"#

%

"#

I <AIHANDBOOK>LENAT'S-IDEAS,JI TUE H-OCT-75 S:I«PM

13-OCT-75 iai2s:l3-PDT,ia37|oooo000 00 000Date! 13 OCT 1975 1425-POTFrom: LENATSubject: MISCTot FEIGENBAUM, EAF at SAILcci LARSON

MY MAILBOXES ON THE NET ARE:LENAT#SUMEXLENAT»SRIDBL#SAIL

MY PHYSICAL MAILBOXES ARE AT THE AI LAB AND (I ASSUME] NOW AT SERRA,

I LOG INTO EACH OF SRI, SAIL, SUMEX A FEW TIMES A WEEK TO CHECK MESSAGES,ALMOST EVERY DAY, I LOG INTO SUMEX, SO THAT IS THE BEST PLACE TO SENDMSG,

THE PASSWORD TO 1225,EAF) WAS SET UP AS AI

THE FILE 818t225,EAF) IS JUST YOUR LIST OF REFERENCES (PASSED OUT IN CLASS ATHE FILE 8182 IS ALL THAT, PLUS I HAVE APPENDED THE SYLLABUS OF THE LAST

AI QUAL [INCLUDING A LONG LIST OF REFERENCES)THE FILE HANDB.OOK IS NOW BLANK, AND CAN BE USED LATER FOR CLASSMATERIALS.

IF YOU NEED MORE DETAILS, JUST ASK,

I WILL TENTATIVELY COMPOSE A BRIEF ARTICLE, BEFORE CLASS THIS WEEK,IN THE FORMAT THAT I THINK MIGHT BE SUITABLE, IF OTHERS DO THESAME, WE CAN LOOK AT THEM AND PICK A SINGLE FORMAT FOR OUR WORK,I AM A FIRM BELIEVER IN GETTING A ONE-QUARTER SEMINAR OFF TO AFAST START, SINCE THE NUMBER OF WEEKS IS SO SMALL,ONE NEW IDEA IS THEFOLLOW JNG: PERHAPS WE CAN JUST CATALOG THEAI FIELD THIS QUARTER, WITH JUST A FEW LINES IN EACHARTICLE; THEN, NEXT QUARTER, WE AND/OR OTHERS CAN BEGIN TO EXPANDTHESE INTO A FEW PARAGRQPHS EACH, WHILE SIMULTANEOUSLY ANOTHERGROUP CAN WORK OUT THE SEMANTIC NET,

'TILL WEDNESDAY'S LUNCH,CHEERS,

DOUG

iO-OCT-75 iai 55 J35-PDT, 2205;000000000000Mall from SU-AI rcvd at 10-OCT-75 1«55-PDTDate: 10 OCT 1975 1 439-PDTPromt Doug Lenat (DBL " SU-AI)Tot EAF, FEIGENBAUM " SUMEX-AIM, LARSON " SUMEX-AIMTot LENAT 9 SRI-AI

H1Yesterday I created ht\ area (225,EAF3 here at SAIL, and today Iyou have 1n fact transferred a couple files to It, Good,

that

I have taken the liberty of shortening the name of the bibliography file toand have created a new file, 8182, which contains 818 olus the entire AI Quasyllabus that Terry and I prepared last Spring, Both the organization and thspecific references may be of interest,

"

0

dm

#

0

I «AIHANDBQOK>LENAT'S-IDEAS,»I TUE 14-OCT-75 SU4PM%

l personally am Interested In writing some of the articles on representationof knowledge,mPerhaps one way to organize tMI handbook is as follows*

0What AI Researchers Have Done

Problems worked onPrograms writtenLanguages developed

Sk

The Ideas of AIol Knowledge Engineering

Probl em-Sol vl ngFormal, abstract ways of representing and solving problems

gk Information Processing Psychology (and possibly cybernetics)

Each given node will be pointed to by several others, and will in turn pointuser to various new nodes, depending on his needs. Of course, one "outline"be given preferential status 1n that the pages 1n the book will be ordered b

00 We must recognize the possible needs of the user of our handbookl1) He has heard of "ancestor filtering" and wants to know what It m2) His program 1s taking too long and he wants to see how other res

in AI have been able to dramatically speed up similar progra5) Before writing his theorem-prover, he wants to survey what Ideas,

and techniques others have already tried, and how they fare«) He wants to study the AI ideas themselves, not the details of sp

111programs? he hopes that he can learn an Idea and use It Incontext (e,g,, speed up theorem-prover by adding multiple ksources) ,

This particular orientation is hard but valuable

There might be a separate color link for each purpose, etc,

" Well* this 1s enough ramblinq for now, If there are any comments, send themto LENAT#SUMEX or DBL»SAIL,

Cheers, -- Doug

" ■»- .«-

#

* <AIHANDBOOK>FEIGENBAUM'B-OUTLINE-OF-A1,225?i TUE 14-OCT-75 StSBPM

10/9/75" Felgenbaum's Proposed Structuring of A, I, Field (Draft)

A.I,

The processes of oroblem solving

" abstract - without world knowledge (e,g, propositionalcalculus machines, predicate calculusinference schemes, GPS core)

knowledge Intensive (e.g., DENDRAL).

" Knowledge-based systems (alias Semantic Information Processing,World Models, Meaning, applied epl stomol ogy, knowledge engineering)

Knowledge acquisition

Smooth Interaction (e.g., MYCIN, fancy displays, time-sharing

" techniquesExpert-CS "custom crafting"Inductive Inference on empirical dataAnalogizing, e,g,» MERLIN" Reading text, e.g., Charnlak

V Knowledge deployment

One-level integration " ala DENDRAL, HEARSAY IMulti-level Integration » ala HEARSAY 11, vlsiHEARSAY 11, vision

Knowledge representation

" Formal systemsProcedural systemsProduction systems - recognl ze-act cycle; COND; S-A|

Event-drl yenSemantic theories - Schank " MargieSemantic networks - QuilHan? Norman et, al,

Sensory Information Processing

" Soeech - speech machine goals? Projects? issues: the powerof semantic support? how to achieve processing efficiency

VisionSpecialized molecules " e.g,, edge finders, region growersHypothesis formation strategies: world knowledge?

specially engineered "visual" representationsTheory of the common-sense 3-D world

Other signals

* Perception 9 Cognition Viewpoint (Hypothesl s)

Effectors

" Problem solving? plans, plan monitoring, uncertainty

" m-

<r

""""""""#

V

""""""«""

; <AIHANOBOOK»FEIGENBAUM'S-OUTtINE-OF"AI 1 225H TUE I««OCT«p7S 5158PM

Robot Hardware

Information Processing Psychology; "classical" osychjtheoretical psych; AI observing nature

Problem Solving! N&S tasks and methodology; other small tasksMemory; EPAM and variants; Qullllan; Anderson % BowerIntellectual Development; child serlatlon behaviorPattern Induction; Function Induction; Intelligence Test

PatternsVisual Imagery; Moran; other CMU workNeurosl sArtistic behavior; music; drawingPerception; Spoehr (EYEBALL.)

" -.1 <AIHANDBOOK>NILSSON-PAPER-OUTLINE/PHILLIPS,DONE;I TUE U»OCWS 5;27PM

" ARTIFICIAL INTELLIGENCE By Nils Nilsson; A sketch of the paper,

AI

" Core of AI

Reasoning

" Mainly concerned with the use of knowledge, 3 major subtopics;

Puzzle SolvingReferences to;" Heuristic Search

Problem spaces and problem statesOperators" Goal "SubgoalBackwards reasoningMeans»ends analysis" Relevant programs; GPS* Logic Theorist

" Question AnsweringLogical inference mechanismsAssociative retrieval of data

" Relevant programs; The QAI series (i»l*2*3*4)

Common Sense ReasoningAttempts for universalityDerivation of answers" Domain and Domain usage knowledgePl annl ngRobotics researchKnowledge based programs

" Modelling and Representation of Knowledge

Assertlonal vs ProceduralFirst order predicate caleulus (Green)Concept nodes (Quinian)" CPneept structures (Schank)Pattern directed Invocation* assertions (Hewitt)Discrimination nets (Rulifson)" Productions (Newell)Frames (Minsky)

" Representation of time* etcFrame problemQual l f l cat lon probl emWork by McCarthy, Haves* Hendrlks & Bruce

Heuristic Search

Idea of a heuristic« State space approach, problem reductionGraphs* Trees* AND»QR graphsAlpha-BetaMi nl maxEvaluation Functions

#

" - i

<AIHANDBOOK>NILSSON-PAPER-OUTLINE/PHILLIPS,DONE;I TUE H»OCT-75 5;27PM1

"""""Jfc F

"#

Jll

""""""#

1

"#

Optimal search algorithms

AI Systems and Languages

Symbolic ManipulationIPL-V, LISPExtensi ens

SearchExpression retrievalPattern Matching

irst Level Applications

Game Playing

Checkers (Samuel)Chess (Greenblatt)Others; Kalah* Go

Math, Seience and Engineering Aids

Symbolic Integration (Moses)Theory Formation and rule extraction from experts (DENDRAL)Rule based operation (MYCIN)

Automatic Theorem Proving

Techniques; ResolutionEfficiency improving strategiesProgram CorrectnessGeometry Drovers (Gelernter)Knowledge based specialists (Bledsoe)

Automatic Programming

Program proving (Floyd)Program Generation (HACKER, Buchanan & Luckham)Aids (INTERLISP)

Robots

Reasoning* locomotive skills* perceptual abilities* natural languag(Edinburgh* MIT* Stanford* HITAC* SRI systems)

Industrial Automation (ALi SRI*)

Vi Sion

Two dimensional Image understandingPattern recognition and classification (Duda & Hart)PerceptronsWork by Roberts* Guzman, Huffman, Waltz* Shiral* Agin* Binford*

Yakimovsky, Feldman, Brlce* ete,Work by Tenenbaum on use of multiple knowledge sources In scene

anal ysi s

" «»' ■ *; «AIHANDBOOK>NILSSON-PAPER-OUTLINE/PHILLIPS,DONEf I TUE 14-OCT-75 5;27PM

Natural Language Understanding

" Winograd's SHRDLUSchank's schemesSpeech Understanding (Newell)

Information Processing Psychology

EPAM* HAMAssociative memory and retrievalBehaviour graphs* Production systemsSemantic NetsInduct ionPARRY, ELINOR, GPS

""

"1

"

&*

(»

"

BREAKDOWN OF ARTIFICIAL INTELLIGENCE , p^j [J,*//, ,\ 5

[This is a composition of a number of other papers which were freelyborrowed from. References will be added later. Few, if any, branchesgo down to the terminal nodes. Most branches go down as far as myknowledge or time to research the subject allowed. Uhen I had vaguenotions about what should be in the subtree, there is an asteriskfollowed by informal comments. Uhen I had idea3 for some but not allof the branches from a point, "there is an "ETC." after the last branch.This is a very rough draft and comments, suggestions, and additions areencouraged at all levels. fV\\ +o Peto a* SAIL.This file is OUTLIN.AI I22S.EAF] and <AIHANDBOOK>BREAKDOUN-OF-AI/UILKINS.]

I. Artificial Intelligence from a global viewpoint[how much of this section, if any, do we wish to include in the handbook?]

A. Phi losophy*perhaps Turing's Test article, Dreyfus's Uhat Computers Can t Do with

Papert's reply, Anderson's Minds and Machines, mind-body problem, .definitions of artificial intelligence and intelligence, etc. Li*lM/itI(A

B. Relationship to society1) Science fiction2) Popular misconceptions3) Home terminals (McCarthy has articles)ETC.

C. Hi story♦cybernetics, etc. Ueiner's Human Uses of Human Beings: Cybernetics

and Soci ty

]£. Funding1) ARPA2) other sources of funding3) proposal writing

E. Conferences and. publ icat ions * .Journa l of AI .SIGART.SIGCAS

■

Machine Intel l igenceIJCAI proceedingsCACM (computer science, some AI)

JACM (computer science, some AI)Cognitive Psychology (some Al)

American Journal of Computational Linguistics (some AI)

Special interest conferences: cybernetics, natural language, robotics. et

11. Artificial Intel l igence Methodologies and TechniquesA. Knowledge representation

[pointer to memory and learning sect ion of information processing psychology]A 1) Production systems

♦Waterman, Newell and Simon, Dendral, Mycin, etc.2) Frames

*Minsky, Uinograd, etc.3) Conceptual Dependency

♦Schank and students4) Semant ic Nets

"♦

2

"♦Qui I l ian, etc.5) Formal systems

a) Predicate calculus[insert El-Masri's outline here]

b) Higher-order logicc) LCFd) Hoare's logicETC.

6) Procedural representationsa) Pattern directed invocationb) Assertionsc) DemonsETC.♦Winograd, Hewitt, etc.

7) Representation of timea) Frame problemb) qual if i cat ion problem

and Hayes, Hendricks and Bruce, etc.8) Di scr i mi nation nets£l\F ♦EPAM, etc.9) Mi seel laneous

♦Hewitt's Actors, etc.

Knowledge Acquisition _[pointer to memory and learning section of information1) User interaction

B psychology]processing

2) CS Custom Crafting3) Reading text4) Inductive inference

♦Dendral system, etc.[the following comes from Aikin's learning outline)

5) Planning--"milepost" paradigm for plans

6) Reasoning by Analogy7) Learning—evaluation functions

--generator functions—CHECKERS (Samuel; 1959,1367)—Hewitt-functional abstraction (19G8 et. seq.)--Learning as heuristic development__sel f-af fecting programs

8) Waterman's ideas—heuristic rule3—heuristic definitions—heuristic blocks— deci sion matrix

9) Samuel's ideas—parametric functions— signature types and tables--rote learning'— learning by generalization—book learning

VIS (Moran, 1973); ACT (Anderson, 137G) ;LISP7O (Tesler, 1373); MYCIN; DENDRALControl elements (Anderson, 197G) "C. Reasoning and problem solving

1) Search methods[pointer to Heuristic Search section)

10) Product ion Systems—productions „„-,,—PSG (Newell, 1973); PASII (Waterman, 1974);

i

9

3"

D

E

" 111

a) Generate and testb) Hill cl imbingc) Means end analysisETC.

2) Theorem proving and proof finding proceduresa) ResolutionETC.

3) Planninga) Subgoalingb) AbstractionETC. . .

4) Dealing with uncertainty♦Sproull, Fledman, The Advice Taker, etc.

5) Reasoning by analogy

*Evans,

MIT, etc.

Heuristic Search1) Combinatorial problems2) Strategies

a) Breadth firstb) Depth firstc) Branch and boundETC.

3) Search spacesa) Graphsb) Treesc) AND-OR graphsETC

4) Heuristics and techniquesa) flinimaxb) Alpha-betac) Killer heuristicd) Evaluation functione) Generator functionETC.

5) Optimal i ty and efficiency

AI Languages1) List processing2) String processing3) Production systems _4) Data structures and retreival

a) Non-associative

♦Sets,

bags, list, tuples, etcb) contexts, pattern matching retreival, etc

5) Control structuresa) Control contextsb) Backtrackingc) Co-routinesd) Multiprocessinge) Demons . . „..„f ) Pattern directed function invocationg ) User specified mechanisms (Conniver)

Problem Domains of Al Research:„„' rthp followinq is WTL's outline]

A ' Au .?m?mrqAS°SPEC?RCATION TECHN QUES9

" PR ° T) NATURAL LANGUAGE TO SPECIFY ALGORITHMSi) HEIDORN'S SYSTEM

. .H

"

0[The

"

ii ) OULiii) ISI WORK

B) EXAMPLESi) HARDY'S WORKii) SHAW'S AND SWARTOUT'S WORKiii) SIKLOSSY'S WORK

C) TRACESi) BIERMAN'S WORKii) SIKLOSSY'S WORKiii) BAUER'S WORK

D) VERY HIGH LEVEL LANGUAGESi) EXTENDABLE LANGUAGESii) SET ORIENTED LANGUAGES

E) PREDICATE LOGICII) PROBLEM TRANSFORMATION TECHNIQUES (AUTOMATIC CODING)

A) THEOREM PROVINGB) STANDARD PROBLEM SOLVING TECHNIQUESC) DEBUGGING TECHNIQUES (SUSSMAN)

III) PROGRAM TRANSFORMATION TECHNIQUESA) DARLINGTON'S & BURSTALL'S WORKB) JIM LOW'S WORK

IV) LEARNING SYSTEMSV) UNDERSTANDING SYSTEMSVI) PROGRAMMER'S AID

EXPERIMENTAL AND COGNITIVE PSYCHOLOGYfollowing is Perry Thorndyke's outline,]

I . Perception(Information sources: Neisser, COGNITIVE PSYCHOLOGY

Norman, MEMORY AND ATTENTIONLindsay & Norman, HUMAN INFORMATION PROCESSINGHaber, INFORMATION PROCESSING APPROACHES TO VISUAL

PERCEPTIONHaber, CONTEMPORARY THEORY AND RESEARCH IN VISUAL

PERCEPTION

Chase,

VISUAL INFORMATION PROCESSINGA. Attention

1. Selective attention; Cocktail Party phenomenon2. Fi I ter models »

B. Visual Perception: Iconic Storage and Coding1. Transient iconic memory2. Masking3. Verbal coding

C. Pattern Recognition1. Displacement and rotation2. Template matching3. Feature analysis: Pandemonium et al.4. Analysis-by-synthesis

D. Auditory Perception1. Speech perception2. Echoic Memory and auditory attention

E. Appl ied Percept i on1. Chess:

Simon,

and Chase & Simon2. Semantic coding of visual percepts: Clark comprehension

mode I

11. Memory and Learning(Information sources: Lindsay & Norman, HUMAN INFORMATION PROCESSING

Kintsch, LEARNING, MEMORY, AND CONCEPTUAL PROCESSESTulving & Donaldson, ORGANIZATION AND MEMORYNorman, MODELS OF HUMAN MEMORYAnderson & Bower, HUMAN ASSOCIATIVE MEMORY

.

t

5

"

"

Kintsch, THE REPRESENTATION OF MEANING IN MEMORYPaivio, IMAGERY AND VERBAL PROCESSESNorman SRumelhart, EXPLORATIONS IN COGNITION

A. Structures and processes1. Short-term memory2. Long-term memory3. Rehearsal4. Chunking5. RecognitionG. Retrieval , recal l7. Inference and question-answering8. Semantic memory vs. episodic memory9. Interference and forgetting

10. Type nodes vs. token nodesB. Memory Models and Knowledge Representations

1. Associative memory modelsa. Semantic Memory and Teachable Language Comprehender

(Qu iI I i an and Co ll i ns & Qu i I I i an)

b. HAM (Anderson & Bower)c. ELINOR (Lindsay, Norman, S Rumelhart)

d. Conceptual Dependency (Schank)e. EPAM

2. Procedural representationsa. Production systems

"

b. PLANNER3. Non-l inguistic representations

a. Imagery and analog representations4. Other representations

a. Frame systems (Minsky, Winograd, et al.)

b. Augmented Transition Networks

111. Language Comprehension/Psycholinguist cs(Information sources: Fodor, Sever, & Garrett, THE PSYCHOLOGY OF

LANGUAGEAnderson & Bower, HUMAN ASSOCIATIVE MEMORYMinsky, SEMANTIC INFORMATION PROCESSINGSchank & Colby, COMPUTER MODELS OF THOUGHT AND

LANGUAGECarroll & Freedle, LANGUAGE COMPREHENSION AND

THE ACQUISITION OF KNOWLEDGEClark & Clark, PSYCHOLINGUISTICS (forthcoming)

A. Concepts to be coped with1. Competence vs. performance models ;;2. Phonology vs. syntax vs. semantics vs. pragmatics3*. Surface structure vs. deep structure4. Taxonomic grammars, generative grammars, transformational

grammars5. Phrase-structure rules, transformation rulesG*. Constituents, lexical entries7. Parsing vs. generation8. Context-free vs. Context-sensitive grammars

B. Computational Linguistics: Language understanding systems1. Augmented Transition Networks (Woods, Kaplan)2. Procedure-based systems (Winograd. and descendants)

3. Logic-based systemsa. SIR (Raphael)b. Coles' systemc. QA3d. STUDENT (Bobrow)

4. Conceptual Dependency (Schank et al.)5. Machine Translation (Wi Iks) .S. Semantic Networks (Simmons, Qui I Man, ELINUH)

«

A

G

"

C

"Scientific Applications [ideas from Fried land's outline]D

7. SCHOLAR: tutorial dialoguesC. Areas of psychological experimentation

1. Acquisition and language development2. Memory for sentences

a. Transformation hypothesisb. Deep structure vs. surface structurec. Implications and abstract ideas

3. Comprehension of sentencesa. Negationb. Comparativesc. Actives and passivesd. Markednesse. Ambiguityf. Anamolous sentencesg. Influences of imageryh. Influences of lexical complexityi. Transient memory load (click studies)

4. Semantic memory5. Discourse structure and memory for prose

a. Effects of content variablesb. Effects of structure variablesc. Effects of instructional variables

IV. Problem Solving(Information sources: Newell & Simon. HUMAN PROBLEM SOLVING

Norman & Rumelhart, EXPLORATIONS IN COGNITIONErnst & Newel I, GPSKleinmuntz, PROBLEM SOLVING

A. Model ing Game Playing1. Chess2. Cryptari thmetic3. Go, Go-mo-ku

B. Solving Logic problemsC. Solving algebra and other mathematical problemsD. Concept formation and identification

V. Behavioral Modeling(Information sources: Schank & Abelson, COMPUTER MODELS OF THOUGHT AND,

LANGUAGEA. Belief Systems and Implication Molecules

(Abel son)B. Conversational Postulates (Grice)

C. TutorialDialogues (SCHOLAR)

D. Parry (The paranoid patient)

Game playing1) Checkers

♦outl ine of Samuel's work2) Chess

a) Shannon's ideasb) Turing's programc) Berstein's Programd) NSS programe) Greenblatt's Programf ) Ber l iner' s workg) Russian's work (KAISSA)

3) Go and Go-mo-ku4) Kalah

1

r

"

E

F

" G

""I-f

1) Chemistrya) Organic synthesis

♦Corey, Wipke, Gelerntsr, Ugib) Pattern recognitionc) Mass spectrometry

♦Dendrald) Protein crystallography

2) Medicinea) MYCINb) DIALOGc) CASNETd) Pauker MIT work

3) Psychology and psychiatry♦PARRY, BELIEVER, etc.

4) Matha) MATHLABb) SAINTETC.

ETC.

Theorem Proving1) Predicate calculus

[pointer to pc section of knowledge representation]2) Herbrand's theorem3) Resolution

a) Ground resolutionb) General resolutionc) Uni f i cat ion

4) Search strategies and efa) Uni t preferenceb) Tautalogy eliminationc) Factoringd) Subsumptione) Hyperresolut i onf) Set of supportg) SL resolutionETC.

efficiency heuristics

ETC.

Robots1 ) Locomot ive ski II s

♦Servoing, etc.2) Reasoning

♦Planning with uncertainty, etc.3) Perceptual abilities

♦Machine vision work4) Hand-eye coordination

♦Bol les and Paul , etc.5) Industrial automationG) Robots

a) SHAKEY, SRIb) FREDDY, Edinburghc) Stanford hand-eyed) MIT hand

Vision [ideas from Arnold's outline]1) Hardware2) Image representation

a) Line descriptions

«.

t

M

"b) Shape descriptions3) Scene analysis (could break down into 2D and 3D)

a) Edge enhancement, spatial differentiationb) Noise removal, spatial smoothingc) Template matchingd) Region analysise) Contour fol lowingf) Perspective considerationsg) Stereo vi si onETC.

4) Multi sensory images♦fenenbaum

5) PerceptronsS) Programs

a) Robertsb) Guzmanc) Falkd) Huffmane) Clowesf) Waltzg) Kel leyh) Shiraii) Barrow and Popplestonej) Feldman and Yakimovsky

ETC.

H. Natural Language Understanding1) Ear ly work

♦Minsky's review in Semantic Information Processing, ELIZA, etc.2! Syntax

a) Transformational grammarsb) Augmented transition netsc) Systemic grammarsd) Context free grammarse) Context sensitive grammarsETC.

3) Semantic theories♦Schank, Winograd, Simmons, etc.

4) Stor i es and Pel ief♦Charniak, Thorndyke, Levy, McDermott, etc.

5) Translation♦Wi Iks

6) Speech understanding♦Reddy, Woods, Walker, Newell, HERESAY, etc.

/

i. M\\c<*ii

"

I , OYW

«V

"

a. LISPi PLA#*<e£c. Cofi/txllij^J. -/VPfo (.

"41. _-Rwt, Yrtt.%

A- S hkft _

ft. '-""""iirt*,fV__J

# I. DeprU hWrc . £*-«.JtL £^p<*" Vturii^'is

e. */_

h C*r*atr, %«/?»* »/ rl 1- -fOr- *. />*.,<//"«>«

X. Tt>»l3 ▼" I <-ckn,^v»_s

I. / <cL*h^ja.. L___A_«____! «___.W.4.4 ,

*»«

2. La*.-,, 4 ,

D. I *_4e«-A ft<Co«»..V.*> * -_ ce*,_ r\<nfl>*slrC. L. y«rv>*r.e* "" L f?

«/-!."»«

' na*pt.iH! <*._ fvnij

J). I*l »cL\»~a " V-t or<s*x*\~s.k\**J

i. f/_K*.Ai * /__dL_____h__

T- C

Ov» £«.#

-fv_l i'CjJ _______rf

5"- rv«n*»«£

1 . 5 ewrxVi"*>»'_*(_

►-2.

"

" T/T . P^eUew. Poih9,'^

#

f- tv-I—.b** ■hi*<-^'»»/a. |4.H c(^.«* C a»r_<|.*v.rj -fccL*. ,jts

Ox.. l I «*."»"<*»

i. Rfc^eV,"^

3. Tnfor*tn r<^»*v^u C r**oi>f r%mlx>~* n*i"L»aj

4. flvfamo}-*. \\c*,iAOy. T e«rc.* tv»*s

p. ______! C»n«v»««_"_" ->""»,,. I f r»tos*»

". Na^iw«[ Lo»s»u*,. S^sfdw-.s

7. S_t*»*«Kc TW.^^3J- I lTq»iU4-..

rt

T, Sotecl\ L/~«l«/-jr-^d.~

5. M &"H* *«M at",

e*

I T~^*o'--»»» P|-»^v *n

C. Oft.W\_ fl»M'-0

P. ttr>*oA JtlUfc f-eixjort- i» - r»T»"le«n -Jol</»«*-E. H*cL.»v_ V\ j:«a

P- fUulS.t*.! Pft«vZM*)>".» f7 foto+Jt-J

7. rt-UUK,6 ■ rl i»to !_____ rV&a in_

""»»»«"*

t

-;3j

"

v

j <AIHANDBOOK>THORNDYKE-OUTIINE-OF-IPP,|i TUE ia-OCT-75 6tOBPM

I have put together a list of topic areas within psychology whichhave Potential relevance to AI, These should be considered for coverageIn your HANDBOOK OF AI, This first Pass of a list Is a breadt h-f | rateffortf I have tried to cover all possible topic areas without going Intoany of them In too much depth. One problem I'm having Is Inferring precise!what the entries In the handbook will consist oft programming teehnlgues,a glossary of terms and concepts relevant to the field, concise andexplanatory summaries of the concepts (like an encyc l oped! a) » or some orall of the above. It would make my Job of listing relevant concepts easierIf I knew exactly what you had In mind. Anyway, here's a first pass list.

Could you mall to me a list of class discussion/lecturetopics for your seminar as soon as It Is available? I am antici-pating comlnq up for some elass meetings like Bob and Jim did lastyear, so It would be helpful to know the schedule, I will arrangetrios to correspond to other appointments In the Stanford area, too,so the schedule would be useful as soon as 1t 1s available.

Perry

" DOMAINS IN EXPERIMENTAL AND COGNITIVE PSYCHOLOGYOF POTENTIAL RELEVANCE TO ARTIFICIAL INTELLIGENCE

I, Perception

(Information sources: Neisser, COGNITIVE PSYCHOLOGYNorman, MEMORY AND ATTENTIONLindsay „ Norman, HUMAN INFORMATION PROCESSINGHaber, INFORMATION PROCESSING APPROACHES TO VISUAL

PERCEPTIONHaber, CONTEMPORARY THEORY AND RESEARCH IN VISUAL

PERCEPTION

Chase,

VISUAL INFORMATION PROCESSING

A, At tent lon1. Selective attention: Cocktail Party phenomenon2, Fl l ter model s

B, Visual Perception: Iconic Storage and Coding1, Transient Iconic memory2, Mask! ng3, Verbal coding

C, Pattern Recognition" 1 , Displacement and rotation2, Template matching3, Feature analysis! Pandemonium et al ,4, Anal ysl s-by-synthesl s

1 -OCT-75 13:28:01-PDT,8a68;0 00000 0 00000Wlall from RAND-RCC rcvd at 1-OCT-75 1327-PDT

Ed,

<AIHANDBOOK>THORNDYKE-OUTLINE-OF-IPP, jI TUE ia-OCT-75 6:OBPM'.

"11.

"

111

"

D t Auditory Perception1, Speech perception2, Echoic Memory and auditory attention

E, Applied Perception1, Chess: Simon, and Chase 8. Simon2, Semantic coding of visual percepts: Clark comprehension

model

Memory and Learning

(Information sources: Lindsay & Norman, HUMAN INFORMATION PROCESSINGKlntsch, LEARNING, MEMORY, AND CONCEPTUAL PROCESSESTulvlng & Donaldson, ORGANIZATION AND MEMORYNorman, MODELS OF HUMAN MEMORYAnderson - Bower, HUMAN ASSOCIATIVE MEMORYKlntsch, THE REPRESENTATION OF MEANING IN MEMORYPalvlo, IMAGERY AND VERBAL PROCESSESNorman & Rumelhart, EXPLORATIONS IN COGNITION

A, Structures and processes1, Short-term memory2, Long-term memory3, RehearsalU, Chunking5, Recognition6, Ret r] eval , recal l7, Inference and question-answering8, Semantic memory vs, episodic memory9, Interference and forgetting

10, Type nodes vs, token nodes

B f Memory Models and Knowledge Representations1, Associative memory models

a. Semantic Memory and Teachable Language Comprehender(QulHlan and Collins & Quilllan)

b, HAM (Anderson & Bower)c, ELISOR (Lindsay, Norman, & Rumelhart)d, Conceptual DeDendency (Schank)e, EPAM

2, Procedural representationsa. Production systemsb, PLANNER

3, Non-1 I ngul st l c reoresent at l onsa t Imagery and analog representations

4, Other representationsa. Frame systems (Minsky, Winograd, et al,)b. Augmented Transition Networks

, Language ComDrehens! on/Psychol l ngul st l cs(Information sources: Fodor, Bever, & Garrett, THE PSYCHOLOGY OF

LANGUAGEAnderson _ Bower, HUMAN ASSOCIATIVE MEMORYMinsky, SEMANTIC INFORMATION PROCESSINGSchank & Colby, COMPUTER MODELS OF THOUGHT AND

<AIHANDBOOK>THORNDYKE-OUTLINE-OF-IPP,»I TUE H-OCT-75 6:OBPM;

\

"

"

IV,

"

LANGUAGECarroll R Freedle, LANGUAGE COMPREHENSION AND

THE ACQUISITION OF KNOWLEDGEClark % Clark, PSYCHOLINGUISTICS ( f or t hcom l no)

A, Concepts to be cooed with1, Competence vs, performance models2, Phonoloqy vs, syntax vs, semantics vs, pragmatics3, Surface structure vs, deeo structure4, Taxonomlc grammars, generative grammars, transformational

grammars5, Phrase-structure rules, transformation rules6, Constituents, lexical entries7, Parsing vs, generation8, Context-free vs. Context-sensitive grammars

B, Computational Linguistics: Language understanding systems1, Augmented Transition Networks (Woods, Kaplan)2, Procedure-based systems (Winograd and descendants)3, Logic-based systems

a, SIR (Raphael)b, Coles' systemc, QA3d. STUDENT (Bobrow)

il, Conceptual Dependency (Schank et al,)5, Machine Translation (Wllks)6, Semantic Networks (Simmons, Qullllan, ELINOR)7, SCHOLAR: tutorial dialogues

C, Areas of psychological exoer 1 mentat l on1, Acgulsltlon and language development2, Memory for sentences

a, Transformation hypothesisb, Deep structure vs, surface structurec, Imollcatlons and abstract Ideas

3, Comprehension of sentencesa. Negationb, Comparativesc, Actives and passivesd, Markednesse. Ambiguityf, Anamolous sentencesg, Influences of Imageryh, Influences of lexical complexity1, Transient memory load (click studies)

li. Semantic memory5, Discourse structure and memory for prose

a. Effects of content variablesb. Effects of structure variablesc. Effects of instructional variable

Problem Solving

(Information sources: Newell & Simon, HUMAN PROBLEM SOLVINGNorman 8, Rumelhart, EXPLORATIONS IN COGNITIONErnst & Newell, GPSKlelnmunt2, PROBLEM SOLVING

<AIHANDBOOK>THORNDYKE-OUTLINE-OF-IPP,f I TUE 14-OCT-7i 6:OBPM»

»

"V,

OF THOUGHT AND

"

mm^

A, Modeling Game Playing1, Chess2, Cryptarl thmet l c3, Go, Go-mo-ku

B, Solving Logic problemsC, Solving algebra and other mathematical problemsD. Concept formation and Identification

Behavl oral Model I ng

(Information sources: Schank & Abelson, COMPUTER MODELSLANGUAGE

A, Belief Systems and Implication Molecules(Abel son)

B, Conversational Postulates (Grice)C, Tutorial Dialogues (SCHOLAR)D, Parry (The paranoid patient)

lZti*h)d&Or€/<f2/?Z (<r** /\x225, Efif

"_S Jl^ /J*.-/**M QUoifa /^k%c^k>£ d s,^%J^"S

s £2~y ojr <*4 /ft/y

t vL k_v£ l/l4yLud yZ. Uajl £>£*&sl A- u^ . i

Cux

?/

?*iV_s^ /'/W-5 /^w___c y „„/a ?

!

S^Ti/j 'a4- l\W $t}J>Usa^ r-*tf*" tC S(?Y^W" l/Ju*»e i

i^A. #^

-fa?l

drvjr~ J yyt

MX- /QBnes>

1/ /yritS.

" 2U

/Oote^ c^ HWt^o^-- /0/Vtt

fJ£t+

"

PAGE 1<LARSON>CS.22S;3 THU 9-OCT-75 12:40PMI

*

"

"

"

10/9/75

Feigenbaum ' s Proposed Structuring of A.I. Field (Draft)

A.I

The processes of problem solving

abstract - without world knowledge (e.g. propositionalcalculus macnines, predicate calculusinference schemes, GPS core)

knowledge intensive (e.g., DENDRAL)

Knowledge-based systems (alias Semantic Information Processing,K-orld Models, Meaning, applied epistomology , knowledge engineering)

Knowledge acauisition

Smooth interaction (e.g., MYCIN, fancy displays, time-sharingtechniques

Expert-CS "custom crafting"Inductive Inference on empirical dataAnalogizing, e.g., MERLINReading text, e.g., Charniak

Knowledge deployment

One-level integration - ala DENDRAL, HEARSAY IMulti-level integration - ala HEARSAY 11, vision

Knowledge representation

Formal systemsProcedural systemsproduction systems - recognize-ac t cycle; COND; S-A

Event-dr ivenSemantic theories - Schank - MargieSemantic networks - Quillian; Norman et. al .

Sensory Information Processing

Speech - speech machine goals; Projects; issues: the powerof semantic support;, how to achieve processing efficiency

VisionSpecialized molecules - e.g., edge finders, region growersHypothesis formation strategies: world knowledge;

specially engineered "visual" representationsTheory of the common-sense 3-D world

Other signals

Perception = Cognition Viewpoint (Hypothesis)

I

I

"

"

tasks

"

Effectors

Problem solving: plans, plan monitoring, uncertaintykobot hardware

information Processing Psychology: "classical" psych;theoretical psych; AI observing nature

Problem Solving: N&S tasks and methodology; other smallMemory: EPAM and variants; Quillian; Anderson & BowerIntellectual Development: child seriation behaviorPattern Induction: Function Induction; Intelligence Test

PatternsVisual Imagery: Moran; other CMU workNeurosisArtistic behavior: music; drawingPerception: Spoehr (EYEBALL)

§

fc \ooi±> 4 L&sj^cj*

«

-»

"

"

"

PAGE 1<LARSON>CS22S.BIBLIO;I THU 9-OCT-75 12:16PMi

"Artificial Intelligence"Marvin Minsky and Seymour Papert

"On Machine Intelligence"Donald Michie

"Explorations in Cognition"Donald A. Norman and David E. Rumelhart

Revolution"Perspectives on the ComputerZenon W. Pylyshyn

"The Sciences of the ArtificialHerbert A. Simon

& Thought""Pattern Recognition, Learning,Leonard Uhr

"Knowledge and Cognition"Lee W. Gregg

"A Computer Model of Skill Acquisition"Gerald Jay Sussman

"Human Associative Memory"John R. Anderson and Gordon H. Bower

"Visual Information ProcessingWilliam G. Chase, editor

"Introduction to Artificial Intelligence-Philip C. Jackson

"Artificial IntelligenceEarl B. Hunt

"Problem-Solving Methods in Artificial Intelligence-Nils J. Nilsson

"Computer Models of Thought and Language"Roger C. Schank and Kenneth Mark Colby, editors

"International Joint Conference on Artificial IntelligenceProceedings", first through fourth

SIGART Newsletters, continuing

"Theoretical Issues in Natural Language Processing"

An Interdisciplinary 'workshop in Computational Linguistics,Psychology, Linguistics, Artificial Intelligence, 6/10-13/75Cambridge, Massachusetts

"Pattern Classification and Scene Analysis"Richard 0. Duda and Peter E. Hart

"Theoretical Approaches to Non-Numerical Problem Solving"K. BAnerji and M.D. Mesarovic, editors

"Human Problem Solving"Allen Newell and Herbert A. Simon

. <LARSON>CS22S.BIBLIO;I THU 9-OCT-75 12:16PM PAGE 4■

"

"

"

"Artificial Intelligence Progress Report

Marvin Minsky and Seymour Papert

"Artificial Paranoia"Kenneth Mark Colby

"A Psychology of Computer Vision"Patrick Winston

"New Progress in Artificial IntelligencePatrick Winston, editor

"Artificial Intelligence-1974 IFIPNils Nilsson

Lecture"

it Where is it going's"Artificial Intelligence: What isEdward Feigenbaum

"Artificial Intelligence Journal" continuing

"Artificial IntelligenceJames Slagle

"Computers and Thought"Edward Feigenbaum and Julian Feldman"Machine Intelligence Series, I-VII"

"MIT-AI Lajoratory Series of Reports" continuing

"Stanford Univer sity-AI Laboratory Series of Reports" continuing

"Carnegie-Mellon University Computer Science SEries of Reports"continuing

"SRI-AI Laboratory Series of Reports" continuing

"SEmantic Information ProcessingMarvin Minsky, editor

</(«/- AJ