences available and possible ways of constructing, changing,) 50.88 205.23 P
(storing, or searching structures.) 50.88 191.23 T
4 14 Q
(\245) 50.88 175.23 T
1 12 Q
0.08 (Some formalisms are suf) 62.22 175.23 P
0.08 (\336ciently abstract to allow a very wide variety of media and mechanisms to) 181.82 175.23 P
(support them. \050The same language can be spoken, written, put into morse code, etc.\051) 62.22 161.23 T
4 14 Q
(\245) 50.88 145.23 T
1 12 Q
-0.04 (Often the medium in which a formalism is implemented supports a richer form of variation than the) 62.22 145.23 P
0.25 (formalism requires: e.g. sequences of characters on a 2-D surface use only a subset of the potential) 62.22 131.23 P
0.3 0 (inherent in the surface. 2-D structure in tables makes more use of the medium than ordinary prose.) 62.22 117.23 B
0.3 0.13 (One form of creativity is extending a formalism to use more of the power of the medium, as has) 62.22 103.23 B
FMENDPAGE
%%EndPage: "23" 25
%%Page: "24" 25
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(24) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.3 0.04 (happened often in the history of mathematics. The result is sometimes an extensions that works in) 62.22 774.23 B
0.3 0.27 (one medium but not another) 62.22 760.23 B
0.3 0.27 (, e.g. written mathematical notations for which there is no spoken) 205.2 760.23 B
(equivalent. Some poets have also used 2-D format.) 62.22 746.23 T
4 14 Q
(\245) 50.88 730.23 T
1 12 Q
0.3 0.07 (The main medium available in computers is essentially a very lar) 62.22 730.23 B
0.3 0.07 (ge one dimensional array of bits) 381.95 730.23 B
0.3 0.22 (with a set of operations that can be used to implement a very wide variety of additional virtual) 62.22 716.23 B
-0.16 (media, including, for example, N-dimensional arrays of bit-patterns for N = 2, 3, etc. Although such) 62.22 702.23 P
0.3 0.12 (mechanisms are very general indeed they are not totally general and in particular cannot support) 62.22 688.23 B
0.3 0.01 (continuous variation, or arbitrary rotations, though such things can be approximated. It is not clear) 62.22 674.23 B
(what the implications of these limitations are.) 62.22 660.23 T
4 14 Q
(\245) 50.88 644.23 T
1 12 Q
0.3 0.03 (The mechanism may support distributed or superimposed states \050like superimposed wave forms in) 62.22 644.23 B
0.16 (multiplexed signals\051. Superimposed structures may be completely separable, or not; i.e. with cross-) 62.22 630.23 P
0.3 0.04 (talk or without. Sometimes cross-talk between structures provides a useful form of generalisation:) 62.22 616.23 B
(storing information about a particular case can change the stored information about similar cases.) 62.22 602.23 T
4 14 Q
(\245) 50.88 586.23 T
1 12 Q
0.3 0.34 (The medium may support direct jumps between arbitrary states, or only transitions via many) 62.22 586.23 B
0.3 0.4 (intermediate states. Conventional computers have the disadvantage that transitions between) 62.22 572.23 B
0.3 0.03 (arbitrary states generally require lar) 62.22 558.23 B
0.3 0.03 (ge numbers of bits in memory to be changed, and this requires) 235.56 558.23 B
0.3 0.22 (many intermediate states in which dif) 62.22 544.23 B
0.3 0.22 (ferent portions of memory are changed. The intermediate) 251.68 544.23 B
0.3 0.11 (states may be meaningless relative to the control-function of the mechanism. An implementation) 62.22 530.23 B
0.24 (using lar) 62.22 516.23 P
0.24 (ge numbers of neurons, or lar) 103.88 516.23 P
0.24 (ge numbers of computers, can support a much lar) 246.11 516.23 P
0.24 (ger number) 484.66 516.23 P
0.29 (direct state transitions, as well as being more robust. Nevertheless if the computers are fast enough) 62.22 502.23 P
0.3 0.03 (they can match the functionality of the parallel implementation, insofar as they can achieve all the) 62.22 488.23 B
0.21 (same state-transitions in the same times. However) 62.22 474.23 P
0.21 (, if neurons support continuous variation that can) 303.18 474.23 P
(only be approximately matched.) 62.22 460.23 T
5 F
(2. Levels of syntax \050allowable forms of variation) 50.88 438.23 T
327.6 436.93 323.61 436.93 2 L
V
0.59 H
0 Z
N
1 F
(\051) 323.61 438.23 T
4 14 Q
(\245) 50.88 422.23 T
1 12 Q
0.17 (There can be dif) 62.22 422.23 P
0.17 (ferent levels of structure within a formalism. E.g. at one level the syntax of written) 140.76 422.23 P
0.3 0.06 (English is a set of short strokes. At another level there is a sequence of letters, punctuation marks) 62.22 408.23 B
0.3 0.27 (and spaces. At another level there are nouns, verbs, phrases, clauses sentences, etc. Discourse) 62.22 394.23 B
(grammars and story grammars encode attempts to characterise still higher levels of structure.) 62.22 380.23 T
4 14 Q
(\245) 50.88 364.23 T
1 12 Q
0.11 (Whether the variability supported at a particular syntactic level is signi\336cant for the control state or) 62.22 364.23 P
0.3 0.14 (is a mere implementation detail depends on whether that variability has a pragmatic or semantic) 62.22 350.23 B
-0.05 (role. For example in many contexts the thickness of strokes forming printed characters is irrelevant,) 62.22 336.23 P
0.3 0.38 (but in some contexts it can be used for emphasis, or to indicate switching between dif) 62.22 322.23 B
0.3 0.38 (ferent) 509.96 322.23 B
0.3 0.13 (notations with dif) 62.22 308.23 B
0.3 0.13 (ferent interpretations \050as in many printed manuals\051. Often 2-D layout of text is) 149.63 308.23 B
0.3 0.35 (used to help the reader parse what has been written, i.e. the layout has a pragmatic but not a) 62.22 294.23 B
(semantic role.) 62.22 280.23 T
4 14 Q
(\245) 50.88 264.23 T
1 12 Q
0.3 0.17 (Each syntactic level def) 62.22 264.23 B
0.3 0.17 (ines a set of possible structures and a sort of topology def) 180.82 264.23 B
0.3 0.17 (ined by which) 469.24 264.23 B
0.3 0.01 (transitions between permitted structures are minimal transitions. Each item has a set of neighbours) 62.22 250.23 B
0.3 0.42 (reachable by minimal transitions. Where there is continuous variation there are no minimal) 62.22 236.23 B
0.3 0.06 (transitions. In some cases more than one ordering of the same set of structures can be useful for a) 62.22 222.23 B
(particular application: for example there is an ordering of pairs of integers based on their sum, e.g.) 62.22 208.23 T
4 F
( \0501 1\051 \0501 2\051 \0502 1\051 \0502 2\051 \0501 3\051 \0503 1\051 \0502 3\051 \0503 2\051 \0501 4\051 \0504 1\051 \0503 3\051) 62.22 194.23 T
1 F
( etc) 407.94 194.23 T
(and a dif) 62.22 180.23 T
(ferent ordering based on their ratios, e.g.:) 103.96 180.23 T
4 F
( ... \0501 32\051 ... \0505 16\051 ... \0503 8\051 ... \0501 2\051 ... \0505 8\051 ... \05013 16\051 ... \05031 32\051 ...) 62.22 166.23 T
1 F
0.3 0.31 (the latter maps onto the natural ordering of the set of rational numbers and admits no nearest) 62.22 152.23 B
(neighbours.) 62.22 138.23 T
4 14 Q
(\245) 50.88 122.23 T
1 12 Q
0.3 0.26 (Whether a particular topology is useful will depend on the pragmatic and semantic roles to be) 62.22 122.23 B
(associated with the states.) 62.22 108.23 T
FMENDPAGE
%%EndPage: "24" 26
%%Page: "25" 26
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(25) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
4 14 Q
0 X
(\245) 50.88 774.23 T
1 12 Q
0.11 (An important feature of the lower level syntax is its generative power for producing new notational) 62.22 774.23 P
0.1 (elements \050e.g. words\051. Contrast the generative power of sequences of letters, and collections of 2-D) 62.22 760.23 P
-0.04 (vectors, or bitmaps. For a given level of generative power in a sequence of symbols there is a trade-) 62.22 746.23 P
0.12 (of) 62.22 732.23 P
0.12 (f between the number of primitive symbols available and the permitted sequence lengths. Human) 71.99 732.23 P
0.3 0.06 (languages might have used a two-letter alphabet but the words would have been much longer and) 62.22 718.23 B
(processing requirements on brains \050e.g. short-term memory requirements\051 quite dif) 62.22 704.23 T
(ferent.) 459.38 704.23 T
4 14 Q
(\245) 50.88 688.23 T
1 12 Q
0.3 0.45 (The syntax of a notation may be more or less closely related to the structures of the things) 62.22 688.23 B
0.3 0.01 (represented, e.g. 2-D pictures of 3-D scenes. It is important not to confuse structural relations with) 62.22 674.23 B
0.3 0.12 (isomorphism. Many people wrongly think that pictures are isomorphic with what they represent.) 62.22 660.23 B
(Y) 62.22 646.23 T
(et it is clear that a 2-D picture cannot be isomorphic with a 3-D object that it depicts.) 69.68 646.23 T
4 14 Q
(\245) 50.88 630.23 T
1 12 Q
-0.14 (Syntax is relative to mechanisms and procedures: it\325) 62.22 630.23 P
-0.14 (s not) 311.43 630.23 P
3 F
-0.14 (physical) 337.14 630.23 P
1 F
-0.14 ( structure that matters, but) 377.12 630.23 P
3 F
-0.14 (used) 505.02 630.23 P
1 F
-0.14 ( or) 527 630.23 P
3 F
(interpr) 62.22 616.23 T
(eted) 95.09 616.23 T
1 F
( structure.) 115.07 616.23 T
5 F
(3. Procedures for operating on and transforming syntactic structures.) 50.88 594.23 T
1 F
1.26 (The mechanisms that support the representing medium will provide a set of operations that can be) 50.88 574.23 P
(performed on states, in order to create or modify representing structures.) 50.88 560.23 T
4 14 Q
(\245) 50.88 544.23 T
1 12 Q
0.3 0.17 (Syntax depends on \324permitted\325 operations e.g. substitution, combination, translation, stretching,) 62.22 544.23 B
0.3 0.27 (rotation, insertion, searching, matching, etc. \050E.g. allowing superscripts changes the syntax of) 62.22 530.23 B
0.3 0.49 (numeric expressions.\051 Operations can be discrete or continuous, quantitative or structural,) 62.22 516.23 B
(destructive or constructive, etc. E.g. conventional vs digital thermostats.) 62.22 502.23 T
4 14 Q
(\245) 50.88 486.23 T
1 12 Q
0.3 0.21 (Procedures are used for creating instances, storing instances, searching for instances, matching) 62.22 486.23 B
0.3 0.31 (instances, transforming instances, describing instances \050e.g. building parse trees\051 interpreting) 62.22 472.23 B
0.23 (instances. Syntactic and semantic operations may overlap. E.g. \336nding edges in images is syntactic) 62.22 458.23 P
0.15 (insofar as it involves analysis of image structures and semantic insofar as it is part of a process that) 62.22 444.23 P
(maps pixel arrays onto con\336gurations of \050possibly continuous\051 curves.) 62.22 430.23 T
4 14 Q
(\245) 50.88 414.23 T
1 12 Q
0.3 0.1 (Signif) 62.22 414.23 B
0.3 0.1 (icant syntax need not map onto underlying physical structure. E.g. as previously remarked,) 91.49 414.23 B
0.3 0.14 (sparse array can have) 62.22 400.23 B
186.22 398.93 172.48 398.93 2 L
V
0.59 H
0 Z
N
0.3 0.14 (far) 172.48 400.23 B
0.3 0.14 ( more components than the transistors implementing it. Compare lazily) 186.22 400.23 B
(evaluated in\336nite datastructures and dynamic lists in Pop2 and Pop1) 62.22 386.23 T
(1.) 389.24 386.23 T
4 14 Q
(\245) 50.88 370.23 T
1 12 Q
0.3 0.14 (Some mechanisms support only local operations, e.g. use of a pencil on paper or most computer) 62.22 370.23 B
0.2 (operations on datastructures. In those cases global operations \050e.g. changing the stroke thickness of) 62.22 356.23 P
0.3 0.11 (all the text in a particular page, or replacing all occurrences of one symbol with another\051 may be) 62.22 342.23 B
0.3 (relatively cumbersome and time-consuming. Some control systems \050brains?\051 may need both global) 62.22 328.23 P
0.3 0.14 (changes and intricate local changes to be made simultaneously) 62.22 314.23 B
0.3 0.14 (, e.g. a pianist smoothly changing) 372.61 314.23 B
(speed and dynamics at the same time as playing many notes.) 62.22 300.23 T
5 F
(4. Semantics and meta-semantics.) 50.88 278.23 T
1 F
0.65 (By \322meta-semantics\323 I mean a set of principles for using notations to refer to, denote, express, store) 50.88 258.23 P
0.94 (information about, various things. Reasoning based on such principles could justify the selection of) 50.88 244.23 P
(particular formalisms for particular tasks.) 50.88 230.23 T
4 14 Q
(\245) 50.88 213.23 T
1 12 Q
-0.17 (An arbitrary formula in predicate calculus, such as \322P\050a,b,c\051\323 does not have a semantics, but there is) 62.22 213.23 P
0.3 0.34 (a meta-semantics which specif) 62.22 199.23 B
0.3 0.34 (ies the sorts of things the symbols can refer to, i.e. properties,) 219.34 199.23 B
(relationships, individual objects, etc.) 62.22 185.23 T
4 14 Q
(\245) 50.88 169.23 T
1 12 Q
0.3 0.15 (Similarly) 62.22 169.23 B
0.3 0.15 (, the meta-semantics of 2-D line drawings requires syntactic relations in the pictures to) 107.4 169.23 B
0.3 0.28 (express some geometrical or topological relationships between things depicted. But the meta-) 62.22 155.23 B
0.3 0.21 (semantics does not require isomorphism: 2-D can represent 3-D. Moreover) 62.22 141.23 B
0.3 0.21 (, in pictures like the) 439.25 141.23 B
0.3 0.25 (following, there need not be any unique decomposition of either picture of thing depicted into) 62.22 127.23 B
FMENDPAGE
%%EndPage: "25" 27
%%Page: "26" 27
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(26) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.14 (corresponding parts, whereas most logical notations require a unique decomposition of the relevant) 62.22 774.23 P
(bit of the world.) 62.22 760.23 T
4 14 Q
(\245) 50.88 664.17 T
1 12 Q
0.3 0.25 (Some syntax/semantics relations are general purpose \050e.g. logic\051 some ad-hoc \050e.g. using digit) 62.22 664.17 B
0.3 0.04 (concatenation in arabic numerals to mean \322multiply by 10 and add\323\051. Musical notations and many) 62.22 650.17 B
(maps use a mixture of arbitrary and principled relationships.) 62.22 636.17 T
4 14 Q
(\245) 50.88 620.17 T
1 12 Q
0.3 0.43 (Choice of syntax and semantics is determined not only by what is represented, but how the) 62.22 620.17 B
0.09 (information is to be used. \050Both maps and lists of towns with their coordinates are useful in atlases,) 62.22 606.17 P
(though they are used for dif) 62.22 592.17 T
(ferent purposes.\051) 194.9 592.17 T
301.23 559.35 50.88 559.35 2 L
V
1.02 H
0 Z
N
2 14 Q
(\05018\051 Dimensions of information states) 50.88 560.83 T
1 12 Q
1.86 (W) 50.88 541.17 P
1.86 (e can now provisionally characterise information states in a behaving system in many dif) 61.24 541.17 P
1.86 (ferent) 511.89 541.17 P
(dimensions, including the following:) 50.88 527.17 T
4 14 Q
(\245) 50.88 510.17 T
1 12 Q
(Their structure or syntax, which depends on the kinds of variation they can support, such as:) 62.22 510.17 T
(- continuous or discrete variability) 72.14 494.17 T
(,) 236.58 494.17 T
(- one or many dimensions of variation,) 72.14 478.17 T
-0.23 (- whether the variation is structural, allowing changes in complexity) 72.14 462.17 P
-0.23 (, or simply variation in a \336xed) 396.35 462.17 P
(number of linear dimensions,) 79.51 448.17 T
(- whether the complexity of syntax has an upper bound or not,) 72.14 432.17 T
0.96 (- whether the space of possibilities is homogeneous \050e.g. the space of N dimensional vectors is) 72.14 416.17 P
-0.24 (homogeneous, whereas the space of unbounded binary trees is not, since at some locations in the) 79.51 402.17 P
0.34 (latter space there are far more neighbours, that is trees that can be accessed by a single change,) 79.51 388.17 P
(than at other locations in the space.\051) 79.51 374.17 T
(- whether the Fregean function/ar) 72.14 358.17 T
(gument syntactic form is supported) 232.45 358.17 T
(- whether higher order function forms are supported \050e.g. quanti\336ers, lambda operators\051,) 72.14 342.17 T
(- etc.) 72.14 326.17 T
4 14 Q
(\245) 50.88 310.17 T
1 12 Q
(What functional role they ful\336l:) 62.22 310.17 T
(- recording input \050at various levels of abstraction\051) 72.14 294.17 T
(- controlling behaviour \050e.g. representing goals, plans, control-signals\051) 72.14 278.17 T
(- whether the control function is direct or indirect, short term or long term, etc.) 72.14 262.17 T
(- long term vs short term information stores) 72.14 246.17 T
(- high level vs low level control of behaviour) 72.14 230.17 T
(, etc.) 288.17 230.17 T
(- storing general vs speci\336c \324facts\325 about some domain) 72.14 214.17 T
(- representing suppositions or possibilities as opposed to facts) 72.14 198.17 T
(- performing inferences) 72.14 182.17 T
(- expressing preferences, goals, instructions, plans, partial plans, etc.) 72.14 166.17 T
4 14 Q
(\245) 50.88 150.17 T
1 12 Q
(Among the functions required in intelligent agents are:) 62.22 150.17 T
0.03 (-) 72.14 134.17 P
132.45 132.86 79.16 132.86 2 L
V
0.59 H
N
0.03 (Desire-like) 79.16 134.17 P
0.03 ( control states, which tend to initiate changes \050under certain conditions\051. These could) 132.45 134.17 P
0.62 (include include a wide variety of types of motivational states, including general principles and) 79.51 120.17 P
(ideals, as well as speci\336c desires and goals.) 79.51 106.17 T
50.88 85.75 539.85 782.23 C
141.56 676.17 449.17 756.23 C
141.56 676.17 449.17 756.23 R
7 X
0 K
V
163.15 685.37 209.22 731.43 R
V
3 H
2 Z
0 X
N
187.96 703.08 234.02 749.15 R
N
164.6 732.47 185.85 749.35 2 L
N
165.23 686.85 187.96 703.08 2 L
N
209.22 731.43 232.1 748.1 2 L
N
208.01 685.87 232.45 702.54 2 L
N
252.73 679.76 415.23 748.1 R
7 X
V
5 10 Q
0 X
0.44 (The number of line segments and) 252.73 741.43 P
5.82 (junctions seen her) 252.73 729.43 P
5.82 (e will dif) 351.84 729.43 P
5.82 (fer) 402.84 729.43 P
0.08 (accor) 252.73 717.43 P
0.08 (ding to whether it is seen as) 281.63 717.43 P
0.63 (a 2-D patter) 252.73 705.43 P
0.63 (n or as a pictur) 311.8 705.43 P
0.63 (e of a) 386.19 705.43 P
(wir) 252.73 693.43 T
(e-frame cube.) 266.24 693.43 T
145.23 680.26 445.23 751.93 R
0.5 H
N
50.88 85.75 539.85 782.23 C
-2.27 32.1 609.73 824.1 C
FMENDPAGE
%%EndPage: "26" 28
%%Page: "27" 28
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(27) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.34 (-) 72.14 774.23 P
130.76 772.93 79.47 772.93 2 L
V
0.59 H
0 Z
N
0.34 (Belief-like) 79.47 774.23 P
0.34 ( control states, which tend to be changed under \324external\325 in\337uences, and which also) 130.76 774.23 P
1.88 (determine how desire-like states produce changes. This would include general beliefs about) 79.51 760.23 P
1.48 (classes of objects and events as well as speci\336c beliefs, e.g. about a particular property of a) 79.51 746.23 P
(particular object.) 79.51 732.23 T
0.41 (-) 72.14 716.23 P
158.84 714.93 79.54 714.93 2 L
V
N
0.41 (Supposition-like) 79.54 716.23 P
0.41 ( control states, i.e. states that are merely used to represent possibilities without) 158.84 716.23 P
0.14 (any commitment to their being actual \050such states are important in the conditions of conditional) 79.51 702.23 P
0.12 (instructions or rules, and in the process of creating plans by exploring possible consequences of) 79.51 688.23 P
(dif) 79.51 674.23 T
(ferent actions\051.) 92.62 674.23 T
0.46 (-) 72.14 658.23 P
122.89 656.93 79.58 656.93 2 L
V
N
0.46 (Plan-like) 79.58 658.23 P
0.46 ( control states, i.e. states that store information about steps required to achieve certain) 122.89 658.23 P
0.46 (ends, or types of ends. These could include very detailed stored plans \050like the pianist who has) 79.51 644.23 P
0.14 (memorised a Beethoven sonata\051 or very abstract schematic plans whose details need to be \336lled) 79.51 630.23 P
(in during execution.) 79.51 616.23 T
2.44 (-) 72.14 600.23 P
171.63 598.93 81.57 598.93 2 L
V
N
2.44 (Association stores) 81.57 600.23 P
2.44 (, i.e. stored mappings of associations between patterns of various kinds.) 171.63 600.23 P
0.11 (Neural nets are one time of mechanism that can store associations though there are many others) 79.51 586.23 P
1 (that have been explored in computer science and AI, including simple linear association lists,) 79.51 572.23 P
1.21 (indexed association tables to reduce the need for searching, including hash-coded association) 79.51 558.23 P
(tables.) 79.51 544.23 T
0.81 (-) 72.14 528.23 P
111.26 526.93 79.94 526.93 2 L
V
N
0.81 (Re\337ex) 79.94 528.23 P
0.81 ( control states, i.e. states that get triggered by detection of some pattern in sensory data) 111.26 528.23 P
-0.26 (and then automatically send signals to motors to perform some action. The mechanisms required) 79.51 514.23 P
0.27 (for such control states must be able to store associations between patterns. It seems very likely) 79.51 500.23 P
-0.12 (that some internal processes also require the use of innate or learned re\337exes:) 79.51 486.23 P
536.85 484.93 452.03 484.93 2 L
V
N
-0.12 (cognitive re\337exes) 452.03 486.23 P
-0.12 (.) 536.85 486.23 P
(\050A deeper analysis would produce a more detailed taxonomy of causal roles.\051) 72.14 470.23 T
4 14 Q
(\245) 50.88 454.23 T
1 12 Q
(Whether they have a semantic function and if so what sort) 62.22 454.23 T
(- whether there\325) 72.14 438.23 T
(s a generative semantic relation or a \336xed set of mappings) 148.07 438.23 T
0.03 (- what kind of mapping is used, e.g. applicative semantics, analogical, systematic, ad-hoc, mixed,) 72.14 422.23 P
(etc.) 79.51 408.23 T
(- the ontology of the domain represented) 72.14 392.23 T
3.09 (- whether only particular positive facts or also general, negative, disjunctive, implicational) 72.14 376.23 P
(propositions can be expressed,) 79.51 362.23 T
(- whether nonexistent objects and states of af) 72.14 346.23 T
(fairs can be represented,) 288.09 346.23 T
3.98 (- in the case of procedural semantics, which kinds of procedures can be speci\336ed \050e.g.) 72.14 330.23 P
(conditionals, loops, recursion, etc.\051) 79.51 316.23 T
4 14 Q
(\245) 50.88 299.23 T
1 12 Q
0.3 0.22 (What kinds of processes they support, e.g. creating representations, changing them, comparing) 62.22 299.23 B
0.3 0.12 (them, storing them, searching for them in various ways, transforming them \050inference\051, etc. This) 62.22 285.23 B
(includes such things as:) 62.22 271.23 T
0.45 (- How they are created \050contrast: perception, reasoning, motive generation, planning, \336ne tuning) 72.14 255.23 P
(in motor control, outputs of neural nets, short term memory in a production system, etc.\051) 79.51 241.23 T
0.62 (- How they interact with other states \050changing them, modifying their in\337uences, deleting them,) 72.14 225.23 P
(etc.\051) 79.51 211.23 T
4 14 Q
(\245) 50.88 194.23 T
1 12 Q
0.3 0.03 (Whether dif) 62.22 194.23 B
0.3 0.03 (ferent items of information are separately represented or superimposed and distributed) 120.21 194.23 B
-0.13 (over many implementation units. This distinction covers a variety of dif) 62.22 180.23 P
-0.13 (ferent cases. For example in) 405.48 180.23 P
0.3 0.05 (a logical database all the theorems are distributed over the axioms needed for their derivation in a) 62.22 166.23 B
0.01 (manner that is dif) 62.22 152.23 P
0.01 (ferent from the distribution of information over weights in a neural net. In the two) 146.31 152.23 P
0.3 0.19 (co-ordinates of a chess board location,) 62.22 138.23 B
4 F
0.3 0.19 (rank) 259.34 138.23 B
1 F
0.3 0.19 ( and) 284.78 138.23 B
4 F
0.3 0.19 (f) 309.67 138.23 B
0.3 0.19 (ile) 313.2 138.23 B
1 F
0.3 0.19 ( there is also information distributed about) 325.78 138.23 B
0.3 0.04 (which two diagonals the location lies on, one computed as) 62.22 124.23 B
4 F
0.3 0.04 (rank+f) 350.61 124.23 B
0.3 0.04 (ile) 385.71 124.23 B
1 F
0.3 0.04 ( the other as) 397.83 124.23 B
4 F
0.3 0.04 (rank-f) 460.94 124.23 B
0.3 0.04 (ile) 493.19 124.23 B
1 F
0.3 0.04 (. There) 505.32 124.23 B
(are many similar forms of distributed representation in conventional computing systems.) 62.22 110.23 T
4 14 Q
(\245) 50.88 94.23 T
1 12 Q
(How all this is achieved) 62.22 94.23 T
FMENDPAGE
%%EndPage: "27" 29
%%Page: "28" 29
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(28) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
(- what mechanisms are used to implement the relevant virtual machine) 72.14 774.23 T
(- how semantic mappings are set up) 72.14 758.23 T
(- which algorithms are used in the virtual machines) 72.14 742.23 T
4 14 Q
(\245) 50.88 725.23 T
1 12 Q
0.3 0.12 (Other dimensions correspond to engineering requirements \050e.g. speed, robustness, completeness,) 62.22 725.23 B
0.3 0.03 (soundness of inferences, etc.\051: usually not all can be met simultaneously) 62.22 711.23 B
0.3 0.03 (, so that trade-of) 412.67 711.23 B
0.3 0.03 (fs have to) 492.4 711.23 B
(be assessed by designers.) 62.22 697.23 T
140.36 675.93 72.14 675.93 2 L
V
0.59 H
0 Z
N
0.3 0.43 (Accessibility) 72.14 677.23 B
0.3 0.43 ( of information is important in intelligent systems. If nothing can access the) 140.36 677.23 B
0.17 (information represented in a sub-state, it might as well not be there. Thus every object potentially has) 50.88 663.23 P
0.3 0.09 (complete information about itself embedded in itself: but usually most of that information is totally) 50.88 649.23 B
0.3 0.01 (inaccessible to the system: there is no part of the system that can use the state information to vary its) 50.88 635.23 B
-0.06 (own behaviour) 50.88 621.23 P
-0.06 (. Contrast the way in which a CPU in a computer can access the states of bit-patterns in) 121.77 621.23 P
0.3 0.34 (the virtual memory) 50.88 607.23 B
0.3 0.34 (, or the way in which a system containing a neural net can implicitly access) 148.76 607.23 B
0.3 0.14 (information stored in the weights in the connections by feeding in a vector and seeing what vector) 50.88 593.23 B
(comes out.) 50.88 579.23 T
0.3 0.03 (Any physical state can in principle be treated as an information store, e.g. a store of information) 72.14 559.23 B
0.3 0.17 (about the system itself \050many animals use the environment as a representation of itself - e.g. trail-) 50.88 545.23 B
0.3 0.22 (blazing as an alternative to memorising a route\051. But often the information will not be) 50.88 531.23 B
540.07 529.93 489.24 529.93 2 L
V
N
0.3 0.22 (accessible) 489.24 531.23 B
0.3 0.12 (unless there are specif) 50.88 517.23 B
0.3 0.12 (ic mechanisms that enable interaction with the structure. \050There are dif) 160.19 517.23 B
0.3 0.12 (ferent) 511.27 517.23 B
0.3 0.52 (kinds of access: updating, interrogating, changing, obeying, comparing, etc.\051 Access can be) 50.88 503.23 B
0.3 0.04 (continuous or discrete, high-bandwidth or low-bandwidth, serial or parallel, etc. Access to computer) 50.88 489.23 B
-0.19 (memories is discrete, \050relatively\051 low bandwidth, parallel \050insofar as several bits of information can be) 50.88 475.23 P
(retrieved simultaneously\051.) 50.88 461.23 T
130.99 439.93 72.14 439.93 2 L
V
N
0.3 0.02 (Explicitness) 72.14 441.23 B
3 F
0.3 0.02 (vs) 134.31 441.23 B
205.84 439.93 147.65 439.93 2 L
V
N
1 F
0.3 0.02 (implicitness) 147.65 441.23 B
0.3 0.02 (. This is a supposed distinction that can easily cause great confusion,) 205.84 441.23 B
0.3 0.3 (since what is clearly explicit in a virtual machine may be implicit in a lower level medium. For) 50.88 427.23 B
0.14 (example for a system that treats sparse arrays just like ordinary arrays the contents of the sparse array) 50.88 413.23 P
0.3 0.11 (are explicit: they can be examined or changed in exactly the same way as the contents of any other) 50.88 399.23 B
0.3 0.06 (array) 50.88 385.23 B
0.3 0.06 (. At the implementation level there is a dif) 75.05 385.23 B
0.3 0.06 (ference insofar as most of the locations in the sparse) 282.94 385.23 B
0.29 (array are not explicitly represented, and their values are computed as required. In this case the access) 50.88 371.23 P
-0.28 (time is bounded and the dif) 50.88 357.23 P
-0.28 (ference in access times between sparse and ordinary arrays may be so small) 180.18 357.23 P
0.08 (as not to be noticeable. If a logical database is used some of whose contents are explicitly represented) 50.88 343.23 P
-0.24 (while others are derived on demand things are more complex. If all the items accessed over a period of) 50.88 329.23 P
0.27 (time are so quickly derivable that it is dif) 50.88 315.23 P
0.27 (\336cult to distinguish the database from one in which they are) 250.02 315.23 P
0.3 0.34 (all pre-stored, then at the level of a machine that uses the database there need be no functional) 50.88 301.23 B
-0.19 (distinction between the stored and the derived items. If the amount of search required varies then there) 50.88 287.23 P
-0.17 (can be gradual degradation as harder and harder questions are asked of the database. Here, instead of a) 50.88 273.23 P
0.3 0.26 (sharp functional distinction between what is explicit and what is implicit we f) 50.88 259.23 B
0.3 0.26 (ind a dif) 446.88 259.23 B
0.3 0.26 (ference in) 489.59 259.23 B
-0.03 (degree of accessibility) 50.88 245.23 P
-0.03 (. However this can be something that varies dynamically) 157.28 245.23 P
-0.03 (, for instance if derived) 428.4 245.23 P
-0.26 (items are explicitly stored in a cache for a time: then access times will vary depending on how recently) 50.88 231.23 P
-0.02 (an item was last requested. If the cache stores not only items explicitly sought but \322neighbours\323 based) 50.88 217.23 P
0.3 0.07 (on some heuristic neighbourhood metric, and if the cache includes a gradual decay mechanism then) 50.88 203.23 B
-0.22 (explicit/implicit distinction is even less relevant. A full overview of useful forms of representation and) 50.88 189.23 P
0.3 0.58 (implementation mechanisms would show that there is far more variety than common-sense) 50.88 175.23 B
0.3 0.05 (distinctions can cope with. Unfortunately many people who discuss representations, including many) 50.88 161.23 B
(philosophers and psychologists, have not been exposed to this variety) 50.88 147.23 T
(.) 383.89 147.23 T
FMENDPAGE
%%EndPage: "28" 30
%%Page: "29" 30
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(29) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
457.55 771.42 50.88 771.42 2 L
0 X
V
1.02 H
0 Z
N
2 14 Q
(\05019\051 Ontological, epistemological and heuristic requirements.) 50.88 772.9 T
1 12 Q
2.98 (Dif) 50.88 753.23 P
2.98 (ferent kinds of representations have a syntax and semantics that implicitly assume dif) 66.65 753.23 P
2.98 (ferent) 511.88 753.23 P
1.1 (ontologies. E.g. predicate calculus presupposes an ontological carving up of the world into objects,) 50.88 739.23 P
0.12 (properties, relations, etc. plus a pair of truth values. Whether a pictorial notation does this depends on) 50.88 725.23 P
0.23 (the detailed form and use of the notation. For example, use of a 2-D array each location of which can) 50.88 711.23 P
-0.17 (contain only a list of names of objects \050as in Glasgow 1993\051 presupposes a dif) 50.88 697.23 P
-0.17 (ferent ontology from the) 422.76 697.23 P
-0.18 (use of an array of pixels representing a sampling of intensity values in the optic array from a particular) 50.88 683.23 P
2.27 (viewpoint \050Sloman, 1989\051. Many maps and atlases use a mixed ontology) 50.88 669.23 P
2.27 (, with names and other) 421.87 669.23 P
0.46 (symbols representing distinct objects superimposed on analogical depictions of continuously varying) 50.88 655.23 P
(contours, coast-lines, rivers, roads, etc.) 50.88 641.23 T
0.3 0.18 (As McCarthy and Hayes pointed out in 1969, there are several dif) 72.14 621.23 B
0.3 0.18 (ferent questions that can be) 402.54 621.23 B
0.3 0.12 (asked in relation to a representational formalism and the world it is used to represent. For example) 50.88 607.23 B
(there are questions about:) 50.88 593.23 T
4 14 Q
(\245) 50.88 577.23 T
1 12 Q
0.3 0.35 (Metaphysical \050or ontological\051 scope/adequacy relative to a world: can the notation represent) 62.22 577.23 B
(everything that can exist in the world?) 62.22 563.23 T
4 14 Q
(\245) 50.88 547.23 T
1 12 Q
0.3 0.23 (Epistemological scope/adequacy relative to an agent: can the notation represent everything the) 62.22 547.23 B
0.3 0.14 (agent needs to know about the world, or everything the agent needs to be able to suppose or ask) 62.22 533.23 B
(about the world?) 62.22 519.23 T
4 14 Q
(\245) 50.88 503.23 T
1 12 Q
0.3 0.21 (Heuristic scope/adequacy relative to an agent, a set of purposes, in a particular world: does the) 62.22 503.23 B
0.3 0 (notation \050with its associated mechanisms\051 allow the agent to solve problems in order to fulf) 62.22 489.23 B
0.3 0 (il those) 504.54 489.23 B
(purposes and do so in a reasonable time, with reasonable accuracy?) 62.22 475.23 T
(In addition to their questions we can ask others from the standpoint of biology or cognitive science:) 50.88 455.23 T
4 14 Q
(\245) 50.88 439.23 T
1 12 Q
(Could the notation have evolved from earlier forms?) 62.22 439.23 T
4 14 Q
(\245) 50.88 423.23 T
1 12 Q
0.3 0.29 (Could it be extended as required to cope with new developments in the environment and new) 62.22 423.23 B
(purposes?) 62.22 409.23 T
-0.2 (These can be broken down into further sub-questions. For example questions about heuristic adequacy) 50.88 389.23 P
(can break down into questions concerning the following:) 50.88 375.23 T
4 14 Q
(\245) 50.88 359.23 T
1 12 Q
-0 (Learnability: could the notation be learnt? Does it support learning of information that it expresses?) 62.22 359.23 P
4 14 Q
(\245) 50.88 343.23 T
1 12 Q
(Storage, searching, matching: e.g. what are the space or time costs and trade-of) 62.22 343.23 T
(fs) 441.7 343.23 T
4 14 Q
(\245) 50.88 326.23 T
1 12 Q
0.3 0.01 (Size of search spaces: does the notation allow irrelevant nodes to be generated in searching so that) 62.22 326.23 B
0.3 0.22 (they have to be explicitly rejected, or does it inherently constrain searching to what is relevant) 62.22 312.23 B
(\050Sloman 1971\051.) 62.22 298.23 T
4 14 Q
(\245) 50.88 281.23 T
1 12 Q
0.3 0.01 (Expressive power relative to solving new problems. E.g. does the notation make it easy to match a) 62.22 281.23 B
0.19 (new problem against stored information about previously solved problems? Does it make it easy to) 62.22 267.23 P
0.3 0.04 (explore dif) 62.22 253.23 B
0.3 0.04 (ferent representations of the same problem to f) 115.03 253.23 B
0.3 0.04 (ind one that facilitates solution? \050\322Easy\323) 342.51 253.23 B
0.3 0.33 (may itself be relative to a mechanism, as previously explained, rather than being an intrinsic) 62.22 239.23 B
(property of the formalism.\051) 62.22 225.23 T
4 14 Q
(\245) 50.88 209.23 T
1 12 Q
0.3 0.05 (Sensitivity to changes in the \324world\325: can f) 62.22 209.23 B
0.3 0.05 (ine dif) 270.11 209.23 B
0.3 0.05 (ferences between objects, or between viewpoints) 301.53 209.23 B
0.3 0.05 (be expressed? \050E.g. without this kind of sensitivity in visual images, binocular disparity could not) 62.22 195.23 B
(be a cue to depth.\051) 62.22 181.23 T
4 14 Q
(\245) 50.88 165.23 T
1 12 Q
0.14 (Stability: does the representation remain invariant when the context or object depicted changes in a) 62.22 165.23 P
0.3 0.17 (way that is irrelevant to current purposes \050e.g. an object centred coordinate frame can provide a) 62.22 151.23 B
0.3 0.05 (representation that is invariant with respect to viewing direction and distance\051. \050Note that stability) 62.22 137.23 B
(and sensitivity can be in opposition.\051) 62.22 123.23 T
FMENDPAGE
%%EndPage: "29" 31
%%Page: "30" 31
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(30) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
4 14 Q
0 X
(\245) 50.88 774.23 T
1 12 Q
0.3 0.09 (Ease of construction, and ease of derivation from originally available data. E.g. enhanced images) 62.22 774.23 B
0.3 0.36 (are easy to derive from visual data by f) 62.22 760.23 B
0.3 0.36 (iltering processes, whereas descriptions of 3-D scene) 265.39 760.23 B
(structure can be very hard to derive.) 62.22 746.23 T
4 14 Q
(\245) 50.88 730.23 T
1 12 Q
0.3 0.4 (Flexibility and multiplicity of use. E.g. Marr) 62.22 730.23 B
0.3 0.4 (\325) 296.58 730.23 B
0.3 0.4 (s hierarchical representations of 3-D structure) 300.31 730.23 B
0.3 0.18 (simplif) 62.22 716.23 B
0.3 0.18 (ied the task of recognition, made it hard to represent or compute some potentially useful) 96.83 716.23 B
(relationships, such as that a person\325) 62.22 702.23 T
(s \336nger tip was on his nose.) 232.1 702.23 T
4 14 Q
(\245) 50.88 686.23 T
1 12 Q
(Ability to be tailored to requirements of particular applications) 62.22 686.23 T
4 14 Q
(\245) 50.88 670.23 T
1 12 Q
(Use as a substratum for other \324virtual machines\325) 62.22 670.23 T
4 14 Q
(\245) 50.88 654.23 T
1 12 Q
0.3 0.14 (Extendability: this depends on whether the formalism uses some medium that has more richness) 62.22 654.23 B
0.3 0.25 (than the formalism already exploits. For example, English does not use all the sets of possible) 62.22 640.23 B
(character sequences corresponding to normal word lengths. E.g. \322glutterzank\323 remains unused.) 62.22 626.23 T
4 14 Q
(\245) 50.88 610.23 T
1 12 Q
0.3 0.01 (Use for transmitting information This involves such things as bandwidth required, ease of parsing/) 62.22 610.23 B
0.3 0.57 (decoding/interpreting by the receiver) 62.22 596.23 B
0.3 0.57 (, ability to make use of shared knowledge, etc. \050But) 261.11 596.23 B
(communication is not the sole or primary function of presentation.\051) 62.22 582.23 T
4 14 Q
(\245) 50.88 566.23 T
1 12 Q
(Robustness E.g. do minor deformations during storage or transmission lead to undetectable errors.) 62.22 566.23 T
(and no doubt many more!) 50.88 546.23 T
0.09 (Contrary to how such issues are sometimes discussed, these properties may not be inherent in the) 72.14 526.23 P
0.3 0.01 (syntax of a formalism, but may be relative both to a class of problems and also to an implementation) 50.88 512.23 B
(mechanism.) 50.88 498.23 T
0.3 0.25 (It is important to note that features that may be nice from a mathematical point of view) 72.14 478.23 B
0.3 0.25 (, e.g.) 515 478.23 B
0.07 (economy of basic symbols and syntactic simplicity or elegance, may be undesirable from other points) 50.88 464.23 P
0.3 0.13 (of view) 50.88 450.23 B
0.3 0.13 (. Features that are desirable from the point of view of a designer \050at compile time\051 may not) 87.63 450.23 B
0.3 0.27 (necessarily be desirable at run time. For example, economy of primitives leads to complexity of) 50.88 436.23 B
0.3 0.29 (representations, complexity of matching and complexity of searching for solutions to problems.) 50.88 422.23 B
0.3 0.45 (\050Compare building an aeroplane out of general-purpose Lego and building one out of a Lego) 50.88 408.23 B
0.3 0.03 (aeroplane kit.\051 Syntactic simplicity of a notation \050as in Lisp\051 can imply considerable local ambiguity) 50.88 394.23 B
0.2 (that needs global context for disambiguation, and therefore make the parsing and interpreting process) 50.88 380.23 P
0.16 (dif) 50.88 366.23 P
0.16 (\336cult for an agent with limited short term memory) 63.99 366.23 P
0.16 (. This is no problem for computing systems with) 305.69 366.23 P
(a very lar) 50.88 352.23 T
(ge stack.) 95.95 352.23 T
79.65 319.42 50.88 319.42 2 L
V
1.02 H
0 Z
N
2 14 Q
(\05020\051) 50.88 320.9 T
362.65 319.42 79.65 319.42 2 L
V
N
(What sort of underlying engine is needed?) 79.65 320.9 T
1 12 Q
0.76 (Much of this is neutral as to what) 50.88 301.23 P
279.49 299.93 220.2 299.93 2 L
V
0.59 H
N
0.76 (mechanisms) 220.2 301.23 P
0.76 ( are used to implement the various kinds of substates) 279.49 301.23 P
-0.27 (and causal linkages. They might be neural mechanisms or some other kind. As in circuit design, global) 50.88 287.23 P
-0.28 (properties of the architecture are more important than which particular mechanisms are used:) 50.88 273.23 P
539.85 271.93 496.5 271.93 2 L
V
N
-0.28 (when the) 496.5 273.23 P
299.67 257.93 50.88 257.93 2 L
V
N
2.26 (design is right architecture dominates mechanism) 50.88 259.23 P
2.26 (. The detailed mechanisms make only mar) 299.67 259.23 P
2.26 (ginal) 515.87 259.23 P
(dif) 50.88 245.23 T
(ferences as long as they provide the following:) 63.99 245.23 T
4 14 Q
(\245) 50.88 228.23 T
1 12 Q
0.3 0.22 (suf) 62.22 228.23 B
0.3 0.22 (f) 77.33 228.23 B
0.3 0.22 (icient structural variability to meet task requirements \050e.g. a representational system that is) 80.88 228.23 B
-0.03 (incapable of taking on more than N distinct states cannot distinguish more than N dif) 62.22 214.23 P
-0.03 (ferent states of) 469.31 214.23 P
(the environment\051.) 62.22 200.23 T
4 14 Q
(\245) 50.88 183.23 T
1 12 Q
(suf) 62.22 183.23 T
(\336cient architectural richness in the notation and the mechanisms using it) 76.66 183.23 T
(- number of independently variable components) 62.22 169.23 T
(- functional dif) 62.22 155.23 T
(ferentiation of components) 133.95 155.23 T
(- variety of causal linkages) 62.22 141.23 T
4 14 Q
(\245) 50.88 124.23 T
1 12 Q
(suf) 62.22 124.23 T
(\336cient speed of operation) 76.66 124.23 T
-0.27 (\322V) 50.88 104.23 P
-0.27 (irtual\323 machines in computers seem to have many of these features for a wide range of tasks. This is) 64.14 104.23 P
(one of the reasons why general purpose computers are used so widely) 50.88 90.23 T
(.) 384.83 90.23 T
FMENDPAGE
%%EndPage: "30" 32
%%Page: "31" 32
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(31) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
0.09 (But we don\325) 50.88 774.23 P
0.09 (t yet know enough about requirements, nor about available mechanisms, to be able to say) 110.15 774.23 P
0.64 (with any con\336dence which infrastructure could and which could not work for implementing human-) 50.88 760.23 P
1.25 (like systems. E.g. it could turn out that, in our universe, only a mixture of electrical pathways and) 50.88 746.23 P
0.8 (chemical soup could provide the right combination of \336ne-grained control, structural variability and) 50.88 732.23 P
-0.22 (global control to support human-like intelligence. If that were so, we might not be able to base human-) 50.88 718.23 P
0.31 (like intelligence on systems built from computers as we currently know them. Of course if additional) 50.88 704.23 P
1.31 (mechanisms are required they will be added to computers of the future, and that will change what) 50.88 690.23 P
2.03 (computers can and cannot do! \050Such changes may not all preserve the mathematical equivalence) 50.88 676.23 P
(between computers and T) 50.88 662.23 T
(uring Machines. Compare Penrose 1989.\051) 174.03 662.23 T
0.3 0.06 (W) 72.14 642.23 B
0.3 0.06 (e need new thinking tools to help us grasp all this complexity) 82.56 642.23 B
0.3 0.06 (. AI has provided many detailed) 382.76 642.23 B
0.11 (concepts for thinking about and modelling processes in perception, planning, reasoning, learning. But) 50.88 628.23 P
0.3 0.07 (we still lack more \322global ideas\323 to help us think about the architecture of the whole mind: how the) 50.88 614.23 B
(bits \336t together) 50.88 600.23 T
(, and this paper is intended as a small contribution to \336lling that need.) 123.03 600.23 T
0.3 0.3 (Our key notion that the mind is a very complex control system, whose functioning depends) 72.14 580.23 B
0.3 0.24 (crucially on the creation and manipulation of information-rich states, using a variety of forms of) 50.88 566.23 B
-0.02 (representation requires us to generalise standard notions of control systems and their states, as already) 50.88 552.23 P
0.3 0.11 (mentioned. In particular) 50.88 538.23 B
0.3 0.11 (, it is important to allow not only \322atomic\323 states in a phase space, but also) 169.11 538.23 B
0.3 0.05 (\322molecular\323 states which have their own internal, changing, structure. \050For more on this see Sloman) 50.88 524.23 B
0.3 0.17 (1993b.\051 The latter seems to be needed for understanding intelligence. A full theory would need to) 50.88 510.23 B
0.3 0.35 (include a more detailed survey of types of uses of information-rich control states in intelligent) 50.88 496.23 B
(systems.) 50.88 482.23 T
0.3 0.02 (A full survey would require an analysis of the types of architectures available and the functional) 72.14 462.23 B
0.3 0.03 (roles they allocate to dif) 50.88 448.23 B
0.3 0.03 (ferent kinds of components and their states. Although there is a rich body of) 168.57 448.23 B
0.3 0.1 (mathematics of control systems, most of it appears to be incapable of coping with the requirements) 50.88 434.23 B
0.3 0.01 (listed here. For example, no f) 50.88 420.23 B
0.3 0.01 (ixed set of dif) 192.92 420.23 B
0.3 0.01 (ferential equations can cope with structural change, either) 260.05 420.23 B
0.3 0.21 (local or global. Such equations do not represent changes in sets of equations. Some other kind of) 50.88 406.23 B
(mathematics is needed for the study of systems whose topology can change.) 50.88 392.23 T
160.51 359.42 50.88 359.42 2 L
V
1.02 H
0 Z
N
2 14 Q
(\05021\051 Conjectures) 50.88 360.9 T
1 12 Q
(I\325ll end with a few conjectures and some unanswered questions. \336rst the conjectures.) 50.88 341.23 T
4 14 Q
(\245) 50.88 324.23 T
1 12 Q
0.3 0.43 (Major steps in evolution involved the development of new representational formalisms and) 62.22 324.23 B
(mechanisms for manipulating them.) 62.22 310.23 T
4 14 Q
(\245) 50.88 293.23 T
1 12 Q
0.3 0.09 (Unfortunately such things don\325) 62.22 293.23 B
0.3 0.09 (t leave fossil records. So evidence is going to be very indirect and) 215.05 293.23 B
(only on the basis of a good theoretical framework can we hope to interpret it.) 62.22 279.23 T
4 14 Q
(\245) 50.88 262.23 T
1 12 Q
-0.11 (A major gap in our ability concerns representation of spatial structure adequate to support the antics) 62.22 262.23 P
0.3 0.14 (of birds, squirrels, and other animals that interact in a quick and f) 62.22 248.23 B
0.3 0.14 (luent manner with an intricate,) 387.67 248.23 B
(changing, collection of spatial structures.) 62.22 234.23 T
4 14 Q
(\245) 50.88 217.23 T
1 12 Q
0.3 0.05 (It may turn out that the ef) 62.22 217.23 B
0.3 0.05 (fective representation of shape is not a representation of structure but of) 187.84 217.23 B
0.3 0.06 (\322af) 62.22 203.23 B
0.3 0.06 (fordances\323 \050Gibson 1979\051. So a cube is not \050just\051 a collection of edges and surfaces \050like CAD) 76.82 203.23 B
-0.28 (representations\051, but a collection of) 62.22 189.23 P
307.37 187.93 233.35 187.93 2 L
V
0.59 H
N
-0.28 (possibilities for) 233.35 189.23 P
-0.28 ( and) 307.37 189.23 P
396.13 187.93 330.12 187.93 2 L
V
N
-0.28 (restrictions of) 330.12 189.23 P
-0.28 ( motion, deformation, change,) 396.13 189.23 P
0.3 0.15 (etc. \050Sloman 1989\051 Intelligent agents need a way of generating all this information very quickly) 62.22 175.23 B
0.3 0.04 (from structural information in images, and making it accessible for) 62.22 161.23 B
418.9 159.93 392.09 159.93 2 L
V
N
0.3 0.04 (many) 392.09 161.23 B
0.3 0.04 ( dif) 418.9 161.23 B
0.3 0.04 (ferent uses, including) 435.47 161.23 B
(control of movement, planning, recognition, etc.) 62.22 147.23 T
4 14 Q
(\245) 50.88 130.23 T
1 12 Q
0.3 0.16 (If this is right, representation of) 62.22 130.23 B
228.69 128.93 224.54 128.93 2 L
V
N
0.3 0.16 (f) 224.54 130.23 B
240.5 128.93 228.04 128.93 2 L
V
N
0.3 0.16 (lat) 228.04 130.23 B
0.3 0.16 ( surfaces,) 240.5 130.23 B
328.34 128.93 291.1 128.93 2 L
V
N
0.3 0.16 (straight) 291.1 130.23 B
0.3 0.16 ( edges, etc. may use notations required for) 328.34 130.23 B
104.69 114.93 62.22 114.93 2 L
V
N
0.3 0.13 (arbitrary) 62.22 116.23 B
0.3 0.13 ( curves and shapes, and representation of) 104.69 116.23 B
338.27 114.93 312.16 114.93 2 L
V
N
0.3 0.13 (static) 312.16 116.23 B
0.3 0.13 ( structure may use notations required for) 338.27 116.23 B
FMENDPAGE
%%EndPage: "31" 33
%%Page: "32" 33
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(32) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
103.55 772.93 62.22 772.93 2 L
0 X
V
0.59 H
0 Z
N
0.3 0 (arbitrary) 62.22 774.23 B
0.3 0 ( motion. Many computer models of visual processing assume it\325) 103.55 774.23 B
0.3 0 (s the other way round, i.e.) 413.66 774.23 B
(the more complex items are represented in terms of the less complex.) 62.22 760.23 T
4 14 Q
(\245) 50.88 744.23 T
1 12 Q
0.3 0.03 (Many aspects of human experience, for instance emotional states, are not concerned so much with) 62.22 744.23 B
0.3 0.19 (what we can represent and what inferences or algorithms we can apply) 62.22 730.23 B
0.3 0.19 (, as with what the global) 416.66 730.23 B
0.3 0.03 (architecture is and what sorts of causal interactions can occur) 62.22 716.23 B
0.3 0.03 (. For example, many of the states we) 360.4 716.23 B
0.3 0.22 (call emotional have a characteristic aspect that includes partial loss of control of one\325) 62.22 702.23 B
0.3 0.22 (s thought) 493.5 702.23 B
(processes and attention.) 62.22 688.23 T
0.3 0.05 (There is much that we still don\325) 72.14 668.23 B
0.3 0.05 (t understand about the representational powers of human beings) 227.91 668.23 B
0.3 0.66 (and other animals. That is partly because we do not yet have a good understanding of the) 50.88 654.23 B
0.3 0.02 (requirements, including, for example, the requirements for a human-like visual system. For instance:) 50.88 640.23 B
0.3 0.05 (what are the af) 50.88 626.23 B
0.3 0.05 (fordances in a piece of crumpled paper? How arbitrary shapes and motions should be) 123.24 626.23 B
0.3 0.12 (represented is still an unanswered question. Y) 50.88 612.23 B
0.3 0.12 (et many animals must have solutions implicit in their) 276.94 612.23 B
0.3 0.11 (design, since they cope so well with varied and richly structured spatial environment. Similarly we) 50.88 598.23 B
0.3 0.02 (don\325) 50.88 584.23 B
0.3 0.02 (t yet know what the design requirements are underlying human motivational and af) 72.74 584.23 B
0.3 0.02 (fective states) 476.98 584.23 B
(\050Beaudoin and Sloman 1993\051.) 50.88 570.23 T
232.82 537.42 50.88 537.42 2 L
V
1.02 H
N
2 14 Q
(\05022\051 Unanswered questions) 50.88 538.9 T
1 12 Q
0.5 (Apart from denotational or model-theoretic semantics how many other kinds are there? In particular) 50.88 519.23 P
0.5 (,) 536.85 519.23 P
2.51 (what should be said about \324intensional\325 semantics, according to which knowing the meaning of) 50.88 505.23 P
1.22 (something need not involve knowing what it refers to but does involve knowing how to determine) 50.88 491.23 P
0.99 (whether something is or isn\325) 50.88 477.23 P
0.99 (t referred to. Perhaps a clue lies in the fact that besides the association) 191.21 477.23 P
0.47 (between representations and what they refer to or depict there is also an association between \050simple) 50.88 463.23 P
3.24 (or complex\051 symbols and procedures for associating those symbols with other things \050objects,) 50.88 449.23 P
1.64 (properties, relations, actions, questions, goals, etc.\051. The objects denoted \050the extension\051 would be) 50.88 435.23 P
0.18 (distinct from the procedures and mechanisms that de\336ne the association \050the intension\051. It is not clear) 50.88 421.23 P
3.21 (whether) 50.88 407.23 P
3.21 (, and to what extent, such distinctions can also be applied to pictorial representations,) 89.03 407.23 P
(information states distributed over the weights in a neural net, and other implicit information stores.) 50.88 393.23 T
0.3 0.19 (I suspect that contrary to the standard approach to analysing properties of a language, like its) 72.14 373.23 B
0.3 0.26 (semantic and pragmatic properties, by simply writing down sentences and formulas relating that) 50.88 359.23 B
0.28 (language to things in the world, a truly general theory would have to de\336ne semantics in terms of the) 50.88 345.23 P
0.3 0.23 (architecture of the control system that makes use of the particular information-rich state, and the) 50.88 331.23 B
(functional roles of such states within the architecture.) 50.88 317.23 T
463.06 292.42 127.67 292.42 2 L
V
N
0 14 Q
(Philosophy needs to become a branch of engineering design.) 127.67 293.9 T
FMENDPAGE
%%EndPage: "32" 34
%%Page: "33" 34
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(33) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
330.71 769.87 260.02 769.87 2 L
0 X
V
1.17 H
0 Z
N
0 16 Q
(References) 260.02 771.57 T
1 12 Q
0.14 (The following items illustrate ways in which topics mentioned in this paper can be developed. This is) 50.88 750.23 P
(not an exhaustive bibliography) 50.88 736.23 T
(.) 198.34 736.23 T
3.77 (BEAUDOIN, L.P) 50.88 714.23 P
3.77 (. and A.SLOMAN, \0501993\051 \324A study of motive processing and attention\325, in) 138.93 714.23 P
3.1 (A.Sloman, D.Hogg, G.Humphreys, D. Partridge, A. Ramsay \050eds\051) 79.22 700.23 P
3 F
3.1 (Pr) 424.15 700.23 P
3.1 (ospects for Arti\336cial) 435.7 700.23 P
(Intelligence) 79.22 686.23 T
1 F
(, IOS Press, Amsterdam, pp 229-238) 135.85 686.23 T
4.73 (BOBROW) 50.88 664.23 P
4.73 (, D.G. \0501975\051 Dimensions of representation in D.G.Bobrow and A Collins, \050eds\051) 102.41 664.23 P
3 F
(Repr) 79.22 650.23 T
(esentation and Understanding) 102.1 650.23 T
1 F
( New Y) 247.35 650.23 T
(ork: Academic Press, pp1-34) 283.45 650.23 T
(DENNETT) 50.88 628.23 T
(, D.C., \0501978\051) 105.28 628.23 T
3 F
(Brainstorms) 174.9 628.23 T
1 F
( Bradford Books and Harvester Press.) 234.21 628.23 T
(FUNT) 50.88 606.23 T
(, B. V) 81.31 606.23 T
(. \0501983\051. A parallel-process model of mental rotation.) 108.41 606.23 T
3 F
(Cognitive Science,) 364.89 606.23 T
1 F
( 7\0501\051: 67-93.) 454.16 606.23 T
1.65 (GAZDAR, G.J.M. \0501979\051) 50.88 584.23 P
3 F
1.65 (Pragmatics: Implicatur) 182.74 584.23 P
1.65 (e, Pr) 296.88 584.23 P
1.65 (esupposition and Logical Form,) 321.4 584.23 P
1 F
1.65 ( New York,) 480.27 584.23 P
(London: Academic Press) 79.22 570.23 T
(.) 199.81 570.23 T
0.09 (GIBSON, J. J. \0501979\051.) 50.88 548.23 P
3 F
0.09 (The Ecological Appr) 161.17 548.23 P
0.09 (oach to V) 260.85 548.23 P
0.09 (isual Per) 306.11 548.23 P
0.09 (ception) 349.4 548.23 P
1 F
0.09 (. Lawrence Erlbaum Associates,) 384.71 548.23 P
(Hillsdale, NJ, reprinted 1986.) 79.22 534.23 T
0.21 (GLASGOW) 50.88 512.23 P
0.21 (, J.I \0501993\051 The Imagery Debate Revisited: A Computational Perspective.) 109.73 512.23 P
3 F
0.21 (Computational) 467.88 512.23 P
(Intelligence) 79.22 498.23 T
1 F
( V) 135.85 498.23 T
(ol. 9 No 4, pp 300-435) 145.96 498.23 T
0.94 (HA) 50.88 476.23 P
0.94 (YES, P) 67.09 476.23 P
0.94 (. J. \0501974\051. Some problems and non-problems in representation theory) 102.03 476.23 P
0.94 (. In) 446.19 476.23 P
3 F
0.94 (Pr) 467.07 476.23 P
0.94 (oceedings of) 478.62 476.23 P
0.45 (the AISB 1974 Summer Confer) 79.22 462.23 P
0.45 (ence) 228.47 462.23 P
1 F
0.45 (, University of Sussex. Reprinted in) 250.45 462.23 P
3 F
0.45 (Readings in knowledge) 427.7 462.23 P
-0.19 (r) 79.22 448.23 P
-0.19 (epr) 83.44 448.23 P
-0.19 (esentation) 98.99 448.23 P
1 F
-0.19 (. Edited by R.J. Brachman and H.J. Levesque. Mor) 148.29 448.23 P
-0.19 (gan Kaufmann, Los Altos, CA,) 391.05 448.23 P
(1985, pp. 4-21.) 79.22 434.23 T
0.3 (HA) 50.88 412.23 P
0.3 (YES, P) 67.09 412.23 P
0.3 (.J. \0501993\051 I Can\325) 101.38 412.23 P
0.3 (t Quite See What Y) 180.99 412.23 P
0.3 (ou Mean,) 274.91 412.23 P
3 F
0.3 (Computational Intelligence) 323.8 412.23 P
1 F
0.3 (, Special issue on) 455.68 412.23 P
(Computational Imagery) 79.22 398.23 T
(, V) 192.71 398.23 T
(ol. 9 No 4, pp 381-386) 205.81 398.23 T
(MARR, D. \0501982\051.) 50.88 376.23 T
3 F
(V) 144.82 376.23 T
(ision.) 151.26 376.23 T
1 F
( W) 177.59 376.23 T
(.H. Freeman, San Francisco, CA.) 190.8 376.23 T
2.66 (MCCAR) 50.88 354.23 P
2.66 (THY) 93.48 354.23 P
2.66 (, J. and P) 116.58 354.23 P
2.66 (.J. HA) 166.88 354.23 P
2.66 (YES. \0501969\051. Some philosophical problems from the standpoint of) 199.41 354.23 P
-0.22 (arti\336cial intelligence. In) 79.22 340.23 P
3 F
-0.22 (Machine Intelligence 4) 196.49 340.23 P
1 F
-0.22 (. Edited by B. Meltzer and D. Michie. Edinbur) 306.63 340.23 P
-0.22 (gh) 527.86 340.23 P
(University Press.) 79.22 326.23 T
1.13 (NARA) 50.88 304.23 P
1.13 (Y) 83.75 304.23 P
1.13 (ANAN. N.H. \050editor\051 \0501993\051) 91.08 304.23 P
3 F
1.13 (T) 236.47 304.23 P
1.13 (aking Issue/Forum: Computational Intelligence) 242.03 304.23 P
1 F
1.13 ( V) 474.27 304.23 P
1.13 (ol. 9 No 4,) 485.5 304.23 P
(Nov) 79.22 290.23 T
(. 1993, pp 300-435 Blackwell.) 99.1 290.23 T
2.57 (NEWELL Allen, and H.A.Simon, \0501981\051 \324Computer science as empirical enquiry: Symbols and) 50.88 268.23 P
2.63 (Search\325 in J. Haugeland \050ed\051) 79.22 254.23 P
3 F
2.63 (Mind Design: Philosophy) 232.24 254.23 P
2.63 (, Psychology) 360.08 254.23 P
2.63 (, Arti\336cial Intelligence) 423.34 254.23 P
1 F
2.63 (,) 536.85 254.23 P
(Cambridge Mass: Bradford Books, MIT Press. pp 35-66) 79.22 240.23 T
(PENROSE, R) 50.88 218.23 T
3 F
(, The Emper) 118.18 218.23 T
(or) 176.7 218.23 T
(\325) 187.81 218.23 T
(s New Mind) 190.25 218.23 T
1 F
( Oxford: Oxford University Press, 1989) 247.56 218.23 T
0.41 (PETERSON, D.M., \0501994\051 \322Re-representation and Emer) 50.88 196.23 P
0.41 (gent Information in Three Cases of Problem) 325.53 196.23 P
(Solving\323, in) 79.22 182.23 T
3 F
(Arti\336cial Intelligence and Cr) 140.2 182.23 T
(eativity) 278.67 182.23 T
1 F
(, T) 313.2 182.23 T
(. Dartnall ed., Kluwer [forthcoming 1994]) 325.64 182.23 T
0.36 (SLOMAN, A. \0501971\051. Interactions between philosophy and A.I.: the role of intuition and non-logical) 50.88 160.23 P
2.77 (reasoning in intelligence. In) 79.22 146.23 P
3 F
2.77 (Pr) 227.21 146.23 P
2.77 (oceedings of the Second International Joint Confer) 238.76 146.23 P
2.77 (ence on) 500.12 146.23 P
1.57 (Arti\336cial Intelligence) 79.22 132.23 P
1 F
1.57 (.Reprinted in) 183.06 132.23 P
3 F
1.57 (Arti\336cial Intelligence) 251.82 132.23 P
1 F
1.57 ( 1 pp 209-225, 1971, and in I) 355.66 132.23 P
3 F
1.57 (mages,) 506.21 132.23 P
(per) 79.22 118.23 T
(ception, and knowledge) 94.77 118.23 T
1 F
(. Edited by J.M. Nicholas. Reidel, Dordrecht-Holland, 1977.) 208.36 118.23 T
FMENDPAGE
%%EndPage: "33" 35
%%Page: "34" 35
595.3 839.08 0 FMBEGINPAGE
50.88 80 539.85 80 2 L
7 X
0 K
V
0.25 H
2 Z
0 X
N
50.88 55.2 539.85 73.15 R
7 X
V
1 12 Q
0 X
(Jan 1994) 50.88 65.15 T
(34) 288.99 65.15 T
-0.27 ( Representations as control sub-states) 355.6 65.15 P
50.88 85.75 539.85 782.23 R
7 X
V
0 X
2.71 (SLOMAN, A. \0501975\051. Obituary notice: analogical representations.) 50.88 774.23 P
3 F
2.71 (AISB Quarterly) 391.27 774.23 P
1 F
2.71 (. Reprinted as) 468.15 774.23 P
3.99 (\324Afterthoughts on Analogical Representation,\325 in R. Schank and B. Nash-W) 79.22 760.23 P
3.99 (ebber \050eds\051,) 479.25 760.23 P
3 F
0.26 (Theor) 79.22 746.23 P
0.26 (etical Issues in Natural Language Pr) 107.43 746.23 P
0.26 (ocessing) 285.85 746.23 P
1 F
0.26 (Proceedings of TINLAP conference held at) 330.41 746.23 P
2.47 (MIT in June 1975, and In) 79.22 732.23 P
3 F
2.47 (Readings in knowledge r) 219.63 732.23 P
2.47 (epr) 345.53 732.23 P
2.47 (esentation) 361.08 732.23 P
1 F
2.47 (. R.J. Brachman and H.J.) 410.38 732.23 P
(Levesque \050eds\051 Mor) 79.22 718.23 T
(gan Kaufmann, Los Altos, CA, 1985, pp. 432-439.) 175.59 718.23 T
0.15 (SLOMAN, A. \0501978\051.) 50.88 696.23 P
3 F
0.15 (The Computer Revolution in Philosophy: Philosophy of Science and Models of) 160.59 696.23 P
(Mind) 79.22 682.23 T
1 F
(. Harvester Press and Humanities Press) 104.54 682.23 T
0.93 (SLOMAN, A. \0501985a\051. Why we need many knowledge representation formalisms. In) 50.88 660.23 P
3 F
0.93 (Resear) 473.76 660.23 P
0.93 (ch and) 506.62 660.23 P
(Development in Expert Systems) 79.22 646.23 T
1 F
(. Ed M. Bramer) 230.77 646.23 T
(. Cambridge University Press, pp. 163-183.) 305.06 646.23 T
-0.22 (SLOMAN, A. \0501985b\051, What enables a machine to understand? in) 50.88 624.23 P
3 F
-0.22 ( Pr) 366.63 624.23 P
-0.22 (oceedings 9th International Joint) 380.95 624.23 P
(Confer) 79.22 610.23 T
(ence on AI,) 112.1 610.23 T
1 F
( pp 995-1001, Los Angeles, August 1985) 166.38 610.23 T
1.28 (SLOMAN, A. \0501989\051. On designing a visual system: towards a Gibsonian computational model of) 50.88 588.23 P
(vision.) 79.22 574.23 T
3 F
(Journal of Experimental and Theor) 114.54 574.23 T
(etical Arti\336cial Intelligence,) 284 574.23 T
1 F
( 1\0504\051: 289-337.) 418.91 574.23 T
2.66 (SLOMAN, A. \0501993a\051 \324V) 50.88 552.23 P
2.66 (arieties of formalisms for knowledge representation\325 in) 181.77 552.23 P
3 F
2.66 (Computational) 467.88 552.23 P
(Intelligence) 79.22 538.23 T
1 F
(, Special issue on Computational Imagery) 135.85 538.23 T
(, V) 335.61 538.23 T
(ol. 9, No. 4, November 1993) 348.72 538.23 T
1.42 (SLOMAN, A \0501993b\051 \324The Mind as a Control System\325 in) 50.88 516.23 P
3 F
1.42 (Philosophy and the Cognitive Sciences,) 344.61 516.23 P
1 F
(\050eds\051 C. Hookway and D. Peterson, Cambridge University Press, pp 69-1) 79.22 502.23 T
(10 1993) 429.88 502.23 T
-0.17 (SLOMAN, A. \0501993c\051. \324Prospects for AI as the general science of intelligence\325 in A.Sloman, D.Hogg,) 50.88 480.23 P
1.22 (G.Humphreys, D. Partridge, A. Ramsay \050eds\051) 79.22 466.23 P
3 F
1.22 (Pr) 308.02 466.23 P
1.22 (ospects for Arti\336cial Intelligence,) 319.57 466.23 P
1 F
1.22 ( IOS Press,) 483.79 466.23 P
(Amsterdam, pp1-10) 79.22 452.23 T
FMENDPAGE
%%EndPage: "34" 36
%%Trailer
%%BoundingBox: 0 0 607.45 856.2
%%Pages: 35 1
%%DocumentFonts: Helvetica-Narrow-Bold
%%+ Times-Roman
%%+ Helvetica-Bold
%%+ Times-Italic
%%+ Times-Bold
%%+ AvantGarde-Demi
%%+ Courier-Bold