F L 86-36 I GJ i1VUlIJGJ i1 b 3 BI JWUV i1lli 3GJ B lDrnU 15lIJrnUl15 OO15UJlD15 f ooUJrn bUJGJlll lBl IlUml1mrn i J1 b 0 0 WAjSAW PACT I I I j TO O MANY GARBLES oooooooooo oo o RE- PSYCHLING THE CODE CLERK MACHINE INTELLIGENCE PROMISE OR DELUSION -o o oo RAP I DTRAN 1 I NEWS BRIEFS CMI CLA CAMINO I 1 1 k 1 I 1 l 3 L 86-36 4 5 9 13 - ------------------' 'fillS DOCtJMRN'f CON'fAINS CODRWORD 1A 'fRRIAL 11B8ihllll DIK il f N illi ill SSM 123 2 Hxempt ' om 8B8 Be 11852 eaRgo Beclmsif I JpMl No ilea ioll h ' the 9rili lllto Declassified and Approved for Release by NSA on '10-'1 '1- 20'1 2 pursuant to E O '135 26 vl DR Case # 54778 DOCID 4009722 Published Monthly by PI Techniques and Standards for the Personnel of Operations VOL II NO 7 JULY 1975 PUBLISHER WILLIAM LUTWINIAK BOARD OF EDITORS Editor in Chief ooo Arthur J Salemme 56425 Collect ion o 1L Cryptanalysis Jlf351tsl P L IC80ZSs 1 Language o o Emery W Tetrault 5236s Machine Support 1 3321s 1 Special Research Vera R Filby 7119s Traffic Analysis Frederic O Masqn Jr 41425 Production Manager o o o 1L - IC4998s For individual subscriptions send name and organizational designator to CRYPTOLOG PI 86-36 DOCID 4009722 g CR T TOP Ul 'IHRA l' '-i 1 FED REP Njjrnbera 'OF I L In recent years the Warsaw Pact has reco nized the need for July 75 OP CRYPTOLOG Page 1 SEEURRE'f lJMBRA EO 1 4 c P L 86-36 86-36 DOCID 4009722 'fOP SECRET UMBRA CI EO 1 4 c P L July 75 -CRYPTOLOG Page 2 TOP SECRET UMBRA 86-36 DOCID CONFIDENTIAl 4009722 Y G T EO 1 4 c P L 86-36 5 SOO MAN ARBLEJt A Mora I Tale for Cryptanalysts Byl ' I P L I July 75 CRYPTOLOG Page 3 CONFIDENTIAl UMIS15E VIA 68 1HlT 611AIHlHS 8 l15Y 86-36 DOCID 4009722 Eo 1 4 'fOP SECRE'f UMBRA P L RE-PSYCHLING THE CODE CLERK ft C I p _ L I Byl - 86-'----36 twho L ouM WO IW --- 6u bmU ted the notlowing e-6pon-6e to the edito ' 6 nv - tion to matc h MJt Ca i 1 maho 6' WIllI ob 6 vatiOn-6 the ApJUi 1975 Mu e Throughout the 1960's the Vietnamese Communists were straightforward in their cryptographic habits and could be relied u on to be fair1 secure However the VC code clerks n o t-u n l e t e lr apanesecounterparts in WWII as discussed by Mr Ca11imahos had a peculiarity that gave the cryptanalyst that helpful boost While there were indeed other features exploited by those engaged in the long-term indepth analysis of these communications this particular one proved especially interesting to those of us en a ed in the initial anal sis of Solution tol IRadiote1ePhone in June 1975 Issue i ------ Puzzle July 75 CRYPTOLOG Page 4 TOP CRg'f UMBRA EO 1 4 c P L 86-36 c 86-36 DOCID 4009722 UNCLASSIFIED MAeHtNE INTELLIGENCE PROMISE 8R DELUSION P L IP15 L - -_ _ The following paper was written as part of a larger project that will include a selective review of ARPA-funded ArtificiaZ Intelligence research since 1967 in the light of the criticisms raised by Dreyfus and others The results of this study will be published as a P1 Report in the near future with a human being on a topic of interest and importance An excellent recent book that tells the story of A I from the point of view of a prominent worker in the field is Michie 1974 Unlike so much writing on technical topics it is clear readable concise and even enjoyable and I strongly recommend it to anyone interested in a good survey from a friendly position What Computers Can't Do Hubert L Dreyfus presents a strong and well-reasoned Dreyfus Harper and Row New York claim that we will never be able to get digi1972 tal computers to do any of the above-mentioned Hubert Dreyfus is a professional philosopher kinds of intelligent things nearly as well as who has been a persistent gadfly to Artificial people do them He shows quite convincingly Intelligence A I researchers for several that the research he has reviewed covering years He published an earlier strongly crit- the years 1957 through 1967 has failed in all ical paper Dreyfus 1965 whose points were all but a few relatively simple highly formal and promptly and emphatically brushed aside by restricted tasks He bases his case on purely spo esmen for the A I community Now he has philosophical grounds on logic and on a demwritten a book which I believe cannot and onstration of the essential inadequacy of those should not be so summarily dismissed by any often unspoken assumptions about the brain thinking person whatever his attitude toward the mind human knowledge and the nature of computers reality that underlie not only A I research but in fact our whole Western scientific For those readers who may not be familiar world-view with that branch of computer science called Artificial Intelligence it involves the attempt Intelligent human activities are assigned to program or build a digital computer capable by Dreyfus to four classes of producing behavior that a human being would accept as truly intelligent A I researchers ASSOCIATIONISTIC ACTIVITIES innate or have been trying with varying degrees of suclearned by repetition such as memory cess to get computers to do such things as games maze problems word-for-word subplaying chess and Go proving logical and mathstitution and responses to rigid unvaryematical theorems understanding and translating patterns or stimuli ing written and spoken language controlling a SIMPLE FORMAL ACTIVITIES learned by robot vehicle or an artificial arm and hand rule such as simple games Nim Tic-tacperceiving and understanding visual scenes and toe simple combinatorial problems patterns and carrying out a sensible dialog July 75 CRYPTOLOG UNCLASSIFIED Page 5 86-36 DOClD UNCLASSIFIED 4009722 mechanIcal theorem proving and recognition of small sets of clear-cut patterns e g typed character fonts which treats man as a device an object responding to the influence of other objects according to universal laws or rules p 144 He states the following stern judgment upon COMPLEX FORMAL ACTIVITIES learned by this philosophy the goal of the philorule and practice such as noncomputable sophical tradition embedded in our culture is games chess Go complex combinatorial to eliminate all risk moral intellectual problems nonmechanical theorem proving and practical and recognition of complex patterns in noise Dreyfus points out that this reductionist viewpoint on the nature of man and human inNONFORMAL ACTIVITIES learned by examtelligence however ordinary and familiar it ple and insight such as ill-defined may sound to us is only one possible viewpoint games riddles open-structured problems and has led us to make some assumptions which translation of natural language and recogare not justified outside the realm of physics nition of varied and distorted patterns and the hard sciences He makes explicit such as speech four major assumptions underlying A I research A digital computer program as Dreyfus readily in particular admits can carry out activities in Classes I A BIOLOGICAL ASSUMPTION that the human and 2 quite adequately and even well Class 3 brain at some basic physiological level activities are theoretically capable of being employs discrete values in the same way as spelled out in a set of instructions but in a digital computer practice the resulting description is far too long and complex for brute-force i plementation A PSYCHOLOGICAL ASSUMPTION that the in any existing or presently conceIvable compuhuman mind can be viewed as a device ter These activities are apparently accomplished operating on discrete bits of information by people through the use of powerful conceptual according to formal rules again like a shortcuts and insights heuristics which prodigital computer vide an immediate and dramatic reduction in comAN EPISTEMOLOGICAL ASSUMPTION that all Plexity While Dreyfus does not commit himse lf knowledge can b formalized and whatever on the question of how far computers can go In heuristic attacks on Class 3 problems his tone can be understood can be expressed in terms is decidely pessimistic Activities in Class 4 of logical relations are in a completely different world Dreyfus AN ONTOLOGICAL ASSUMPTION about the nature claims separated from Classes I through 3 by a of- reality and basic to the other three major discontinuity that cannot be bridged by any that all relevant information about the computer program These activities are inherently world everything essential to the producindescribable in terms of any formal language tion of intelligent behavior must be ana such as mathematics logic and programming lyzable as oo a set of indifferent passive languages and hence aannot be programmed They facts each logically independent of all exist at a level totally outside of or prior the others to the reasoning digital linguistic descripIf these mechanistic assumptions are not tion-generating capacities of our minds adequate to explain human behavior then what What picture of the world permits A I reare the capacities that permit people to do searchers in the face of frequent setbacks to the intelligent things that computers cannot remain so confident that computers can reproduce do Dreyfus suggests an attractive alternative human intelligence In studying this question viewpoint based on Gestalt Psychology and the Dreyfus makes some of his most provocative and philosophical school of Phenomenology __ interesting points He eloquently illuminates fields which his arguments should motivate any the philosophical bases of modern science tech- thoughtful reader to examine more closely He nology computer science and industry and remakes a good case for a primary nondigital or veals the assumptions and biases culminating in analog capability which allows the human the methods of A I research Tracing this cast mind to accomplish what it does To paraphrase of thought back to Plato he characterizes it Dreyfus' definitions a digital device repr e thus o what cannot be stated explicitly in sents data by discrete states e g a switch precise instructions -- all areas of human assuming two or more distinct positions and thought which require skill intuition or a gets a result by counting An analog device sense of tradition -- are relegated to some kind uses physical quantities voltage duration of arbitrary fumbling p xvi He finds it angle of rotation of a disc proportional to again in Hobbes' Leviathan When a man reasons the value to be represented then combines he does nothing else but conceive a sum total these quantities physically and measures the from addition of parcels for REASON is result He shows that rnanyneurologists and nothing but reckoning p xvii He conpsychologists lean increasingly toward postucludes that Western thought has already comlating analog devices in the human brain and mitted itself to what would count as an explanaa continuous field-like mode of functioning tion of human behavior oo a theory of practice in human perception and cognition July 75 CRYFTOLOG Page 6 UNCLASSIFIED t t DOCID 4009722 UNClASSIFIED Four maj or nondigital features are presented by Dreyfus as characteristic of and essential to human information processing These are FRINGE CONSCIOUSNESS -- the ability to use the background structure of a gestalt within which objects of our conscious attention are embedded For example a chess player says of his opponent I notice that one of his pieces is not defended the rook In so doing he does not resort to a systematic inventory of all the pieces on the board but perceives a high-level field of relationships all at once at a glance crystallizing around a single piece -- the rook -- to which his eye is drawn AMBIGUITY TOLERANCE or CONTEXT DEPENDENCE -- the ability to deal with ambigu- t ous situations without having to transform them by substituting a precise description People are able to use complex coordinated schemata of knowledge about the world stored and retrieved in some unknown manner to rapidly resolve ambiguity in any given situation ESSENTIAL-INESSENTIAL DISCRIMINATION or INSIGHT -- the ability to select a constantly chan in set of thin s in our co - nitive world as important to us with respect to our current and ever-changing motivations and goals This is in fact what a human programmer does for the computer when he plans a program he prearranges matters so that only those objects and relations needed for the purpose of the program are singled out and described leaving thema chine with a simple formal Class 2 problem to solve PERSPICUOUS GROUPING -- the ability to perceive structured wholes as elements rather than exhaustively and mechanically searching through lists that spell out a higher-level pattern in terms of many atomic traits Psychologists have been learning more about human and animal perception and are demonstrating that we do not build up percepts additively from tiny units such as individual points of light on the retina Instead the visual mechanism works with much more complex elements edges moving horizontal or vertical bars small or large areas of light or dark that expand or contract etc We are capable further of recognizing distorted or varying patterns directly and not by successively and mechanically transforming what we see until it matches a stored template or ideal form We can even recognize a family resemblance among objects even though they may have no explicit single trait in common Dreyfus' last and most interesting point about human information processing concerns July 75 the crucial importance of our bodies and bodily skills the fact that living in our bodies we have built up a motor space An essential difference between metal machines and meat machines as some over-enthusiastic modern reductionists have calledusJ is that we are embodied in a way no digital computer can ever be Thus our experience of a tool we are using is essentially different from our experience of an object the tool is an extension of our body a transparent access to the objects we touch with it Dreyfus suggests that we learn to speak our native language in the very same way -- as a bodily skill and this would certainly explain the relative difficulty of second-language learning which must be accomplished long after our basic body schemata have been formed Thanks to these bodily skills man is at home in his world has it comfortably wrapped around him so to speak Human beings are somehow situated in such a way that what they need in order to cope with things is distributed around them where they need it not packed away in a trunk full of objects or even carefully indexed in a filing cabinet This system of relations which makes it possible to discover objects where they are needed is our home or our world p 172 It is instructive to contrast our experience when we are forced to orient ourselves in a world where we are not yet at home -- learning a new game or skill finding our way about under water orin a foreign city For a while in the new situation we too must operate hesitantly slowly formally and often quite ineffectually like the digital computer Dreyfus is highly pessimistic about the degree of success attainable by A I researchers who base their work upon the set of assumptions he criticizes He concedes that an artificial embodied agent might someday be developed but only if it relied extensively on analog techniques not presently imaginable to us He very briefly considers the possibility of a learning device which might start out with an initial minimal body schema like that of a newborn baby a tiny store of facts and some basic motivations as elementary as the baby's predilection for nipples and smiles It might then build up a respectable world in time under the tutelage of patient human mentors I do not feel that he accords enough serious consideration to this possibil i ty which seems well worth pursuing and in fact will undoubtedly be pursued in some form within the next few years He dismisses it rather arbitraril y in the following words Computers can only deal with facts but man -- the source of facts -- is not a fact or set of facts but a being who creates himself and the world of facts in the process of living in the world There is no reason to suppose that a world organized in terms of these fundamental human capacities should be accessible by other means p 203 CRYPTOLOG UNCLASSIFIED Page 7 -- - - iE'lED '- '-' _ _ '''''''''A ''''''''' __ ''' DOCID 4009722 UNCLASSIFIED The patient reader having stayed with this review this far may be asking himself Why should I care What does all this matter to me in my job or anywhere else First I would like to convince you the reader that it all matters quite a lot to us both as NSA employees and also in a wider context as thinking people caught up in the frenetic and often inhumane activities of a Western industrial society We care at NSA because many of the things we will need to do in the next five or ten years will undoubtedly call for solution of some problems verging on the realm of Artificial Intelligence Our requirements have al ways pushed the state of the art to its limits and they may soon include such things as speech and language understanding optical pattern recognition and the handling of very large organized knowledge bases We are faced with essentially the same barriers of size and complexity as are plaguing A I research ARPA the Advanced Research Projects Agency of the Defense Department is in fact the maj or source of funding for A I research and is at present under attack from several directions because its projects are not considered to be paying off as rapidly as they should be for the expenditures involved The more pragmatic of Dreyfus' criticisms even though they may be dismissed out of hand by A I workers are echoed by others not so easily brushed aside for example J Lighthill 1973 speaking for the Science Research Council in England If enough critics succeed in discouraging ARPA research and A I work in general some of the potential benefits for NSA that might have been around the corner may never materialize One promising spin-off of A I technology possibly within the next five to ten years is a superintelligent terminal based on a number of techniques whose feasibility has been demonstrated by current ARPA-funded projects Such super-terminals might make our jobs a lot easier in the near-future world of computer networks and agglomerations of data bases Finally I feel that aside from the immediate practical feasibility or consequences of A I research all of us should care about the more basic points Dreyfus has made Our socieity bases all of its ways of doing things on the set of assumptions whose limitations Dreyfus so clearly exposes A number of other books have been appearing lately that attack the scientific world-view and its philosophic bases two that I strongly recommend to the brave reader are Roszak 1972 and Schumacher It is easy to dismiss such writers 1973 as irrational romantic impractical or mystical I find myself somewhat to my sorrow agreeing wholeheartedly with most of their fundamental points of view I still cannot accept Dreyfus' pessimistic conclusion that an artificial embodied agent or even as a more immediately attainable goal an intelligent digital assistant is forever impossible I am also not convinced that such an undertaking must necessarily be a bad or perverse nne as Dreyfus obviously feels The criticism Dreyfus and others have raised should certainly not be used as an excuse to shut off A I research neither should they simply be brushed aside Instead we should heed the clear warnings that our world is much too narrow that we are in danger of leaving out all the truly important things and that we must begin empirical explorations of some alternative approaches to human nature and human thought to supplement the digital computer model In any case it seems clear that the thoughtful reader whatever his attitude toward computers and technology can profit by conscientiously exploring Dreyfus' well-reasoned case for What Computers Can't Do Dreyfus Hubert L 1965 Alchemy and Artificial Intelligence RAND Corporation Paper P3244 AD 625 719 December 1965 Lighthill Donald 1973 Artificial Intelligence A General Summary Artificial Intelligence A Paper Symposium Science Research Council Pamphlet Science Research Council State House High Holborn London April 1973 Michie Donald 1974 On Machine Intelligence Wiley New York N Y 1974 Q335 M58 Roszak Theodore 1972 Where the Wasteland Ends Politics and Transcendence in Postindustrial Society Doubleday and Co Inc Garden City N Y 1972 Schumacher E F 1973 Small Is Beautiful Economics As If People Mattered Harper and Row New York N Y 1973 Te _ L -- COMPUTER coNS oUDATED IBM XEROX @ nCIi INC TEKTRONIX TyAS INSB NTS unications81 INCORPORATED July 75 CRYPTOLOG UNCLASSIFIED Page 8 o DOCID L 4009722 TOP 8ECftET U IBftA 1 4 c 86-36 86-36 G 5 At the first meeting of the Crypto-Linguistic Association's newly-establishp-d Special Interest Group on Translation SIGTRAN Ion 22 JanUarrO 9 5Gr m s i P m Free-Lance Translation During the course of his talk he made the following comments The real truth on what you earn as a free-lance translator is not what you get per 1000 words It's how fast you translate If you write out everything in longhand obviously you're not going to make much money Even if you type your translation directly you're not going to make much money This fiscal fact of life plus pressures at home led him to an interesting solution I was forced to mechanize And I soon found that you can double triple and quadruple your translating speed simply by reading into the dictaphone Then he added I have often wondered why people in this Agency who translate things that remain constant P L After getting good enough the typist can even proof for you I know of no instance in which that has been tried I know of a number of specific areas where I believe it would work L- While we do talk and properly so of the problem of having management understand us linguists I don't know of any managers who have as they perhaps should challenged our production as linguists If a linguist produces three translations or eight translations all day the usual manager doesn't know enough about the linguist's problems or procedures so as to be able to make an effective judgment on whether that is a proper output in terms of time e f July 75 Freeing the SIGINT translator from the constraints of a pencil or typewriter was apparently attempted as early as 1946 by an enter- n Ru an bih r o a who dec1 e to 1ctate 1S trans at10nsto a stenographer He estimates that his output doubled the total achieved by the tfaditional method CRYPTOLOG Page 9 tFOP SBEURRBT UMRIU EO 1 4 c P L 86-36 86-36 TOP SECRET UMBRA DOCiD 4009f-2-2 CRYPTOLOG Page TOP SECRET Ul'tIBRA July 75 10 P L 86-36 EO 1 4 c DOCID 4009722 'fOP SECRB'F UMBRA r __ L - July 75 CRYPTOLOG ---- ----------Page 'FOP SBCRB'F UMBRA i P L 86- 3 6 EO 1 4 c DOCID 4009722 'FOP SECRE'f Ul fBRA Io I - -C-Rtyp-TO-LOG-- -p-ag-e-12-- ' - - - - - - - - - - - - - - J U - l - y-7-5 'fOP SEEURRE'I' UMRRA ii P L 86-36 EO 1 4 c J t DOCID 4009722 o NEWS SECREt ' eMI BRIEFS FROM THE At their respective annual banquets held in May the Cryptomathematics Institute and the Cryptolinguistic Association announced the following winners in their 1975 essay contests CMI E6 oay Con tu t 1st Prize $100 --LI ---- ---- R51 Cycles from Nonlinear Shift Registers UNCLASSIFIED 2nd Prize $50 --I G93 Linear Programming Applied to Manual Cryptosystems l'fap SI3EURFlB't EURa13 laF l3 AND eLA Except for Mr Gurj 's article which was published in NSA C yp ologie Spectnum all the prize-winning articles were published in NSA Teehnicai o nal ' P L 86-36 The CLA b st owed i ts T h o o r o Ariri'aClh eydfn ey Jaffe ward uponl J 1 I search and Support ti o r q_a_n_i z a_t_i b_n_ _ I i b t 3rd Prize $25 --I 1 A54 Symmetric Approximations to Boolean Functions ISI3EURM ' o CLA E ay COn tu t 1st Prize $100 -- Doris E Miller P16 Language and the COMINT Production Process fOP 3ECM Y eorJEWeM'J 2nd Prize $50 -- Emery W Tetrault P16 u S Linguists and Language Capability Overseas ISBSFlI3 ' el3flflneM'J 3rd Prize $25 -- Jacob Gurin R54 Examining Some Myths About Language IS13EURFll3'f large Russian language output of the Agency over the past 15 years is inestimable He has insisted on maintaining the highest standards in language-based product in the face of a growing trend toward larger and more rapid production At the same time he has moved to improve the rapidity of language support by mechanizing research files combining them with those of collaborating centers and preparing his own staff to deal with more difficult and more specialized language texts EO 1 4 c P L 86-36 nVEURSa o iCAMINO EN INGLES There lysts to ects ofElectronic Warfare For the off-line IBM-370 files of which we now have a considerable number a special program has been developed to provide a capability of this kind The program retrieves and prints all entries whose English meanings contain one or more of a set of retrieval words specified by the user The program was developed in response to a request for any terms referring to various as- The program can be run against any CAMINO file in the IBM-370 format False hits can in most cases be eliminated by inspection of the full entry and any remaining doubts can be resolved by the file executive Users interested in trying out this English term retrieval capability are invited to contact ---_ PIp x3045s P L 86-36 REMEMBER 15 AUGUSTI t Not necessarily as the anniversary of the founding of Asuncion Paraguay 1537 but definitely as the deadline for submitting your article to CRYPTOLOG for possible inclusion in the special Vietnam Wrap-Up issue to appear in October Send your article to CRYPTOLOG Editor Pl U July 75 CRYPTOLOG S BEURRET Page 13 IlM4BbE VIA eeMI 'f eltAHl4Eb5 SUbY PI-JL 75-53-24000 This document is from the holdings of The National Security Archive Suite 701 Gelman Library The George Washington University 2130 H Street NW Washington D C 20037 Phone 202 994-7000 Fax 202 994-7005 nsarchiv@gwu edu