/LMAR=0/XLINE=3/FONT#0=BAXL30/FONT#1=BAXM30/FONT#2=BAXB30/FONT#3=SUB/FONT#4=SUP/FONT#5=BASL35/FONT#6=NGR25/FONT#7=MATH30/FONT#8=FIX25/FONT#9=GRK30



␈↓ ↓H␈↓Commentary by John McCarthy, Artificial Intelligence Laboratory, Stanford, CA 94305.

␈↓ ↓H␈↓␈↓ α_A␈α
comparison␈α
of␈α
animals␈α
and␈α
machines␈α
or␈α
computer␈α
programs␈α
may␈α
suggests␈α
changing␈α
the
␈↓ ↓H␈↓emphasis␈α
from␈α
the␈α
command␈α
neuron␈α
to␈α
the␈α
behavior␈α
commanded.␈α
 A␈α
␈↓↓unisource␈↓␈α
behavior␈αmight␈α
be
␈↓ ↓H␈↓defined␈αas␈α
one␈αcaused␈α
by␈αan␈αevent␈α
in␈αa␈α
small␈αpart␈α
of␈αa␈αmachine␈α
or␈αanimal␈α
whenever␈αthe␈α
rest␈αof
␈↓ ↓H␈↓the␈α
system␈αsatisfies␈α
certain␈αgeneral␈α
conditions␈αthat␈α
prevail␈αan␈α
interesting␈αfraction␈α
of␈αthe␈α
time.␈α In
␈↓ ↓H␈↓designing␈α⊂a␈α⊃machine␈α⊂or␈α⊂computer␈α⊃program␈α⊂one␈α⊃often␈α⊂faces␈α⊂decisions␈α⊃about␈α⊂whether␈α⊃a␈α⊂certain
␈↓ ↓H␈↓behavior␈α
should␈α
be␈α
produced␈α
in␈α
a␈α
uni-source␈α
way,␈α
e.g.␈α
by␈α
a␈α
specific␈α
subroutine,␈α
or␈α
whether␈αit␈α
must
␈↓ ↓H␈↓arise␈α⊂from␈α⊂the␈α⊂interaction␈α⊂of␈α⊂many␈α⊂parts␈α⊂of␈α⊂the␈α⊂machine.␈α⊂ Thus␈α⊂Fortran␈α⊂compilers␈α⊃give␈α⊂error
␈↓ ↓H␈↓messages␈α∪on␈α∪meeting␈α∪ungrammatical␈α∪statements,␈α∪e.g.␈α∪a␈α∪compiler␈α∪may␈α∪print␈α∪"PARENTHESIS
␈↓ ↓H␈↓ERROR␈α
IN␈αSTATEMENT␈α
45".␈α In␈α
some␈αcompilers,␈α
the␈αprinting␈α
of␈α
"PARENTHESIS␈αERROR
␈↓ ↓H␈↓IN␈αSTATEMENT"␈αis␈αa␈α
␈↓↓unisource␈↓␈αevent␈αproduced␈αby␈αa␈α
routine␈αthat␈αdetects␈αparenthesis␈αerrors.␈α
 In
␈↓ ↓H␈↓other␈α
compilers,␈α
it␈α
will␈αbe␈α
a␈α
multi-source␈α
event,␈αnamely␈α
"ERROR␈α
IN␈α
STATEMENT"␈αis␈α
␈↓↓unisource␈↓,
␈↓ ↓H␈↓being␈α
produced␈αby␈α
a␈αgeneral␈α
part␈αof␈α
the␈αerror␈α
routine␈αwhile␈α
the␈αadjectival␈α
"PARENTHESIS"␈αis
␈↓ ↓H␈↓generated␈α
by␈αthe␈α
part␈αthat␈α
detected␈αthe␈α
parenthesis␈αerror.␈α
 In␈αno␈α
reaonable␈αcompiler␈α
will␈αthe␈α
whole
␈↓ ↓H␈↓event␈α
including␈α
printing␈α
"45"␈α
be␈α
␈↓↓unisource␈↓,␈α
because␈α
no-one␈α
would␈α
make␈α
a␈α
subroutine␈α
specialized␈α
to
␈↓ ↓H␈↓detecting␈αparenthesis␈αerrors␈αin␈αstatement␈α45.␈α If␈αone␈αfound␈αthe␈αword␈α"PARENTHESIS"␈α
sometimes
␈↓ ↓H␈↓misspelled␈αin␈α
an␈αerror␈α
message,␈αone␈α
would␈αinfer␈α
that␈αthere␈α
were␈αat␈α
least␈αtwo␈α
different␈αroutines␈α
that
␈↓ ↓H␈↓handled parenthesis errors, although the inference would not be completely conclusive.

␈↓ ↓H␈↓␈↓ α_A␈α∂second␈α∞example␈α∂comes␈α∞from␈α∂cryptography.␈α∞ If␈α∂one␈α∞successfully␈α∂concludes␈α∞that␈α∂a␈α∞certain
␈↓ ↓H␈↓feature␈α
of␈α
cryptograms␈αin␈α
a␈α
given␈α
cipher␈αis␈α
the␈α
invariable␈α
consequence␈αof␈α
the␈α
presence␈α
of␈αa␈α
certain
␈↓ ↓H␈↓stock phrase in the message, one is well on the way to solving the cryptogram.

␈↓ ↓H␈↓␈↓ α_It␈αwould␈αseem␈αthat␈α
the␈αdiscovery␈αthat␈αa␈α
given␈αbehavior␈αin␈αa␈α
mollusc␈αhas␈αa␈αsingle␈αsource,␈α
e.g.
␈↓ ↓H␈↓is␈α
triggered␈α
by␈α
the␈α
firing␈αof␈α
a␈α
particular␈α
neuron,␈α
is␈αimportant.␈α
 It␈α
raises␈α
the␈α
question␈α
of␈αwhether
␈↓ ↓H␈↓␈↓↓unisource␈↓ behaviors in molluscs are usually triggered by single neurons.