/LMAR=0/XLINE=14/FONT#0=BASL30[ IJ,DBL]/FONT#1=BASB30/FONT#2=NGR25/FONT#3=BASI30[ IJ,DBL]/FONT#4=NGR40/FONT#5=FIX25/FONT#6=NGR20/TMAR=100/PMAR=2220/BMAR=100



␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α¬␈↓∧  BEINGS:  KNOWLEDGE AS INTERACTING EXPERTS␈↓





␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α≡␈↓↓Douglas B. Lenat
␈↓"	␈↓ x␈↓↓␈α?␈α?␈α?␈α?␈α?␈α?␈α+Stanford Artificial Intelligence Laboratory
␈↓"	␈↓ x␈↓↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈αλStanford, California␈↓










␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α0␈↓∧␈↓&ABSTRACT␈↓)αβ␈↓

␈↓"%␈↓ x␈↓␈↓ ↓XKnowledge␈α
may␈αbe␈α
organized␈α
 as␈α a␈α
community␈α
 of␈α interacting␈α
 modules.␈α
Each␈αmodule␈α
is␈α
granted␈αa
␈↓ x␈↓complex␈αstructure,␈αto␈αsimulate␈αa␈αparticular␈αexpert␈αin␈α
some␈αsmall␈αdomain.␈α An␈αextended␈αanalogy␈αis␈αdrawn␈αto␈α
a
␈↓ x␈↓group␈α∞of␈α∞cooperating␈α∞human␈α∞specialists.␈α∞ Based␈α∞on␈α∞this,␈α∞an␈α∞internal␈α∞constraint␈α∞is␈α∞imposed␈α∞on␈α∂the␈α∞modules:
␈↓ x␈↓Their␈α∃structure␈α∀ must␈α∃be␈α∀ standard␈α∃ over␈α∀ the␈α∃ entire␈α∀community.␈α∃ Some␈α∀advantages␈α∃ of␈α∃ a␈α∀ uniform
␈↓ x␈↓formalism␈α
are␈α
thereby␈α
preserved.␈α
 An␈α
 experimental␈α
 community␈α
was␈α
implemented␈α
 for␈α
 the␈α
task␈α
domain␈αof
␈↓ x␈↓automatic␈α⊗programming.␈α⊗ It␈α⊗has␈α⊗managed␈α⊗to␈α⊗synthesize␈α⊗a␈α⊗few␈α⊗inductive␈α⊗inference␈α⊗ LISP␈α⊗ programs,
␈↓ x␈↓nonformally,  from  specific  restricted dialogues with a human user.
␈↓ x␈↓␈α?␈α?␈α?␈α?␈α¬␈↓∧  BEINGS:  KNOWLEDGE AS INTERACTING EXPERTS␈↓

␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α≡␈↓↓Douglas B. Lenat
␈↓"	␈↓ x␈↓↓␈α?␈α?␈α?␈α?␈α?␈α?␈α+Stanford Artificial Intelligence Laboratory
␈↓"	␈↓ x␈↓↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈αλStanford, California␈↓
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α∞␈↓∧␈↓&1. Experts and BEINGs␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XConsider␈α∪an␈α∩interdisciplinary␈α∪enterprise,␈α∩attempted␈α∪by␈α∩a␈α∪community␈α∩of␈α∪human␈α∩experts␈α∪who␈α∩are
␈↓ x␈↓specialists␈α⊂in␈α⊂--␈α⊂and␈α⊂only␈α⊂in␈α⊂--␈α⊂their␈α⊂own␈α⊂fields.␈α⊂ What␈α⊂modes␈α⊂of␈α⊂interactions␈α⊂will␈α⊂be␈α⊃productive?␈α⊂ The
␈↓ x␈↓dominant␈α∂paradigm␈α∂might␈α⊂well␈α∂settle␈α∂into␈α⊂␈↓βquestioning␈α∂and␈α∂answering␈↓␈α∂each␈α⊂other.␈α∂ Instead␈α∂of␈α⊂a␈α∂chairman,
␈↓ x␈↓suppose␈αthe␈αgroup␈αadopts␈αrules␈αfor␈αgaining␈αthe␈α
floor,␈αwhat␈αa␈αspeaker␈αmay␈αdo,␈α and␈αhow␈αto␈α
resolve␈αdisputes.
␈↓ x␈↓When␈αa␈αtopic␈αis␈αbeing␈αconsidered,␈αone␈αor␈αtwo␈αexperts␈α
might␈αrecognize␈αit␈αand␈αspeak␈αup.␈αIn␈αthe␈αcourse␈αof␈α
their
␈↓ x␈↓exposition␈α∞they␈α∞might␈α∞need␈α
to␈α∞call␈α∞on␈α∞other␈α∞specialists.␈α
This␈α∞might␈α∞be␈α∞by␈α
name,␈α∞by␈α∞specialty,␈α∞or␈α∞simply␈α
by
␈↓ x␈↓posing␈α∀a␈α∪new␈α∀sub-question␈α∀and␈α∪hoping␈α∀someone␈α∀could␈α∪recognize␈α∀his␈α∀own␈α∪relevance␈α∀and␈α∀volunteer␈α∪a
␈↓ x␈↓suggestion.␈α
 Such␈α
transfers␈α
would␈α
be␈α
more␈α
common␈α∞at␈α
the␈α
beginning,␈α
when␈α
the␈α
task␈α
is␈α
(by␈α∞assumption)␈α
too
␈↓ x␈↓general␈α∀for␈α∀any␈α∀one␈α∀member␈α∀to␈α∀comprehend.␈α∃ As␈α∀the␈α∀questions␈α∀focus␈α∀on␈α∀more␈α∀specific␈α∃issues,␈α∀single
␈↓ x␈↓individuals␈α⊂should␈α⊂be␈α⊂able␈α⊂to␈α⊂supply␈α⊂complete␈α⊂solutions.␈α⊂ If␈α⊂the␈α⊂task␈α⊂is␈α⊂to␈α⊂construct␈α⊂something,␈α⊃then␈α⊂the
␈↓ x␈↓activities␈αof␈αthe␈αexperts␈α
should␈αnot␈αbe␈αstrictly␈α
verbal.␈α Often,␈αone␈αwill␈α
recognize␈αhis␈αrelevance␈αto␈α
the␈αcurrent
␈↓ x␈↓situation and ask to ␈↓βdo␈↓ something: clarify or modify or (rarely) create.
␈↓"%␈↓ x␈↓␈↓ ↓XWhat␈α∩would␈α∩it␈α∩mean␈α⊃to␈α∩␈↓βsimulate␈↓␈α∩the␈α∩above␈α∩activity?␈α⊃ Imagine␈α∩several␈α∩little␈α∩programs,␈α∩each␈α⊃one
␈↓ x␈↓modelling␈αa␈αdifferent␈αexpert.␈αWhat␈αshould␈αeach␈αprogram,␈αcalled␈αa␈α␈↓βBEING␈↓,␈αbe␈αcapable␈αof?␈α It␈αmust␈αpossess␈αa
␈↓ x␈↓corpus␈α⊃of␈α⊃specific␈α⊃facts␈α⊃and␈α⊃strategies␈α⊃for␈α⊃its␈α⊃designated␈α⊃speciality.␈α⊃It␈α⊃must␈α⊃interact␈α⊃via␈α⊃questioning␈α⊂and
␈↓ x␈↓answering␈αother␈αBEINGs.␈αEach␈αBEING␈αshould␈αbe␈αable␈αto␈αrecognize␈αwhen␈αit␈αis␈αrelevant.␈α It␈αmust␈αset␈αup␈αand
␈↓ x␈↓alter structures, just as the human specialists do.
␈↓"%␈↓ x␈↓␈↓ ↓XLet␈αus␈αreturn␈αto␈αour␈αmeeting␈αof␈αhuman␈αexperts.␈α To␈αbe␈αmore␈αconcrete,␈αsuppose␈αtheir␈αtask␈αis␈αto␈αdesign
␈↓ x␈↓and␈α⊂code␈α⊃a␈α⊂large␈α⊃computer␈α⊂program:␈α⊂a␈α⊃concept␈α⊂formation␈α⊃system[2].␈α⊂Experts␈α⊂who␈α⊃will␈α⊂be␈α⊃useful␈α⊂include
␈↓ x␈↓scientific␈α
programmers,␈α
non-programming␈α
psychologists,␈α∞system␈α
hackers,␈α
and␈α
management␈α∞personnel.␈α
 What
␈↓ x␈↓happens␈α
in␈αthe␈α
ensuing␈αsession?␈α
 When␈α
an␈αexpert␈α
participates,␈αhe␈α
will␈α
either␈αbe␈α
aiding␈αa␈α
colleague␈α
in␈αsome
␈↓ x␈↓difficulty␈αor␈αelse␈αtransferring␈αa␈αtiny,␈αcustomized␈αbit␈α
of␈αhis␈αexpertise␈α(facts␈αabout␈αhis␈αfield)␈αinto␈αa␈α
programmed
␈↓ x␈↓function␈αwhich␈αcan␈αdo␈αsomething.␈α The␈αfinal␈αcode␈αreflects␈αthe␈αmembers'␈αknowledge,␈αin␈αthat␈αsense.␈α One␈αway
␈↓ x␈↓the␈α∂session␈α∂might␈α∞proceed␈α∂is␈α∂for␈α∞the␈α∂specialists␈α∂to␈α∂actually␈α∞␈↓βdo␈↓␈α∂the␈α∂concept␈α∞formation␈α∂task.␈α∂As␈α∂they␈α∞become
␈↓ x␈↓familiar␈α∞with␈α∞what␈α
part␈α∞of␈α∞their␈α
own␈α∞expertise␈α∞is␈α
being␈α∞called␈α∞upon,␈α
and␈α∞in␈α∞what␈α
ways,␈α∞they␈α∞can␈α∞begin␈α
to
␈↓ x␈↓isolate␈α∂it.␈α∞ When␈α∂it␈α∞is␈α∂clear␈α∞precisely␈α∂what␈α∂each␈α∞is␈α∂doing,␈α∞they␈α∂can␈α∞take␈α∂their␈α∞extracted␈α∂bits␈α∂of␈α∞knowledge,
␈↓ x␈↓organize␈α∞them,␈α
formalize␈α∞them,␈α∞and␈α
program␈α∞them.␈α∞ {A␈α
conscious␈α∞effort␈α
along␈α∞these␈α∞lines␈α
was␈α∞made␈α∞in␈α
[7],
␈↓ x␈↓where␈α⊃experts␈α⊂gradually␈α⊃replaced␈α⊃themselves␈α⊂by␈α⊃programs.␈α⊃ Instead␈α⊂of␈α⊃discussing␈α⊃how␈α⊂to␈α⊃write␈α⊃a␈α⊂speech
␈↓ x␈↓program,␈α∞they␈α∞␈↓βdid␈↓␈α
speech␈α∞recognition,␈α∞until␈α
each␈α∞one␈α∞could␈α
introspect␈α∞sufficiently␈α∞into␈α
his␈α∞own␈α∞activities␈α
to
␈↓ x␈↓formalize␈αthem.}␈α For␈αour␈αtask,␈αone␈αexpects␈αthe␈αpsychologists␈αto␈αdominate␈αthe␈αearly␈αdiscussions,␈αlater␈αyielding
␈↓ x␈↓to␈α⊃programmers.␈α⊃ The␈α⊃project␈α⊃sponsor␈α⊃might␈α⊃be␈α⊃passive,␈α⊃submitting␈α⊃a␈α⊃single␈α⊃specification␈α⊃order␈α⊃for␈α⊃the
␈↓ x␈↓program,␈α∪or␈α∪active,␈α∀participating␈α∪in␈α∪the␈α∪work␈α∀as␈α∪a␈α∪(somewhat␈α∪priveleged)␈α∀member␈α∪of␈α∪the␈α∀team.␈α∪This
␈↓ x␈↓individual is the one who wants the final product, hence will be called the ␈↓βuser␈↓.
␈↓"%␈↓ x␈↓␈↓ ↓XHow␈α⊃could␈α⊃BEINGs␈α⊃do␈α⊃this?␈α⊃There␈α⊂would␈α⊃be␈α⊃some␈α⊃little␈α⊃program␈α⊃containing␈α⊃information␈α⊂about
␈↓ x␈↓␈↓αCONCEPT-FORMATION␈↓␈α∂(much␈α⊂more␈α∂than␈α∂would␈α⊂be␈α∂used␈α⊂in␈α∂writing␈α∂any␈α⊂single␈α∂concept␈α⊂formation␈α∂program),
␈↓ x␈↓another␈αBEING␈α
who␈αknows␈αhow␈α
to␈αmanage␈αa␈α
group␈αto␈α␈↓αWRITE-PROGRAMS␈↓,␈α
and␈αmany␈α
lower-level␈αspecialists,
␈↓ x␈↓for␈α→example␈α→␈↓αINFO-OBTAINER,␈α→TEST,␈α→MODIFY-DATA-STRUCTURE,␈α→UNTIL-LOOP,␈α~VISUAL-PERCEPTION,␈α→AVOID-
␈↓ x␈↓αCONTRADICTION,␈α
PROPOSE-PLAUSIBLE-NAME␈↓.␈α
 Like␈α
the␈α
human␈αspecialists,␈α
the␈α
BEINGs␈α
would␈α
contain␈α
far␈αtoo
␈↓ x␈↓much␈α∂information,␈α∂far␈α∂too␈α∂inefficiently␈α∂represented,␈α∂to␈α∂be␈α∂able␈α∂to␈α∂say␈α∂"we␈α∂ourselves␈α∂constitute␈α⊂the␈α∂desired
␈↓ x␈↓program!"␈αThey␈αwould␈αhave␈αto␈αdiscuss,␈αand␈αperhaps␈αcarry␈αout,␈αthe␈αconcept␈αformation␈αtask.␈αThey␈αwould␈αwrite

␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α␈↓ε1␈↓␈↓ λ
␈↓ x␈↓specialized␈αversions␈αof␈α
themselves,␈αprograms␈αwhich␈α
could␈αdo␈αexactly␈α
what␈αthe␈αBEINGs␈α
did␈αto␈αcarry␈α
out␈αthe
␈↓ x␈↓task,␈α⊂no␈α⊂more␈α⊂nor␈α⊂less␈α⊃(although␈α⊂they␈α⊂would␈α⊂hopefully␈α⊂take␈α⊃much␈α⊂less␈α⊂time,␈α⊂be␈α⊂more␈α⊃customized).␈α⊂ This
␈↓ x␈↓activity␈αis␈αreferred␈αto␈αin␈αthe␈αsequel␈αas␈α␈↓βautomatic␈αprogramming␈↓.␈α Some␈αBEINGs␈α(e.g.,␈α␈↓αTEST␈↓)␈αmay␈αhave␈αseveral
␈↓ x␈↓distinct,␈αstreamlined␈αfractions␈αof␈αthemselves␈αin␈αthe␈αfinal␈αprogram.␈αBEINGs␈αwhich␈αonly␈αaided␈αother␈αBEINGs
␈↓ x␈↓(e.g., ␈↓αPROPOSE-PLAUSIBLE-NAME␈↓) may not have ␈↓βany␈↓ correlates in the final synthesized code.
␈↓"%␈↓ x␈↓␈↓ ↓XAn␈αexperimental␈αsystem,␈αPUP6,␈αwas␈αdesigned␈α
and␈αpartially␈αimplemented.␈αPUP6␈αsynthesized␈αa␈α
concept
␈↓ x␈↓formation␈αprogram␈α(similar␈αto␈α[6]),␈αbut␈αthe␈αuser,␈αwho␈αis␈αhuman,␈α must␈αcome␈αup␈αwith␈αcertain␈αspecific␈αanswers
␈↓ x␈↓to␈α⊃some␈α⊃of␈α⊂the␈α⊃BEINGs'␈α⊃critical␈α⊂queries.␈α⊃ A␈α⊃grammatical␈α⊂inference␈α⊃program␈α⊃and␈α⊂a␈α⊃ simple␈α⊃property␈α⊂list
␈↓ x␈↓maintenance␈αroutine␈αwere␈αalso␈αgenerated.␈α
Only␈αa␈αfew␈αnew␈αBEINGs␈αhad␈α
to␈αbe␈αadded␈αto␈αPUP6␈α
to␈αsynthesize
␈↓ x␈↓them, but communication flexibility problems existed.
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈αnext␈αsection␈αillustrates␈α
how␈αthe␈αexperts␈αmight␈α
have␈αcooperated␈αon␈αthe␈α
task␈αof␈αwriting␈αthe␈α
concept
␈↓ x␈↓formation␈α⊂program.␈α⊂Section␈α⊂3␈α⊂describes␈α⊂the␈α⊂program␈α⊂they␈α⊂produced.␈α⊂Next␈α⊂comes␈α⊂the␈α⊃BEING␈α⊂hypothesis:
␈↓ x␈↓complex␈α∪but␈α∀standard␈α∪anatomy.␈α∪ Later␈α∀sections␈α∪explain␈α∀this,␈α∪both␈α∪theoretically␈α∀and␈α∪by␈α∀examining␈α∪the
␈↓ x␈↓behavior of the actual PUP6 pool of 100 BEINGs.
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α⊗␈↓∧␈↓&2. Experts Interacting␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈αinput/output␈αbehavior␈αof␈αthe␈αdesired␈αconcept␈αformation␈αprogram␈αis␈αspecified␈αin␈αthis␈αsection,␈αand
␈↓ x␈↓we␈αeavesdrop␈αon␈α
a␈αsimulated␈αgroup␈α
of␈αspecialists␈α as␈α
they␈αget␈αto␈α
work␈αon␈αwriting␈α
it.␈α As␈αthe␈α
presentation␈αof
␈↓ x␈↓the␈α
experts'␈αactivities␈α
becomes␈α
more␈αspecific,␈α
the␈α
reader's␈αcurrently␈α
vague␈α
conception␈αof␈α
BEINGs␈α
will␈αbe␈α
made
␈↓ x␈↓less␈α⊂amorphous␈α⊂(because␈α⊂BEINGs␈α⊃are␈α⊂constrained␈α⊂to␈α⊂carry␈α⊂on␈α⊃approximately␈α⊂the␈α⊂same␈α⊂discussion␈α⊃as␈α⊂the
␈↓ x␈↓experts below do).
␈↓"%␈↓ x␈↓␈↓ ↓XExternally,␈αthe␈αconcept␈α
formation␈αtask␈αcan␈αbe␈α
specified␈αas␈αfollows:␈αpictures␈α
of␈αstructures␈α(built␈α
out␈αof
␈↓ x␈↓simple␈α
geometrical␈α∞shapes)␈α
will␈α
be␈α∞presented␈α
one␈α
after␈α∞another.␈α
For␈α
each␈α∞such␈α
scene,␈α
the␈α∞concept␈α
formation
␈↓ x␈↓program,␈α
call␈αit␈α
CF,␈αmust␈α
guess␈αits␈α
name.␈αThe␈α
presenter␈αwill␈α
then␈αreveal␈α
the␈αcorrect␈α
name␈αof␈α
the␈αstructure.␈α
CF
␈↓ x␈↓must␈αquickly␈αlearn␈α
to␈αidentify␈αsimple␈αstructures␈α
(ARCH,␈αTOWER),␈αand␈αmust␈α
never␈αmake␈αthe␈α
same␈αmistake
␈↓ x␈↓twice in a row.
␈↓"%␈↓ x␈↓␈↓ ↓XOur␈α
group␈αof␈α
experts␈αare␈α
given␈αthis␈α
specification␈α
for␈αCF.␈α
Assume␈αthat␈α
the␈αuser␈α
(the␈αfinancial␈α
sponsor)
␈↓ x␈↓is␈αavailable␈αfor␈αresolving␈αimportant␈αquestions,␈αvia␈αmessenger,␈αand␈αhe␈αmay␈αin␈αfact␈αask␈αquestions␈αof␈αthe␈αgroup.
␈↓ x␈↓Whenever␈αan␈αexpert␈α
speaks,␈αalmost␈αall␈αthe␈α
others␈αin␈αthe␈αroom␈α
hear␈αhim.␈αUsually␈αonly␈α
a␈αfew␈αcan␈αbenefit␈α
from
␈↓ x␈↓what␈α⊂he␈α⊂says,␈α⊂and␈α⊂fewer␈α⊂still␈α⊂care␈α⊂to␈α⊂react.␈α∂ The␈α⊂conversation␈α⊂in␈α⊂the␈α⊂room␈α⊂might␈α⊂go␈α⊂something␈α⊂like␈α∂the
␈↓ x␈↓following (the suggestive names of the experts are of course coincidental):
␈↓"∃␈↓ x␈↓GENL-MANAGER:␈α∃Can␈α⊗anybody␈α∃here␈α⊗figure␈α∃out␈α⊗what␈α∃to␈α∃do,␈α⊗what␈α∃the␈α⊗user's␈α∃saying?␈α⊗(waves␈α∃the
␈↓"	␈↓ x␈↓input/output specifications in the air)
␈↓"∃␈↓ x␈↓PGM-MANAGER:␈α∞I␈α∞can.␈α∞He␈α∞wants␈α
a␈α∞computer␈α∞program␈α∞to␈α∞be␈α
written.␈α∞If␈α∞somebody␈α∞will␈α∞explain␈α∞the␈α
task
␈↓"	␈↓ x␈↓"con-cept-for-ma-tion" to me a little clearer, I'll delegate it properly.
␈↓"∃␈↓ x␈↓PSYCH: Permit me to assist you. I know all about concept formation.  In fact, my master's thesis...
␈↓"∃␈↓ x␈↓PGM-MANAGER: Wait, let me tell the user that we'll be able to handle the job.
␈↓"∃␈↓ x␈↓MESSENGER: Here, I can take that message. Go on with your work.
␈↓"∃␈↓ x␈↓PGM-MANAGER: We need a name for this program. Somebody get one, please.
␈↓"∃␈↓ x␈↓NAMER: How about "CONCEPT"? Maybe just "CF". Let's ask the user to decide.
␈↓"∃␈↓ x␈↓MESSENGER: (panting) I just came back from there! Alright, I'm going...  User says to call it CF.
␈↓"∃␈↓ x␈↓PGM-MANAGER:␈α∩Now␈α∩then,␈α∩I␈α∩have␈α⊃four␈α∩people␈α∩in␈α∩mind,␈α∩one␈α∩of␈α⊃whom␈α∩must␈α∩take␈α∩over␈α∩now␈α∩in␈α⊃an
␈↓"	␈↓ x␈↓important way. Each of them always wants to do something different.
␈↓"∃␈↓ x␈↓CHOOSER: Give me their names and I'll choose the best one for you.
␈↓"∃␈↓ x␈↓PGM-MANAGER: They are INFO-OBTAINER, INFO-USER, PGMMER, and ANALYST.
␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α␈↓ε2␈↓␈↓ λ
␈↓ x␈↓CHOOSER: You four experts each must justify to me why you should take control now.
␈↓"∃␈↓ x␈↓␈↓ ↓(INFO-OBTAINER: We already have some unused information, so I guess I can wait.
␈↓"∃␈↓ x␈↓␈↓ ↓(INFO-USER: Let me have the floor! I know what to do with the user's specification.
␈↓"∃␈↓ x␈↓␈↓ ↓(PGMMER: Well, I suppose I could set up some "empty" top-level function, er...
␈↓"∃␈↓ x␈↓␈↓ ↓(ANALYST: Marginal profits accruing from my working at this point in time do not warrant...
␈↓"∃␈↓ x␈↓CHOOSER: Okay, INFO-USER, you seem to be the only relevant expert.
␈↓"∃␈↓ x␈↓INFO-USER: First thing is for PSYCH to tell us how to actually ␈↓βdo␈↓ concept formation.
␈↓"∃␈↓ x␈↓PSYCH:␈α
There␈α
are␈αseveral␈α
decisions␈α
to␈αbe␈α
made,␈α
depending␈αon␈α
what␈α
your␈αindependent␈α
variables␈α
and␈αyour
␈↓"	␈↓ x␈↓experimental␈αsetup␈αare.␈αFor␈α
example,␈αare␈αwe␈αstudying␈α
one␈αindividual,␈αor␈αmust␈α
our␈αprogram␈αadapt␈αto␈α
simulate
␈↓"	␈↓ x␈↓many␈α∞different␈α
subjects?␈α∞Must␈α
stimulus␈α∞items␈α
be␈α∞just␈α
classified,␈α∞or␈α
classified␈α∞and␈α
ordered,␈α∞or␈α∞classified␈α
and
␈↓"	␈↓ x␈↓ordered␈αand␈αmetrized?␈αAre␈αthe␈αscenes␈αleft␈αin␈αview␈α
indefinitely,␈αor␈αjust␈αthe␈αcurrent␈αscene,␈αor␈αis␈αeven␈α
that␈αone
␈↓"	␈↓ x␈↓just␈α
flashed␈α
before␈α
the␈αsubject␈α
for␈α
a␈α
limited␈α
time?␈αMay␈α
the␈α
subject␈α
make␈αwritten␈α
notes?␈α
Can␈α
he␈α
learn␈αfrom
␈↓"	␈↓ x␈↓positive instances of the concepts?  from negative instances?  Is there to be any positive transfer effect...?
␈↓"∃␈↓ x␈↓DEFER:␈α
Hold␈α
on␈α
here;␈α
we␈αcan't␈α
swamp␈α
the␈α
user␈α
with␈α
lots␈αof␈α
unnecessary␈α
questions.␈α
 Let's␈α
go␈α
through␈αeach
␈↓"	␈↓ x␈↓one␈α⊃and␈α⊃see␈α∩if␈α⊃it␈α⊃can␈α∩wait.␈α⊃ Consider␈α⊃the␈α∩ decision␈α⊃about␈α⊃the␈α∩task␈α⊃being␈α⊃classificatory,␈α∩comparative,␈α⊃or
␈↓"	␈↓ x␈↓metrical. What is the first situation in which we really use the result of that choice?
␈↓"∃␈↓ x␈↓WHEN-NEXT: This involves a list of alternatives. Who knows about that?
␈↓"∃␈↓ x␈↓ALTERNATIVES:␈α∂I␈α∂know␈α∂a␈α∂few␈α∂tricks␈α∂here.␈α∂If␈α∂all␈α∂the␈α∂alternative␈α∂sequences␈α∂of␈α∂activities␈α∂have␈α∂the␈α∞same
␈↓"	␈↓ x␈↓initial subsequence, then do that common subsequence before making the decision.
␈↓"∃␈↓ x␈↓UTILIZE: In this case, all three begin by partitioning a domain of elements.
␈↓"∃␈↓ x␈↓DEFER:␈αTemporarily,␈αreplace␈αthe␈αtask␈αCF␈αby␈αthe␈αtask␈αof␈αpartitioning␈αa␈αdomain.␈α Now␈αconsider␈αthe␈αdecision
␈↓"	␈↓ x␈↓about subject-specific behavior being required.
␈↓"∃␈↓ x␈↓PSYCH: This involves periodically inputting a description of the human subject.
␈↓"∃␈↓ x␈↓CODER: That would mean adjusting the algorithms based on a vector of parameters.
␈↓"∃␈↓ x␈↓WHEN-NEXT: This would probably affect the entire code.
␈↓"∃␈↓ x␈↓DEFER: I can't defer this decision. Someone had better take care of it.
␈↓"∃␈↓ x␈↓RESOLVE: Ask the user about it.
␈↓"∃␈↓ x␈↓ASK-USER: Phrase this as a yes/no question. Explain each alternative to the user.
␈↓"∃␈↓ x␈↓MESSENGER: That's my job... User says no, don't simulate different people.
␈↓"∃␈↓ x␈↓DEFER: The next decision...
␈↓"∃␈↓ x␈↓  ␈↓βEleven decisions are ultimately proposed by PSYCH, and all but one are deferred.␈↓
␈↓"∃␈↓ x␈↓INFO-USER: I have no objections now if someone wants the floor.
␈↓"∃␈↓ x␈↓PGM-MANAGER: Do any of the other three experts I mentioned earlier want to speak now?
␈↓"∃␈↓ x␈↓PGMMER: Yes. The top-level function CF can now be coded.
␈↓"∃␈↓ x␈↓CODER: Give me the arguments and the body of the code, please.
␈↓"∃␈↓ x␈↓PGMMER: There are no known arguments. The body is a call on ␈↓βPARTITION-DOMAIN␈↓.
␈↓"∃␈↓ x␈↓CODER:␈α∞Okay.␈α
I␈α∞will␈α∞precede␈α
that␈α∞with␈α∞a␈α
call␈α∞to␈α∞an␈α
␈↓βINITIALIZE␈↓␈α∞function,␈α∞and␈α
follow␈α∞it␈α∞with␈α
a␈α∞call␈α∞to␈α
a
␈↓"	␈↓ x␈↓␈↓βFINALIZE␈↓␈α
function,␈αwhich␈α
are␈αboth␈α
defined␈α
as␈αNIL␈α
for␈αnow.␈α
 Is␈α␈↓βPARTITION-DOMAIN␈↓␈α
simple␈α
enough␈αto
␈↓"	␈↓ x␈↓be composed right now and filled in here?
␈↓"∃␈↓ x␈↓MATHEMATICIAN: No way. While conceptually elegant in its simplicity, any realizate...
␈↓"∃␈↓ x␈↓CODER: Uh, thanks. There. The function CF is defined as
␈↓"∃␈↓ x␈↓␈↓ ↓X␈↓α(LAMBDA ( ) (INITIALIZE) (PARTITION-DOMAIN) (FINALIZE)).␈↓
␈↓"∃␈↓ x␈↓ANALYST:␈α∞Remind␈α∞me␈α∞to␈α
examine␈α∞the␈α∞initialization␈α∞and␈α
finalization␈α∞functions␈α∞at␈α∞the␈α
end␈α∞of␈α∞our␈α∞task.␈α
If
␈↓"	␈↓ x␈↓either function is still null, it will be deleted.
␈↓"∃␈↓ x␈↓WARNER: I have just put that note into the code for CF, as a comment.
␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α␈↓ε3␈↓␈↓ λ
␈↓ x␈↓PGMMER: Can someone advise me of what else to do to finish defining this function?
␈↓"∃␈↓ x␈↓PGM-MANAGER:␈α
Each␈αfunction␈α
should␈αhave␈α
a␈αproper␈α
name.␈αShow␈α
the␈αuser␈α
the␈αnames␈α
you␈α
have␈αpicked,
␈↓"	␈↓ x␈↓and let him choose other ones if he prefers.
␈↓"∃␈↓ x␈↓MESSENGER: Okay...  The user agrees to all three names for the function calls.
␈↓"∃␈↓ x␈↓INFO-USER: Somebody, please tell the group how to ␈↓βdo␈↓ partitioning of a space of examples.
␈↓"%␈↓ x␈↓␈↓ ↓XA␈αcomplete␈αscript,␈αlike␈αthe␈αabove,␈αwas␈αconstructed␈αby␈αhand.␈α In␈αthe␈αsequel,␈αthis␈αwill␈αbe␈αreferred␈αto␈αas
␈↓ x␈↓the␈α␈↓βprotocol␈↓.␈α In␈αall,␈α87␈α
different␈αexperts␈αwere␈αcalled␈αfor:␈α
17␈αspecificly␈αdealing␈αwith␈αinductive␈α
inference␈αtasks,
␈↓ x␈↓and␈α
70␈α
dealing␈α
with␈α
programming,␈α
managing␈α
workers,␈αand␈α
communicating␈α
with␈α
the␈α
user.␈α
 Near␈α
the␈α
end␈αof
␈↓ x␈↓the␈α∂protocol,␈α∂the␈α⊂user␈α∂is␈α∂asked␈α⊂which␈α∂of␈α∂the␈α⊂three␈α∂types␈α∂of␈α⊂concept␈α∂formation␈α∂CF␈α⊂is␈α∂supposed␈α∂to␈α⊂do.␈α∂He
␈↓ x␈↓responds␈α∞"␈↓βCLASSIFICATORY␈α∞only␈↓",␈α∞and␈α∞the␈α∞experts␈α∞discover␈α∞that␈α∞they␈α∞are␈α∞finished.␈α∞All␈α∞the␈α∞newly␈α
created
␈↓ x␈↓code␈α∞is␈α∞dumped␈α∞out␈α∞onto␈α∞a␈α∞fresh␈α∞file:␈α∞After␈α∞hundreds␈α∞of␈α∞pages,␈α∞a␈α∞concept␈α∞formation␈α∞program␈α∂meeting␈α∞the
␈↓ x␈↓user's specifications had been written.  The next section will describe that program in detail.
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α≤␈↓∧␈↓&3. The Program the Experts Wrote␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XOne␈αof␈αthe␈αexperts␈αat␈α
the␈αsimulated␈αmeeting␈αmust␈αhave␈α
read␈αP.␈αWinston's␈αdissertation[6],␈αbecause␈α
CF,
␈↓ x␈↓the␈α
synthesized␈α
concept␈α
formation␈α
program,␈α
was␈α
remarkably␈α
similar␈α
to␈α
the␈α
one␈α
therein␈α
described.␈α
CF␈α
has␈αa
␈↓ x␈↓much␈αsimpler␈α
graph-matching␈αalgorithm,␈α
and␈αrelations␈α
on␈αrelations␈α
are␈αstored␈α
in␈αa␈α
different␈αway␈αthan␈α
simple
␈↓ x␈↓relations␈αon␈αobjects.␈α Since␈αCF␈αwas␈αlater␈αsynthesized␈αby␈αPUP6,␈αthe␈αprogrammed␈αpool␈αof␈αBEINGs,␈αit␈αis␈αworth
␈↓ x␈↓detailing here.
␈↓"%␈↓ x␈↓␈↓ ↓XCF␈α∞repeatedly␈α∞scans␈α∞a␈α∞scene␈α∞and␈α∞tries␈α∞to␈α∞name␈α∞it.␈α∞As␈α∞a␈α∞first␈α∞step,␈α∞the␈α∞scene␈α∞is␈α∞broken␈α∞into␈α∞a␈α∞set␈α∞of
␈↓ x␈↓objects␈α
and␈α
a␈α
set␈α
of␈αfeatures␈α
(relations␈α
on␈α
those␈α
objects).␈αCF␈α
maintains␈α
a␈α
model␈α
for␈α
each␈αdifferently-named
␈↓ x␈↓scene␈αit␈αhas␈αencountered.␈αA␈αmodel␈αcontains␈αa␈αdescription␈αof␈αthe␈αobjects␈αone␈αexpects␈αin␈αsuch␈αa␈αstructure,␈αa␈αset
␈↓ x␈↓of␈αfeatures␈αwhich␈α␈↓βmust␈↓␈αbe␈αpresent␈αin␈αany␈αscene␈αhaving␈αthis␈αname,␈αa␈αset␈αof␈αfeature␈αwhich␈α␈↓βmust␈αnot␈↓␈αbe␈αpresent
␈↓ x␈↓if␈αthe␈αscene␈αis␈αto␈αhave␈αthis␈αname,␈αand␈αa␈αset␈αof␈αfeatures␈αwhich␈α␈↓βmay␈↓␈αbe␈αpresent␈αor␈αabsent.␈αThus␈αa␈αmodel␈αis␈αan
␈↓ x␈↓archtypical scene plus a name.  For example, part of a scene might be:
␈↓"	␈↓ x␈↓        OBJECTS a,b,c,d
␈↓"	␈↓ x␈↓        RELATIONS  (GREEN a) (BLUE c) (TOUCHES c d) (SUPPORTS a c) (SUPPORTS b c)
␈↓"	
␈↓"	␈↓ x␈↓CF's current model for an arch might be:           
␈↓"	␈↓ x␈↓        NAME    ARCH
␈↓"	␈↓ x␈↓        OBJECTS a,b,c
␈↓"	␈↓ x␈↓        MUST    (SUPPORTS a c) (SUPPORTS b c)
␈↓"	␈↓ x␈↓        MUSTNOT (TOUCHES a b)
␈↓"	␈↓ x␈↓        MAY     (GREEN a) (WEDGE c) (PRISM a) (BLOCK b) (PARALLEL a b)
␈↓"%␈↓ x␈↓␈↓ ↓XEach␈αtime␈αit␈αis␈αconfronted␈αby␈αa␈αnew␈αscene,␈αCF␈α must␈αscan␈αits␈αmodels␈αuntil␈αit␈αfinds␈αone␈αwhich␈αmatches
␈↓ x␈↓it.␈αA␈αmodel␈αis␈αsaid␈αto␈αmatch␈α
a␈αscene␈αif␈αall␈αthe␈αMUST␈αfeatures␈α
associated␈αwith␈αthat␈αmodel␈αare␈αobserved␈αin␈α
the
␈↓ x␈↓scene,␈α
and␈α
all␈α
the␈α
MUSTNOT␈α
 features␈α
 are␈α
absent␈α from␈α
 the␈α
  scene.␈α
CF␈α
informs␈α
the␈α
user␈α
of␈α
this␈αguess,␈α
and
␈↓ x␈↓accepts␈α
the␈α
proper␈α
 name.␈α
If␈α
it␈α
 guessed␈α
 incorrectly,␈α
 it␈α
modifies␈α
its␈α
models.␈α
The␈α
wrong-guess␈α
model␈α
may␈α
have
␈↓ x␈↓features␈α
added␈α
to␈α
its␈α
MUST␈α
or␈α
MUSTNOT␈α
sets.␈α
 P.␈α
Gadwa␈α
showed␈α
this␈α
is␈α
sufficient␈α
to␈α
prevent␈α∞CF␈α
from
␈↓ x␈↓making␈αthe␈αsame␈αwrong␈αguess␈α
twice␈αin␈αsuccession.␈α The␈αcorrect-name␈αmodel␈α
may␈αhave␈α to␈αbe␈α modified␈αor␈α
(if
␈↓ x␈↓it's␈α∞a␈α∞new␈α∞name)␈α∞created␈α∞and␈α∞inserted␈α∞into␈α∞the␈α∞list␈α∞of␈α∞models,␈α∞to␈α∞ensure␈α∞that␈α∞CF␈α∞will␈α∞eventually␈α∞learn␈α∞that
␈↓ x␈↓concept.  The genesis of the idea of a MUST/MUSTNOT/MAY division is simulated in Section 5.3.
␈↓"%␈↓ x␈↓␈↓ ↓XSuppose␈α∂ that␈α∞the␈α∂target␈α∞program␈α∂reads␈α∞in␈α∂the␈α∞above␈α∂scene␈α∞fragment␈α∂and␈α∞tries␈α∂to␈α∞match␈α∂it␈α∂to␈α∞ the
␈↓ x␈↓above␈α⊃ ARCH␈α∩model.␈α⊃ The␈α∩ MUST␈α⊃relations␈α∩ should␈α⊃ all␈α⊃ be␈α∩ present.␈α⊃  Yes,␈α∩ the␈α⊃ scene␈α∩ does␈α⊃contain
␈↓ x␈↓(SUPPORTS␈αa␈αc)␈αand␈α(SUPPORTS␈αb␈αc).␈αNext,␈αthe␈αMUSTNOT␈α relations␈α must␈αbe␈αabsent␈αfrom␈αthe␈αscene.
␈↓ x␈↓Sure␈α∩enough,␈α∩(TOUCHES␈α∪a␈α∩b)␈α∩isn't␈α∩there.␈α∪ So␈α∩the␈α∩model␈α∩and␈α∪scene␈α∩are␈α∩consistent,␈α∩and␈α∪the␈α∩program
␈↓ x␈↓announces␈αthat␈αits␈αguess␈αis␈αARCH.␈α If␈αthe␈αuser␈αverifies␈αthis␈αguess,␈αthen␈α the␈αMAY␈αset␈αof␈αthe␈αARCH␈αmodel␈α
is
␈↓ x␈↓augmented with the relations (BLUE c) and (TOUCHES c d), and the OBJECTS set is augmented with "d."
␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α␈↓ε4␈↓␈↓ λ
␈↓ x␈↓␈↓ ↓XIf␈αthe␈αuser␈αdenies␈αthat␈αthe␈αscene␈αis␈αan␈α
arch,␈αCF␈αsees␈αif␈αthere␈αare␈αany␈αrelations␈αin␈αthe␈α
ARCH␈αmodel's
␈↓ x␈↓MAY␈α
set␈α
which␈α
do␈α
not␈α
occur␈α
in␈α
the␈α
scene.␈α
If␈αso,␈α
one␈α
of␈α
them␈α
(e.g.,␈α
(PARALLEL␈α
a␈α
b))␈α
will␈α
be␈αtransferred␈α
from
␈↓ x␈↓the␈αMAY␈αto␈αthe␈αMUST␈αset.␈α If␈αno␈αsuch␈αfeature␈αexisted,␈αthe␈αprogram␈αwould␈αlook␈αfor␈αa␈αfeature␈αpresent␈αin␈αthe
␈↓ x␈↓scene␈α⊂but␈α⊂not␈α∂mentioned␈α⊂in␈α⊂any␈α⊂set␈α∂of␈α⊂the␈α⊂ARCH␈α⊂model␈α∂(e.g.,␈α⊂(TOUCHES␈α⊂c␈α⊂d)),␈α∂and␈α⊂insert␈α⊂it␈α⊂into␈α∂the
␈↓ x␈↓MUSTNOT␈α∞set.␈α∞ In␈α
either␈α∞case,␈α∞the␈α
user␈α∞would␈α∞be␈α
asked␈α∞what␈α∞the␈α
true␈α∞name␈α∞was,␈α
and␈α∞that␈α∞model␈α
would
␈↓ x␈↓have␈α∂its␈α∂MAY␈α∂set␈α∂augmented␈α∂by␈α∂any␈α∂new␈α∂features␈α∂in␈α∂the␈α∂scene.␈α∂Any␈α∂features␈α∂on␈α∂ the␈α∂true-name␈α∂model's
␈↓ x␈↓MUST or MUSTNOT sets which contradicted the scene would be transferred to the MAY set.
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α∧␈↓∧␈↓&4. Anatomy of Synergetic Cooperation␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XConsider␈α∂the␈α∂birth␈α∞of␈α∂one␈α∂small␈α∞idea␈α∂necessary␈α∂in␈α∂the␈α∞writing␈α∂of␈α∂CF,␈α∞that␈α∂of␈α∂classifying␈α∂a␈α∞model's
␈↓ x␈↓features␈α∞into␈α∂three␈α∞categories␈α∞(MUST,␈α∂MUSTNOT,␈α∞MAY).␈α∞Later,␈α∂this␈α∞genesis␈α∞will␈α∂be␈α∞presented,␈α∞as␈α∂if␈α∞the
␈↓ x␈↓group␈αof␈αexperts␈αwere␈αdiscussing␈αit.␈αNo␈αsingle␈αspecialist␈αat␈αthe␈αmeeting␈αcould␈αhave␈αhad␈αthis␈αidea␈αby␈αhimself.
␈↓ x␈↓How␈α⊂do␈α∂intellects␈α⊂mesh,␈α∂effectively␈α⊂communicate,␈α∂and␈α⊂unite␈α∂their␈α⊂powers?␈α∂ A␈α⊂tentative␈α⊂mechanism,␈α∂which
␈↓ x␈↓barely␈α⊃scratches␈α⊂the␈α⊃surface␈α⊂of␈α⊃this␈α⊂mystery,␈α⊃will␈α⊂be␈α⊃hypothesized.␈α⊂The␈α⊃BEINGs␈α⊂in␈α⊃PUP6␈α⊃ embody␈α⊂this
␈↓ x␈↓concept, and are able to reproduce both the experts' discussion and the actual working CF program.
␈↓"%␈↓ x␈↓␈↓ ↓XViewing␈α
the␈α
group␈α
of␈α
experts␈α
as␈α
a␈α
single␈α
entity,␈α
what␈α
makes␈α
it␈α
productive?␈α
The␈α
members␈α
must␈α
be␈α
very
␈↓ x␈↓different␈αin␈αabilities,␈αin␈α
order␈αto␈αhandle␈αsuch␈α
a␈αcomplex␈αtask,␈αyet␈αsimilar␈α
in␈αbasic␈αcognitive␈αstructure␈α
(in␈αthe
␈↓ x␈↓anatomy␈αof␈αtheir␈αminds)␈αto␈αpermit␈αfacile␈αcommunications␈αto␈αflow.␈α For␈αexample,␈αeach␈αspecialist␈αknows␈αhow␈αto
␈↓ x␈↓direct␈α
a␈α
programmer␈α
to␈α
do␈α
some␈α
of␈α
the␈α
things␈α
he␈α
can␈α
do,␈α
but␈α
the␈α
specific␈α
facts␈α
each␈α
expert␈α
has␈α
under␈α
this
␈↓ x␈↓category␈αmust␈αbe␈αquite␈αunique.␈αSimilarly,␈αeach␈αmember␈αmay␈αhave␈αa␈αset␈αof␈αstrategies␈αfor␈αrecognizing␈αhis␈αown
␈↓ x␈↓relevance␈α
to␈α
a␈α∞proposed␈α
question,␈α
but␈α
the␈α∞␈↓βcontents␈↓␈α
of␈α
that␈α
knowledge␈α∞varies␈α
from␈α
individual␈α∞to␈α
individual.
␈↓ x␈↓The␈α∂hypothesis␈α∞is␈α∂that␈α∂all␈α∞the␈α∂experts␈α∞can␈α∂be␈α∂said␈α∞to␈α∂consist␈α∞of␈α∂categorized␈α∂information,␈α∞where␈α∂the␈α∂set␈α∞of
␈↓ x␈↓categories␈αis␈α
fairly␈αstandard,␈α
and␈αindicates␈αthe␈α
␈↓βtypes␈↓␈αof␈α
questions␈αany␈αexpert␈α
can␈αbe␈α
expected␈αto␈α
answer.␈αAn
␈↓ x␈↓expert␈α⊂is␈α⊂considered␈α⊂␈↓βequivalent␈↓␈α⊃to␈α⊂his␈α⊂answers␈α⊂to␈α⊂several␈α⊃standard␈α⊂questions.␈α⊂ Each␈α⊂expert␈α⊂has␈α⊃the␈α⊂same
␈↓ x␈↓mental␈α⊃"parts",␈α∩it␈α⊃is␈α∩only␈α⊃the␈α∩values␈α⊃stored␈α∩in␈α⊃these␈α∩parts,␈α⊃their␈α∩contents,␈α⊃which␈α∩distinguish␈α⊃him␈α∩as␈α⊃an
␈↓ x␈↓individual.
␈↓"%␈↓ x␈↓␈↓ ↓XArmed␈α
with␈α
this␈α
dubious␈αview␈α
of␈α
intelligence,␈α
let␈αus␈α
return␈α
to␈α
the␈αdesign␈α
of␈α
BEINGs.␈α
Each␈αBEING
␈↓ x␈↓shall␈αhave␈α
many␈αparts,␈α
each␈αpossessing␈α
a␈αname␈α
(a␈αquestion␈α
it␈αdeals␈α
with)␈αand␈α
a␈αvalue␈α
(a␈αprocedure␈αcapable␈α
of
␈↓ x␈↓answering␈α∞that␈α
question).␈α∞ Henceforth,␈α∞"␈↓βpart␈↓"␈α
will␈α∞be␈α∞used␈α
in␈α∞this␈α
technical␈α∞sense.␈α∞ When␈α
a␈α∞BEING␈α∞asks␈α
a
␈↓ x␈↓question,␈αit␈αis␈αreally␈αjust␈αone␈αpart␈αwho␈αis␈αasking.␈αIn␈αfact,␈αit␈αmust␈αbe␈αthat␈αthe␈α␈↓βvalue␈↓␈αsubpart␈αof␈αsome␈αpart␈αcan't
␈↓ x␈↓answer␈α
␈↓βhis␈↓␈α
question␈α
without␈αfurther␈α
assistance.␈α
He␈α
may␈αnot␈α
know␈α
enough␈α
to␈αcall␈α
on␈α
specific␈α
other␈αBEINGs
␈↓ x␈↓(let␈αanyone␈αrespond␈αwho␈αfeels␈αrelevant),␈αbut␈αhe␈αshould␈α␈↓βalways␈↓␈αspecify␈αwhat␈αBEING␈α␈↓βpart␈↓␈αthe␈αquestion␈αshould
␈↓ x␈↓be␈α
answered␈α
by.␈α
 By␈α
analogy␈α∞with␈α
the␈α
experts,␈α
each␈α
BEING␈α
will␈α∞have␈α
the␈α
same␈α
"universal"␈α
set␈α
of␈α∞types␈α
of
␈↓ x␈↓parts␈α
(will␈αanswer␈α
the␈α
same␈αkinds␈α
of␈α
queries),␈αand␈α
this␈α
uniformity␈αshould␈α
permit␈αpainless␈α
intercommunication.
␈↓ x␈↓Since␈αthe␈αparadigm␈αof␈αthe␈αmeeting␈αis␈αquestioning␈αand␈α
answering,␈αthe␈αnames␈αof␈αthe␈αparts␈αshould␈αcover␈αall␈α
the
␈↓ x␈↓types␈α
of␈α
questions␈α
one␈α
expert␈α
wants␈α
to␈α
ask␈α
another.␈α
Each␈α
part␈α
of␈α
each␈α
BEING␈α
will␈α
have␈α
implicit␈α
access␈αto␈α
this
␈↓ x␈↓list:␈α
it␈α
may␈α
ask␈α
only␈α
these␈α
types␈α
of␈α
questions.␈α
Each␈α
BEING␈α
should␈α
␈↓βnot␈↓␈α
have␈α
access␈α
to␈α
the␈α
list␈α
of␈α
all␈αBEINGs␈α
in
␈↓ x␈↓the␈α⊂system:␈α⊂requests␈α⊂should␈α⊂be␈α⊂phrased␈α⊂in␈α⊂terms␈α⊂of␈α⊂what␈α⊂is␈α⊂wanted;␈α⊂rarely␈α⊂is␈α⊂the␈α⊂name␈α⊂of␈α⊃the␈α⊂answerer
␈↓ x␈↓specified␈αin␈α
advance.␈α (By␈αanalogy:␈α
the␈αhuman␈αspeaker␈α
is␈αnot␈α
aware␈αof␈αprecisely␈α
who␈αis␈αin␈α
the␈αroom;␈αwhen␈α
he
␈↓ x␈↓feels␈α
inadequate,␈α
he␈α
asks␈α
for␈α
help␈α
and␈α
hopes␈α
someone␈α
responds).␈α
 Another␈α
point␈α
is␈α
that␈α
BEINGs␈α
are␈α
not␈α
a
␈↓ x␈↓recursive␈α⊂concept␈α⊂(like␈α⊂ACTORs[3]␈α⊂are):␈α⊂a␈α⊂part␈α⊂of␈α⊂a␈α⊂BEING␈α⊂is␈α⊂a␈α⊂brief␈α⊂collection␈α⊂of␈α⊃knowledge␈α⊂(usually
␈↓ x␈↓procedural),␈α
not␈αanother␈α
BEING;␈αa␈α
collection␈α
of␈αBEINGs␈α
(also␈αcalled␈α
a␈α
community,␈αa␈α
pool,␈αthe␈α
system,␈α
or␈αa
␈↓ x␈↓group) is also not itself a BEING. There are no ␈↓βstructured␈↓ clusters of BEINGs.
␈↓"%␈↓ x␈↓␈↓ ↓XOnce␈α
again:␈α
the␈α
concept␈α
of␈α
a␈α
pool␈α
of␈α
BEINGs␈α
is␈α
that␈α
many␈α
entities␈α
coexist,␈α
each␈α
having␈α
a␈α
complex
␈↓ x␈↓structure,␈αbut␈αthat␈αstructure␈αdoes␈αnot␈αvary␈αfrom␈α
BEING␈αto␈αBEING.␈α This␈αidea␈αhas␈αanalogues␈αin␈αmany␈α
fields:
␈↓ x␈↓transactional analysis in psychology, anatomy in medicine, modular design in architechture.

␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α␈↓ε5␈↓␈↓ λ
␈↓ x␈↓␈↓ ↓XHow␈α∞can␈α∞we␈α∞test␈α∞out␈α∞this␈α∞idea?␈α∞We␈α∞must␈α
build␈α∞a␈α∞pool␈α∞of␈α∞BEINGs,␈α∞a␈α∞modular␈α∞program␈α∞which␈α
will
␈↓ x␈↓interact␈α∞with␈α∞a␈α∂human␈α∞user␈α∞and␈α∂generate␈α∞the␈α∞CF␈α∂program.␈α∞ Recasting␈α∞the␈α∂idea␈α∞into␈α∞operational␈α∂terms,␈α∞we
␈↓ x␈↓arrive at this procedure for writing a pool of BEINGs:
␈↓"%␈↓ x␈↓(1) Study the task which the pool is to do. See what kinds of questions are asked by simulated experts.
␈↓"%␈↓ x␈↓(2)␈αDistill␈α
this␈αinto␈αa␈α
core␈αof␈α
simple␈αquestions,␈αQ,␈α
in␈αsuch␈α
a␈αway␈αthat␈α
each␈αinter-expert␈α
question␈αor␈αtransfer␈α
of
␈↓ x␈↓control␈αcan␈αbe␈αrephrased␈αin␈αterms␈αof␈αQ.␈αThe␈αsize␈αof␈αQ␈αis␈αvery␈αimportant.␈α If␈αQ␈αis␈αtoo␈αlarge,␈αaddition␈αof␈αnew
␈↓ x␈↓BEINGs␈αwill␈αdemand␈αeither␈αgreat␈αeffort␈αor␈αgreat␈αintelligence␈α(an␈αexample␈αof␈αa␈αsystem␈αlike␈αthis␈αis␈αACTORS).
␈↓ x␈↓If␈α
Q␈α
is␈αtoo␈α
small,␈α
all␈α
the␈αnon-uniformity␈α
is␈α
simply␈α
pushed␈αdown␈α
into␈α
the␈α
values␈αof␈α
one␈α
or␈α
two␈αgeneral␈α
catchall
␈↓ x␈↓questions (all first-order logical languages do this).
␈↓"%␈↓ x␈↓(3)␈α
List␈α
all␈α
the␈α
BEINGs␈α
who␈αwill␈α
be␈α
present␈α
in␈α
the␈α
pool,␈α
and␈αfill␈α
in␈α
their␈α
parts.␈α
The␈α
time␈α
to␈αencode␈α
knowledge
␈↓ x␈↓into␈α∞many␈α∞simple␈α∞representation␈α∞schemes␈α∞is␈α∞proportional␈α∞to␈α∞the␈α∞square␈α∞of␈α∞(occasionally␈α∞exponential␈α∂in)␈α∞the
␈↓ x␈↓amount␈α∂of␈α∂interrelated␈α∂knowledge␈α∂(e.g.,␈α∂consider␈α∂the␈α⊂frame␈α∂problem).␈α∂ The␈α∂filling␈α∂in␈α∂of␈α∂a␈α∂new␈α⊂BEING␈α∂is
␈↓ x␈↓␈↓βindependent␈↓␈α∃of␈α∀the␈α∃number␈α∀of␈α∃BEINGs␈α∀already␈α∃in␈α∀the␈α∃pool,␈α∀because␈α∃BEINGs␈α∀can␈α∃communicate␈α∀via
␈↓ x␈↓nondeterministic␈α
goal␈α
mechanisms,␈α
and␈αnot␈α
have␈α
to␈α
know␈αthe␈α
names␈α
of␈α
the␈αBEINGs␈α
who␈α
will␈α
answer␈αtheir
␈↓ x␈↓queries.␈αThis␈α
filling␈αin␈α
is␈α ␈↓βlinear␈↓␈αin␈α
the␈αnumber␈α
of␈αBEING␈α
parts␈αlisted␈αin␈α
Q;␈αall␈α
parts␈αof␈α
all␈αBEINGs␈αmust␈α
be
␈↓ x␈↓(at least, should be) filled in.
␈↓"%␈↓ x␈↓(4) The human user interacts with the completed BEING community, until the desired task is complete.
␈↓"%␈↓ x␈↓␈↓ ↓XIn␈αsection␈α5,␈αthe␈αgain␈αfrom␈αconstraining␈αthat␈αQ␈αbe␈αconstant␈α(over␈αall␈αthe␈αBEINGs␈αin␈αthe␈αsystem)␈αwill
␈↓ x␈↓be␈αclarified.␈αTheoretical␈αaspects␈αof␈αBEING␈αsystems␈αfollow,␈αin␈αsection␈α6.␈αNext␈αcomes␈αan␈αevaluation␈αof␈αPUP6's
␈↓ x␈↓behavior.  The uses and the problems with BEINGs are summed up in the final section.
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α>␈↓∧␈↓&5. The parts of a BEING␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XA␈αset␈α
of␈α29␈αubiquitous␈α
questions␈αwere␈αchosen,␈α
representing␈αeverything␈αone␈α
expert␈αmight␈αwant␈α
to␈αask
␈↓ x␈↓another.␈αAt␈αleast,␈α
they␈αnaturally␈αencompass␈α
those␈αquestions␈αwhich␈αwere␈α
asked␈αduring␈αthe␈α
simulated␈αmeeting,
␈↓ x␈↓hence␈α∞should␈α∂be␈α∞sufficient␈α∞for␈α∂generating␈α∞CF.␈α∞ Q,␈α∂this␈α∞universal␈α∞set␈α∂of␈α∞BEING␈α∞parts,␈α∂is␈α∞listed␈α∞in␈α∂the␈α∞box
␈↓ x␈↓below.  The percentage of the experimental pool  of BEINGs which actually needed each part is also noted.

␈↓"¬␈↓ x␈↓α␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α9␈↓↓␈↓&BEING Parts␈↓)αβ␈↓α

␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓WHAT␈↓α 82%  A brief summary of the global purpose, a template for an English phrase.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓WHY␈↓α 77%  A justification of the BEING's existence, partly filled in by the BEING who called it.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓HOW␈↓α 72%  A summary of the methods the BEING intends to employ. Global strategies.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓IDEN␈↓α 54%  How this BEING is referenced in English phrases. Implemented as productions.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓WHEN␈↓α 19%  Why this BEING should be given control now. The sum of weighted factors.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓ARGS␈↓α 63%  Number of arguments; which are optional; names of any local variables.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓ARG-CHECK␈↓α 81%  Predicates which examine each argument for suitability.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓EVAL-ARGS␈↓α  4%  Which arguments are to be evaluated, and which quoted.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓REQUISITES␈↓α 10%  If this BEING is actually chosen, what must be made true prior to (pre)
␈↓"¬␈↓ x␈↓α␈↓ ↓_        during (co), and after (post) execution.  Work to make these true (unlike ARG-CHECK).
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓DEMONS␈↓α 7%  Set of demons to be kept active while the BEING is in control.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓INHIBIT-DEMONS␈↓α  5%  A lock/unlock mechanism, useful when handling demonic interrupts.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓EFFECTS␈↓α 27%  What will probably be true after this BEING is through. What it achieves.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓RESULTS␈↓α 15%  How many values this returns. What domain each is in. Side effects.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓META-CODE␈↓α 70%  The body of the executable code, but with uninstantiated subparts.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓COMMENTS␈↓α 16%  Hints for filling in undefined sections of other BEING parts.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓STRUCTURE␈↓α 4% Viewing this BEING as a data structure, the operations doable to it.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓AFFECTS␈↓α 14%  Other BEINGs which might be called by this BEING, and why.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓COMPLEXITY␈↓α 92%  A vector of utility measures, including probable ease of calling,
␈↓"¬␈↓ x␈↓α␈↓ ↓_        success, (calling)* itself, time/space cost, efficiency of its output results.
␈↓"∧␈↓ x␈↓α␈↓ ↓_␈↓↓GENERALIZATIONS␈↓α 27%  Some other BEINGs, encompassing this one.␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α␈↓ε6␈↓␈↓ λ
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓ALTERNATIVES␈↓α 16%  Some equivalent BEINGs (to try if this one fails).
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓SPECIALIZATIONS␈↓α 40%  How to write a streamlined, special-case version of this BEING.
␈↓"¬␈↓ x␈↓α␈↓ ↓_␈↓↓ENCODABLE␈↓α 9%  How to order the evaluation of the other parts for self-streamlining.
␈↓ x␈↓␈↓ ↓XEach␈αof␈αthe␈α100␈αBEINGs␈αin␈αPUP6␈αshould␈αhave␈αhad␈αa␈αvalue␈αfor␈αeach␈αpart␈α(in␈αreality,␈αonly␈α40%␈αwere
␈↓ x␈↓filled␈α∞in;␈α∞only␈α∂30%␈α∞were␈α∞actually␈α∞necessary␈α∂to␈α∞generate␈α∞CF).␈α∞A␈α∂value␈α∞for␈α∞a␈α∞part␈α∂is␈α∞simply␈α∞a␈α∂LISP␈α∞program
␈↓ x␈↓which␈α
can␈αanswer␈α
that␈αquestion,␈α
often␈αby␈α
asking␈αquestions␈α
of␈α
the␈αsame␈α
BEING,␈αof␈α
other␈αBEINGs,␈α
and␈αof␈α
the
␈↓ x␈↓user.␈α
A␈α
part␈α
may␈α
also␈α
assert␈α
some␈α
fact,␈α
create␈α
or␈α
modify␈α
some␈α
structure␈α
(including␈α
demons,␈α
BEINGs,␈αand␈α
parts
␈↓ x␈↓of␈α∂BEINGs).␈α∂ Appendix␈α∂1␈α∂shows␈α∂the␈α∂values␈α∂stored␈α∂under␈α∂each␈α∂part␈α∂for␈α∂a␈α∂typical␈α∂BEING.␈α∂The␈α⊂reader␈α∂is
␈↓ x␈↓invited to glance over this table now.
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈α
set␈α
of␈α
parts␈α
breaks␈α
into␈α
three␈α
rough␈α
categories:␈α
those␈α
parts␈α
which␈α
are␈α
useful␈α
in␈α
deciding␈αwhich
␈↓ x␈↓BEING␈αgets␈αcontrol,␈αthose␈αuseful␈αonly␈αto␈αanswer␈αthe␈αuser's␈αquestions␈αand␈αkeep␈αhim␈αoriented,␈αand␈αthose␈αused
␈↓ x␈↓once the BEING gains control.  Each of these three groups of parts will be described.
␈↓"%␈↓ x␈↓␈↓↓␈↓&5.1. When a BEING Gains Control␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XAt␈α⊃the␈α⊃humans'␈α⊂meeting,␈α⊃only␈α⊃one␈α⊂expert␈α⊃spoke␈α⊃at␈α⊂a␈α⊃time;␈α⊃in␈α⊂the␈α⊃BEINGs␈α⊃community,␈α⊃only␈α⊂one
␈↓ x␈↓BEING␈α∞has␈α∞control␈α∞at␈α∞any␈α∞given␈α∞moment.␈α∞He␈α∞uses␈α∞his␈α∞parts␈α∞to␈α∞do␈α∞things␈α∞(ask,␈α∞create,␈α∞modify),␈α∞and␈α∞yields
␈↓ x␈↓control either voluntarily or through interruption.
␈↓"%␈↓ x␈↓␈↓ ↓XIn␈α
slightly␈α
more␈α
procedural␈α
terms,␈α
the␈α
scenario␈α
is␈α
as␈α
follows.␈α
One␈α
part␈α
of␈α
a␈α
BEING␈α
senses␈α
its␈α
relevance
␈↓ x␈↓(often␈α
the␈αIDEN␈α
or␈α
EFFECTS␈αparts,␈α
which␈α
are␈αunited␈α
with␈α
all␈αsuch␈α
parts␈α
to␈αform␈α
a␈α
large␈αproduction␈α
system).
␈↓ x␈↓If␈α∃more␈α∃than␈α∃one␈α∃BEING␈α∀wants␈α∃control␈α∃at␈α∃any␈α∃time,␈α∀a␈α∃special␈α∃BEING,␈α∃CHOOSER,␈α∃seizes␈α∀control
␈↓ x␈↓momentarily.␈αHe␈α
asks␈αeach␈α
competing␈αBEING␈α
to␈αevaluate␈αits␈α
WHEN␈αpart,␈α
to␈αsee␈α
how␈αseriously␈α
it␈αneeds␈αto␈α
go
␈↓ x␈↓immediately.␈α
If␈α∞some␈α
BEINGs␈α
are␈α∞still␈α
tied␈α∞for␈α
first␈α
place,␈α∞he␈α
asks␈α
them␈α∞to␈α
evaluate␈α∞their␈α
COMPLEXITY
␈↓ x␈↓parts,␈αto␈αsee␈αwhich␈αis␈αthe␈α
simplest.␈αIf␈αany␈α␈↓βstill␈↓␈αtie␈αfor␈αtop,␈α
one␈αis␈αrandomly␈αchosen.␈αIn␈αany␈αcase,the␈α
winner␈αis
␈↓ x␈↓then␈αpassed␈αcontrol.␈αOnce␈αin␈αcontrol,␈αa␈αBEING␈αarranges␈αsome␈αof␈αits␈αparts␈αin␈αsome␈αorder␈αand␈αevaluates␈αthem.
␈↓ x␈↓For␈α
example,␈α
the␈α
ARGS␈αpart␈α
might␈α
be␈α
first;␈α
if␈αit␈α
asks␈α
for␈α
some␈αarguments␈α
which␈α
no␈α
BEING␈α
has␈αsupplied,
␈↓ x␈↓then␈α
the␈αwhole␈α
BEING␈αmight␈α
decide␈α
to␈αfail.␈α
Some␈α parts,␈α
when␈α
evaluated,␈αmight␈α
create␈αa␈α
new␈αBEING,␈α
might
␈↓ x␈↓ask␈α∀questions␈α∀which␈α∪require␈α∀this␈α∀whole␈α∀process␈α∪to␈α∀repeat␈α∀recursively,␈α∪etc.␈α∀This␈α∀"asking"␈α∀really␈α∪means
␈↓ x␈↓broadcasting␈α∂a␈α∂request␈α∞to␈α∂one␈α∂or␈α∂two␈α∞parts␈α∂of␈α∂every␈α∂BEING;␈α∞for␈α∂example␈α∂"Is␈α∂there␈α∞a␈α∂known␈α∂fast␈α∂way␈α∞of
␈↓ x␈↓gronking␈α∂toves?"␈α∂would␈α∂be␈α∂asked␈α∂as␈α∂a␈α∂search␈α∂for␈α∂a␈α∂BEING␈α∂whose␈α∂COMPLEXITY␈α∂indicated␈α∂speed,␈α∂and
␈↓ x␈↓whose␈α⊂EFFECTS␈α∂part␈α⊂contained␈α∂a␈α⊂production␈α∂with␈α⊂a␈α∂template␈α⊂matching␈α∂"gronking␈α⊂toves".␈α∂ A␈α⊂list␈α⊂of␈α∂the
␈↓ x␈↓responders␈αwould␈αbe␈αreturned.␈α(Incidentally,␈αGERUND␈αwould␈αrecognize␈αthis,␈αbut␈αlater␈αgive␈αup␈αwhen␈αno␈αone
␈↓ x␈↓could␈α∂recognize␈α∂"gronk␈α∂toves".)␈α∂The␈α∂questioner␈α∂might␈α∞pose␈α∂some␈α∂new␈α∂questions␈α∂directly␈α∂to␈α∂these␈α∞BEINGs,
␈↓ x␈↓might␈α
turn␈α
control␈α
over␈α
to␈α
them␈α
directly,␈αetc.␈α
One␈α
way␈α
or␈α
another,␈α
the␈α
BEING␈α
eventually␈αrelinquishes␈α
control.
␈↓ x␈↓If␈αit␈αhad␈αno␈αdirect␈α
successor␈αin␈αmind,␈αall␈αthe␈α
BEINGs␈αare␈αasked␈αif␈αthey␈α
want␈αto␈αtake␈αover.␈α There␈αwill␈α
always
␈↓ x␈↓be␈α⊂␈↓βsome␈↓␈α∂BEING␈α⊂who␈α∂will␈α⊂take␈α∂over;␈α⊂the␈α∂general␈α⊂management␈α∂types␈α⊂of␈α∂BEINGs␈α⊂are␈α∂always␈α⊂able␈α⊂ --␈α∂but
␈↓ x␈↓reluctant  -- to do so.
␈↓"%␈↓ x␈↓␈↓ ↓XHow␈α∞does␈α∞each␈α∞BEING␈α∞decide␈α∞which␈α∞parts␈α∞to␈α∞evaluate,␈α∞and␈α∞in␈α∞which␈α∞order,␈α∞once␈α∞it␈α∞gains␈α
control?
␈↓ x␈↓The␈α∂answer␈α∂might␈α∂seem␈α∂to␈α∂be␈α∂difficult␈α∂or␈α∂tedious␈α∂for␈α∂whoever␈α∂writes␈α∂BEINGs,␈α∂since␈α∂it␈α∂might␈α∂vary␈α∂from
␈↓ x␈↓BEING␈αto␈αBEING.␈α
In␈αfact,␈αit␈αdoesn't!␈α
 The␈αcommitment␈αto␈α
a␈αuniversal␈αset␈αof␈α
BEING␈αparts␈αis␈α
inefficient␈αin
␈↓ x␈↓some␈αways␈α(each␈αBEING␈α␈↓βneeded␈↓␈αonly␈αa␈αthird␈αof␈αall␈αthe␈αparts)␈αbut␈αallows␈αfor␈αsome␈αsimplifications␈αright␈αhere.
␈↓ x␈↓What␈α
parts␈α
should␈α
be␈α
evaluated,␈α
and␈α
in␈α
what␈α
order,␈α
when␈α
a␈α
BEING␈α
gains␈α
control?␈α
This␈α
decision␈α
depends
␈↓ x␈↓primarily␈αon␈αthe␈α␈↓βtypes␈↓␈αof␈αparts␈αpresent␈αin␈αthe␈αBEING,␈αnot␈αon␈αtheir␈α␈↓βvalues␈↓.␈α But␈αevery␈αBEING␈αhas␈αthe␈αsame
␈↓ x␈↓anatomy,␈α∪so␈α∩one␈α∪single␈α∩algorithm␈α∪can␈α∩assemble␈α∪any␈α∩BEING's␈α∪parts␈α∩into␈α∪an␈α∩executable␈α∪LISP␈α∩function.
␈↓ x␈↓Moreover,␈αthis␈αassemby␈αcan␈αbe␈αdone␈α
when␈αthe␈αsystem␈αis␈αfirst␈αloaded␈α
(or␈αwhen␈αa␈αnew␈αBEING␈αis␈αfirst␈α
created),
␈↓ x␈↓and␈αneed␈αonly␈αbe␈αredone␈αfor␈αa␈αBEING␈αwhen␈α
the␈αvalues␈αof␈αits␈αparts␈αchange.␈αSuch␈αchanges␈αare␈α
rare:␈αexperts
␈↓ x␈↓are␈α⊂not␈α⊂often␈α⊂open-minded.␈α⊂The␈α⊂precise␈α⊂algorithm␈α⊃is␈α⊂sketched␈α⊂in␈α⊂the␈α⊂box␈α⊂below.␈α⊂The␈α⊂parts␈α⊃useful␈α⊂here
␈↓ x␈↓include ARGS, DEMONS, META-CODE, COMMENTS, ARG-CHECK, and REQUISITES.



␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α␈↓ε7␈↓␈↓ λ
␈↓ x␈↓α␈α?␈α?␈α?␈α?␈α~␈↓↓␈↓&Algorithm for assembling a BEING into an executable function␈↓)αβ␈↓α
␈↓"∩␈↓ x␈↓α␈↓ ↓_When␈αa␈αBEING␈α
␈↓βB␈↓α␈αfirst␈αgains␈α
control,␈αits␈αEXPLICIT-ARGS␈α
are␈αbound.␈αThe␈α
IMPLICIT-ARGS␈αare␈αinitialized,␈αthe␈α
name
␈↓"π␈↓ x␈↓α␈↓βB␈↓α␈α∂is␈α∂pushed␈α∞onto␈α∂the␈α∂BEING␈α∞control␈α∂stack,␈α∂and␈α∂any␈α∞newly-activated␈α∂DEMONS␈α∂are␈α∞so␈α∂tagged.␈α∂The␈α∂BEING␈α∞who
␈↓"π␈↓ x␈↓αcalled␈α␈↓βB␈↓α␈αshould␈αhave␈αexplained␈αhis␈αreasons␈αby␈αassigning␈αsome␈αphrase␈αto␈αthe␈αvariable␈αBECAUSE.␈α This␈αreason␈αis
␈↓"π␈↓ x␈↓αnow␈α
stored␈α
as␈α
a␈α
special␈α
sub-part␈α
of␈α
the␈α
WHY␈α
part␈α
of␈α
␈↓βB␈↓α.␈α
 BECAUSE␈α
is␈α
rebound␈α
periodically␈α
in␈α
the␈αMETA-CODE
␈↓"π␈↓ x␈↓αand␈αCOMMENTS␈αparts,␈αto␈αkeep␈αcurrent␈αthe␈αexplanation␈αof␈αeach␈αcall␈αthat␈α␈↓βB␈↓α␈αmakes.␈αEach␈αARG-CHECK␈αpredicate␈αis
␈↓"π␈↓ x␈↓αevaluated.␈α∞If␈α∂any␈α∞returns␈α∞NIL,␈α∂the␈α∞entire␈α∂BEING␈α∞reports␈α∞that␈α∂it␈α∞has␈α∞failed;␈α∂otherwise,␈α∞the␈α∂PRE-REQUISITES␈α∞are
␈↓"π␈↓ x␈↓αexamined.␈αEffort␈α
is␈αexpended␈α
to␈αmake␈α
them␈αtrue,␈α
if␈αthey␈α
are␈αcurrently␈α
not␈αsatisfied.␈α
Each␈αCOMMENT␈αis␈α
evaluated,
␈↓"π␈↓ x␈↓αthen␈α∂the␈α∂CO-REQUISITES,␈α∂META-CODE,␈α∞and␈α∂the␈α∂current␈α∂demons␈α∞are␈α∂executed␈α∂in␈α∂pseudo-parallel.␈α∂ Each␈α∞POST-
␈↓"π␈↓ x␈↓αREQUISITE␈αis␈αthen␈α
examined,␈αand␈αan␈α
effort␈αmade␈αto␈αsatisfy␈α
it.␈α The␈αnewly-activated␈α
demons␈αare␈αexorcized,␈α
␈↓βB␈↓α␈αis
␈↓"π␈↓ x␈↓αpopped from the BEING control stack, and the value computed by the META-CODE is returned.
␈↓"∩␈↓ x␈↓α␈↓ ↓_Some␈α
heuristics␈α
were␈α
devised␈α
to␈α
take␈α
advantage␈α
of␈α
the␈αfact␈α
that␈α
the␈α
BEINGs␈α
often␈α
didn't␈α
need␈α
many␈α
of␈αthe
␈↓"π␈↓ x␈↓αstandard␈α
parts.␈α
For␈α
example,␈α
INFO-OBTAINER␈α
has␈α
no␈α
new␈α
demons␈α
or␈α
co-requisites,␈α
so␈α
no␈α
parallel␈α
processing␈α
need
␈↓"π␈↓ x␈↓αbe␈α
simulated.␈α∞ No␈α
results␈α∞are␈α
explicitly␈α∞produced,␈α
so␈α∞the␈α
value␈α∞returned␈α
by␈α∞the␈α
META-CODE␈α∞is␈α
not␈α∞saved␈α
and
␈↓"π␈↓ x␈↓αlater␈α
returned.␈α
 Here␈α
is␈α
the␈α
way␈α
that␈α
BEING␈αactually␈α
appeared␈α
after␈α
assembly;␈α
contrast␈α
this␈α
with␈α
the␈α
values␈αof␈α
its
␈↓"π␈↓ x␈↓αparts, as presented in Appendix 1.  Notice that BECAUSE is set to the reason that CHOOSE-FROM is called.

␈↓"π␈↓ x␈↓α␈↓ ↓8(INFO-OBTAINER
␈↓"π␈↓ x␈↓α␈↓ ↓8   (LAMBDA  (U)
␈↓"π␈↓ x␈↓α␈↓ ↓8        (AND
␈↓"π␈↓ x␈↓α␈↓ ↓8           (PUSH  INFO-OBTAINER  BEING-STACK)
␈↓"π␈↓ x␈↓α␈↓ ↓8           (PUT-APPEND  INFO-OBTAINER  WHY  BECAUSE)
␈↓"π␈↓ x␈↓α␈↓ ↓8           (EVERY  CURRENT-DEMONS  APPLY*)
␈↓"π␈↓ x␈↓α␈↓ ↓8           (SETQQ BECAUSE (WE CAN ONLY TRY TO OBTAIN USABLE INFO IN ONE WAY AT A TIME))
␈↓"π␈↓ x␈↓α␈↓ ↓8           (CHOOSE-FROM
␈↓"π␈↓ x␈↓α␈↓ ↓8                     (GET-NEW-INFORMATION U)
␈↓"π␈↓ x␈↓α␈↓ ↓8                     (TRANSLATE U)
␈↓"π␈↓ x␈↓α␈↓ ↓8                     (ANALYZE-IMPLICATIONS U)
␈↓"π␈↓ x␈↓α␈↓ ↓8                     (EXTRACT-RELEVANT-SUBSET U))
␈↓"π␈↓ x␈↓α␈↓ ↓8           (POP   BEING-STACK))))

␈↓"%␈↓ x␈↓␈↓↓␈↓&5.2. Keeping the User Informed␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XIn␈α~the␈α~earlier␈α→conversation␈α~excerpts,␈α~the␈α→simulated␈α~human␈α~user␈α→had␈α~no␈α~trouble␈α→whatever
␈↓ x␈↓understanding␈α
what␈α
the␈α
simulated␈α∞experts␈α
asked␈α
him.␈α
In␈α
the␈α∞actual␈α
programmed␈α
PUP6␈α
system,␈α∞the␈α
human
␈↓ x␈↓who␈αwas␈αsitting␈αat␈αthe␈αteletype␈αquite␈α␈↓βrarely␈↓␈αunderstood␈αwhat␈αwas␈αwanted␈αby␈αthe␈αBEINGs.␈αHe␈αfrequently␈αhad
␈↓ x␈↓to␈α
interrupt␈α
them␈α
and␈α
ask␈α
them␈α
questions␈α
about␈α
who␈αwas␈α
in␈α
control,␈α
why,␈α
what␈α
he␈α
was␈α
trying␈α
to␈α
do,␈αwhat␈α
had
␈↓ x␈↓recently␈α∞transpired,␈α∂etc.␈α∞These␈α∂ideally␈α∞can␈α∞be␈α∂phrased␈α∞as␈α∂simple␈α∞retrievals␈α∞and␈α∂EVALs␈α∞of␈α∂active␈α∞BEINGs'
␈↓ x␈↓parts.␈α
 The␈α
BEING␈α
parts␈α
most␈α∞often␈α
called␈α
for␈α
by␈α
the␈α∞user␈α
are␈α
the␈α
simple␈α
one-line␈α∞"orientation"␈α
templates.
␈↓ x␈↓These␈α⊗include␈α↔WHAT,␈α⊗HOW,␈α⊗WHY,␈α↔and␈α⊗AFFECTS.␈α⊗ For␈α↔theoretical␈α⊗reasons␈α⊗explained␈α↔later,␈α⊗the
␈↓ x␈↓synthesized␈α∂program,␈α∞CF,␈α∂was␈α∂writen␈α∞as␈α∂a␈α∂pool␈α∞of␈α∂BEINGs␈α∂itself␈α∞(by␈α∂PUP6,␈α∂but␈α∞not␈α∂during␈α∂the␈α∞protocol.
␈↓ x␈↓Actually,␈α⊃a␈α⊃fortuitous␈α⊃"bug"␈α⊂in␈α⊃PUP6␈α⊃created␈α⊃this␈α⊂intriguing␈α⊃situation.)␈α⊃Although␈α⊃its␈α⊂question-answering
␈↓ x␈↓ability␈αis␈αinferior␈αto␈αPUP6,␈αthe␈αfact␈αthat␈αit␈αhas␈α␈↓βany␈↓␈αsuch␈αpower␈αwas␈αsurprising␈αto␈αthe␈αauthor.␈αIn␈αother␈αwords,
␈↓ x␈↓one␈α
can␈α
interrupt␈α
the␈α
target␈α
program␈α
as␈α
it␈α
is␈α
running␈α
and␈α
ask␈α
questions.␈α
Any␈α
BEING␈α
on␈α
the␈α
control␈α
stack␈α
will
␈↓ x␈↓provide␈α
fully␈αinstantiated␈α
answers␈α
to␈αany␈α
of␈α
its␈α29␈α
allowable␈α
queries␈α(its␈α
parts);␈α
all␈αother␈α
BEINGs␈αwill␈α
provide
␈↓ x␈↓only␈α∞hypothetical␈α
answers.␈α∞As␈α
an␈α∞example,␈α
consider␈α∞this␈α
actual␈α∞excerpt␈α
of␈α∞a␈α
human␈α∞using␈α
the␈α∞CF␈α
program
␈↓ x␈↓synthesized␈αby␈αPUP6.␈α (Some␈αliberty␈αhas␈αbeen␈αtaken␈αwith␈αthe␈αEnglish;␈αe.g.,␈αthe␈αuser␈αreally␈αtypes␈α␈↓βWHAT?␈↓,␈αnot
␈↓ x␈↓␈↓βWhat␈αare␈αyou␈αdoing?␈↓)␈α"???"␈αsimply␈αmeans␈α"guess␈αthe␈αname␈αof␈αthe␈αscene␈αwith␈αthese␈αobjects␈αand␈αrelations".␈α CF
␈↓ x␈↓types in boldface, the user in italics.

␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α␈↓ε8␈↓␈↓ λ
␈↓ x␈↓↓READY TO ACCEPT BRAND NEW SCENE:   ␈↓β(??? (A B) (Block A) (Green B) (Touches A B))␈↓↓
␈↓"	␈↓ x␈↓↓NOT AN ARCH. NOT A TOWER. MIGHT BE A ROW.  MY GUESS IS ROW.
␈↓"	␈↓ x␈↓↓WHAT IS IT REALLY?                 ␈↓βNothing␈↓↓
␈↓"	␈↓ x␈↓↓TRANSFER FROM MAY TO MUST SETS OF ROW: (TOUCH A B)
␈↓"	␈↓ x␈↓↓READY TO ACCEPT BRAND NEW SCENE:
␈↓"%␈↓ x␈↓␈↓ ↓XHere␈α∃is␈α⊗the␈α∃␈↓βsame␈↓␈α∃excerpt,␈α⊗but␈α∃with␈α∃the␈α⊗user␈α∃interrupting␈α∃occasionally.␈α⊗ "␈↓βU:␈↓"␈α∃signifies␈α⊗he␈α∃has
␈↓ x␈↓interrupted, and "␈↓βOK.␈↓" signifies he has told CF "␈↓βOK, go back to work now.␈↓"

␈↓"	␈↓ x␈↓↓        ␈↓βU:  What are you doing?␈↓↓
␈↓"	␈↓ x␈↓↓CF:  Divide a domain into subdomains, using specialized version of PARTITION-DOMAIN.
␈↓"	␈↓ x␈↓↓        ␈↓βHow?␈↓↓
␈↓"	␈↓ x␈↓↓CF:  By building up the partition gradually, by adding one <scene, name> pair.    ␈↓βOK.␈↓↓
␈↓"∃␈↓ x␈↓↓READY TO ACCEPT BRAND NEW SCENE:   ␈↓β(??? (A B) (Block A) (Green B) (Touches A B))␈↓↓
␈↓"∃␈↓ x␈↓↓        ␈↓βU:  What are you doing?␈↓↓
␈↓"	␈↓ x␈↓↓CF:  Break OBJECTS piece off (??? (A B) (BLOCK A) (GREEN B) (TOUCHES A B)) to examine.
␈↓"	␈↓ x␈↓↓        ␈↓βWhat code is the current BEING going to execute?␈↓↓
␈↓"	␈↓ x␈↓↓CF:  The meta-code of SOME-PART-OF␈↓#v1␈↓#␈↓#v4␈↓# is (CADR arg1). Evals to (A B).    ␈↓βOK.␈↓↓
␈↓"∃␈↓ x␈↓↓NOT AN ARCH.
␈↓"	␈↓ x␈↓↓        ␈↓βU:  Why?␈↓↓
␈↓"	␈↓ x␈↓↓CF: The ARCH MUST-relation (SUPPORTS A B) is contradicted in the SCENE.    ␈↓βOK.␈↓↓
␈↓"∃␈↓ x␈↓↓NOT A TOWER. MIGHT BE A ROW.  MY GUESS IS ROW.
␈↓"	␈↓ x␈↓↓WHAT IS IT REALLY?                 ␈↓βU:  Why are you asking?␈↓↓
␈↓"	␈↓ x␈↓↓CF: CF cannot resolve it, and must at this time, so CF asks USER
␈↓"	␈↓ x␈↓↓        ␈↓βOK. What you wanted is:   Nothing␈↓↓
␈↓"∃␈↓ x␈↓↓        ␈↓βU:  What are you doing?␈↓↓
␈↓"	␈↓ x␈↓↓CF: Go through space of concept names, and make user aware of changes.
␈↓"	␈↓ x␈↓↓        ␈↓βWhat BEINGs might this affect?␈↓↓
␈↓"	␈↓ x␈↓↓CF: MESSAGE is possibly called; some version of TEST is possibly called.     ␈↓βOK.␈↓↓
␈↓"∃␈↓ x␈↓↓TRANSFER FROM MAY TO MUST SETS OF ROW: (TOUCH A B)
␈↓"	␈↓ x␈↓↓        ␈↓βU:  Why?␈↓↓
␈↓"	␈↓ x␈↓↓CF: Change model of ROW until it contradicts the previous SCENE.           ␈↓βOK.␈↓↓
␈↓"∃␈↓ x␈↓↓READY TO ACCEPT BRAND NEW SCENE:

␈↓"%␈↓ x␈↓␈↓↓␈↓&5.3. Passing Control in the System␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XRather␈α⊃than␈α⊃delve␈α⊃into␈α⊃detailed␈α⊃discussions␈α∩about␈α⊃each␈α⊃BEING␈α⊃part␈α⊃which␈α⊃is␈α⊃useful␈α∩in␈α⊃passing
␈↓ x␈↓control,␈αanother␈αconversation␈αexample␈αwill␈αbe␈αpresented,␈αannotated␈αas␈αto␈αwhy␈αeach␈αBEING␈αwas␈αable␈αto␈αgain
␈↓ x␈↓control␈α∂just␈α∂when␈α∂it␈α∂did.␈α∂ The␈α∂BEINGs␈α∂in␈α∂PUP6␈α∂have␈α∂the␈α∂same␈α∂names,␈α∂and␈α∂carry␈α∂on␈α∂the␈α∂same␈α∂kind␈α∂of
␈↓ x␈↓discussion,␈α
as␈α
our␈αearlier␈α
experts␈α
do.␈α Below,␈α
they␈α
work␈α
together␈αto␈α
create␈α
the␈αidea␈α
of␈α
structuring␈αCF's␈α
internal
␈↓ x␈↓models'␈α
features␈α
into␈α
three␈α
disjoint␈α
sets:␈α
MUSTNOT,␈α∞MUST,␈α
and␈α
MAY.␈α
 After␈α
each␈α
speaker's␈α
name␈α∞is,␈α
in
␈↓ x␈↓brackets,␈α
a␈αbrief␈α
sketch␈αof␈α
the␈αparts␈α
involved␈αin␈α
this␈αparticular␈α
interaction.␈αThese␈α
parts␈αare␈α
typically␈αIDEN,
␈↓ x␈↓WHEN,␈α∂COMPLEXITY,␈α∂EFFECTS,␈α∂and␈α⊂DEMONS.␈α∂It␈α∂is␈α∂advised␈α∂that␈α⊂the␈α∂excerpt␈α∂be␈α∂read␈α⊂twice:␈α∂once,
␈↓ x␈↓ignoring␈α∀the␈α∀information␈α∀in␈α∀the␈α∀brackets,␈α∀[...],␈α∀to␈α∀understand␈α∀␈↓βwhat␈α∀the␈α∀BEINGs␈α∀are␈α∀saying␈↓␈α∃(how␈α∀the
␈↓ x␈↓MUST/MUSTNOT/MAY idea evolves), and once to attain a feel for how the control-passing parts work.
␈↓"%␈↓ x␈↓␈↓ ↓XAssume␈α
the␈α
"meeting"␈α
of␈α
BEINGs␈αhas␈α
progressed␈α
to␈α
the␈α
stage␈αwhere␈α
they␈α
know␈α
that␈α
CF␈αmust␈α
compare
␈↓ x␈↓the␈α
inputted␈α
scene␈αagainst␈α
internally␈α
stored␈αmodels␈α
of␈α
named␈αscenes.␈α
 When␈α
a␈αmodel␈α
is␈α
found␈α
which␈αpasses
␈↓ x␈↓that␈αcomparison,␈αits␈αname␈αis␈αreported␈αas␈αCF's␈αguess␈αfor␈αthe␈αscene.␈αA␈αpredicate,␈αa␈αtest,␈αis␈αdesired.␈αThe␈αexperts
␈↓ x␈↓ask each other and the user ␈↓βHow do we know when a model has passed or failed the test?␈↓
␈↓"∃␈↓ x␈↓INFO-OBTAINER:␈α∞[␈↓¬Final␈α∂course␈α∞of␈α∂action␈α∞in␈α∞META-CODE␈α∂part.␈↓]␈α∞I␈α∂am␈α∞going␈α∞to␈α∂consult␈α∞the␈α∂user␈α∞on
␈↓"	␈↓ x␈↓this matter. Someone take this message, please.
␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α␈↓ε9␈↓␈↓ λ
␈↓ x␈↓MESSENGER:␈α∂[␈↓¬IDEN␈α∂part␈α∂recognizes␈α∂relevance;␈α∞META-CODE␈α∂part␈α∂types␈α∂on␈α∂teletype.␈↓]␈α∂ User,␈α∞how
␈↓"	␈↓ x␈↓do we know whether the model we're examining is suitable?
␈↓"∃␈↓ x␈↓USER: Fail if any feature of the model is in contradiction to the features of the scene.
␈↓"∃␈↓ x␈↓LOOP:␈α⊗[␈↓¬The␈α∃IDEN␈α⊗part␈α⊗template␈α∃"←act␈α⊗if␈α∃any␈α⊗←x␈α⊗ε␈α∃←X␈α⊗satisfies␈α∃P(x)"␈α⊗matches␈α⊗the␈α∃last
␈↓"	␈↓ x␈↓¬sentence.␈α⊂"act"␈α⊂is␈α⊂bound␈α∂to␈α⊂"fail",␈α⊂"x"␈α⊂to␈α∂"feature",␈α⊂etc.␈↓]␈α⊂The␈α⊂outer␈α∂structure␈α⊂called␈α⊂for␈α⊂is␈α∂a
␈↓"	␈↓ x␈↓loop through the model's features, terminated when some test fails.
␈↓"∃␈↓ x␈↓CONTRADICTION:␈α⊃[␈↓¬The␈α⊃IDEN␈α⊃part␈α⊃can␈α⊃handle␈α⊃"←x␈α⊂is␈α⊃in␈α⊃contradiction␈α⊃to␈α⊃the␈α⊃set␈α⊃←Y"␈↓.]␈α⊂This
␈↓"	␈↓ x␈↓failure␈αshould␈αoccur␈αprecisely␈αwhen␈αthe␈αmodel␈αdemands␈αsome␈αproperty␈αof␈αa␈αscene␈αfeature␈αwhich␈αthat␈αfeature
␈↓"	␈↓ x␈↓doesn't possess.
␈↓"∃␈↓ x␈↓SCENE:␈α∪[␈↓¬The␈α∪IDEN␈α∪part␈α∪prods␈α∪the␈α∀STRUCTURE␈α∪part,␈α∪which␈α∪already␈α∪understands␈α∪what␈α∀a␈α∪scene
␈↓"	␈↓ x␈↓¬is␈α⊗composed␈α⊗of,␈α∃in␈α⊗particular␈α⊗what␈α∃constraints␈α⊗are␈α⊗imposed␈α∃on␈α⊗the␈α⊗features.␈↓]␈α⊗The␈α∃only
␈↓"	␈↓ x␈↓properties␈αthe␈αscene's␈αfeatures␈α
can␈αpossess␈αare␈αthat␈α
of␈αabsence␈αand␈αpresence.␈α
The␈αuser␈αhas␈αalready␈α
ruled␈αout
␈↓"	␈↓ x␈↓negation and combination of features.
␈↓"∃␈↓ x␈↓TRANSLATE:␈α∂[␈↓¬EFFECTS␈α∂include␈α∂uniting␈α⊂two␈α∂related␈α∂sentences,␈α∂using␈α∂substitution.␈↓]␈α⊂Thus␈α∂the
␈↓"	␈↓ x␈↓predicate␈αshould␈αreturn␈αfailure␈αif␈αthe␈αmodel␈αspecifies␈αthat␈αa␈αfeature␈αbe␈αpresent␈αand␈αit␈αis␈αactually␈αabsent,␈αor␈αif
␈↓"	␈↓ x␈↓the model specifies a feature be absent when it is actually present in the scene.
␈↓"∃␈↓ x␈↓PGMMER:␈α
[␈↓¬WHEN␈α
part␈α
has␈α
a␈α
very␈α
high␈α
rating␈α
now.␈↓]␈α
I␈α
believe␈α
that,␈α
with␈α
LOOP's␈α
help,␈α∞we␈α
understand
␈↓"	␈↓ x␈↓this enough to program it.
␈↓"∃␈↓ x␈↓CODER:␈α∪[␈↓¬EFFECTS␈α∪indicate␈α∩this␈α∪is␈α∪the␈α∪only␈α∩BEING␈α∪who␈α∪should␈α∪be␈α∩active␈α∪now.␈↓]␈α∪Give␈α∪me␈α∩the
␈↓"	␈↓ x␈↓arguments and the body of the code.
␈↓"∃␈↓ x␈↓LOOP:␈α∩[␈↓¬DEMON␈α∩recently␈α⊃set␈α∩up␈α∩just␈α⊃for␈α∩this␈α∩occasion␈α⊃is␈α∩triggered.␈α∩ SPECIALIZATIONS␈α⊃part
␈↓"	␈↓ x␈↓¬now␈α⊂produces␈α∂streamlined␈α⊂version␈α∂of␈α⊂LOOP␈α∂BEING␈α⊂itself.␈↓]␈α∂The␈α⊂arguments␈α∂are␈α⊂two␈α∂pointers:␈α⊂to␈α∂the
␈↓"	␈↓ x␈↓features␈αof␈αa␈αmodel␈αand␈αto␈αthe␈αfeatures␈αof␈αa␈αscene.␈α The␈αouter␈αstructure␈αis␈αto␈αstep␈αthrough␈αthe␈αmodel␈αfeatures
␈↓"	␈↓ x␈↓until the test returns failure. If this never happens, then report success.
␈↓"∃␈↓ x␈↓PGMMER:␈α→[␈↓¬DEMON␈α→triggered␈α_here␈α→also.␈α→COMMENTS␈α→part␈α_contributes␈α→knowledge␈α→of␈α→how␈α_to
␈↓"	␈↓ x␈↓¬handle␈α∂cases␈α∂at␈α∂this␈α∂level.␈↓]␈α∂The␈α∂testing␈α∂function␈α∂must␈α∂check␈α∂each␈α∂model␈α∂feature.␈α∂If␈α∂it␈α∂is␈α⊂one␈α∂which
␈↓"	␈↓ x␈↓must␈αbe␈αpresent␈αin␈αthe␈αobserved␈αscene,␈αthen␈αcheck␈αfor␈αthat␈αpresence.␈αIf␈αit␈αis␈αone␈αwhich␈αmust␈αbe␈αabsent,␈αcheck
␈↓"	␈↓ x␈↓that it is in fact absent.
␈↓"∃␈↓ x␈↓CODER:␈α↔[␈↓¬Summoned␈α⊗by␈α↔its␈α⊗EFFECTS␈α↔part.␈α↔META-CODE␈α⊗part␈α↔suggests␈α⊗using␈α↔COND␈α↔for␈α⊗cases.
␈↓"	␈↓ x␈↓¬COMMENTS␈α∩part␈α⊃warns␈α∩that␈α⊃this␈α∩only␈α⊃works␈α∩if␈α∩case␈α⊃tests␈α∩are␈α⊃disjoint.␈↓]␈α∩I␈α⊃want␈α∩to␈α∩embed␈α⊃the
␈↓"	␈↓ x␈↓tests in a COND. Can a model feature demand itself be both present and absent from a scene?
␈↓"∃␈↓ x␈↓SCENE: [␈↓¬IDEN grabs this. STRUCTURE part indicates that they are disjoint.␈↓] No.
␈↓"∃␈↓ x␈↓CODER: (scribbling)
␈↓"	␈↓ x␈↓α (LAMBDA   (M-FEATURES  S-FEATURES)
␈↓"	␈↓ x␈↓α   (FOR-EVERY  M-FEATURE  IN  M-FEATURES
␈↓"	␈↓ x␈↓α     (COND
␈↓"	␈↓ x␈↓α       ((IS-OF-TYPE  M-FEATURE  MUST-BE-PRESENT)   (MEMBER  M-FEATURE  S-FEATURES))
␈↓"	␈↓ x␈↓α       ((IS-OF-TYPE  M-FEATURE  MUST-BE-ABSENT)    (NOT (MEMBER  M-FEATURE  S-FEATURES)))
␈↓"	␈↓ x␈↓α       (T     ...))))
␈↓"∃␈↓ x␈↓What happens if the model feature doesn't have to be present ␈↓βor␈↓ absent in the scene?
␈↓"∃␈↓ x␈↓␈↓ ↓_CONTRADICTION:␈α≥[␈↓¬IDEN␈α≡part␈α≥recognizes␈α≡relevance,␈α≥WHAT␈α≡part␈α≥easily␈α≡answers␈α≥the
␈↓"	␈↓ x␈↓¬question.␈↓] Then there can be no contradiction. (Blank stare from CODER.)
␈↓"∃␈↓ x␈↓␈↓ ↓_PGMMER:␈α∃[␈↓¬IDEN␈α⊗part␈α∃is␈α∃not␈α⊗anxious,␈α∃but␈α∃CODER's␈α⊗is␈α∃even␈α∃lower-valued.␈α⊗Examines␈α∃the
␈↓"	␈↓ x␈↓¬CODE part of the TRANSLATE BEING.␈↓] He means, the test function should return T.
␈↓"∃␈↓ x␈↓CODER:␈α∞[␈↓¬Regains␈α∞control␈α∞hierarchically.␈α∞A␈α∞REQUISITE␈α∞must␈α∞now␈α∞be␈α∞satisfied.␈↓]␈α∞Got␈α∞it...␈α∞ This␈α∞is
␈↓"	␈↓ x␈↓still very tentative; does anyone see any undesired properties of this piece of code?
␈↓"∃␈↓ x␈↓CASE-SPLIT:␈α∞[␈↓¬IDEN␈α∞part␈α∂responds,␈α∞WHEN␈α∞part␈α∞is␈α∂high␈α∞enough␈α∞to␈α∞take␈α∂control.␈↓]␈α∞That␈α∞call␈α∂on␈α∞IS-
␈↓"	␈↓ x␈↓OF-TYPE␈αis␈αtoo␈αgeneral,␈α
would␈αtake␈αtoo␈αmuch␈αtime.␈α
I␈αknow␈αa␈αgreat␈αdeal␈α
about␈αcase␈αtesting.␈αExplain␈α
to␈αme
␈↓"	␈↓ x␈↓what exactly is needed, and I will replace each IS-OF-TYPE call by a much more efficient one.
␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α∧␈↓ε10␈↓␈↓ λ
␈↓ x␈↓␈↓ ↓_GENL-MANAGER:␈α$[␈↓¬IDEN␈α%reluctant;␈α$must␈α$fall␈α%back␈α$on␈α$very␈α%general␈α$management
␈↓"	␈↓ x␈↓¬techniques.␈↓] No one seems to understand you. Go through your bag of "tricks" one by one, in detail.
␈↓"∃␈↓ x␈↓CASE-SPLIT:␈α⊂[␈↓¬WHAT␈α⊂and␈α⊂HOW␈α⊂parts␈α⊂are␈α⊂presented␈α⊂to␈α⊂the␈α⊂pool;␈α⊂hoping␈α⊂for␈α⊃recognition.␈↓]␈α⊂Many
␈↓"	␈↓ x␈↓of␈αthe␈αtricks␈αdeal␈αwith␈αchanging␈αthe␈αinternal␈αrepresentation␈αof␈αsets,␈αto␈αmake␈αsearches␈αon␈αthem␈αmore␈αefficient.
␈↓"	␈↓ x␈↓Is␈α
their␈α
any␈α
constraint␈α
on␈α∞how␈α
the␈α
features␈α
of␈α
a␈α
model␈α∞be␈α
stored?␈α
 (pause)␈α
Well,␈α
the␈α
biggest␈α∞savings␈α
would
␈↓"	␈↓ x␈↓result␈αfrom␈α
replacing␈αIS-OF-TYPE␈αby␈α
the␈αfast␈αfunction␈α
MEMBER.␈αThe␈α
next␈αbiggest␈αgain␈α
would␈αbe␈αto␈α
store
␈↓"	␈↓ x␈↓all the "must be present" features before all the...
␈↓"∃␈↓ x␈↓STRUCTURE-INDUCER:␈α∪[␈↓¬IDEN␈α∪indicates␈α∩very␈α∪high␈α∪relevance.␈α∩META-CODE␈α∪is␈α∪able␈α∪to␈α∩handle
␈↓"	␈↓ x␈↓¬one␈α∩of␈α⊃the␈α∩"tricks"␈α⊃all␈α∩by␈α∩itself.␈↓]␈α⊃Wait.␈α∩I␈α⊃don't␈α∩see␈α∩anything␈α⊃wrong␈α∩with␈α⊃your␈α∩first␈α∩idea.␈α⊃ The
␈↓"	␈↓ x␈↓model's␈α
features␈α
will␈αbe␈α
divided␈α
into␈α
4␈αdisjoint␈α
sets:␈α
those␈αwhich␈α
must␈α
be␈α
present,␈αthose␈α
which␈α
must␈αbe␈α
absent,
␈↓"	␈↓ x␈↓those in both, and those in neither class.
␈↓"∃␈↓ x␈↓SCENE:␈α∩[␈↓¬DEMON␈α∩part␈α∩says␈α∩that␈α∩an␈α∩inefficiency␈α∩was␈α∩just␈α∩overlooked;␈α∩COMPLEXITY␈α∩says␈α∩it␈α⊃is
␈↓"	␈↓ x␈↓¬urgent.␈↓] There can be no features which must be both present and absent.
␈↓"∃␈↓ x␈↓STRUCTURE-INDUCER:␈α
[␈↓¬Continues␈αon␈↓]␈α
So␈αCF␈α
maintains␈α
3␈αdifferently-named␈α
sets␈αof␈α
features␈α
for␈αeach
␈↓"	␈↓ x␈↓model.
␈↓"∃␈↓ x␈↓PGM-MANAGER:␈α
[␈↓¬COMPLEXITY␈α
has␈α
just␈α
risen␈αpast␈α
danger␈α
level.␈↓]␈α
This␈α
seems␈αlike␈α
too␈α
big␈α
a␈α
step␈αto
␈↓"	␈↓ x␈↓go unreported to the user.
␈↓"∃␈↓ x␈↓MESSENGER:␈α⊃[␈↓¬EFFECTS␈α⊃indicate␈α∩this␈α⊃is␈α⊃the␈α⊃only␈α∩BEING␈α⊃to␈α⊃call␈α⊃on␈α∩now.␈↓]␈α⊃I␈α⊃will␈α⊃tell␈α∩him␈α⊃what
␈↓"	␈↓ x␈↓was done. If he asks any questions, I'll relay them back to you.
␈↓"∃␈↓ x␈↓PGM-MANAGER:␈α∩[␈↓¬REQUISITE␈α∩part␈α∪warns␈α∩that␈α∩this␈α∪must␈α∩be␈α∩done␈α∪soon.␈↓]␈α∩We␈α∩must␈α∪have␈α∩proper
␈↓"	␈↓ x␈↓names for the three types of sets.
␈↓"∃␈↓ x␈↓NAMER:␈α⊗[␈↓¬EFFECTS␈α⊗part␈α⊗shows␈α⊗this␈α⊗is␈α⊗the␈α⊗ideal␈α⊗BEING␈α⊗to␈α⊗take␈α⊗over␈α⊗now.␈α⊗CODE␈α↔does␈α⊗the
␈↓"	␈↓ x␈↓¬simple␈α≡initials/mainwords/concat␈α≡operations␈α≡to␈α≡produce␈α≡mnemonic␈α≡labels.␈↓]␈α≡How␈α≡about
␈↓"	␈↓ x␈↓"PRESENT, ABSENT, EITHER"?
␈↓"∃␈↓ x␈↓USER: No. Call them "MUST-BE-PRESENT, MUSTNOT-BE-PRESENT, MAY-BE-PRESENT".
␈↓"∃␈↓ x␈↓NAMER: [␈↓¬Length DEMON awakens␈↓.] Too long. We shall call them just "MUST, MUSTNOT, MAY".
␈↓"∃␈↓ x␈↓WARNER:␈α∪[␈↓¬WHEN␈α∪sees␈α∪relevance,␈α∪META-CODE␈α∪says␈α∪to␈α∪locate␈α∪areas␈α∪which␈α∪will␈α∪possibly␈α∩need
␈↓"	␈↓ x␈↓¬revision.␈↓] All previous accesses to models' features have been located and tagged.  Now what?
␈↓"∃␈↓ x␈↓STRUCTURE-INDUCER:␈α∃[␈↓¬Specialized␈α∃facts␈α∃about␈α∀the␈α∃effect␈α∃of␈α∃substitutions␈α∃on␈α∀earlier
␈↓"	␈↓ x␈↓¬code.␈↓]␈αEach␈αsuch␈αaccess␈αmust␈αeither␈αbe␈αreplaced␈αby␈αAPPEND␈αof␈αthree␈αaccesses,␈αto␈αthe␈αthree␈αnew␈αnames␈αfor
␈↓"	␈↓ x␈↓sets of model features, or else the statement it is in must be replaced by three similar statements.
␈↓"∃␈↓ x␈↓ANALYST:␈α⊂[␈↓¬DEMON␈α∂part␈α⊂wakes␈α∂up␈α⊂at␈α∂this␈α⊂blunder.␈↓]␈α⊂Not␈α∂always:␈α⊂not␈α∂all␈α⊂three␈α∂types␈α⊂of␈α⊂features␈α∂may
␈↓"	␈↓ x␈↓need to be accessed each time.
␈↓"%␈↓ x␈↓␈↓ ↓XConsidering␈α⊃the␈α∩thousands␈α⊃of␈α∩inter-expert␈α⊃transfers␈α∩which␈α⊃preceded␈α∩completion␈α⊃of␈α∩CF,␈α⊃it␈α∩is␈α⊃not
␈↓ x␈↓surprising␈αthat␈α
all␈αthese␈αbrief␈α
excerpts␈αseem␈αsomewhat␈α
defocussed.␈α This␈αprice␈α
is␈αtolerable␈αsince␈α
the␈αBEING
␈↓ x␈↓execution␈α∞times␈α∂are␈α∞small,␈α∂and␈α∞since␈α∞the␈α∂uniform␈α∞structuring␈α∂(which␈α∞causes␈α∞this␈α∂lack␈α∞of␈α∂direction)␈α∞benefits
␈↓ x␈↓more than it hurts (the system actually ␈↓βworks␈↓).
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α≤␈↓∧␈↓&6. Theory of Pure BEINGs Systems␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XWe␈α∀now␈α∀discuss␈α∀the␈α∪constraints␈α∀each␈α∀BEING,␈α∀and␈α∀each␈α∪group␈α∀of␈α∀BEINGs,␈α∀must␈α∀conform␈α∪to.
␈↓ x␈↓Hopefully, ideas will be separated from implementation details, prejudices from  plausible features.
␈↓"%␈↓ x␈↓␈↓ ↓XIt␈α∞would␈α∞be␈α∞aesthetically␈α
pleasing␈α∞to␈α∞restrict␈α∞all␈α
entities␈α∞in␈α∞the␈α∞system␈α
to␈α∞be␈α∞BEINGs.␈α∞However,␈α
this
␈↓ x␈↓would␈α
cause␈α
an␈α
infinite␈α
regress,␈α
as␈α
each␈α
part␈α
of␈αeach␈α
BEING␈α
would␈α
have␈α
parts␈α
which␈α
had␈α
parts...␈α
To␈αstop
␈↓ x␈↓this,␈αone␈αcan␈αassert␈αthat␈αat␈αsome␈αfinite␈αlevel,␈αall␈αconstructs␈αare␈αprimitive.␈αACTORs,␈αfor␈αexample,␈αset␈αthis␈α
level
␈↓ x␈↓to␈αzero;␈α
BEINGs␈αset␈α
it␈αto␈αone.␈α
ACTORs␈αthemselves␈α
are␈αprimitive,␈α
but␈αonly␈α␈↓βparts␈↓␈α
of␈αBEINGs␈α
can␈αbe.␈αFor␈α
this
␈↓ x␈↓reason, BEINGs can not be viewed as a convergent recursive definition, while ACTORs can be.
␈↓"%␈↓ x␈↓␈↓ ↓XSuppose it were decreed that the only autonomous entities possessing control abilities were BEINGs.


␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α∧␈↓ε11␈↓␈↓ λ
␈↓ x␈↓␈↓ ↓XIn␈αparticular,␈αwe␈αforbid␈αany␈αplain␈α␈↓βfunctions␈↓␈αto␈αexist.␈α In␈αthe␈αcase␈αof␈αan␈αautomatic␈αprogramming␈αtask,
␈↓ x␈↓the␈α∞BEINGs␈α∞would␈α∞have␈α
to␈α∞write␈α∞new␈α∞BEINGs,␈α
not␈α∞new␈α∞LISP␈α∞functions.␈α
The␈α∞target␈α∞program␈α∞would␈α
thus
␈↓ x␈↓itself␈αbe␈αa␈αcommunity␈αof␈αBEINGs.␈αIn␈αorder␈αto␈αfill␈αin␈αall␈αthe␈αparts,␈αa␈αvast␈αamount␈αof␈αsuperfluous␈αinformation
␈↓ x␈↓would␈α∃be␈α∀collected.␈α∃These␈α∃supplementary␈α∀facts␈α∃can␈α∃be␈α∀viewed␈α∃as␈α∃a␈α∀standardized,␈α∃organized␈α∃body␈α∀of
␈↓ x␈↓␈↓βdocumentation␈↓, a formatted system of comments tacked onto each BEING produced.
␈↓"%␈↓ x␈↓␈↓ ↓XWhich␈αBEINGs␈α
would␈αwrite␈α
the␈αnew␈αBEINGs?␈α
 Looking␈αback␈α
at␈αour␈α
interdisciplinary␈αexperts,␈αwe␈α
see
␈↓ x␈↓that␈α⊂each␈α⊂expert␈α⊂is␈α⊂responsible␈α⊂for␈α⊂distilling␈α⊂his␈α⊂own␈α⊂essential␈α⊂contribution,␈α⊂which␈α⊂is␈α⊂then␈α⊂encoded␈α⊂by␈α∂a
␈↓ x␈↓programmer.␈α Perhaps␈αeach␈αBEING␈αshould␈αbe␈αable␈αto␈αdirect␈αconstruction␈αof␈αnew,␈αspecialized␈αBEINGs␈αwhich
␈↓ x␈↓relate␈αto␈αit.␈α
If␈αno␈αBEING␈α
relates␈αto␈αa␈α
task,␈αthen␈αit␈αcan't␈α
be␈αcoded;␈αif␈α
several␈αrespond,␈αthey␈α
should␈αcooperate.
␈↓ x␈↓This␈α∞ability␈α
is␈α∞in␈α
reality␈α∞the␈α
SPECIALIZATIONS␈α∞part␈α
of␈α∞each␈α
BEING␈α∞(see␈α
box,␈α∞Section␈α
5).␈α∞The␈α
BEING
␈↓ x␈↓which␈α∂actually␈α∞does␈α∂the␈α∞creation␈α∂(␈↓αCODER␈↓)␈α∞in␈α∂the␈α∞experimental␈α∂system␈α∞is␈α∂almost␈α∞trivial,␈α∂getting␈α∂very␈α∞precise
␈↓ x␈↓instructions from other BEINGs.
␈↓"%␈↓ x␈↓␈↓ ↓XSince␈αthe␈αpool␈αmust␈αcommunicate␈αwith␈α
the␈αuser,␈αsome␈αBEINGs␈αmust␈αtranslate␈α
quasi-English␈αphrases
␈↓ x␈↓into␈α
calls␈α
on␈α
BEINGs.␈α
 Drawing␈α
again␈α
on␈α∞our␈α
experts␈α
analogy,␈α
we␈α
require␈α
that␈α
each␈α
BEING␈α∞recognize␈α
his
␈↓ x␈↓own␈αrelevance.␈α
So␈αtranslation␈α
is␈αmerely␈α
the␈αact␈α
of␈αasking␈α
the␈αwhole␈α
pool␈α"Who␈α
can␈αrecognize␈αthis...",␈α
collecting
␈↓ x␈↓the␈αresponders,␈αhaving␈α␈↓βthem␈↓␈αdecide␈αwho␈αshould␈αtake␈αcontrol,␈αand␈αletting␈αthe␈αwinner␈αdo␈αthe␈αtranslation.␈αMost
␈↓ x␈↓communication is done as if it, too, were such a translation activity.
␈↓"%␈↓ x␈↓␈↓ ↓XOne␈αbias␈αis␈αthe␈αrejection␈αof␈αdebugging␈αas␈αa␈α
fundamental␈αprogramming␈αtool.␈α It␈αis␈αfelt␈αto␈αbe␈αworth␈α
the
␈↓ x␈↓extra␈α∂effort␈α∂to␈α∂make␈α∂the␈α∂system's␈α∂internal␈α∂model␈α∂of␈α∂the␈α∂current␈α∂partial␈α∂target␈α∂program␈α⊂␈↓βcorrect␈↓.␈α∂Debugging
␈↓ x␈↓demands␈α
detective␈α
work,␈α
examing␈α
one's␈α
earlier␈α
efforts␈α
for␈α
flaws,␈α
for␈α
details␈α
which␈α
have␈α
been␈αoverlooked.␈α
 Any
␈↓ x␈↓tireless␈α∞system␈α
 should␈α∞not␈α∞ignore␈α
details,␈α∞but␈α∞rather␈α
defer␈α∞them,␈α∞asserting␈α
a␈α∞warning␈α∞to␈α
this␈α∞effect␈α∞when␈α
it
␈↓ x␈↓does␈αso.␈αProcrastination␈α
is␈αquite␈αvaluable;␈αin␈α
PUP6,␈αmuch␈αeffort␈αis␈α
spent␈αdeferring␈αany␈αunresolvable␈α
decision.
␈↓ x␈↓Undeferrable␈α
unresolvable␈α
decisions␈α
must␈α
cause␈α
a␈αbacktrack␈α
 point␈α
to␈α
be␈α
reluctantly␈α
set␈α
up.␈αAnother␈α
prejudice
␈↓ x␈↓is␈α
that␈α
most␈α
carelessness␈α
bugs␈α
can␈α
be␈αeliminated␈α
by␈α
this␈α
deferral,␈α
feed-forward,␈α
and␈α
precise␈αrecord-keeping.
␈↓ x␈↓Humans␈αdepend␈αon␈αtheir␈αadaptability␈αto␈αcompensate␈αfor␈αlimitations␈αin␈αtheir␈αbrain␈αhardware,␈αbut␈αthere␈αis␈αno
␈↓ x␈↓need␈α∩for␈α∩an␈α∩␈↓βautomatic␈↓␈α∩programming␈α∩system␈α∩to␈α∩do␈α⊃so.␈α∩ These␈α∩biasses␈α∩are␈α∩not␈α∩inherent␈α∩in␈α∩the␈α⊃BEINGs
␈↓ x␈↓formulation, but only in the design of the PUP6 system (and in the mind of the author).
␈↓"%␈↓ x␈↓␈↓ ↓XTo␈αclarify␈αwhat␈αBEINGs␈αare␈α
and␈αare␈αnot,␈αthey␈αare␈α
contrasted␈αwith␈αsome␈αother␈αideas.␈αFRAMES[4]␈α
are
␈↓ x␈↓sufficiently␈α
amorphous␈α∞to␈α
subsume␈α∞BEINGs.␈α
In␈α∞philosophy,␈α
FRAMES␈α∞are␈α
meant␈α∞to␈α
model␈α∞perception,␈α
and
␈↓ x␈↓intentionally␈αrely␈αon␈αimplicit␈αdefault␈αvalues;␈α
BEINGs␈αintentionally␈αavoid␈αmaking␈αdecisions␈αby␈α
default.␈α This
␈↓ x␈↓is␈αalso␈αthe␈αdifference␈α between␈αHACKER␈αand␈α
PUP6.␈α Since␈αPUP6␈αwrites␈αstructured␈αprograms,␈αit␈α
should␈αbe
␈↓ x␈↓distinguished␈αfrom␈αmacro␈αexpansion.␈αMacro␈αprocedures␈αexpand␈αmechanically:␈α␈↓↓expand(sequence␈α  m␈↓#v1␈↓#␈α m␈↓#v2␈↓#)␈α=
␈↓ x␈↓↓(sequence␈α∞ expand(m␈↓#v1␈↓#)␈α∞ expand(m␈↓#v2␈↓#)))␈↓.␈α
BEINGs␈α∞could␈α∞use␈α
information␈α∞gleaned␈α∞during␈α
expansion␈α∞of␈α∞m␈↓#v1␈↓#␈α
to
␈↓ x␈↓improve␈α
the␈αway␈α
m␈↓#v2␈↓#␈αwas␈α
handled.␈α ACTORs[3],␈α
unlike␈αBEINGs,␈α
have␈αno␈α
fixed␈αstructure␈α
imposed,␈α
and␈αdo
␈↓ x␈↓not broadcast their messages (they  specify who gets each message, by name, to a bureaucracy).
␈↓"%␈↓ x␈↓␈↓ ↓XBEINGs␈αsubsume␈α(inefficiently)␈αmany␈αpopular␈αAI␈αfeatures;␈αthe␈αdemonstration␈αwill␈αbe␈αbrief:␈α
A␈α␈↓βdemon␈↓
␈↓ x␈↓could␈αbe␈αreplaced␈αby␈αa␈αBEING␈αwhose␈αARG-CHECK␈αpredicate␈αwas␈αthe␈αtriggering␈αpredicate,␈αwhose␈αWHEN
␈↓ x␈↓part␈α∂was␈α∂high␈α∂enough␈α∂to␈α∂ensure␈α∂frequent␈α∂attention,␈α∂and␈α∂whose␈α∂META-CODE␈α∂part␈α∂was␈α∂the␈α∂body␈α∂of␈α∞the
␈↓ x␈↓demon.␈α
An␈α∞␈↓βassertion␈↓␈α
in␈α∞an␈α
associative␈α∞data␈α
network␈α∞could␈α
be␈α
a␈α∞BEING␈α
with␈α∞only␈α
an␈α∞IDEN␈α
part␈α∞filled␈α
in;
␈↓ x␈↓when␈α⊂it␈α⊂recognizes␈α⊃its␈α⊂relevance,␈α⊂a␈α⊂fully␈α⊃instantiated␈α⊂assertion␈α⊂is␈α⊂returned.␈α⊃A␈α⊂␈↓βfunction␈↓␈α⊂is␈α⊂equivalent␈α⊃to␈α⊂a
␈↓ x␈↓BEING␈α∞with␈α∞only␈α∞a␈α∞META-CODE,␈α∞ARGS,␈α∂and␈α∞EVAL-ARGS␈α∞parts;␈α∞one␈α∞knows␈α∞almost␈α∞nothing␈α∂about␈α∞it
␈↓ x␈↓before␈α
executing␈α
it.␈α
 The␈α
inefficiencies␈α
should␈α
be␈α∞clear:␈α
whenever␈α
a␈α
BEING␈α
throws␈α
a␈α
question␈α
open␈α∞to␈α
the
␈↓ x␈↓floor,␈α
"Who␈α
can...",␈α
it␈α∞takes␈α
an␈α
amount␈α
of␈α∞time␈α
proportional␈α
to␈α
the␈α
number␈α∞of␈α
BEINGs␈α
in␈α
the␈α∞system.␈α
One
␈↓ x␈↓would introduce this huge time factor by replacing any of the above mechanisms by BEINGs.
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈α␈↓βnumber␈↓␈αof␈αBEING␈αparts␈α seems␈αto␈αindicate␈αthe␈αbalance␈αbetween␈αuniformity␈αand␈αstructure␈αin␈αthe

␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α∧␈↓ε12␈↓␈↓ λ
␈↓ x␈↓community.␈α⊃ This␈α⊃was␈α⊂touched␈α⊃on␈α⊃as␈α⊂point␈α⊃(2)␈α⊃of␈α⊂Section␈α⊃4.␈α⊃ A␈α⊂small␈α⊃universal␈α⊃set␈α⊂of␈α⊃BEING␈α⊃parts␈α⊂is
␈↓ x␈↓necessary␈α
to␈αpreserve␈α
some␈α
of␈αthe␈α
advantages␈α
of␈αuniformity␈α
(easy␈αaddition␈α
of␈α
knowledge␈αto␈α
the␈α
system,␈αeasy
␈↓ x␈↓inter-BEING␈α
communication).␈αThis␈α
demands␈α
that␈αthe␈α
number␈αof␈α
parts␈α
of␈αeach␈α
BEING␈αbe,␈α
say,␈α
under␈α100.
␈↓ x␈↓But␈α∂it␈α∂is␈α∂the␈α∂complex␈α∞structure␈α∂of␈α∂a␈α∂BEING␈α∂which␈α∞makes␈α∂complex␈α∂behaviors␈α∂feasable,␈α∂including␈α∞flexible
␈↓ x␈↓communication␈α
as␈α∞well␈α
as␈α
viable␈α∞final␈α
products.␈α
So␈α∞each␈α
BEING␈α
should␈α∞have␈α
many␈α
parts,␈α∞say␈α
at␈α∞least␈α
ten.
␈↓ x␈↓This␈α
range,␈α
10␈α
to␈α
100,␈α
is␈α
wide␈α
for␈α
the␈α
domain␈α
of␈α
automatic␈α
programming.␈α
In␈α
other␈α
domains,␈α
it␈α
may␈αbe␈α
narrow
␈↓ x␈↓or disappear altogether; this would indicate that BEINGs could ␈↓βnot␈↓ be used effectively for those tasks.
␈↓"	

␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈απ␈↓∧␈↓&7. Experimental Results␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XTwo␈α∂of␈α∂the␈α∂most␈α∂significant␈α∂questions␈α∂for␈α∂automatic␈α∂programming␈α∂systems␈α∂are␈α∂what␈α⊂programs␈α∂are
␈↓ x␈↓synthesized, and what kinds of things the user must tell the system.
␈↓"%␈↓ x␈↓␈↓↓␈↓&7.1. Character of the Dialogue between User and PUP6␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XBy␈α⊃lumping␈α⊂all␈α⊃the␈α⊂parts␈α⊃of␈α⊂the␈α⊃PUP6␈α⊂system␈α⊃together␈α⊂conceptually,␈α⊃the␈α⊂interaction␈α⊃is␈α⊂seen␈α⊃as␈α⊂a
␈↓ x␈↓␈↓βdialogue␈↓␈α
between␈α
a␈α
human␈α
user␈α
and␈α
an␈α
automatic␈α
programming␈α
system.␈α
 The␈α
level␈α
of␈α
detail␈α∞maintained␈α
in
␈↓ x␈↓the␈αearlier␈αpolylogues␈αobscured␈αwhat␈αthe␈αhuman␈αactually␈αsaw␈αand␈αresponded.␈αBelow␈αare␈αthose␈αsame␈αexcerpts,
␈↓ x␈↓but as the ␈↓βuser␈↓ would see them.
␈↓"%␈↓ x␈↓␈↓ ↓XHere␈α∂are␈α⊂the␈α∂initial␈α∂few␈α⊂lines␈α∂of␈α∂an␈α⊂actual␈α∂300-page␈α∂long␈α⊂dialogue.␈α∂In␈α∂section␈α⊂2,␈α∂this␈α⊂excerpt␈α∂was
␈↓ x␈↓presented the way experts (and the same way the PUP6 BEINGs) discussed it.

␈↓"	␈↓ x␈↓␈↓ ↓XPUP: I want any task.
␈↓"	␈↓ x␈↓␈↓ ↓XUSER: Write a program which does concept formation
␈↓"	␈↓ x␈↓␈↓ ↓XPUP: I am about to write a program for ␈↓αspecialized CONCEPT-FORMATION␈↓.
␈↓"	␈↓ x␈↓␈↓ ↓XPUP: Give me a proper name for this program. (I suggest CF or CONCEPT)
␈↓"	␈↓ x␈↓␈↓ ↓XUSER: CF
␈↓"	␈↓ x␈↓␈↓ ↓XPUP: Must CF mimic different human subjects? (respond YES, NO, ?, or DETAILS)
␈↓"	␈↓ x␈↓␈↓ ↓XUSER: NO
␈↓"	␈↓ x␈↓␈↓ ↓XPUP: I am temporarily replacing the task "␈↓αspecialized form of CONCEPT-FORMATION␈↓"
␈↓"	␈↓ x␈↓␈↓ ↓X                by "␈↓αspecialized form of PARTITION-A-DOMAIN␈↓".
␈↓"%␈↓ x␈↓␈↓ ↓XFrom␈α
the␈α
above,␈α
one␈α
gets␈α
the␈α
true␈α
impression␈α
that␈α
the␈α
dialogue␈α
was␈α
fairly␈α
tedious:␈α
choosing␈α
names␈α
and
␈↓ x␈↓resolving␈α
design␈α
decisions␈α
␈↓βwas␈↓␈α
the␈α
user's␈α
main␈α
role.␈αThere␈α
were,␈α
in␈α
sum,␈α
about␈α
300␈α
such␈α
responses␈αinvolved␈α
in
␈↓ x␈↓generating␈αCF.␈α13␈αcrucial␈αresponses␈α
were␈αalso␈αnecessary,␈αwhere␈αthe␈α
user␈αsupplies␈αneeded␈αinformation␈αthat␈α
was
␈↓ x␈↓not␈αexplicitly␈αrequested␈αby␈αthe␈αBEINGs.␈α A␈αfew␈αof␈αthese␈αkey␈αlines␈αare:␈αWrite␈αa␈αprogram␈αwhich␈αdoes␈αconcept
␈↓ x␈↓formation;␈α
an␈α
"element␈α
of␈α
the␈α
domain␈αof␈α
stimuli"␈α
is␈α
a␈α
static␈α
scene;␈αthe␈α
guessing␈α
process␈α
is␈α
one␈α
of␈αsearching,␈α
not
␈↓ x␈↓computing;␈αterminate␈α
the␈αloop␈α
iff␈αany␈α
model␈αfeature␈αis␈α
incompatible␈αwith␈α
the␈αscene's␈α
features.␈αIn␈αaddition,␈α
the
␈↓ x␈↓user␈α≡must␈α≡check␈α≡over␈α≡each␈α≡decision␈α≡to␈α≡transfer␈α≡between␈α≡the␈α≡seven␈α≡current␈α≡sets␈α≡of␈α≡features
␈↓ x␈↓(MAY/MUST/MUSTNOT␈α∀of␈α∃the␈α∀guessed␈α∀and␈α∃the␈α∀correct␈α∀model␈α∃name;␈α∀SCENE␈α∀features):␈α∃what␈α∀gets
␈↓ x␈↓transferred,␈α
when,␈αand␈α
why.␈α
{There␈αought␈α
to␈αbe␈α
ANALOGY␈α
BEINGs␈αwhich␈α
learn␈α
from␈αthe␈α
first␈αcouple␈α
such
␈↓ x␈↓tedious constructions, and answer for the user during the writing of the last several transfer functions.}
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈αtranslation␈α
mechanism␈αis␈αsimply␈α
the␈αIDEN␈α
parts␈αof␈αall␈α
the␈αBEINGs.␈αEach␈α
such␈αpart␈α
contains␈αa
␈↓ x␈↓few␈α
productions,␈α
which␈α
are␈α
united␈αwith␈α
those␈α
of␈α
the␈α
other␈α
BEINGs␈αinto␈α
one␈α
large␈α
production␈α
system.␈αThe␈α
left
␈↓ x␈↓side␈α∂of␈α∂a␈α∂production␈α∂is␈α∂a␈α∞template␈α∂which␈α∂is␈α∂matched␈α∂against␈α∂the␈α∞given␈α∂phrase.␈α∂The␈α∂right␈α∂side␈α∂is␈α∂a␈α∞small
␈↓ x␈↓program␈α∞to␈α∞be␈α∞executed␈α∞in␈α∞case␈α∞the␈α∂phrase␈α∞is␈α∞recognized.␈α∞Generally,␈α∞that␈α∞program␈α∞simply␈α∞says␈α∂to␈α∞translate
␈↓ x␈↓some␈α
subphrase␈α
of␈α
the␈α
phrase,␈α∞and␈α
then␈α
(probably)␈α
pass␈α
control␈α
to␈α∞the␈α
BEING␈α
which␈α
this␈α
is␈α
in␈α∞the␈α
IDEN
␈↓ x␈↓part␈α∞of.␈α∞ While␈α∞not␈α∞up␈α∞to␈α∞the␈α∞"state␈α∞of␈α∞the␈α∞art"␈α∞in␈α∞natural␈α∞language␈α∞processing,␈α∞this␈α∞is␈α∞both␈α∂adequate␈α∞and
␈↓ x␈↓faithful␈α∂to␈α∂the␈α∂spirit␈α∂of␈α∂distributing␈α∂problems␈α∂over␈α∂all␈α∂the␈α∂BEINGs,␈α∂by␈α∂making␈α∂them␈α∂␈↓βparts␈↓,␈α∂so␈α∂it␈α∂is␈α∞each
␈↓ x␈↓BEING's duty to know a little about them.
␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α∧␈↓ε13␈↓␈↓ λ
␈↓ x␈↓␈↓ ↓XIn␈α⊂the␈α∂excerpt␈α⊂presented␈α∂in␈α⊂section␈α∂5.3,␈α⊂again␈α∂only␈α⊂the␈α∂first␈α⊂user␈α∂comment␈α⊂is␈α∂other␈α⊂than␈α⊂a␈α∂simple
␈↓ x␈↓choice.  Consider that excerpt from the point of view of the user:

␈↓"	␈↓ x␈↓␈↓ ↓8PUP: Please type in a logical expression which is true iff we terminate the loop
␈↓"	␈↓ x␈↓␈↓ ↓8USER: Any feature in model-features is incompatible with scene-features
␈↓"	␈↓ x␈↓␈↓ ↓8PUP: PUP wants USER to type in name for specialized version of CONTRADICTS
␈↓"	␈↓ x␈↓␈↓ ↓8USER: Call it IMPOSS
␈↓"	␈↓ x␈↓␈↓ ↓8PUP: Move cursor to IMPOSS type of CONTRADICTS. I can no longer defer this decision:
␈↓"	␈↓ x␈↓␈↓ ↓8           A... Contradiction arises from features who have probability=1 of occurring
␈↓"	␈↓ x␈↓␈↓ ↓8           B... Contradiction arises from features who have probability=0 of occurring
␈↓"	␈↓ x␈↓␈↓ ↓8           C... Contradiction arises from features who have probability>0&<1 of occurring
␈↓"	␈↓ x␈↓␈↓ ↓8        Please type some sequence of some of the letters A,B,C; or type ?, DEFER, or DETAILS.
␈↓"	␈↓ x␈↓␈↓ ↓8USER: A, B, and C are all relevant.
␈↓"	␈↓ x␈↓␈↓ ↓8PUP: I recommend that each model's features be structured (along 1 dimension) into these 3 pieces:
␈↓"	␈↓ x␈↓␈↓ ↓8        PRESENT, which will hold features whose occurrence should have probability=1
␈↓"	␈↓ x␈↓␈↓ ↓8        ABSENT,  which will hold features whose occurrence should have probability=0
␈↓"	␈↓ x␈↓␈↓ ↓8        EITHER,  which will hold features whose occurrence should have any other probability
␈↓"	␈↓ x␈↓␈↓ ↓8     Please type back YES, NO, UNSURE, DETAILS, or 3 new names.
␈↓"	␈↓ x␈↓␈↓ ↓8USER: YES. Call them MUST-BE-PRESENT, MUSTNOT-BE-PRESENT, MAY-BE-PRESENT.
␈↓"	␈↓ x␈↓␈↓ ↓8PUP: Too long. Nicknamed MUST, MUSTNOT, MAY.
␈↓"	␈↓ x␈↓␈↓ ↓8PUP: In all code generated to date, three references to model's features have been located.
␈↓"	␈↓ x␈↓␈↓ ↓8PUP: In all cases, reference has been replaced by (APPEND MUST MUSTNOT MAY).
␈↓"%␈↓ x␈↓␈↓ ↓XBecause␈αof␈αits␈αgenesis␈αfrom␈αa␈αsingle␈α"experts␈αmeeting"␈αpolylogue,␈αhereafter␈αreferred␈αto␈αas␈αthe␈α␈↓βprotocol␈↓,
␈↓ x␈↓the␈α∞pool␈α∂of␈α∞BEINGs␈α∂was␈α∞simply␈α∂not␈α∞capable␈α∞of␈α∂widely␈α∞varied␈α∂dialogues␈α∞with␈α∂the␈α∞user.␈α∂ Two␈α∞assumptions
␈↓ x␈↓implicit␈αin␈αusing␈αthe␈αsimulated␈αexperts'␈αdiscussion␈αas␈αa␈αliteral␈αmodel␈αturned␈αout␈αto␈αbe␈αserious:␈αthe␈αabilities␈αof
␈↓ x␈↓any␈α
actual␈α
user␈α
of␈α
PUP6␈α
must␈α
coincide␈α
with␈α
those␈α∞of␈α
the␈α
user␈α
who␈α
was␈α
simulated␈α
in␈α
the␈α
protocol,␈α∞and␈α
the
␈↓ x␈↓order␈α∞of␈α∞and␈α
wording␈α∞of␈α∞the␈α∞actual␈α
user's␈α∞comments␈α∞must␈α
closely␈α∞adhere␈α∞those␈α∞of␈α
the␈α∞user␈α∞in␈α∞the␈α
protocol.
␈↓ x␈↓Such␈αdifficulties␈αmust␈αbe␈αovercome␈αin␈αany␈αsystem␈αdesigned␈αfor␈αwide␈αusership,␈αbut␈αwere␈αconsidered␈αignorable
␈↓ x␈↓in the specific automatic code generation task at hand.
␈↓"%␈↓ x␈↓␈↓ ↓XAlso␈αas␈αa␈αresult␈αof␈αthis␈αapproach␈αto␈αsystem␈αspecification,␈αeach␈αBEING␈αhad␈αonly␈αthose␈αparts␈αspecified
␈↓ x␈↓which␈α∞it␈α∞actually␈α∞would␈α∞need␈α∞in␈α∞the␈α∞ensuing␈α∞dialogue.␈α∞ Part␈α∞of␈α∞the␈α∞difficulty␈α∞with␈α∞new␈α∞dialogues␈α∞stemmed
␈↓ x␈↓from␈α∩this␈α∩minimal␈α∩completion.␈α∩ In␈α∩the␈α∩protocol,␈α∩when␈α∩a␈α∩decision␈α∩was␈α∩made␈α∩by␈α∩experts,␈α∩the␈α∩knowledge
␈↓ x␈↓necessary␈αto␈αfollow␈α
the␈α␈↓βother␈↓␈αalternative␈α
branch␈αwas␈αnot␈αused,␈α
nor␈αwere␈αsuch␈α
superfluous␈αfacts␈αsupplied␈αto␈α
the
␈↓ x␈↓BEINGs␈α
in␈α∞PUP6.␈α
Thus␈α∞the␈α
user␈α
of␈α∞PUP6␈α
must␈α∞almost␈α
always␈α
resolve␈α∞each␈α
choice␈α∞the␈α
way␈α∞the␈α
simulated
␈↓ x␈↓(protocol)␈α
user␈α
did.␈α
 It␈α
is␈αfelt␈α
that␈α
if␈α
all␈α
the␈αparts␈α
of␈α
all␈α
the␈α
BEINGs␈αhad␈α
been␈α
faithfully␈α
filled␈α
in,␈αthis␈α
problem
␈↓ x␈↓would␈α∂have␈α∂subsided.␈α∂Basically,␈α∂the␈α∂difficulty␈α∂is␈α⊂one␈α∂of␈α∂modelling␈α∂all␈α∂the␈α∂possibly␈α∂relevant␈α⊂knowledge␈α∂an
␈↓ x␈↓expert has, rather than (as was done) just capturing enough of his knowledge to do a few given tasks.
␈↓"%␈↓ x␈↓␈↓ ↓XWhile␈αall␈αthe␈α
BEINGs'␈αinteractions␈αwere␈α
invisible␈αto␈αthe␈α
user,␈αthe␈αsystem␈α
still␈αswamped␈αhim␈αwith␈α
data
␈↓ x␈↓about␈α∞what␈α
was␈α∞going␈α
on.␈α∞For␈α∞example,␈α
most␈α∞of␈α
the␈α∞entities␈α
he␈α∞was␈α∞asked␈α
to␈α∞name␈α
were␈α∞never␈α∞referred␈α
to
␈↓ x␈↓again␈αby␈αname.␈αThe␈αconverse␈αproblem␈αexisted␈αas␈αwell:␈αit␈αwas␈αnecessary␈αto␈αinclude␈αa␈αBEING␈αwhich␈αsimulated
␈↓ x␈↓forgetfulness,␈α∞to␈α∞prevent,␈α
e.g.,␈α∞anaphora␈α∞spanning␈α
minutes␈α∞of␈α∞real␈α
time.␈α∞ Orienting␈α∞the␈α
user␈α∞was␈α∞not␈α
solved
␈↓ x␈↓satisfactorally.␈αPointers␈αinto␈αa␈αgraph␈αof␈αgenerated␈αcode␈αwere␈αsimulated,␈αbut␈αoften␈αa␈αuser␈αwished␈αto␈αrefer␈αto␈αa
␈↓ x␈↓piece of code not by name or by pointing, but by some brief meaningful (to him only!) phrase.
␈↓"%␈↓ x␈↓␈↓↓␈↓&7.2. The Range of Programs Synthesized by PUP6␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈α
system,␈α
PUP6,␈α
did␈αeventually␈α
synthesize␈α
CF,␈α
the␈αtarget␈α
concept␈α
formation␈α
program.␈α
 PUP6␈αwas
␈↓ x␈↓200␈α∞pages␈α∞of␈α∞INTERLISP[5],␈α∞ CF␈α∞was␈α∞30␈α∞pages␈α∞long␈α∞(6␈α∞pages␈α∞when␈α∞coded␈α∞by␈α∞hand␈α∞during␈α∂the␈α∞protocol).
␈↓ x␈↓This␈α∩was␈α∩generated␈α∩in␈α∩60␈α∩cpu␈α∩minutes␈α∩(compiled,␈α∩PDP-10␈α∩TENEX).␈α∩The␈α∩dialogue␈α∩consisted␈α∪of␈α∩300K
␈↓ x␈↓characters typed by PUP6, and 4K by the user. It occupied 300 pages, and five hours of real time.

␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α∧␈↓ε14␈↓␈↓ λ
␈↓ x␈↓␈↓ ↓XDespite␈α⊃the␈α⊂lack␈α⊃of␈α⊂dialogue␈α⊃flexibility,␈α⊂it␈α⊃␈↓βwas␈↓␈α⊂felt␈α⊃that␈α⊂most␈α⊃of␈α⊂the␈α⊃BEINGs␈α⊂could␈α⊃be␈α⊃useful␈α⊂in
␈↓ x␈↓generating␈α∩other␈α∪programs.␈α∩For␈α∪this␈α∩reason,␈α∪two␈α∩additional␈α∪target␈α∩programs␈α∪were␈α∩specified.␈α∪They␈α∩were
␈↓ x␈↓synthesized with little change to PUP6, but only by someone familiar with the system.
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈α∞second␈α∞target␈α∞program,␈α∞GI,␈α∞is␈α∞a␈α∞grammatical␈α∞inference␈α∞program,␈α∞which␈α∞accepts␈α∂strings␈α∞labelled
␈↓ x␈↓LEGAL,␈αILLEGAL,␈αor␈α??.␈αIn␈αthe␈αlatter␈αcase,␈αGI␈αmust␈αguess␈αthe␈αlegality.␈αInternally,␈αpotential␈αsets␈αof␈αrules␈αare
␈↓ x␈↓maintained.␈α∂Of␈α∞the␈α∂original␈α∞pool,␈α∂46␈α∞out␈α∂of␈α∂70␈α∞"general"␈α∂BEINGs␈α∞were␈α∂used␈α∞in␈α∂synthesizing␈α∂both␈α∞targets.
␈↓ x␈↓Three␈α∞of␈α∞the␈α∂existing␈α∞17␈α∞"domain-specific"␈α∂BEINGs␈α∞also␈α∞were␈α∂useful␈α∞(those␈α∞involving␈α∂partitioning).␈α∞ Four
␈↓ x␈↓totally␈αnew␈αBEINGs␈αhad␈αto␈αbe␈αadded,␈αrelated␈αto␈αformal␈αgrammars␈αand␈αrules.␈α Unfortunately,␈αthe␈αaddition␈αof
␈↓ x␈↓␈↓βany␈↓␈α∂new␈α∂BEINGs␈α∂demands␈α∂that␈α∂the␈α∂user␈α⊂be␈α∂acquainted␈α∂with␈α∂the␈α∂format␈α∂conventions␈α∂of␈α∂PUP6.␈α⊂ The␈α∂GI
␈↓ x␈↓program generated was 20 pages long; a hand-coded version was one-fifth that size.
␈↓"%␈↓ x␈↓␈↓ ↓XPL␈αwas␈αthe␈αfinal␈αtarget␈αprogram␈αattempted,␈αa␈αsimple␈αproperty␈αlist␈αmanipulator.␈α It␈α
repeatedly␈αaccepts
␈↓ x␈↓requests␈αfrom␈αthe␈αuser␈αto␈αinsert,␈αinspect,␈αor␈αdelete␈αsome␈αrecord(s).␈αAny␈αunspecified␈αfields␈αare␈αtreated␈αas␈αdon't-
␈↓ x␈↓cares,␈αso␈αa␈αsimple␈αpattern-matcher␈αhad␈αto␈αbe␈αsynthesized.␈α 39␈α␈↓βgeneral␈↓␈αBEINGs␈αwere␈αused␈αin␈αsynthesizing␈α␈↓βall
␈↓ x␈↓βthree␈↓␈αtargets,␈αbut␈αno␈α␈↓βdomain-specific␈↓␈α
BEINGs␈αwere␈αcommon␈αto␈αsynthesizing␈α
PL␈αand␈αany␈αother␈αprogram.␈α
 One
␈↓ x␈↓general BEING and one domain-specific BEING had to be ␈↓βadded␈↓ to PUP6.
␈↓"%␈↓ x␈↓␈↓ ↓XAs␈αproposed␈αin␈αSection␈α6,␈αthe␈αBEINGs␈αgenerate␈αother␈αBEINGs,␈αnever␈αplain␈αfunctions.␈αThis␈αexplains
␈↓ x␈↓the␈αhuge␈αincreases␈αin␈αtarget␈αcode␈αlengths␈αin␈αthe␈αPUP6␈αversions␈αcompared␈αto␈αthe␈αversions␈αproduced␈αby␈αhand
␈↓ x␈↓when␈αsimulating␈αthe␈αexperts␈α(who␈αwrote␈αthe␈αtarget␈αprograms␈αas␈αfunctions).␈α CF␈αwas␈αa␈αpool␈αof␈α56␈αbrand␈αnew
␈↓ x␈↓BEINGs,␈αGI␈α37,␈αand␈αPL␈α24.␈αAs␈αwith␈αPUP6,␈αone␈αcan␈αinterrupt␈αthe␈αtarget␈αprograms␈αas␈αthey␈αare␈αrunning␈αand
␈↓ x␈↓ask␈α⊂questions.␈α⊂Any␈α⊂BEING␈α⊂on␈α⊂the␈α⊂control␈α⊂stack␈α⊂will␈α⊂provide␈α⊂fully␈α⊂instantiated␈α⊂answers␈α⊂to␈α⊂any␈α⊂of␈α⊂its␈α∂29
␈↓ x␈↓allowable␈α∂queries␈α∂(its␈α∂parts);␈α∂all␈α∂other␈α∂BEINGs␈α∂will␈α∂provide␈α∂only␈α∂hypothetical␈α∂answers.␈α∂Recall␈α⊂the␈α∂excerpt
␈↓ x␈↓from CF itself running, found in section 5.2.
␈↓"%␈↓ x␈↓␈↓ ↓XSome␈αof␈αthe␈αdifficulties␈αstem␈αfrom␈αthe␈αnature␈αof␈αthe␈αtask.␈αIn␈αany␈αlong␈αdialogue,␈αthe␈αuser␈αoften␈αforgets,
␈↓ x␈↓changes␈α
his␈αmind,␈α
errs,␈αetc.␈α
A␈αvery␈α
sophisticated␈α
user␈αmodel␈α
would␈αbe␈α
necessary␈αto␈α
accomodate␈α
this␈αerrorful
␈↓ x␈↓process␈α∞in␈α∞a␈α∞non-debugging␈α∞system.␈α
Without␈α∞such␈α∞abilities,␈α∞the␈α∞system␈α
itself␈α∞may␈α∞be␈α∞led␈α∞into␈α∞error.␈α
 While
␈↓ x␈↓most␈αbugs␈α␈↓βare␈↓␈α
avoidable␈αby␈αcareful␈α
record-keeping,␈αit␈αproved␈α
unrealistic␈αto␈αmake␈α
no␈αprovision␈αfor␈α
debugging
␈↓ x␈↓a new thirty-page program. When a few errors did occur, PUP6 itself had to be altered.
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α_␈↓∧␈↓&8. Conclusions␈↓)αβ␈↓
␈↓"	␈↓ x␈↓␈↓↓␈↓&8.1. About PUP6␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XWhat␈α∪have␈α∪we␈α∪learned␈α∀from␈α∪this␈α∪experimental␈α∪study?␈α∀The␈α∪overall␈α∪feasability␈α∪of␈α∀BEINGs␈α∪was
␈↓ x␈↓demonstrated,␈αbut␈αthe␈α
difficulties␈αof␈αcommunicating␈αwith␈α
the␈αuser␈αmade␈αthe␈α
system␈αalmost␈αimpossible␈αto␈α
work
␈↓ x␈↓with.␈αThe␈αset␈αof␈αquestions␈αthe␈αuser␈αwas␈αexpected␈αto␈αwant␈αto␈αask␈αis␈αthe␈αsame␈αas␈αthe␈αset␈αthat␈αone␈αBEING␈αcan
␈↓ x␈↓ask␈α
another:␈αthe␈α
BEING␈αparts.␈α
When␈αthe␈α
"nice"␈αuser␈α
interrupts,␈αhis␈α
questions␈αare␈α
translated␈αtrivially␈α
into␈αa
␈↓ x␈↓simple retrieval. Real users are seldom nice; the BEINGs generally misunderstood what users asked.
␈↓"%␈↓ x␈↓␈↓ ↓XTo␈α∪modify␈α∪PUP6␈α∪to␈α∪synthesize␈α∪new␈α∪programs,␈α∪it␈α∪was␈α∪necessary␈α∪to␈α∪add␈α∪a␈α∪few␈α∪general-purpose
␈↓ x␈↓programming␈αand␈αcommunication␈αBEINGs,␈αplus␈αadd␈αseveral␈αBEINGs␈αspecific␈αto␈αthe␈αnew␈αprogram's␈αdomain,
␈↓ x␈↓plus␈α∞generalize␈α∂a␈α∞few␈α∞existing␈α∂BEINGs'␈α∞parts.␈α∂ The␈α∞dialogue␈α∞to␈α∂produce␈α∞the␈α∞new␈α∂program␈α∞may␈α∂be␈α∞poorly
␈↓ x␈↓suited to that domain, since most of the recognized phrases stem from a single (CF-producing) protocol.
␈↓"%␈↓ x␈↓␈↓ ↓XTo␈α
improve␈α
PUP6's␈α
performance,␈α
one␈α
could␈α
add␈α
some␈α
debugging␈α
specialist␈α
BEINGs,␈α
some␈αdialogue
␈↓ x␈↓specialists,␈αsome␈αsophisticated␈αuser␈αpsychology␈αexperts␈α(why␈αis␈αthe␈αuser␈αasking␈αme␈αthat␈αquestion,␈αwhat␈αneedn't
␈↓ x␈↓I␈α⊂tell␈α∂him,␈α⊂how␈α∂should␈α⊂I␈α∂direct␈α⊂his␈α∂attention),␈α⊂some␈α∂BEINGs␈α⊂whose␈α∂task␈α⊂is␈α∂to␈α⊂aid␈α∂the␈α⊂untrained␈α⊂user␈α∂in
␈↓ x␈↓inserting new domain-specific BEINGs, and perhaps a whole library of varied specialist BEINGs.


␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α∧␈↓ε15␈↓␈↓ λ
␈↓ x␈↓␈↓↓␈↓&8.2. About BEINGs␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈α⊂performance␈α⊂of␈α⊂the␈α⊂BEINGs␈α⊂representation␈α⊂itself␈α⊂in␈α⊂PUP6␈α⊂is␈α⊂mixed.␈α⊂ Two␈α⊃advantages␈α⊂were
␈↓ x␈↓hoped␈αfor␈αby␈αusing␈αa␈αuniform␈αset␈αof␈αBEING␈αparts.␈α Addition␈αof␈αnew␈αBEINGs␈αto␈αthe␈αpool␈αwas␈αnot␈αeasy␈α(for
␈↓ x␈↓untrained␈α
users)␈α
but␈α
communication␈α
among␈α
BEINGs␈α
␈↓βwas␈↓␈α
easy␈α
(fast,␈α
natural).␈α
Two␈α
advantages␈α
were␈α
hoped␈α
for
␈↓ x␈↓by␈α
keeping␈α
the␈α∞BEINGs␈α
highly␈α
structured.␈α
 The␈α∞interactions␈α
(especially␈α
with␈α
the␈α∞user)␈α
were␈α
brittle,␈α∞but␈α
the
␈↓ x␈↓complex tasks put to the pool ␈↓βwere␈↓ successfully completed.
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈αcrippling␈αproblems␈αare␈α
seen␈αto␈αbe␈αwith␈αuser-system␈α
communication,␈αnot␈αwith␈αthe␈α
BEINGs␈αideas
␈↓ x␈↓themselves.␈α Sophisticated,␈α
bug-free␈αprograms␈α
␈↓βwere␈↓␈αgenerated,␈α
after␈αhours␈α
of␈αfairly␈α
high␈αlevel␈α
dialogue␈αwith
␈↓ x␈↓an␈α⊂active␈α⊃user,␈α⊂after␈α⊃tens␈α⊂of␈α⊂thousands␈α⊃of␈α⊂messages␈α⊃passed␈α⊂among␈α⊂the␈α⊃BEINGs.␈α⊂ Part␈α⊃of␈α⊂this␈α⊃success␈α⊂is
␈↓ x␈↓attributed␈α∂to␈α∂distributing␈α⊂the␈α∂responsibility␈α∂for␈α⊂writing␈α∂code␈α∂and␈α∂for␈α⊂recognizing␈α∂relevance,␈α∂to␈α⊂a␈α∂hundred
␈↓ x␈↓entities,␈α∞rather␈α∞than␈α∞having␈α∞a␈α∞few␈α∞central␈α∞monitors␈α∞worry␈α∞about␈α∞everything.␈α∞ The␈α∞standardization␈α∂of␈α∞parts
␈↓ x␈↓made filling in the BEINGs contents fairly painless.
␈↓"%␈↓ x␈↓␈↓ ↓XAll␈α
this␈α
suggests␈αtwo␈α
possible␈α
continuations,␈α
both␈αof␈α
which␈α
are␈α
underway␈αhere␈α
at␈α
the␈α
Stanford␈αAI␈α
Lab.
␈↓ x␈↓One␈α
is␈α
to␈αrethink␈α
the␈α
communication␈αproblems,␈α
and␈α
develop␈αa␈α
new␈α
system␈αfor␈α
the␈α
concept␈αformation␈α
program
␈↓ x␈↓synthesis␈α
task.␈α
The␈α
earliest␈α
programs␈αby␈α
our␈α
group␈α
[1]␈α
were␈α
aimed␈αat␈α
getting␈α
the␈α
target␈α
program;␈αthis␈α
research
␈↓ x␈↓insisted␈α⊃on␈α⊃getting␈α⊃the␈α⊃target␈α⊃program␈α⊃by␈α∩going␈α⊃through␈α⊃one␈α⊃"proper"␈α⊃sequence␈α⊃of␈α⊃reasoning␈α∩steps;␈α⊃the
␈↓ x␈↓proposed continuation wants several untrained users to  succeed by many different "proper" routes.
␈↓"%␈↓ x␈↓␈↓ ↓X The␈α
other␈α
way␈α
of␈α
continuing␈α∞is␈α
to␈α
find␈α
a␈α
task␈α∞where␈α
BEINGs␈α
are␈α
well-suited,␈α
where␈α∞the␈α
problems
␈↓ x␈↓encountered␈α∞in␈α∞PUP6␈α∞won't␈α∞recur.␈α∞What␈α∞␈↓βare␈↓␈α∞BEINGs␈α∞good␈α∞for?␈α∞ The␈α∞idea␈α∞of␈α∞a␈α∞fixed␈α∞set␈α∞of␈α∂parts␈α∞(which
␈↓ x␈↓distinguishes␈αthem␈αfrom␈αACTORs)␈αis␈αuseful␈α
if␈αthe␈αmass␈αof␈αknowledge␈αis␈α
too␈αhuge␈αfor␈αone␈αindividual␈αto␈α
keep
␈↓ x␈↓"on␈αtop"␈αof.␈α It␈αthen␈αshould␈αbe␈αorganized␈αin␈αa␈αvery␈αuniform␈αway␈α(to␈αsimplify␈αpreparing␈αit␈αfor␈αstorage),␈αyet␈αit
␈↓ x␈↓must␈α∂also␈α∂be␈α∂highly␈α⊂structured␈α∂(to␈α∂speed␈α∂up␈α⊂retrieval).␈α∂In␈α∂fact,␈α∂BEINGs␈α⊂can␈α∂be␈α∂utilized␈α∂within␈α⊂the␈α∂usual
␈↓ x␈↓solution␈αto␈αa␈αlarge␈αproblem:␈αa␈α
group␈αproject.␈αThe␈αmembers␈αsplit␈αapart␈α
after␈αagreeing␈αon␈αwhat␈αthe␈αset␈αof␈α
parts
␈↓ x␈↓is␈α
to␈α
be.␈α
In␈α
selecting␈α
the␈α
continuation,␈α
the␈α
problems␈α
in␈α
this␈α
research␈α
should␈α
be␈α
isolated␈α
from␈α
the␈αstaggering
␈↓ x␈↓complexities␈αof␈αnatural␈α
language␈αhandling.␈αFor␈α
these␈αreasons,␈αthe␈α
author␈αis␈αcurrently␈αinvestigating␈α
elementary
␈↓ x␈↓number␈α
theory␈α
as␈α
a␈α
potential␈α
task␈α
domain.␈α
 BEINGs␈α
are␈α
big␈α
and␈α
slow,␈α
but␈α
valuable␈α
for␈α
organizing␈α
knowledge
␈↓ x␈↓in␈α
ways␈α∞meaningful␈α
to␈α∞how␈α
it␈α∞will␈α
be␈α
used.␈α∞In␈α
future␈α∞systems,␈α
BEINGs␈α∞will␈α
be␈α
one␈α∞--␈α
but␈α∞not␈α
the␈α∞only␈α
--
␈↓ x␈↓internal mechanism for representing and manipulating knowledge.
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α1␈↓∧␈↓&References␈↓)αβ␈↓
␈↓"⊗␈↓ x␈↓[1]␈α∀Green,␈α∀Waldinger,␈α∀Barstow,␈α∀Elschlager,␈α∀Lenat,␈α∪McCune,␈α∀Shaw,␈α∀and␈α∀Steinberg,␈α∀␈↓βProgress␈α∀Report␈α∪on
␈↓"¬␈↓ x␈↓βProgram-Understanding␈α_Systems␈↓,␈α_Memo␈α_AIM-240,␈α_CS␈α_Report␈α→STAN-CS-74-444,Artificial␈α_Intelligence
␈↓"¬␈↓ x␈↓Laboratory, Stanford University, August, 1974.
␈↓"⊗␈↓ x␈↓[2] Hempel, Carl G., ␈↓βFundamentals of Concept Formation in Empirical Science␈↓, Chicago, 1952.
␈↓"⊗␈↓ x␈↓[3]␈α⊃Hewitt,␈α⊃Carl,␈α⊃␈↓βA␈α⊂Universal␈α⊃Modular␈α⊃ACTOR␈α⊃Formalism␈α⊂for␈α⊃Artificial␈α⊃Intelligence␈↓,␈α⊃Third␈α⊂International
␈↓"¬␈↓ x␈↓Joint Conference on Artificial Intelligence, 1973, pp. 235-245.
␈↓"⊗␈↓ x␈↓[4] Minsky, Marvin, ␈↓βFrames␈↓, in ␈↓βPsychology of Computer Vision␈↓, 1974.
␈↓"⊗␈↓ x␈↓[5] Teitelman, Warren, ␈↓βINTERLISP Reference Manual␈↓, XEROX PARC, 1974.
␈↓"⊗␈↓ x␈↓[6]␈α∪Winston,␈α∩Patrick,␈α∪␈↓βLearning␈α∩Structural␈α∪Descriptions␈α∩from␈α∪Examples␈↓,␈α∩Ph.D.␈α∪thesis,␈α∩Dept.␈α∪of␈α∩Electrical
␈↓"¬␈↓ x␈↓Engineering, TR-76, Project MAC, TR-231, MIT AI Lab, September, 1970.
␈↓"⊗␈↓ x␈↓[7]␈α⊃Woods,␈α⊃W.A.,␈α⊃and␈α⊃Makhoul,␈α⊃J.,␈α⊃␈↓βMechanical␈α⊃Inference␈α⊃Problems␈α⊃in␈α⊃Continuous␈α∩Speech␈α⊃Understanding␈↓,
␈↓"¬␈↓ x␈↓Third International Joint Conference on Artificial Intelligence, 1973, pp. 200-207.
␈↓"⊗␈↓ x␈↓The␈α
ideas␈α
and␈α
the␈α
system␈α
described␈α
are␈αbuilt␈α
upon␈α
recent␈α
researches.␈α
Many␈α
hours␈α
of␈α
creative␈αdiscussions␈α
were
␈↓"¬␈↓ x␈↓equally␈αimportant.␈α In␈αparticular,␈αthe␈αauthor␈α acknowledges␈αthe␈αcontributions␈αby␈αC.␈αGreen,␈αR.␈αWaldinger,␈αD.
␈↓"¬␈↓ x␈↓Shaw,␈α∀and␈α∀E.␈α∀Sacerdoti.␈α∃ Computer␈α∀time␈α∀for␈α∀the␈α∀research␈α∃was␈α∀generously␈α∀provided␈α∀by␈α∃the␈α∀Artificial
␈↓"¬␈↓ x␈↓Intelligence Center of SRI.
␈↓"	
␈↓"∧␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α∧␈↓ε16␈↓␈↓ λ