/LMAR=0/XLINE=14/FONT#0=BASL30.FNT[ IJ,DBL]/FONT#1=BASB30.FNT[XGP,SYS]/FONT#2=NGR25.FNT[XGP,SYS]/FONT#3=BASI30.FNT[ IJ,DBL]/FONT#4=NGR40.FNT[XGP,SYS]/FONT#5=FIX25.FNT[XGP,SYS]
␈↓ 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including␈αthat␈α
of␈αcontrol,␈αmay␈α
be␈αorganized␈αas␈αa␈α
community␈αof␈αinteracting␈α
modules␈α(e.g.,
␈↓ x␈↓ACTORS).␈α⊃Each␈α⊃module␈α⊃is␈α⊃granted␈α⊃a␈α⊃complex␈α⊃structure,␈α⊃to␈α⊃simulate␈α⊃a␈α⊃particular␈α⊃expert␈α⊃in␈α⊃some␈α⊂small
␈↓ x␈↓domain.␈α
 By␈α
constraining␈α∞that␈α
this␈α
structure␈α
be␈α∞standard␈α
over␈α
the␈α
entire␈α∞community,␈α
some␈α
advantages␈α∞of␈α
a
␈↓ x␈↓uniform␈α⊗formalism␈α⊗might␈α⊗be␈α⊗preserved.␈α⊗ The␈α⊗most␈α⊗natural␈α⊗task␈α⊗domain␈α⊗is␈α⊗shown␈α⊗to␈α↔be␈α⊗automatic
␈↓ x␈↓programming;␈αan␈αexperimental␈α
system␈αwas␈αpartially␈αimplemented␈α
for␈αthis.␈αIt␈αhas␈α
managed␈αto␈αsynthesize␈αa␈α
few
␈↓ x␈↓inductive␈α
inference␈α∞LISP␈α
programs,␈α∞nonformally,␈α
from␈α∞specific␈α
restricted␈α∞dialogues.␈α
 This␈α∞research␈α
clarified
␈↓ x␈↓some difficulties of using structured modular systems and of automatic programming.
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α$␈↓∧␈↓&1. Motivation␈↓)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⊂ When␈α∂a␈α∂topic␈α⊂is␈α∂being
␈↓ x␈↓considered,␈α∞one␈α
or␈α∞two␈α∞experts␈α
would␈α∞recognize␈α∞it␈α
and␈α∞demand␈α∞the␈α
floor.␈α∞In␈α∞the␈α
course␈α∞of␈α∞their␈α
exposition
␈↓ x␈↓they␈αmight␈αneed␈αto␈αcall␈αon␈αother␈αspecialists.␈αThis␈αmight␈αbe␈αby␈αname,␈αby␈αspecialty,␈αor␈αsimply␈αby␈αposing␈αa␈αnew
␈↓ x␈↓sub-question␈α⊂and␈α⊃hoping␈α⊂someone␈α⊂could␈α⊃recognize␈α⊂his␈α⊃own␈α⊂relevance␈α⊂and␈α⊃volunteer␈α⊂a␈α⊃suggestion.␈α⊂ Such
␈↓ x␈↓transfers␈αwould␈αbe␈αmore␈αcommon␈αat␈αthe␈αbeginning,␈αwhen␈αthe␈αtask␈αis␈α(by␈αassumption)␈αtoo␈αgeneral␈αfor␈αany␈αone
␈↓ x␈↓member␈α∞to␈α∞comprehend.␈α∞ As␈α∞the␈α∞questions␈α∞focus␈α∞on␈α∞more␈α∞specific␈α∞issues,␈α∞single␈α∞individuals␈α∞will␈α∞be␈α∂able␈α∞to
␈↓ x␈↓supply␈αcomplete␈αsolutions.␈α If␈αthe␈αtask␈αis␈αto␈αconstruct␈αsomething,␈αthen␈αthe␈αactivities␈αof␈αthe␈αexperts␈αshould␈αnot
␈↓ x␈↓be␈α
strictly␈αverbal.␈α
 Often,␈α
one␈αwill␈α
recognize␈α
his␈αrelevance␈α
to␈α
the␈αcurrent␈α
situation␈α
and␈αask␈α
to␈α
␈↓βdo␈↓␈αsomething:
␈↓ x␈↓create a new thing, modify one which already exists.
␈↓"%␈↓ 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␈↓corresponding␈α⊂to␈α∂a␈α⊂different␈α∂expert.␈α⊂What␈α∂must␈α⊂each␈α⊂program,␈α∂called␈α⊂a␈α∂␈↓βBEING␈↓,␈α⊂be␈α∂capable␈α⊂of?␈α⊂ It␈α∂must
␈↓ x␈↓possess␈αa␈α
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
␈↓ x␈↓and␈αanswering␈αother␈α
BEINGs,␈αincluding␈αthe␈αability␈α
to␈αrecognize␈αwhen␈αit␈α
is␈αrelevant.␈α BEINGs␈αmust␈α
also␈αbe
␈↓ x␈↓able to set up and to alter structures representing what the humans produce.
␈↓"%␈↓ x␈↓␈↓ ↓XTo␈α
be␈α∞more␈α
concrete,␈α
suppose␈α∞the␈α
humans'␈α
task␈α∞is␈α
to␈α
design␈α∞and␈α
code␈α
a␈α∞large␈α
computer␈α∞program:␈α
a
␈↓ x␈↓concept␈α∂formation␈α∞system[2].␈α∂Some␈α∞experts␈α∂will␈α∞be␈α∂scientific␈α∞programmers,␈α∂non-programming␈α∞psychologists,
␈↓ x␈↓system␈αhackers,␈αmanagement␈αpersonnel,␈αand␈αso␈αon.␈α What␈αhappens␈αin␈αthe␈αensuing␈αsession?␈α When␈αan␈αexpert
␈↓ x␈↓participates,␈α⊂he␈α⊂will␈α∂either␈α⊂be␈α⊂aiding␈α∂a␈α⊂colleague␈α⊂in␈α∂some␈α⊂difficulty␈α⊂or␈α∂else␈α⊂transferring␈α⊂a␈α∂tiny␈α⊂bit␈α⊂of␈α∂his
␈↓ x␈↓expertise␈α∩into␈α∩a␈α∩programmed␈α∩function␈α∩which␈α∩can␈α∩do␈α∩something.␈α∩ The␈α∩final␈α∩code␈α∩reflects␈α∪the␈α∩members'
␈↓ x␈↓knowledge.␈α
 One␈α
way␈α
the␈αsession␈α
might␈α
proceed␈α
is␈αfor␈α
the␈α
specialists␈α
to␈αactually␈α
␈↓βdo␈↓␈α
the␈α
concept␈αformation␈α
task.
␈↓ x␈↓As␈αthey␈αbecome␈αfamiliar␈αwith␈αwhat␈αpart␈αof␈αtheir␈αown␈αexpertise␈αis␈αbeing␈αcalled␈αupon,␈αand␈αin␈αwhat␈αways,␈αthey
␈↓ x␈↓can␈α
begin␈α
to␈α
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
␈↓ x␈↓knowledge,␈α
organize␈αthem,␈α
formalize␈α
them,␈αand␈α
program␈α
them.␈α A␈α
conscious␈α
effort␈αalong␈α
these␈α
lines␈αwas␈α
made
␈↓ x␈↓in␈α
[7],␈αwhere␈α
experts␈α
gradually␈αreplaced␈α
themselves␈αby␈α
programs.␈α
 Instead␈αof␈α
discussing␈α
how␈αto␈α
write␈αa␈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 them.
␈↓"%␈↓ x␈↓␈↓ ↓XHow␈αcould␈αBEINGs␈αdo␈αthis?␈αThere␈αwould␈αbe␈αsome␈αlittle␈αprogram␈αcontaining␈αinformation␈α(much␈αof␈αit
␈↓ x␈↓never␈α
to␈αbe␈α
used)␈α
about␈α␈↓αCONCEPT-FORMATION␈↓,␈α
another␈αBEING␈α
who␈α
knows␈αhow␈α
to␈α
manage␈αa␈α
group␈αto␈α
␈↓αWRITE-
␈↓ x␈↓αPROGRAMS␈↓,␈αand␈αmany␈αlower-level␈αspecialists,␈αfor␈αexample␈α␈↓αOBTAIN-USABLE-INFORMATION,␈αTEST,␈αMODIFY-DATA-
␈↓ x␈↓αSTRUCTURE,␈α≤UNTIL-LOOP,␈α≥VISUAL-PERCEPTION,␈α≤AVOID-CONTRADICTION,␈α≥PROPOSE-PLAUSIBLE-NAME␈↓.␈α≤ Some
␈↓ x␈↓chairman␈αmight␈αbe␈αnecessary␈α(a␈αglobal␈αmonitor),␈αor␈αmerely␈αuniversally␈αaccepted␈αmeans␈αof␈αcommunicating,␈αof
␈↓ x␈↓gaining␈α⊂the␈α⊂floor,␈α⊂and␈α⊂of␈α⊂resolving␈α⊂disputes.␈α∂ The␈α⊂project␈α⊂sponsor␈α⊂might␈α⊂be␈α⊂passive,␈α⊂submitting␈α⊂a␈α∂single
␈↓ x␈↓specification␈αorder␈αfor␈αthe␈αprogram,␈αor␈αactive,␈αparticipating␈αin␈αthe␈αwork␈αas␈αa␈α(somewhat␈αpriveleged)␈αmember
␈↓ x␈↓of␈αthe␈αteam.␈αThis␈αindividual␈αis␈αthe␈αone␈αwho␈αwants␈αthe␈αfinal␈αproduct,␈αhence␈αwill␈αbe␈αcalled␈αthe␈α␈↓βuser␈↓.␈αAlthough
␈↓ x␈↓specialists,␈αthe␈αBEINGs␈αwould␈αcontain␈αfar␈αtoo␈αmuch␈αinformation,␈αfar␈αtoo␈αinefficiently␈αrepresented,␈αto␈αbe␈αable
␈↓ x␈↓to␈α⊂say␈α⊂"we␈α⊂ourselves␈α∂are␈α⊂the␈α⊂desired␈α⊂program!"␈α∂They␈α⊂would␈α⊂have␈α⊂to␈α∂discuss,␈α⊂and␈α⊂perhaps␈α⊂carry␈α⊂out,␈α∂the
␈↓ x␈↓concept␈α⊃formation␈α⊃task.␈α⊃They␈α∩would␈α⊃write␈α⊃specialized␈α⊃versions␈α∩of␈α⊃themselves,␈α⊃programs␈α⊃which␈α∩could␈α⊃do
␈↓ x␈↓exactly␈α
what␈αthe␈α
BEINGs␈αdid␈α
to␈αcarry␈α
out␈αthe␈α
task,␈αno␈α
more␈αnor␈α
less.␈αSome␈α
BEINGs␈α(e.g.,␈α
␈↓αTEST␈↓)␈α
may␈αhave
␈↓ x␈↓several␈α
specialized,␈αstreamlined␈α
fractions␈αof␈α
themselves␈αin␈α
the␈α
final␈αprogram.␈α
BEINGs␈αwhich␈α
only␈αaided␈α
other
␈↓ x␈↓BEINGs (e.g., ␈↓αPROPOSE-PLAUSIBLE-NAME␈↓) will not have ␈↓βany␈↓ correlates in the ultimate synthesized code.
␈↓"%␈↓ x␈↓␈↓ ↓XA␈αsystem␈αwas␈αdesigned␈αand␈αpartially␈αimplemented␈αto␈αstudy␈αthis␈αprocess.␈αA␈αconcept␈αformation␈αprogram
␈↓ x␈↓similar␈α
to␈α
[6]␈αwas␈α
actually␈α
synthesized,␈α
but␈αthe␈α
user␈α
must␈α
come␈αup␈α
with␈α
certain␈α
specific␈αanswers␈α
to␈α
some␈αof␈α
the
␈↓ x␈↓BEINGs'␈αcritical␈αqueries.␈α
 A␈αgrammatical␈αinference␈α
program␈αand␈αa␈α
simple␈αproperty␈αlist␈α
maintenance␈αroutine
␈↓ x␈↓were␈αalso␈αgenerated,␈αbut␈α
similar␈αdialogue␈αproblems␈αplagued␈α
these␈αefforts.␈α Before␈αdescribing␈α
this␈αcommunity
␈↓ x␈↓of BEINGs, individual BEINGs are vivisected.
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α<␈↓∧␈↓&2. External Behavior of a BEING␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓X␈↓αWRITE-PROGRAM␈↓␈α∂is␈α∂a␈α∂high-level,␈α∂programming-knowledge␈α∂BEING.␈α∂It␈α∂can␈α∂answer␈α⊂several␈α∂questions
␈↓ x␈↓relevant␈αto␈α
managing␈αcode␈α
production,␈αand␈α
in␈αthe␈α
course␈αof␈α
answering␈αit␈α
calls␈αon␈α
other␈αBEINGs␈α
when␈αit␈α
feels
␈↓ x␈↓deficient.␈αHere␈αare␈α
some␈αof␈αthe␈α
things␈αwhich␈αanother␈αBEING␈α
--␈αor␈αthe␈α
user␈α--␈αmight␈α
ask␈α␈↓αWRITE-PROGRAM␈↓,
␈↓ x␈↓and its replies:

␈↓"	␈↓ x␈↓Can you recognize this phrase: "Write a program which does frob gyrnation" ?
␈↓"	␈↓ x␈↓␈↓α  If anyone can recognize the act of "frob gyrnation", then I will handle it.␈↓

␈↓"	␈↓ x␈↓What argument(s) do you require?
␈↓"	␈↓ x␈↓␈↓α  One argument, binding (by λ) the variable "task". Two local variables are used.␈↓

␈↓"	␈↓ x␈↓What do you do?
␈↓"	␈↓ x␈↓␈↓α  I write a program which does <value of task goes here>.␈↓
␈↓ x␈↓Why?
␈↓"	␈↓ x␈↓␈↓α  Fundamental drive to serve; code must be written to do <task>; ask <whoever called me>.␈↓

␈↓"	␈↓ x␈↓When should you take control?
␈↓"	␈↓ x␈↓␈↓α  If <task> can already be done in some way, I probably shouldn't speak up.
␈↓"	␈↓ x␈↓α  If the user is aware that I want to go, and does not voice objection, I will go.
␈↓"	␈↓ x␈↓α  If there is any new but unassimilated info around, it should be scrutinized first.␈↓

␈↓"	␈↓ x␈↓What happens when you are in control?
␈↓"	␈↓ x␈↓␈↓α  A program for <task> gets initialized.
␈↓"	␈↓ x␈↓α  Until it is completed, I repeatedly pass control among four BEINGs
␈↓"	␈↓ x␈↓α    (OBTAIN-USABLE-INFO, USE-INFO, FILL-IN-UNDEFINED-PIECE, CLARIFY-IMPROBABLE-PIECE)
␈↓"	␈↓ x␈↓α    They, not I, decide which of them should be in control at each instant.
␈↓"	␈↓ x␈↓α  It is complete when code for <task> exists, there are no coding-warnings, no new info,
␈↓"	␈↓ x␈↓α    no undefined pieces of code, no deferred decisions left unresolved.
␈↓"	␈↓ x␈↓α  When complete, I assert this fact, call on SUPPORT-AND-DUMP to optimize the program
␈↓"	␈↓ x␈↓α    and store it on a new disk file, and relinquish control to whoever wants it most.␈↓

␈↓"	␈↓ x␈↓What must be made true before you start all this?
␈↓"	␈↓ x␈↓␈↓α  Some BEING must make the user aware that I am about to write a program for <task>.
␈↓"	␈↓ x␈↓α  <task> must have a proper name <task-name> (some BEING must bring this effect about).␈↓

␈↓"	␈↓ x␈↓What must be attempted after you are finished?
␈↓"	␈↓ x␈↓␈↓α  Try to make the user aware that I have written the program <task-name> for <task>␈↓

␈↓"	␈↓ x␈↓What should you be on the watch for while you have the floor?
␈↓"	␈↓ x␈↓␈↓α  Activate demons for deferral and reinvestigation of decisions.␈↓

␈↓"	␈↓ x␈↓What are the main effects of your having (successfully) finished your job?
␈↓"	␈↓ x␈↓␈↓α  The user, and any BEING, can do <task>; code has been written for <task>.␈↓

␈↓"	␈↓ x␈↓How complex/costly are you?
␈↓"	␈↓ x␈↓␈↓α  The difficulty to call is 0.3  +  0.7 * (estimated difficulty of <task>)
␈↓"	␈↓ x␈↓α  There is a 0.9 chance that I will be called recursively by myself or a descendant.
␈↓"	␈↓ x␈↓α  There is a 0.7 chance of my success, in general, if I actually gain control.
␈↓"	␈↓ x␈↓α  I will use a disproportionate share of time and space if I am called.
␈↓"	␈↓ x␈↓α  There is generally no reason to absolutely inhibit me.␈↓

␈↓"%␈↓ x␈↓␈↓ ↓XThe␈α
fractional␈α
numbers␈α
in␈α∞the␈α
last␈α
answer␈α
represent␈α∞the␈α
author's␈α
guesses,␈α
on␈α∞a␈α
scale␈α
of␈α
zero␈α∞to␈α
one.
␈↓ x␈↓One␈α∞question␈α∞which␈α∞can't␈α∞be␈α∞answered␈α∞by␈α∂␈↓αWRITE-PROGRAM␈↓␈α∞is␈α∞"How␈α∞do␈α∞you␈α∞write␈α∞a␈α∞specialized␈α∂version␈α∞of
␈↓ x␈↓yourself?"␈α⊂Many␈α⊂BEINGs␈α⊃␈↓βdo␈↓␈α⊂have␈α⊂this␈α⊃ability,␈α⊂and␈α⊂a␈α⊂brief␈α⊃example␈α⊂is␈α⊂now␈α⊃given.␈α⊂ ␈↓αPARTITION-A-DOMAIN␈↓
␈↓ x␈↓contains␈α∂knowledge␈α∂relevant␈α∂to␈α∞many␈α∂kinds␈α∂of␈α∂inductive␈α∞inference␈α∂programs,␈α∂including␈α∂concept␈α∞formation.
␈↓ x␈↓Here are a few of its answers:

␈↓"	␈↓ x␈↓␈↓ ↓XWhat is your basic idea, what do you do?
␈↓"	␈↓ x␈↓␈↓ ↓X␈↓α  Divide a space of elements into named classes, by gradually building up a partition.␈↓
␈↓ x␈↓␈↓ ↓XWhat happens when you are in control?
␈↓"	␈↓ x␈↓␈↓ ↓X␈↓α  Repeatedly: accept an input;
␈↓"	␈↓ x␈↓α␈↓ ↓X        if it is a <element,class-name> pair, merely add it to the current partition;
␈↓"	␈↓ x␈↓α␈↓ ↓X        if it has element but no name, try to guess its name, then check with the user;
␈↓"	␈↓ x␈↓α␈↓ ↓X        if it has name but no element description, guess what element it might be, and check;
␈↓"	␈↓ x␈↓α␈↓ ↓X        if I had to guess, some modification of my guessing strategies might be done.␈↓

␈↓"	␈↓ x␈↓␈↓ ↓XInstead of acting, how can you write a streamlined version of yourself?
␈↓"	␈↓ x␈↓␈↓ ↓X␈↓α  If can guarantee that partition is total, every element must be tied to some class-name.
␈↓"	␈↓ x␈↓α␈↓ ↓X  If can guarantee that partition is strict, no element can belong to >1 class ever.
␈↓"	␈↓ x␈↓α␈↓ ↓X  Only some of the three types of inputs might ever occur, might need to be differentiated.␈↓

␈↓"	␈↓ x␈↓␈↓ ↓XWhat BEINGs should be considered if you fail?
␈↓"	␈↓ x␈↓␈↓ ↓X  ␈↓αDISCRETIZE is an alternative; CREATE-FUNCTION is a generalization.␈↓

␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α<␈↓∧␈↓&3. Internal Structure of a BEING␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XEach␈α
BEING␈α
is␈α
composed␈α
of␈α
a␈α
couple␈α
dozen␈α
parts,␈α
each␈α
of␈α
which␈α
is␈α
the␈α
name␈α
of␈α
a␈α
question␈α
and␈αa
␈↓ x␈↓little␈αprogram␈αfor␈α
how␈αthis␈αBEING␈αshould␈α
answer␈αit.␈αThis␈αanswering␈α
program␈αmay␈αitself␈αpose␈α
questions␈α(to
␈↓ x␈↓other␈α⊂parts␈α⊂of␈α⊂the␈α⊂same␈α∂BEING,␈α⊂to␈α⊂parts␈α⊂of␈α⊂other␈α⊂BEINGs,␈α∂to␈α⊂the␈α⊂user),␈α⊂assert␈α⊂propositions,␈α⊂create␈α∂and
␈↓ x␈↓modify structures (including demons and parts of BEINGs).
␈↓"%␈↓ x␈↓␈↓ ↓XSeveral␈α∞design␈α∞questions␈α∞arise:␈α∞what␈α∞parts␈α∞are␈α∂filled␈α∞in␈α∞for␈α∞each␈α∞BEING,␈α∞how␈α∞does␈α∞any␈α∂part␈α∞know
␈↓ x␈↓what␈αquestions␈αit␈αmay␈αask␈αeach␈αBEING,␈αhow␈αdo␈αthe␈αvalues␈αof␈αthe␈αparts␈αget␈αdetermined,...?␈α Before␈α
discussing
␈↓ x␈↓these␈α
design␈αaspects␈α
of␈αan␈α
interacting␈αcommunity,␈α
the␈α
actual␈αparts␈α
of␈αa␈α
particular␈αBEING␈α
in␈αthe␈α
experimental
␈↓ x␈↓system␈α⊂will␈α∂be␈α⊂presented.␈α∂Consider␈α⊂␈↓αOBTAIN-USABLE-INFORMATION␈↓,␈α∂a␈α⊂BEING␈α∂which␈α⊂is␈α∂independent␈α⊂of␈α∂task
␈↓ x␈↓domain.␈α Below␈αis␈αlisted,␈αfor␈αeach␈αpart,␈αits␈αabbreviated␈αname,␈αa␈αfuller␈αEnglish␈αquestion,␈αthe␈αvalue␈αstored,␈αand
␈↓ x␈↓a brief explanation.


␈↓"	␈↓ x␈↓α␈↓↓WHAT␈↓α        What do you do?          (OBTAIN SOME INFORMATION WHICH CAN BE USED)
␈↓"	␈↓ x␈↓α␈↓↓WHY␈↓α          Why?     (PUP HAS NO MORE INFORMATION THAT IT CAN USE TO PROGRESS)
␈↓"	␈↓ x␈↓α␈↓↓HOW␈↓α          How?     (OBTAIN NEW FACTS ABOUT OLD INFORMATION, OR OBTAIN TOTALLY NEW INFO)
␈↓"%␈↓ x␈↓␈↓ ↓XThese␈αthree␈αparts␈αare␈αprimarily␈αfor␈αthe␈αuser's␈αbenerfit.␈αIn␈αthe␈αcurrent␈αcase,␈αthey␈αdon't␈αeven␈αhave␈αany
␈↓ x␈↓uninstantiated␈α⊃parts.␈α⊃ If␈α⊂the␈α⊃user␈α⊃asks␈α⊂one␈α⊃of␈α⊃these␈α⊃questions␈α⊂when␈α⊃this␈α⊃BEING␈α⊂is␈α⊃in␈α⊃control,␈α⊃then␈α⊂the
␈↓ x␈↓appropriate answer can simply be typed out.


␈↓"	␈↓ x␈↓α␈↓↓IDEN␈↓α           Can you recognize this phrase: "Find out more about frob gyrnation"?
␈↓"	␈↓ x␈↓α((if you see:  (OBTAIN-USABLE-INFORMATION any1)
␈↓"	␈↓ x␈↓α                                then return:(OBTAIN-USABLE-INFORMATION (TRANSLATE any1)))
␈↓"	␈↓ x␈↓α (if you see:  (FIND OUT MORE ABOUT any1)
␈↓"	␈↓ x␈↓α                                then return:(OBTAIN-USABLE-INFORMATION any1)))
␈↓ x␈↓␈↓ ↓XThe␈α∞phrase␈α∂matches␈α∞the␈α∂second␈α∞template.␈α∂␈↓αOBTAIN-USABLE-INFORMATION␈↓␈α∞is␈α∂given␈α∞control,␈α∂with␈α∞"frob
␈↓ x␈↓gyrnation"␈α
as␈α
its␈α
argument.␈αIf␈α
it␈α
does␈α
not␈αfail,␈α
the␈α
value␈α
it␈αreturns␈α
is␈α
taken␈α
to␈αbe␈α
the␈α
translation␈α
of␈αthe␈α
original
␈↓ x␈↓phrase.␈α
 Each␈α
BEING␈α∞is␈α
given␈α
the␈α∞duty␈α
of␈α
recognizing␈α∞and␈α
processing␈α
English␈α∞phrases␈α
related␈α
to␈α∞him.␈α
 In
␈↓ x␈↓practice,␈α
only␈α
a␈α
few␈α
simple␈α
phrase␈αtemplates␈α
were␈α
needed␈α
for␈α
each␈α
BEING;␈α␈↓αOBTAIN-USABLE-INFORMATION␈↓␈α
only
␈↓ x␈↓used␈α⊂two␈α⊂patterns.␈α⊂ All␈α⊃the␈α⊂IDEN␈α⊂parts␈α⊂are␈α⊃collected␈α⊂together␈α⊂into␈α⊂one␈α⊂table.␈α⊃ When␈α⊂a␈α⊂form␈α⊂␈↓βF␈↓␈α⊃must␈α⊂be
␈↓ x␈↓translated,␈α
the␈α∞␈↓αTRANSLATE␈↓␈α
BEING␈α∞takes␈α
control␈α∞and␈α
(by␈α∞searching␈α
this␈α∞table)␈α
asks␈α∞"who␈α
can␈α∞recognize␈α
␈↓βF␈↓?"
␈↓ x␈↓Each␈αIDEN␈αpart␈αruns␈αa␈αlittle␈αprogram␈α(typically␈αa␈αpattern-match),␈αthen␈αreplies␈αeither␈αNO,␈αor␈αelse␈αprovides␈αa
␈↓ x␈↓little␈α∞program␈α∞whose␈α∞value␈α∞should␈α∂be␈α∞the␈α∞translation␈α∞of␈α∞␈↓βF␈↓.␈α∞ If␈α∂there␈α∞is␈α∞more␈α∞than␈α∞one␈α∞responder,␈α∂a␈α∞global
␈↓ x␈↓choice␈α∂BEING,␈α∂␈↓αCHOOSE-FROM␈↓,␈α∂selects␈α∂the␈α∂winner.␈α∂ Often␈α∂the␈α∂program␈α∂which␈α∂is␈α∂returned␈α⊂calls␈α∂␈↓αTRANSLATE␈↓
␈↓ x␈↓recursively on some parts of ␈↓βF␈↓.


␈↓"	␈↓ x␈↓α␈↓↓EXPLICIT-ARGS␈↓α       What argument(s) do you take?           ( U )
␈↓"	␈↓ x␈↓α␈↓↓NLAMBDA␈↓α                 Which are quoted, not evaluated?         NIL
␈↓"	␈↓ x␈↓α␈↓↓IMPLICIT-ARGS␈↓α       What local variables are needed?         NIL
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈αargument␈αto␈αthis␈α
BEING␈αis␈αnever␈αoperated␈α
upon,␈αhence␈αthere␈αis␈α
no␈αneed␈αto␈αunderstand␈α
anything
␈↓ x␈↓about it.  It is simply passed to BEINGs called by ␈↓αOBTAIN-USABLE-INFORMATION␈↓.


␈↓"	␈↓ x␈↓α␈↓↓WHEN␈↓α        When should you take control (justify your answer)?
␈↓"	␈↓ x␈↓α((if T then add in -10 because (THIS IS EXPONENTIALLY-GROWING, GENERALLY UNDESIRABLE))
␈↓"	␈↓ x␈↓α (if NEW-INFO-LIST then add in  (PLUS  100  (LENGTH  NEW-INFO-LIST))
␈↓"	␈↓ x␈↓α        because (WE SHOULD WORK ON UNASSIMILATED NEW  INFORMATION IF THERE IS ANY)))
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈αWHEN␈α
part␈αof␈α
a␈αBEING␈α
is␈αa␈α
collection␈αof␈α
triples:␈αif␈α
<predicate>␈αthen␈α
<value>␈αbecause␈α
<reason>.
␈↓ x␈↓If␈α
the␈α<predicate>␈α
evaluates␈α
to␈αnon-null,␈α
then␈α
the␈α<value>␈α
program␈α
is␈αexecuted.␈α
It␈α
returns␈αa␈α
number,␈αwhich␈α
is
␈↓ x␈↓then␈α⊂added␈α⊂together␈α∂with␈α⊂the␈α⊂other␈α∂factors'␈α⊂numbers␈α⊂to␈α⊂produce␈α∂a␈α⊂rough␈α⊂estimate␈α∂of␈α⊂how␈α⊂a␈α⊂propos␈α∂this
␈↓ x␈↓BEING␈α⊂is␈α⊃to␈α⊂take␈α⊃control␈α⊂right␈α⊃now.␈α⊂ The␈α⊃<reason>␈α⊂evaluates␈α⊃to␈α⊂an␈α⊃English␈α⊂phrase,␈α⊃for␈α⊂the␈α⊃benefit␈α⊂of
␈↓ x␈↓inquisitive␈αusers.␈α This␈αlinear␈αscheme␈αis␈αundesirable␈αbut␈α(sigh)␈αadequate.␈αThe␈αfirst␈αfactor␈αhere␈αsays␈αto␈αalways
␈↓ x␈↓add␈αin␈αthe␈αnumber␈α-10;␈αthe␈αsecond␈αsays␈αif␈αthere␈αis␈αsome␈αnew␈αinformation␈αsitting␈αaround␈αunexploited,␈αto␈αadd
␈↓ x␈↓in␈α100␈αplus␈αthe␈αnumber␈αof␈αsuch␈αpieces.␈α These␈αfactors␈αand␈αtheir␈αweights,␈αlike␈αthe␈αcontents␈αof␈αall␈αthe␈αparts␈αof
␈↓ x␈↓all the BEINGs initially in the experimental system, were decided upon and inserted by hand.


␈↓"	␈↓ x␈↓α␈↓↓META-CODE␈↓α           What happens when you are in control?
␈↓"	␈↓ x␈↓α(DO
␈↓"	␈↓ x␈↓α     (CHOOSE-FROM ((GET-NEW-INFORMATION U)
␈↓"	␈↓ x␈↓α                         (TRANSLATE U)
␈↓"	␈↓ x␈↓α                         (ANALYZE-IMPLICATIONS U)
␈↓"	␈↓ x␈↓α                         (EXTRACT-RELEVANT-SUBSET U)))
␈↓"	␈↓ x␈↓α  BECAUSE
␈↓"	␈↓ x␈↓α      (WE CAN ONLY TRY TO OBTAIN USABLE INFORMATION IN ONE WAY AT A TIME))
␈↓ x␈↓␈↓ ↓XOnce␈αit␈αgains␈αcontrol,␈α
the␈αBEING␈αchooses␈αfrom␈α
the␈αfollowing␈αfour␈αalternatives,␈α
each␈αof␈αwhich␈αis␈α
itself
␈↓ x␈↓a␈α∞BEING:␈α∂Get-Brand-New-Information␈α∞(in␈α∞English,␈α∂from␈α∞the␈α∞user),␈α∂Translate␈α∞something␈α∂(transform␈α∞from
␈↓ x␈↓English␈α⊂to␈α⊂BEING-calls),␈α⊂Analyze-The-Implications␈α∂(of␈α⊂some␈α⊂existing,␈α⊂translated␈α⊂information),␈α∂Extract-a-
␈↓ x␈↓Relevant-Subset (of the existing information) to concentrate upon.


␈↓"	␈↓ x␈↓α␈↓↓MAIN-EFFECTS␈↓α      Can you cause this to occur: "Usable information exists"?
␈↓"	␈↓ x␈↓α((to get (NEW INFORMATION any1)       do (OBTAIN-USABLE-INFORMATION any1))
␈↓"	␈↓ x␈↓α (to get (USABLE INFORMATION any1)   do (OBTAIN-USABLE-INFORMATION any1)))
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈α∂EFFECTS␈α∂parts␈α∂of␈α∂each␈α∂BEING␈α∂are␈α∞collected␈α∂into␈α∂one␈α∂table␈α∂to␈α∂facilitate␈α∂asking␈α∂all␈α∞BEINGs
␈↓ x␈↓simultaneously␈α∂"Can␈α∂you␈α∂cause␈α∂effect␈α∂␈↓βX␈↓␈α∂to␈α∂occur?"␈α∞Each␈α∂EFFECTS␈α∂part␈α∂examines␈α∂␈↓βX␈↓␈α∂and␈α∂the␈α∂world,␈α∞and
␈↓ x␈↓either␈α
replies␈α
␈↓βNo␈↓,␈α∞or␈α
else␈α
returns␈α
a␈α∞little␈α
program␈α
(involving␈α
calls␈α∞and␈α
constants)␈α
which␈α
should␈α∞produce␈α
the
␈↓ x␈↓effect␈α⊂when␈α⊂it␈α⊂is␈α⊂run.␈α⊂ This␈α⊂program␈α⊂generally␈α∂will␈α⊂involve␈α⊂a␈α⊂call␈α⊂to␈α⊂the␈α⊂BEING␈α⊂itself.␈α∂␈↓αOBTAIN-USABLE-
␈↓ x␈↓αINFORMATION␈↓ shows that it should be called to acheive the existence of new or usable information.


␈↓"	␈↓ x␈↓α␈↓↓AFFECTS␈↓α             What other BEINGs might you call on directly?
␈↓"	␈↓ x␈↓α((CHOOSE-FROM is called)
␈↓"	␈↓ x␈↓α (some BEING who can cause (AWARE USER (ABOUT TO OBTAIN USABLE INFO)) is called)
␈↓"	␈↓ x␈↓α (GET-NEW-INFORMATION possibly is called)
␈↓"	␈↓ x␈↓α (TRANSLATE possibly is called)
␈↓"	␈↓ x␈↓α (ANALYZE-IMPLICATIONS possibly is called)
␈↓"	␈↓ x␈↓α (EXTRACT-RELEVANT-SUBSET possibly is called) )
␈↓"%␈↓ x␈↓␈↓ ↓XThis␈αpart␈αis␈αclear;␈α␈↓αCHOOSE-FROM␈↓␈αis␈αdefinitely␈αgoing␈αto␈αbe␈αcalled␈αon,␈αanother␈αone␈αcalled␈αon␈αis␈αspecified
␈↓ x␈↓only by its effects, and each of the other four BEINGs mentioned might be called and might not be.


␈↓"	␈↓ x␈↓α␈↓↓COMPLEXITY-VECTOR␈↓α           How costly are you?             (.5 .5 .9 .5 .1)
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈α
first␈α
component␈α
says␈α
that␈α
␈↓αOBTAIN-USABLE-INFORMATION␈↓␈α
is␈α
of␈α
average␈α
difficulty␈α
to␈α
call.␈αNext,␈α
there
␈↓ x␈↓exists␈α∞a␈α
.5␈α∞chance␈α
that␈α∞some␈α
descendant␈α∞will␈α
call␈α∞it␈α
again.␈α∞The␈α
third␈α∞component␈α
indicates␈α∞that␈α∞this␈α
activity
␈↓ x␈↓almost␈α∞always␈α∞succeeds.␈α∞The␈α∞time/space␈α∞used␈α∞in␈α∞allowing␈α∞this␈α∞BEING␈α∞to␈α∞try␈α∞is␈α∞typical.␈α∞Finally,␈α∞there␈α∞is␈α∞no
␈↓ x␈↓good reason for inhibiting it ever.  In general, each component can be a ␈↓βprogram␈↓, not just a constant.


␈↓"	␈↓ x␈↓α␈↓↓GENERALIZATIONS␈↓α      What BEINGs are more general than you?
␈↓"	␈↓ x␈↓α(WRITE-PROGRAM SERVE-THE-USER)
␈↓"	␈↓ x␈↓α␈↓↓ALTERNATIVES␈↓α          What BEINGs are similar to you, alternatives in case you fail?
␈↓"	␈↓ x␈↓α(USE-INFORMATION FIX-INCORRECT-PIECE OPTIMIZE FILL-IN-UNDEFINED SECTION)
␈↓"%␈↓ x␈↓␈↓ ↓XThese␈αlist␈αthings␈αto␈αtry␈αif␈αthis␈αBEING␈αfails␈αin␈αits␈αmission.␈αThe␈αalternatives␈αshould␈αgenerally␈αbe␈αtried
␈↓ x␈↓before the more general BEINGs.
␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α)␈↓∧␈↓&4. A Universal Set of BEING Parts␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XHaving␈α
decided␈α
to␈α∞represent␈α
knowledge␈α
as␈α
interacting␈α∞modules,␈α
communicating␈α
via␈α∞questioning,␈α
one
␈↓ x␈↓must␈α
now␈α∞examine␈α
how␈α
each␈α∞BEING␈α
knows␈α∞what␈α
questions␈α
it␈α∞can␈α
ask␈α
each␈α∞other␈α
BEING.␈α∞ ACTORS␈α
[3]
␈↓ x␈↓solves␈αthis␈αproblem␈αby␈α
␈↓βhandshaking␈↓;␈αeach␈αmodule␈αmust␈α
be␈αcompatible␈αwith␈αall␈αthe␈α
others␈αit␈αis␈αever␈α
going␈αto
␈↓ x␈↓interact␈α∞with.␈α∞While␈α∞this␈α∞makes␈α
the␈α∞representation␈α∞of␈α∞each␈α∞ACTOR␈α
efficient,␈α∞it␈α∞is␈α∞conceptually␈α∞difficult␈α
to
␈↓ x␈↓add␈αnew␈αACTORS␈αto␈αthe␈αcommunity␈αwithout␈αknowing␈αa␈αgreat␈αdeal␈αabout␈αwho␈αis␈αthere␈αalready.␈αA␈αuniversal
␈↓ x␈↓set␈α
of␈αBEING␈α
parts␈α(=␈α
ACTOR␈α
message␈αformats)␈α
might␈αbe␈α
developed,␈α
in␈αthe␈α
sense␈αthat␈α
each␈α
module␈αmust
␈↓ x␈↓have␈α
each␈α
part␈α
filled␈α
in.␈α
The␈αdanger␈α
here␈α
is␈α
the␈α
huge␈α
number␈αof␈α
parts␈α
which␈α
might␈α
be␈α
necessary,␈α
and␈αthe
␈↓ x␈↓irrelevance␈αof␈α
most␈αof␈α
them␈αto␈α
any␈αgiven␈α
BEING.␈αPerhaps␈α
by␈αrestricting␈α
the␈αtask␈α
that␈αthe␈α
pool␈αof␈αBEINGs␈α
is
␈↓ x␈↓supposed to do, the number of parts necessary might drop to a reasonable figure.
␈↓"%␈↓ x␈↓␈↓ ↓XAn␈α
experimental␈α
pool␈α
of␈αa␈α
hundred␈α
BEINGs␈α
was␈α
designed␈αand␈α
partially␈α
implemented,␈α
to␈α
test␈αboth␈α
the
␈↓ x␈↓hypothesis␈αof␈αa␈αuniversal␈αset␈αof␈αparts␈αand␈αthe␈αfeasiability␈αof␈αBEINGs␈αin␈αgeneral.␈α The␈αsystem,␈αnamed␈αPUP6,
␈↓ x␈↓was␈α⊂aimed␈α∂at␈α⊂synthesizing␈α⊂LISP␈α∂programs␈α⊂from␈α∂restricted␈α⊂discussions␈α⊂within␈α∂itself␈α⊂and␈α∂with␈α⊂the␈α⊂user.␈α∂In
␈↓ x␈↓particular,␈α
the␈α
community's␈αtask␈α
was␈α
to␈αgenerate␈α
a␈α
particular␈αconcept␈α
formation␈α
program,␈αcommunicating␈α
with
␈↓ x␈↓the␈αuser␈αin␈αa␈αminiscule␈αsubset␈αof␈αEnglish.␈α So␈αa␈αfixed␈αset␈αof␈αBEING␈αparts␈αneed␈αonly␈αbe␈αrelevant␈αto␈αwriting␈αa
␈↓ x␈↓small␈α⊃subclass␈α⊃of␈α⊃inductive␈α⊃inference␈α⊃LISP␈α⊃programs,␈α⊂and␈α⊃to␈α⊃handling␈α⊃simple␈α⊃dialogue.␈α⊃Such␈α⊃a␈α⊃set␈α⊂was
␈↓ x␈↓actually chosen, consisting of 29 different parts.
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈αspecific␈αparts␈αdecided␈αupon␈αare␈αrelevant␈αto␈αthe␈αepistemology␈αof␈αprogramming.␈α The␈αuniversal␈αset
␈↓ x␈↓is␈αlisted␈αbelow;␈αrecall␈αthat␈αeach␈αpart␈αof␈αa␈αBEING␈αcorresponds␈αto␈αa␈αquestion␈αwhich␈αis␈αanswered␈αby␈αthat␈αpart's
␈↓ x␈↓value.  The percentage of the experimental pool of BEINGs which actually needed each part is also noted.

␈↓"	␈↓ 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 by the one 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␈↓↓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␈↓α␈↓↓FORM-CHANGING␈↓α 1%  How to perform radical alterations of parts (in special cases).

␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α;␈↓∧␈↓&5. What Happens When a BEING Gains Control␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈αcommitment␈αto␈αa␈αuniversal␈αset␈αof␈αBEING␈αparts␈αis␈αinefficient␈αin␈αsome␈αways␈α(each␈αBEING␈α␈↓βneeded␈↓
␈↓ x␈↓only␈α∞a␈α∞third␈α∞of␈α
all␈α∞the␈α∞parts)␈α∞but␈α∞allows␈α
for␈α∞some␈α∞simplifications.␈α∞What␈α
parts␈α∞should␈α∞be␈α∞evaluated,␈α∞and␈α
in
␈↓ x␈↓what␈αorder,␈αwhen␈αa␈αBEING␈αgains␈αcontrol?␈αThis␈αdecision␈αdepends␈αprimarily␈αon␈αthe␈α␈↓βtypes␈↓␈αof␈αparts␈αpresent␈αin
␈↓ x␈↓the␈α
BEING,␈α
not␈αon␈α
their␈α
␈↓βvalues␈↓.␈α But␈α
every␈α
BEING␈αhas␈α
the␈α
same␈αset␈α
of␈α
parts,␈αso␈α
one␈α
single␈α
algorithm␈αcan
␈↓ x␈↓assemble␈α⊂any␈α⊂BEING␈α⊂into␈α⊂an␈α⊂executable␈α⊂LISP␈α⊂function.␈α⊂Moreover,␈α⊂this␈α⊂assemby␈α⊂can␈α⊂be␈α⊂done␈α⊂when␈α∂the
␈↓ x␈↓system␈αis␈αfirst␈αloaded␈α(or␈αwhen␈αa␈αnew␈αBEING␈αis␈αfirst␈αcreated),␈αand␈αneed␈αonly␈αbe␈αredone␈αfor␈αa␈αBEING␈αwhen
␈↓ x␈↓the values of its parts change. Such changes are rare: experts are not often open-minded.
␈↓"%␈↓ x␈↓␈↓ ↓XWhen␈α
a␈αBEING␈α
␈↓βB␈↓␈α
first␈αgains␈α
control,␈α
its␈α␈↓αEXPLICIT-ARGS␈↓␈α
are␈α
bound.␈αThe␈α
␈↓αIMPLICIT-ARGS␈↓␈αare␈α
initialized,
␈↓ x␈↓the␈α
name␈α
␈↓βB␈↓␈α
is␈α
pushed␈αonto␈α
the␈α
BEING␈α
control␈α
stack,␈α
and␈αany␈α
newly-activated␈α
␈↓αDEMONS␈↓␈α
are␈α
so␈α
tagged.␈αThe
␈↓ x␈↓BEING␈αwho␈αcalled␈α␈↓βB␈↓␈αshould␈αhave␈αexplained␈αhis␈αreasons␈αby␈αassigning␈αthe␈αvariable␈α␈↓αBECAUSE␈↓␈αto␈αsome␈αphrase.
␈↓ x␈↓This␈αreason␈αis␈αnow␈αstored␈αas␈αa␈αspecial␈αsub-part␈αof␈αthe␈αWHY␈αpart␈αof␈α␈↓βB␈↓.␈α ␈↓αBECAUSE␈↓␈αis␈αrebound␈αperiodically␈αin
␈↓ x␈↓the␈α
META-CODE␈α
and␈α
COMMENTS␈α
parts,␈α
to␈α
keep␈αcurrent␈α
the␈α
explanation␈α
of␈α
each␈α
call␈α
that␈α
␈↓βB␈↓␈αmakes.␈α
Each
␈↓ x␈↓␈↓αARG-CHECK␈↓␈α
predicate␈α
is␈α
evaluated.␈α
If␈α
none␈αreturns␈α
NIL,␈α
the␈α
␈↓αPRE-REQUISITES␈↓␈α
are␈α
examined.␈α
Effort␈αis␈α
expended
␈↓ x␈↓to␈α
make␈αthem␈α
true,␈α
if␈αthey␈α
are␈α
currently␈αnot␈α
satisfied.␈αEach␈α
␈↓αCOMMENT␈↓␈α
is␈αevaluated,␈α
then␈α
the␈α␈↓αCO-REQUISITES,
␈↓ x␈↓αMETA-CODE␈↓,␈α∃and␈α∃the␈α∃current␈α∃demons␈α⊗are␈α∃executed␈α∃in␈α∃pseudo-parallel.␈α∃ Each␈α∃␈↓αPOST-REQUISITE␈↓␈α⊗is␈α∃then
␈↓ x␈↓examined,␈αand␈αan␈αeffort␈αmade␈αto␈αsatisfy␈αit.␈α The␈αnewly-activated␈αdemons␈αare␈αexorcized,␈αthe␈αBEING␈αstack␈αis
␈↓ x␈↓popped, and the value computed by the ␈↓αMETA-CODE␈↓ is returned.
␈↓"%␈↓ x␈↓␈↓ ↓XSome␈αheuristics␈αwere␈αdevised␈αto␈αtake␈αadvantage␈αof␈αthe␈αfact␈αthat␈αthe␈αBEINGs␈αoften␈αdidn't␈αneed␈αmany
␈↓ x␈↓of␈αthe␈αstandard␈αparts.␈αFor␈αexample,␈α␈↓αOBTAIN-USABLE-INFORMATION␈↓␈αhas␈αno␈αnew␈αdemons␈αor␈αco-requisites,␈αso␈αno
␈↓ x␈↓parallel␈α
processing␈α
need␈αbe␈α
simulated.␈α
 No␈αresults␈α
are␈α
explicitly␈αproduced,␈α
so␈α
the␈αvalue␈α
returned␈α
by␈αthe␈α
␈↓αMETA-
␈↓ x␈↓αCODE␈↓␈α
is␈α
not␈α
saved␈α
and␈α
later␈α
returned.␈α
 Here␈α
is␈α
the␈α
way␈α
that␈α
BEING␈α
actually␈α
appeared␈α
after␈αassembly;␈α
contrast
␈↓ x␈↓this with the values of its parts, as presented in Section 3.

␈↓"	␈↓ x␈↓α␈↓ ↓X(OBTAIN-USABLE-INFORMATION
␈↓"	␈↓ x␈↓α␈↓ ↓X  (LAMBDA  (U)
␈↓"	␈↓ x␈↓α␈↓ ↓X      (AND
␈↓"	␈↓ x␈↓α␈↓ ↓X        (PUSH  OBTAIN-USABLE-INFORMATION  BEING-STACK)
␈↓"	␈↓ x␈↓α␈↓ ↓X        (PUT-APPEND  OBTAIN-USABLE-INFORMATION  WHY  BECAUSE)
␈↓"	␈↓ x␈↓α␈↓ ↓X        (EVERY  CURRENT-DEMONS  APPLY*)
␈↓"	␈↓ x␈↓α␈↓ ↓X        (SETQQ BECAUSE (WE CAN ONLY TRY TO OBTAIN USABLE INFO IN ONE WAY AT A TIME))
␈↓"	␈↓ x␈↓α␈↓ ↓X        (CHOOSE-FROM
␈↓"	␈↓ x␈↓α␈↓ ↓X                 (GET-NEW-INFORMATION U)
␈↓"	␈↓ x␈↓α␈↓ ↓X                 (TRANSLATE U)
␈↓"	␈↓ x␈↓α␈↓ ↓X                 (ANALYZE-IMPLICATIONS U)
␈↓"	␈↓ x␈↓α␈↓ ↓X                 (EXTRACT-RELEVANT-SUBSET U))
␈↓"	␈↓ x␈↓α␈↓ ↓X        (POP  BEING-STACK))))
␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α&␈↓∧␈↓&6. Character of BEING Interactions␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XTo␈αdemonstate␈αthe␈α
flavor␈αof␈αBEINGs␈αactively␈α
doing␈αsomething,␈αan␈αexcerpt␈α
will␈αbe␈αpresented␈α
from␈αa
␈↓ x␈↓dialogue␈α⊃between␈α⊃a␈α⊃human␈α⊃user␈α⊃and␈α⊃PUP6,␈α⊃the␈α⊃experimental␈α⊃pool␈α⊃of␈α⊃BEINGs.␈α⊃The␈α∩␈↓βexternal␈↓␈α⊃character
␈↓ x␈↓involves␈αconsidering␈αall␈αof␈αPUP6␈αas␈αone␈αentity,␈αnot␈αdistinguishing␈αwhich␈αBEING␈αasked␈αthe␈α
user␈αsomething,
␈↓ x␈↓not␈αobserving␈αthe␈αdiscussion␈αtaking␈αplace␈αstrictly␈αamong␈αBEINGs.␈αAfter␈αthe␈αexample␈αis␈αa␈αfuller␈αdiscussion␈αof
␈↓ x␈↓what␈αwent␈αon␈α
␈↓βinternally␈↓.␈α The␈αintent␈α
of␈αthe␈αuser␈α
in␈αthis␈αdialogue␈α
is␈αto␈αencode␈α
the␈αconcept␈αformation␈α
program.
␈↓ x␈↓Here is the initial and final few lines of that 300-page dialogue:

␈↓"	␈↓ 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␈↓␈↓ ↓X .
␈↓"	␈↓ x␈↓␈↓ ↓X .
␈↓"	␈↓ x␈↓␈↓ ↓XPUP: I can no longer defer this NESTED decision (respond A, B, C, ?, DETAILS)
␈↓"	␈↓ x␈↓␈↓ ↓X   A)  CF does classificatory concept formation
␈↓"	␈↓ x␈↓␈↓ ↓X   B)  CF does comparative and classificatory concept formation
␈↓"	␈↓ x␈↓␈↓ ↓X   C)  CF does metrical, comparative, and classificatory concept formation
␈↓"	␈↓ x␈↓␈↓ ↓XUSER: A
␈↓"	␈↓ x␈↓␈↓ ↓XPUP: I have written the program CF for ␈↓αspecialized CONCEPT-FORMATION␈↓.
␈↓"	␈↓ x␈↓␈↓ ↓XPUP: I want any task.

␈↓"%␈↓ x␈↓␈↓ ↓XThe␈α
BEING␈αinitially␈α
in␈α
control␈αis␈α
␈↓αSERVE-THE-USER␈↓.␈α
Its␈αessential␈α
function␈α
is␈αto␈α
obtain␈αnew␈α
information
␈↓ x␈↓until some of it is "obeyable", then obey it.
␈↓"%␈↓ x␈↓␈↓ ↓XAfter␈α
the␈α
user's␈α
first␈α
order,␈α
the␈α
␈↓αWRITE-PROGRAM␈↓␈α∞BEING␈α
gains␈α
control.␈α
He␈α
asks␈α
for␈α
someone␈α∞to␈α
type
␈↓ x␈↓out␈α⊂his␈α∂intentions,␈α⊂and␈α⊂␈↓αMESSAGE␈↓␈α∂recognizes␈α⊂its␈α∂relevance␈α⊂and␈α⊂performs␈α∂that␈α⊂service.␈α⊂Similarly,␈α∂␈↓αGET-NAME␈↓
␈↓ x␈↓temporarily␈αgains␈αcontrol␈αto␈αobtain␈αa␈αproper␈αname.␈α␈↓αWRITE-PROGRAM␈↓␈αthen␈αenters␈αits␈αREPEATEDLY␈αloop␈α(see
␈↓ x␈↓Section␈α∂2).␈α∞The␈α∂user's␈α∞command␈α∂is␈α∞usable␈α∂information,␈α∂and␈α∞the␈α∂WHEN␈α∞part␈α∂of␈α∞␈↓αUSE-INFO␈↓␈α∂has␈α∂the␈α∞highest
␈↓ x␈↓value␈α∞of␈α∞the␈α∞four␈α∞choices.␈α
 When␈α∞␈↓αUSE-INFO␈↓␈α∞is␈α∞given␈α∞control,␈α
it␈α∞examines␈α∞the␈α∞SPECIALIZATIONS␈α∞part␈α
of
␈↓ x␈↓␈↓αCONCEPT-FORMATION␈↓.␈αEleven␈αdecisions␈αare␈αlisted,␈αwhose␈α
answers␈αdelimit␈αthe␈αtypes␈αof␈αstreamlining␈α
which␈αare
␈↓ x␈↓possible[2].␈α Each␈α
decision␈αis␈α
examined␈αin␈α
turn.␈αIf␈αit␈α
cannot␈αbe␈α
resolved,␈αPUP6␈α
tries␈αto␈α
defer␈αit.␈α For␈α
example,
␈↓ x␈↓the␈α
type␈α∞of␈α
concept␈α∞formation␈α
may␈α
involve␈α∞simple␈α
partitioning,␈α∞partitioning␈α
and␈α
ordering␈α∞of␈α
the␈α∞classes,␈α
or
␈↓ x␈↓partitioning␈αand␈α
ordering␈αand␈α
finding␈αa␈α
metric␈αon␈α
the␈αspace␈α
of␈αclasses␈α
of␈αthe␈α
partition.␈αSince␈α
all␈αthree␈α
choices
␈↓ x␈↓involve␈α
partitioning,␈α
the␈α
decision␈α
may␈α
be␈α
deferred␈α
until␈α
after␈α
code␈α
has␈α
been␈α
written␈α
for␈α
that␈α∞subtask.␈α
 The
␈↓ x␈↓rationale␈α⊂for␈α⊂deferral␈α⊂is␈α⊂that␈α⊃at␈α⊂a␈α⊂later␈α⊂time,␈α⊂PUP6␈α⊂might␈α⊃have␈α⊂enough␈α⊂new␈α⊂information␈α⊂to␈α⊃resolve␈α⊂the
␈↓ x␈↓decision without bothering the user.
␈↓ x␈↓␈↓ ↓XOnly␈α
one␈α
decision␈α
cannot␈α∞be␈α
deferred:␈α
must␈α
the␈α∞program␈α
simulate␈α
various␈α
experimental␈α∞subjects,␈α
or
␈↓ x␈↓just one archtypical concept former? PUP6 has no way of resolving this without asking the user, so it does.
␈↓"%␈↓ x␈↓␈↓ ↓XAfter␈α∞a␈α∞CPU␈α∂hour␈α∞has␈α∞elapsed,␈α∞code␈α∂for␈α∞the␈α∞type␈α∞of␈α∂partitioning␈α∞desired␈α∞has␈α∞been␈α∂synthesized.␈α∞ A
␈↓ x␈↓demon␈α⊃then␈α⊃finds␈α⊃he␈α⊃must␈α⊃ask␈α⊃the␈α⊃user␈α⊃whether␈α⊂anything␈α⊃else␈α⊃need␈α⊃be␈α⊃done␈α⊃by␈α⊃CF.␈α⊃When␈α⊃he␈α⊂replies
␈↓ x␈↓negatively,␈α∃␈↓αWRITE-PROGRAM␈↓␈α∃finds␈α∃itself␈α∃completely␈α∃finished.␈α∃Its␈α∃POST-REQUISITES␈α∃are␈α∃satisfied␈α∃by
␈↓ x␈↓␈↓αMESSAGE␈↓, and it relinquishes control to ␈↓αSERVE-THE-USER␈↓.
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α≤␈↓∧␈↓&7. Theory of Pure BEINGs Systems␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XMany␈α⊂constraints␈α⊃are␈α⊂placed␈α⊃on␈α⊂BEINGs␈α⊃as␈α⊂individuals␈α⊂and␈α⊃in␈α⊂groups.␈α⊃Here␈α⊂these␈α⊃are␈α⊂collected.
␈↓ x␈↓Hopefully,␈α∂the␈α∂ideas␈α∂will␈α∂be␈α∂separated␈α∂from␈α∂the␈α∂implementation␈α∂details,␈α∂the␈α∂prejudices␈α∂from␈α∂the␈α∞plausible
␈↓ x␈↓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,␈α∩one␈α∩can␈α∩assert␈α⊃that␈α∩certain␈α∩primitive␈α∩constructs␈α⊃are␈α∩allowed;␈α∩for␈α∩example,␈α⊃any␈α∩part␈α∩which␈α∩can␈α⊃be
␈↓ x␈↓expressed as a few lines of LISP code involving no auxilliary function calls.
␈↓"%␈↓ x␈↓␈↓ ↓XSuppose␈α⊃there␈α⊂␈↓βwere␈↓␈α⊃only␈α⊂BEINGs␈α⊃--␈α⊃hence␈α⊂no␈α⊃functions␈α⊂--␈α⊃allowed␈α⊃to␈α⊂exist.␈α⊃ In␈α⊂the␈α⊃case␈α⊃of␈α⊂an
␈↓ x␈↓automatic␈αprogramming␈αtask,␈αthe␈αBEINGs␈αwould␈αhave␈αto␈αwrite␈αnew␈αBEINGs,␈αnot␈αnew␈αLISP␈αfunctions.␈αThe
␈↓ x␈↓target␈α∪program␈α∪would␈α∩thus␈α∪itself␈α∪be␈α∩a␈α∪community␈α∪of␈α∩BEINGs.␈α∪What␈α∪effects␈α∩would␈α∪this␈α∪have?␈α∩ Many
␈↓ x␈↓superfluous␈α
parts␈α
would␈α
get␈α
filled␈α
in.␈α
What␈α
would␈α
be␈α
their␈α
significance?␈α
 They␈α
are␈α
organized␈α
␈↓βdocumentation␈↓,␈α
a
␈↓ x␈↓standard 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⊂proposed␈α⊂earlier␈α⊂for␈α⊂each␈α⊂BEING.␈α⊂The␈α∂BEING
␈↓ x␈↓which␈αactually␈αdoes␈αthe␈αcreation␈α(␈↓αMAKE-NEW-BEING␈↓)␈αin␈αthe␈αexperimental␈αsystem␈αis␈αalmost␈αtrivial,␈αgetting␈αvery
␈↓ x␈↓precise 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 responders, having ␈↓βthem␈↓ decide who should take control, and letting the winner do the translation.
␈↓"%␈↓ x␈↓␈↓ ↓XA␈αuniversal␈αset␈αof␈αBEING␈αparts␈αshould␈α
preserve␈αsome␈αof␈αthe␈αadvantages␈αof␈αuniformity␈α(easy␈α
addition
␈↓ x␈↓of␈αknowledge␈α
to␈αthe␈αsystem,␈α
easy␈αinter-BEING␈αcommunication).␈α
The␈αcomplex␈αstructure␈α
of␈αa␈α
BEING␈αshould
␈↓ x␈↓make␈α∞complex␈α∞behaviors␈α∂feasable,␈α∞including␈α∞dialogue␈α∂flexibility␈α∞as␈α∞well␈α∞as␈α∂final␈α∞products.␈α∞ The␈α∂␈↓βnumber␈↓␈α∞of
␈↓ x␈↓BEING␈αparts␈αis␈αa␈αtype␈αof␈αindicator␈αof␈αthe␈αbalance␈αbetween␈αuniformity␈αand␈αstructure␈αin␈αthe␈αcommunity.␈αIf␈αit
␈↓ x␈↓were␈α~1,␈αthere␈α
would␈αbe␈αno␈αstructure␈α
(e.g.,␈αpredicate␈αcalculus);␈αit␈α
it␈αwere␈αthe␈αunion␈α
of␈αall␈αthe␈αBEINGs'␈α
desires,
␈↓ x␈↓~1000, there would be no uniformity (e.g., ACTORS).
␈↓ x␈↓␈↓ ↓XAutomatic␈α
programming␈α
is␈α
distinguished␈α
as␈α
a␈α
task␈α
for␈α
BEINGs␈α
in␈α
the␈α
following␈α
sense.␈α
 Say␈α
a␈αpool␈α
has
␈↓ x␈↓been␈αassigned␈αtask␈α
T.␈α They␈αcreate␈α
new␈αprograms␈αwhich␈α
simulate␈αT.␈α If␈α
the␈αoriginal␈αtask␈α
had␈αbeen␈αto␈αwrite␈α
a
␈↓ x␈↓program␈αwhich␈αdid␈αT,␈α␈↓βthere␈α
would␈αbe␈αno␈αextra␈αeffort␈αnecessary␈↓.␈α
The␈αnew␈αBEINGs␈αwould␈αsimply␈α
be␈αdumped
␈↓ x␈↓onto a file.
␈↓"%␈↓ 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␈α∂overlooked.
␈↓ x␈↓Ideally,␈α∂an␈α∂automatic␈α∂programming␈α∂system,␈α∂including␈α∂BEINGs,␈α∂should␈α∂not␈α∂have␈α∂to␈α∂ignore␈α∂details.␈α∂ At␈α∞the
␈↓ x␈↓least,␈αthey␈α
might␈αdefer␈αthem,␈α
asserting␈αa␈α
warning␈αto␈αthis␈α
effect.␈αIt␈α
is␈αfelt␈αthat␈α
procrastination␈αis␈αquite␈α
valuable;
␈↓ x␈↓in␈α⊂PUP6,␈α⊂much␈α∂effort␈α⊂is␈α⊂spent␈α∂deferring␈α⊂any␈α⊂unresolvable␈α∂decision.␈α⊂ Undeferrable␈α⊂unresolvable␈α∂decisions
␈↓ x␈↓must␈αcause␈α
a␈αbacktrack␈α
point␈αto␈α
be␈αreluctantly␈αset␈α
up.␈αAnother␈α
prejudice␈αis␈α
that␈αmost␈α
carelessness␈αbugs␈αcan␈α
be
␈↓ x␈↓eliminated␈αby␈αthis␈αdeferral,␈αfeed-forward,␈αand␈αprecise␈αrecord-keeping.␈αHumans␈αdepend␈αon␈αtheir␈αadaptability
␈↓ x␈↓to␈α
compensate␈α
for␈α
limitations␈α
in␈α
their␈α
brain␈α
hardware,␈α
but␈α
there␈α
is␈α
no␈α
need␈α
for␈α
an␈α∞␈↓βautomatic␈↓␈α
programming
␈↓ x␈↓system to do so.
␈↓"%␈↓ 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␈↓same␈α
difference␈αexists␈α
between␈αHACKER␈α
and␈αPUP6,␈α
the␈α
experimental␈αpool␈α
of␈αBEINGS.␈α
 Since␈αPUP6␈α
writes
␈↓ x␈↓structured␈α∃programs,␈α∃it␈α⊗appears␈α∃similar␈α∃to␈α∃macro␈α⊗expansion.␈α∃Macro␈α∃procedures␈α⊗expand␈α∃mechanically:
␈↓ x␈↓␈↓↓expand(sequence␈α∂m␈↓#v1␈↓#␈α∂m␈↓#v2␈↓#)␈α∂=␈α∂(sequence␈α∂expand(m␈↓#v1␈↓#)␈α∂expand(m␈↓#v2␈↓#)))␈↓.␈α∂BEINGs␈α∂could␈α∂use␈α⊂information␈α∂gleaned
␈↓ x␈↓during expansion of m␈↓#v1␈↓# to improve the way m␈↓#v2␈↓# was handled.
␈↓"%␈↓ x␈↓␈↓ ↓XBEINGs␈αlinearly␈αsubsume␈αmany␈αpopular␈αAI␈αfeatures;␈αthe␈αdemonstration␈αwill␈αbe␈αbrief:␈αA␈α␈↓βdemon␈↓␈αcould
␈↓ x␈↓be␈α
replaced␈α
by␈α
a␈α
BEING␈α
whose␈α
ARG-CHECK␈α
predicate␈α
was␈α
the␈α
triggering␈α
predicate,␈α
whose␈α
WHEN␈α
part
␈↓ x␈↓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␈αdemon.
␈↓ x␈↓An␈α␈↓βassertion␈↓␈α
could␈αbe␈α
a␈αBEING␈αwith␈α
only␈αan␈α
IDEN␈αpart␈α
filled␈αin;␈αwhen␈α
it␈αrecognized␈α
its␈αrelevance,␈α
a␈αfully
␈↓ x␈↓instantiated␈αassertion␈αis␈αreturned.␈αA␈α␈↓βfunction␈↓␈αis␈α
equivalent␈αto␈αa␈αBEING␈αwith␈αonly␈αa␈α
META-CODE,␈αARGS,
␈↓ x␈↓and␈αNLAMDA␈αparts;␈αone␈αknows␈αalmost␈αnothing␈αabout␈αit␈αbefore␈αexecuting␈αit.␈α While␈αtheoretically␈αclear,␈αthese
␈↓ x␈↓subsumptions␈α
are␈α
very␈α
inefficient.␈α∞ Replacing,␈α
e.g.,␈α
an␈α
assertion␈α∞by␈α
even␈α
a␈α
single␈α∞(nondeterministic)␈α
BEING
␈↓ x␈↓call introduces an extra time factor proportional to the total number of BEINGs in the system.
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α∧␈↓∧␈↓&8. Experimental System␈↓)αβ␈↓
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈αfirst␈αdecision␈αto␈αbe␈αsettled␈αin␈αconstructing␈αan␈αactual␈αpool␈αof␈αBEINGs␈αis␈αwhat␈αtheir␈αtask␈αshall␈αbe.
␈↓ x␈↓The␈α∞choice␈α∞was␈α∞automatic␈α∂code␈α∞generation␈α∞of␈α∞simple␈α∂inductive␈α∞inference␈α∞LISP␈α∞programs,␈α∂from␈α∞discussions
␈↓ x␈↓with an ␈↓βactive␈↓ user.  The universal set of BEING parts listed in Section 4 was constructed by introspection.
␈↓"%␈↓ x␈↓␈↓ ↓XWhat␈α
program␈α∞would␈α
be␈α∞the␈α
specific␈α∞"target"?␈α
 A␈α∞variant␈α
of␈α∞Winston's␈α
structural␈α∞concept␈α
formation
␈↓ x␈↓system␈α
was␈αstudied␈α
and␈αsimplified.␈α
A␈αcomplete␈α
dialogue␈αwas␈α
simulated␈αby␈α
humans,␈αsupposedly␈α
between␈αa␈α
user
␈↓ x␈↓and␈αa␈αdesirable␈αautomatic␈αprogramming␈αsystem.␈αCareful␈αrecords␈αwere␈αkept,␈αand␈αthe␈α"system"␈αwrote␈αdown␈αhis
␈↓ x␈↓reasoning␈αsteps␈αbetween␈αsuccessive␈αoutputs.␈αA␈αclean␈α
structured␈αprogram␈αresulted,␈αand␈αthe␈αgoal␈αof␈αthe␈α
research
␈↓ x␈↓became␈α∩to␈α∩elicit␈α∩that␈α⊃program,␈α∩using␈α∩virtually␈α∩the␈α⊃same␈α∩dialogue,␈α∩from␈α∩a␈α⊃pool␈α∩of␈α∩BEINGs.␈α∩ The␈α⊃most
␈↓ x␈↓important␈α
constraint␈α
was␈α
that␈αthe␈α
BEINGs␈α
actually␈α
go␈α
through␈αstates␈α
corresponding␈α
to␈α
the␈αhuman's␈α
reasoning
␈↓ x␈↓steps.␈α
 Two␈α
restrictions␈α
implicit␈α
here␈α
turned␈α
out␈α
to␈α
be␈α
serious:␈α
the␈α
abilities␈α
of␈α
the␈α
actual␈α
user␈α∞must␈α
coincide
␈↓ x␈↓with␈α⊂those␈α⊂of␈α⊂the␈α⊂human␈α⊂who␈α⊂participated␈α⊂in␈α⊂the␈α∂protocol,␈α⊂and␈α⊂the␈α⊂order␈α⊂of␈α⊂and␈α⊂wording␈α⊂of␈α⊂the␈α∂user's
␈↓ x␈↓comments␈α
must␈α
closely␈αadhere␈α
to␈α
that␈α
single␈αdialogue.␈α
These␈α
difficulties␈α
are␈αcentral␈α
to␈α
any␈α
system␈αdesigned␈α
for
␈↓ x␈↓wide usership, but were considered ignorable in the specific automatic code generation task at hand.
␈↓"%␈↓ x␈↓␈↓ ↓XAt␈α
this␈α
stage,␈α
a␈α
rough␈α
initial␈α
set␈α
of␈αabout␈α
eighty␈α
BEINGs␈α
surfaced.␈α
Each␈α
one␈α
had␈α
not␈α
much␈αmore␈α
than
␈↓ x␈↓a␈α⊂name␈α⊃and␈α⊂a␈α⊃vague␈α⊂description␈α⊃of␈α⊂what␈α⊃it␈α⊂must␈α⊂do.␈α⊃The␈α⊂dialogue␈α⊃was␈α⊂cycled␈α⊃through␈α⊂again,␈α⊃and␈α⊂the
␈↓ x␈↓comments␈αwere␈αreplaced␈αby␈αa␈αsketch␈αof␈αwhich␈αBEINGs␈αwould␈αcall␈αwhich␈αother␈αones,␈αwhy,␈αand␈αthe␈αresults␈αof
␈↓ x␈↓the␈αcall.␈α The␈αnecessary␈αset␈αof␈αBEING␈αparts␈αand␈αthe␈αconstraints␈αon␈αeach␈αBEING␈αthus␈αgrew.␈αOccasionally␈αan
␈↓ x␈↓unexpected␈α⊃BEING␈α⊃or␈α⊂BEING␈α⊃part␈α⊃had␈α⊃to␈α⊂be␈α⊃added␈α⊃to␈α⊃the␈α⊂design.␈α⊃Eventually,␈α⊃17␈α⊃inductive␈α⊂inference
␈↓ x␈↓domain-specific␈αBEINGs␈αwere␈αcalled␈α
for,␈αand␈αanother␈α70␈α
which␈αdealt␈αwith␈αprogramming,␈α
organization,␈αand
␈↓ x␈↓communicating␈α∞with␈α
the␈α∞user.␈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
␈↓ x␈↓those␈α
parts␈αspecified␈α
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
␈↓ x␈↓dialogues stemmed from this minimal completion.
␈↓"%␈↓ 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simulation).
␈↓ 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␈α⊂typed␈α⊂by␈α⊂PUP6,␈α⊃and␈α⊂4K␈α⊂by␈α⊂the␈α⊃user.␈α⊂The␈α⊂mean␈α⊂wait␈α⊃time␈α⊂between␈α⊂user␈α⊂input␈α⊃and␈α⊂machine
␈↓ x␈↓response was several cpu seconds; the longest delay was one cpu minute.
␈↓"%␈↓ x␈↓␈↓ ↓XTwo␈α⊂of␈α⊂the␈α⊂most␈α⊂significant␈α⊂questions␈α⊃for␈α⊂automatic␈α⊂proramming␈α⊂systems[1]␈α⊂are␈α⊂how␈α⊃flexible␈α⊂the
␈↓ x␈↓dialogue␈α∂can␈α∂be,␈α∂and␈α∂how␈α⊂different␈α∂the␈α∂generated␈α∂programs␈α∂can␈α∂be.␈α⊂Because␈α∂of␈α∂its␈α∂genesis␈α∂from␈α⊂a␈α∂single
␈↓ x␈↓protocol,␈α
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
 It␈α
␈↓βwas␈↓␈αfelt
␈↓ x␈↓that␈α∞most␈α∞of␈α∞them␈α∞could␈α
be␈α∞useful␈α∞in␈α∞generating␈α∞other␈α
programs,␈α∞however.␈α∞For␈α∞this␈α∞reason,␈α∞two␈α
additional
␈↓ x␈↓target programs were specified.
␈↓"%␈↓ x␈↓␈↓ ↓XAll␈α∞three␈α∞synthesized␈α∂programs␈α∞are␈α∞now␈α∞discussed.␈α∂ CF␈α∞lacked␈α∞an␈α∞efficient␈α∂heuristic␈α∞graph-matcher
␈↓ x␈↓but␈α∞is␈α∞otherwise␈α∞similar␈α∞to␈α∞[6].␈α∞Repeatedly,␈α∞a␈α∂scene␈α∞(a␈α∞set␈α∞of␈α∞objects␈α∞and␈α∞static␈α∞structural␈α∂relations␈α∞between
␈↓ x␈↓them)␈αis␈αinputted.␈αIf␈αno␈αname␈αis␈αgiven,␈αCF␈αmust␈αguess␈αone␈αand␈αthen␈αcheck␈αwith␈αthe␈αuser.␈αThe␈αguess␈αis␈αbased
␈↓ x␈↓on␈α
comparing␈α
the␈α
scene␈α
with␈α
internal␈α
models␈α
of␈α
previously-named␈α
scenes,␈α
until␈α
one␈α
matches.␈α
After␈α
each␈α
input,
␈↓ x␈↓the␈αmodels␈αwith␈αthe␈αguessed␈αand␈αthe␈αcorrect␈αnames␈αare␈αmodified.␈αAn␈αincorrect␈αguess␈αmust␈αnot␈αbe␈αmade␈αtwice
␈↓ x␈↓in␈αa␈αrow␈αon␈αthe␈αsame␈αscene.␈αFor␈αthis,␈αP.␈αGadwa␈αdemonstrated␈αthat␈αit␈αsuffices␈αto␈αmaintain␈αthree␈αlists␈αfor␈αeach
␈↓ x␈↓named␈αmodel:␈αthose␈αrelations␈αwhich␈αmust␈αbe␈αpresent␈αin␈αthe␈αobserved␈αscene␈αfor␈αit␈αto␈αpossibly␈αhave␈αthis␈αname,
␈↓ x␈↓those␈αrelations␈α
which␈αmust␈α
never␈αoccur␈α
in␈αany␈α
scene␈αhaving␈αthis␈α
name,␈αand␈α
those␈αwhich␈α
may␈αbe␈α
present␈αor
␈↓ x␈↓absent.
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈α
target␈α
program␈α
GI␈α
is␈α
a␈α
grammatical␈α
inference␈α
program,␈α
which␈α
accepts␈α
strings␈α
labelled␈αLEGAL,
␈↓ x␈↓ILLEGAL,␈α∪or␈α∪??.␈α∪In␈α∪the␈α∪latter␈α∪case,␈α∪GI␈α∪must␈α∪guess␈α∪the␈α∪legality.␈α∪Internally,␈α∪potential␈α∪sets␈α∪of␈α∪rules␈α∩are
␈↓ 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␈α
new␈αBEINGs␈α
had␈α
to␈αbe␈α
added,␈α
related␈αto␈α
formal␈α
grammars␈αand␈α
rules.␈α
 While␈αthis␈α
number␈α
is␈αsmall,␈α
the
␈↓ x␈↓addition␈α∞of␈α∞␈↓βany␈↓␈α∂new␈α∞BEINGs␈α∞demands␈α∂that␈α∞the␈α∞user␈α∞be␈α∂very␈α∞familiar␈α∞with␈α∂all␈α∞the␈α∞format␈α∂conventions␈α∞of
␈↓ x␈↓PUP6,␈α
all␈α
the␈α
kluges,␈α
etc.␈αThe␈α
final␈α
GI␈α
program␈α
synthesized␈α
was␈α20␈α
pages␈α
long;␈α
a␈α
hand-coded␈α
version␈αwas
␈↓ x␈↓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␈α∀BEING␈α∀and␈α∀one␈α∀domain-specific␈α∪BEING␈α∀had␈α∀to␈α∀be␈α∀␈↓βadded␈↓␈α∪to␈α∀PUP6␈α∀--␈α∀again,␈α∀this␈α∀is␈α∪both
␈↓ x␈↓encouragingly small yet far beyond the abilities of the intended users.
␈↓"%␈↓ x␈↓␈↓ ↓XBefore␈αdelivering␈αour␈αverdict␈αon␈αthis␈αsystem,␈αone␈αfurther␈αaspect␈αis␈αpresented.␈α As␈αproposed␈αin␈αSection
␈↓ x␈↓7,␈α
the␈α
BEINGs␈α
generate␈α
other␈α
BEINGs,␈α
not␈α
just␈αfunctions.␈α
This␈α
explains␈α
the␈α
huge␈α
increases␈α
in␈α
target␈αcode
␈↓ x␈↓lengths.␈αCF␈αwas␈αa␈αpool␈αof␈α56␈αbrand␈αnew␈αBEINGs,␈αGI␈α37,␈αand␈αPL␈α24.␈αAs␈αwith␈αPUP6,␈αone␈αcan␈αinterrupt␈αthe
␈↓ x␈↓target␈αprograms␈αas␈α
they␈αare␈αrunning␈α
and␈αask␈αquestions.␈αAny␈α
BEING␈αon␈αthe␈α
control␈αstack␈αwill␈α
provide␈αfully
␈↓ x␈↓instantiated␈α⊂answers␈α∂to␈α⊂any␈α⊂of␈α∂its␈α⊂29␈α∂allowable␈α⊂queries;␈α⊂all␈α∂other␈α⊂BEINGs␈α∂will␈α⊂provide␈α⊂only␈α∂hypothetical
␈↓ x␈↓answers.␈αAs␈α
an␈αexample,␈α
consider␈αthis␈αexcerpt␈α
between␈αCF␈α
and␈αthe␈αuser.␈α
 Some␈αliberty␈α
has␈αbeen␈α
taken␈αwith
␈↓ x␈↓the␈αEnglish;␈αe.g.,␈αthe␈αuser␈αreally␈αtypes␈α␈↓βWHAT?␈↓,␈αnot␈α␈↓βWhat␈αare␈αyou␈αdoing?␈↓␈αPUP6␈αtypes␈αin␈αboldface,␈αthe␈αuser␈αin
␈↓ x␈↓italics.

␈↓"	␈↓ 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 PUP6 "␈↓βOK, go back to work now.␈↓"

␈↓"	␈↓ x␈↓↓        ␈↓βU:  What are you doing?␈↓↓
␈↓"	␈↓ x␈↓↓PUP:  Divide a domain into subdomains, using specialized version of PARTITION-DOMAIN
␈↓"	␈↓ x␈↓↓        ␈↓βHow?␈↓↓
␈↓"	␈↓ x␈↓↓PUP:  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␈↓↓PUP:  Break NAME piece off (??? (A B) (BLOCK A) (GREEN B) (TOUCHES A B)) to examine.
␈↓"	␈↓ x␈↓↓        ␈↓βWhat code is the current BEING going to execute?␈↓↓
␈↓"	␈↓ x␈↓↓PUP:  The meta-code of SOME-PART-OF␈↓#v1␈↓#␈↓#v4␈↓# is (CAR arg1). Evals to ???. ␈↓βOK.␈↓↓

␈↓"	␈↓ x␈↓↓NOT AN ARCH.
␈↓ x␈↓↓        ␈↓βU:  Why?␈↓↓
␈↓"	␈↓ x␈↓↓PUP: 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␈↓↓PUP: PUP cannot resolve it, and must at this time, so PUP asks USER
␈↓"	␈↓ x␈↓↓        ␈↓βOK. What you wanted is:   Nothing␈↓↓

␈↓"	␈↓ x␈↓↓        ␈↓βU:  What are you doing?␈↓↓
␈↓"	␈↓ x␈↓↓PUP: Go through space of concept names, and make user aware of changes.
␈↓"	␈↓ x␈↓↓        ␈↓βWhat BEINGs might this affect?␈↓↓
␈↓"	␈↓ x␈↓↓PUP: 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␈↓↓READY TO ACCEPT BRAND NEW SCENE:


␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α_␈↓∧␈↓&9. Conclusions␈↓)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∩a␈α∩new␈α∩program,␈α∩it␈α∩is␈α∩necessary␈α∩to␈α∩add␈α∩a␈α∩few␈α∩general␈α⊃BEINGs,
␈↓ x␈↓generalize␈αa␈αfew␈αexisting␈αBEINGs'␈αparts,␈αand␈αadd␈αseveral␈αBEINGs␈αspecific␈αto␈αthe␈αnew␈αprogram's␈αdomain.␈αIf
␈↓ x␈↓even␈αone␈αnew␈αBEING␈αmust␈αbe␈αadded,␈αthe␈αtask␈αbecomes␈αvery␈αcomplex.␈αAlso,␈αthe␈αdialogue␈αto␈αproduce␈αthe␈αnew
␈↓ x␈↓program␈α∞will␈α
be␈α∞brittle␈α∞and␈α
probably␈α∞not␈α∞well␈α
suited␈α∞to␈α∞that␈α
domain,␈α∞since␈α∞most␈α
of␈α∞the␈α∞recognized␈α
phrases
␈↓ x␈↓stem from a single (CF-producing) dialgoue.
␈↓"%␈↓ 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⊂Often,␈α∂a␈α⊂single␈α⊂message␈α∂would␈α⊂issue␈α⊂simultaneously␈α∂from␈α⊂several␈α⊂BEINGs␈α∂(in
␈↓ x␈↓slightly␈αaltered␈α
formats).␈αThe␈αconverse␈α
problem␈αexisted␈α
as␈αwell:␈αit␈α
was␈αnecessary␈αto␈α
include␈αa␈α
BEING␈αwhich
␈↓ x␈↓simulated␈αforgetfulness,␈αto␈αprevent,␈αe.g.,␈αdelayed␈αanaphora.␈α Orienting␈αthe␈αuser␈αwas␈αnot␈αsolved␈αsatisfactorally.
␈↓ x␈↓Pointers␈αinto␈αa␈αgraph␈αof␈αgenerated␈αcode␈αwere␈αsimulated,␈αbut␈αoften␈αone␈αwished␈αto␈αrefer␈αto␈αa␈αpiece␈αof␈αcode␈αnot
␈↓ x␈↓by name or by pointing, but by some brief meaningful (to him only!) phrase.
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈αproblem␈αof␈αlocating␈αrelevant␈αinformation␈αis␈αdistributed␈αby␈αgiving␈αeach␈αBEING␈αthe␈α
responsibility
␈↓ x␈↓for␈α
recognizing␈α
when␈α
it␈α
is␈α
talked␈α
about.␈α
This␈α
beats␈α
the␈α
combinatorial␈α
nature␈α
of␈α
the␈α
problem␈α
on␈α
--␈α
but␈α␈↓βonly␈↓␈α
on
␈↓ x␈↓-- a computer with as many processors as there are BEINGs.
␈↓"%␈↓ 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sophistocated␈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␈αbugs␈α␈↓βare␈↓␈αavoidable␈αby␈αcareful␈αrecord-keeping,␈αit␈αwas␈αunrealistic␈αto␈αmake␈αno␈αprovision␈αfor␈αdebugging␈αa
␈↓ x␈↓new␈α
thirty-page␈αprogram.␈α
When␈α
a␈αfew␈α
errors␈α
did␈αoccur,␈α
PUP6␈α
itself␈αhad␈α
to␈α
be␈αaltered.␈α
 The␈α
need␈αfor␈α
flexible
␈↓ x␈↓dialogues was underestimated in this domain.
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈αoverall␈αperformance␈α
of␈αBEINGs␈αin␈α
PUP6␈αis␈αmixed.␈α
 Two␈αadvantages␈αwere␈α
hoped␈αfor␈αby␈αusing␈α
a
␈↓ x␈↓uniform␈α
set␈αof␈α
BEING␈αparts.␈α
 Addition␈αof␈α
new␈α
BEINGs␈αto␈α
the␈αpool␈α
was␈αnot␈α
easy␈α(for␈α
untrained␈α
users)␈αbut
␈↓ x␈↓communication␈α∂among␈α∂BEINGs␈α∂␈↓βwas␈↓␈α∂easy␈α∂(fast,␈α∂natural).␈α∂Two␈α∂advantages␈α∂were␈α∂hoped␈α∂for␈α∂by␈α⊂keeping␈α∂the
␈↓ x␈↓BEINGs␈αhighly␈α
structured.␈α The␈αinteractions␈α
(especially␈αwith␈α
the␈αuser)␈αwere␈α
brittle,␈αbut␈α
the␈αcomplex␈αtasks␈α
put
␈↓ x␈↓to them ␈↓βwere␈↓ completed.
␈↓"%␈↓ x␈↓␈↓ ↓XThe␈α∃major␈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
␈↓ x␈↓themselves.␈α∂The␈α∂ideas␈α∞seem␈α∂to␈α∂fit␈α∂together␈α∞viably,␈α∂and␈α∂the␈α∞automatic␈α∂programming␈α∂itself␈α∂went␈α∞effortlessly.
␈↓ x␈↓Sophistocated,␈αbug-free␈αprograms␈α
were␈αgenerated,␈αafter␈αhours␈α
of␈αfairly␈αhigh␈αlevel␈α
dialgue␈αwith␈αan␈αactive␈α
user
␈↓ x␈↓and tens of thousands of messages passed among the BEINGs.
␈↓"%␈↓ x␈↓␈↓ ↓XAll␈α⊃this␈α⊂suggests␈α⊃two␈α⊃possible␈α⊂continuations.␈α⊃One␈α⊂is␈α⊃to␈α⊃rethink␈α⊂the␈α⊃communication␈α⊃problems,␈α⊂and
␈↓ x␈↓develop␈αa␈αbetter␈αsystem␈αfor␈αthe␈αconcept␈αformation␈αprogram␈αgeneration␈αtask.␈αThis␈αis␈αnow␈αunder␈αstudy␈αhere␈αat
␈↓ x␈↓the␈αStanford␈αAI␈αLab.␈α The␈αother␈αcontinuation␈αis␈αto␈αfind␈αa␈αdomain␈αwhere␈αflexible␈αnatural␈αlanguage␈αhandling
␈↓ x␈↓is␈α∞not␈α∞crucial.␈α∞The␈α∞author␈α
is␈α∞currently␈α∞investigating␈α∞elementary␈α∞number␈α
theory␈α∞as␈α∞a␈α∞potential␈α∞task␈α
domain.
␈↓ x␈↓BEINGs␈α
will␈α
be␈α
present␈α
in␈α
that␈α
system,␈α
for␈α
their␈α
organizational␈α
power,␈α
but␈α
so␈α
will␈α
many␈α
simpler␈α
and␈αmore
␈↓ x␈↓efficient internal mechanisms for representing and manipulating knowledge.
␈↓"	
␈↓"	␈↓ x␈↓␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α?␈α␈↓∧␈↓&10. References␈↓)αβ␈↓
␈↓"∃␈↓ x␈↓[1]␈α∞Green␈α∞et␈α∞al.,␈α∞␈↓βProgress␈α∞Report␈α∞on␈α∞Program-Understanding␈α∞Systems␈↓,␈α∞Memo␈α∞AIM-240,␈α∞CS␈α∂Report␈α∞STAN-
␈↓"¬␈↓ x␈↓CS-74-444,Artificial Intelligence Laboratory, Stanford University, August, 1974.
␈↓"∃␈↓ x␈↓[2]␈α∩Hempel,␈α∩Carl␈α∪G.,␈α∩␈↓βFundamentals␈α∩of␈α∪Concept␈α∩Formation␈α∩in␈α∪Empirical␈α∩Science␈↓,␈α∩University␈α∪of␈α∩Chicago,
␈↓"¬␈↓ x␈↓Chicago, Illinois, 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 IJCAI, pp. 200-207.
␈↓"∃␈↓ x␈↓The␈α∩ideas␈α∩and␈α∩the␈α∩system␈α∩described␈α⊃use␈α∩concepts␈α∩from␈α∩ACTORS,␈α∩heterarchy,␈α∩structured␈α⊃programming,
␈↓"¬␈↓ x␈↓assertional␈αdata␈α
bases,␈αflexible␈α
data␈αtypes,␈α
pattern-directed␈αinvocation␈α
of␈αprocedural␈α
knowledge,␈αthe␈α
paradigm
␈↓"¬␈↓ x␈↓of␈α
dialogue,␈α
studies␈αon␈α
program␈α
specification,␈αQLISP,␈α
INTERLISP,␈α
LISP,␈αEnglish,...␈α
In␈α
particular,␈αthe␈α
author
␈↓"¬␈↓ x␈↓drew heavily from creative discussions with C. Green, R. Waldinger, D. Shaw, and E. Sacerdoti.