ARTIFICIAL INTELLIGENCE STEERING GROUP MEETING MINUTES

• Document Type: CREST
• Collection: General CIA Records
• Document Number (FOIA) /ESDN (CREST): CIA-RDP86M00886R000500040011-4
• Release Decision: RIPPUB
• Original Classification: U
• Document Page Count: 54
• Document Creation Date: December 21, 2016
• Document Release Date: September 3, 2008
• Sequence Number: 11
• Publication Date: June 18, 1984
• Content Type: REPORT

Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
,~R OFFICIAL USE ONLY
~,`~ _..;
The Director of Central Intelligence
Intelligence Research &
Development Council
ARTIFICIAL INTELLIGENCE STEERING GROUP
MEETING MINUTES
DATE/PLACE
18 June 1984
NavlT n~nteY _>=or Appl.aed Research
in Artificial Intelligence
Bob Cutter
Jude Franklin
George Forsen
Howard Greyber
Jim Jones
STAT
Mark Macomber
Ken Richhart
Jim Thomas
Bruce Waxman
NEXT MEETING
The next meeting of the AI Steering Group will be held on
Monday, 27 August at 1330 in Dining Room C of the National Bureau of
Standards in Gaithersburg, Maryland. In addition to our regular
business meeting, we will receive a presentation on the Bureau's
Automated Manufacturing Research Facility (machine vision and
robotics).
The American Association for Artificial Intelligence annual
conference will be held 6-10 August 1984 in Austin, Texas. More
information can be obtained from AAAI at [phone].
The Armed Forces Communications and Electronics Association will
hold a symposium on The Role of Knowledge-Based Systems in Command
and Control in Kansas City, Missouri on 17-19 October 1984. More
information is available from the General Chairman Mr. Gary W. Dozier
on [phone].
The First International Workshop on Expert Database Systems
will be held at Kiawah Island, South Carolina on 25-27 October
1984. For more information contact Professor Larry Kerschberg on
[phone].
FOR OFFICIAL USE ONLY ~~~~
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
STAT
STAT
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
An unclassified summary of the FY-84 AI Symposium has been
prepared and is at the printers. Several copies will be sent to
each AISG member in a few weeks. You are asked to pass them along
to interested parties in your organization.
We have begun planning for the FY-85 AI Symposium. We are
targeting the Symposium for March to give more time for
preparation. DIA is considering hosting the Symposium at their
new facility at Bolling Air Force Base. The IC Staff has provided
funds again this year to help with the expenses. This year we
would like greater participation by Government personnel in
discussing actual .:..nt:el~.igencE applications f_or AI~
has agreed to coordinate a technical
seminar on Image Understanding for the Steering Group. The 1/2-1
day session will be held in Rosslyn in the early Fall. You will
be asked to pass the word to other interested parties in your
respective agencies. If you have any suggestions for other
similar seminars which the Steering Group should sponsor, please
contact
The CIA's internal AI Applications Working Group has prepared
a 60-page report on potential applications for AI in the CIA. The
SECRET-level report has been approved by the Agency's Information
Systems Board and copies will be sent to each AI Steering Group
member in the next few weeks (when they are returned from the
printers).
reported that he has received input from
CIA, DIA, DMA, and the three Services for our AISG report. NSA is
still finishing their response on their AI program. No further
action has been taken regarding the draft recommendations which we
discussed at our March meeting. The general sense of the members
present at this meeting was that we should press on to produce a
report in the relatively near future.
Dr. Jude Franklin, NCARAI, mentioned that the FY-84 update to
the Joint Directors of Laboratories report on DoD AI projects will
be out in the next few months. This report typically will exclude
intelligence applications. NRL will also be compiling a library
of all reports on AI ever funded by DoD. A bibliographic index
will be made available over the ARPAnet.
has been giving some thought as to how
to run the propose visory Panel. A strategy paper is
forthcoming. You are requested to submit the names of any of your
cleared contractors whom ou feel would be qualified to serve on
such a panel. an be reached on
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
STAT
STAT
STAT
STAT
SSTAT
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
FOR OFFICIAL USE ONLY
Dr. Jude Franklin is the Director of the Navy's Center for
Applied Research in Artificial Intelligence (NCARAI). The Center
is organizationally attached to the Naval Research Laboratories
and located on Bolling Air Force Base. It was started three years
ago in an effort to bring to fruition 15 years of Navy-sponsored
AI research and to apply it to military problems. Their work is
mostly applied research/exploratory development in providing a
transition from university research to military applications.
Their charter is to establish an AI laboratory of national stature
with an active visiting scientist program throughout the military,
academia, and industry. In 1983 the Joint Directors of
Laboratories endorsed NCARAI as the lead laboratory for military
AI research, and several joint-Service projects have begun.
NCARAI has a few projects under way in three general classes:
Expert Systems, Natural Language, and Distributed Problem
Solving. Dr. Franklin briefly described three of their expert
system projects, and two members of the NCARAI staff described
their work in natural language processing and multi-sensor
fusion. A set of viewgraphs is attached.
The Maintenance and Troubleshooting of Electronic Equipment
Expert System has been developed to provide support for
maintenance personnel in the field and to generate automatic test
equipment (ATE) code automatically. In an effort to resolve the
typical knowledge acquisition bottleneck problem they have
developed a semi-automated rule generator. The system follows a
mixed-initiative approach to diagnose failures in electronic
circuitry and to isolate faults. The system was written in
FRANZLISP and operates on the SUN workstation.
The Combat Management System (BATTLE) was developed for the
Marines to allocate weapons to targets during combat. The system
has been designed as an advisor to the operator and to provide
warnings about the operator's proposed strategy. A particularly
noteworthy point is that by analyzing and improving the heuristics
employed by the system the performance in achieving an allocation
was improved from over 45 minutes to less than 1 second. The
Combat Management System is a general resource allocation system
and NCARAI intends to study applying it to other resource
allocation problems. In Dr. Franklin's view, there is no truely
domain independent expert system available in industry today.
A third project concerns the development of a target
classification system using Inverse Synthetic Aperature Radar
(ISAR). The project attempts to combine AI with existing, more
traditional statistical pattern recognition techniques. The'
prototype system is heavily interactive and leads t]ie operator to
conduct certain signal processing operations. In a recent test
the system identified 84$ of the targets correctly. More tests
under varying imaging conditions are planned.
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
FOR OFFICIAL USE ONLY
Dr. Elaine Marsh described her Message Automation project
which deals with the automated distribution of Casualty Reports by
their content and the generation of message summaries. The
approach uses grammar-based natural language processing techniques
to (1) parse the message narrative fields, (2) regularize the
expressions into a common assertion form, and (3) map the
assertions into a special information format. This format is then
sent to a Dissemination System, written in OPS-5, which uses
production rules to analyze both the formatted and (processed)
narrative fields to determine proper dissemination.
The summarization system operates solely on the narrative
field. It employs three types of rules.: (1) to perform
i:ifei:enca.~ig, (2) to rate or~~:at~ rows a~~cording io their poter~tYal
importance to the summary, and (3) to select the row with the
highest rating to generate the summary. Future plans call for
making the system more robust and using it to control message
entry.
Dr. Y. T. Chen described his project in integrating
multi-sensor information. The application environment is to ~~
support the Tactical Coordinator (TACCO) aboard a submarine-hunting
P3 aircraft. The system addresses sensor utilization, threat
assessment, weapons deployment, situation description, and mission
strategic planning. The knowledge base of the system consists of
object-level knowledge (facts, heuristics, and judgments about
sensors and targets) and meta-level knowledge (which determine the
order in which object-level rules are to be applied). There will be
a hierarchy of sensor specialists, each with a different sensor.
The results of the various experts are then combined by a "Platform
Specialist" which can also direct specific requests back down to the
individual sensor specialist.
Dr. Chen is targeting to finish the knowledge base by the end of
the summer with a demonstration by the end of the year. The system
will be implemented on LMI machines. Dr. Chen has concluded that
OPS-5 does not have an adequate facility for developing a flexible
enough control strategy (meta-rules).
Executive Secretary
AI Steering Group
STAT
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
PRESENTATION TO NAVSEA CORPORATE ~O1~RD
ARTIFICIAL INTELLIGENCE
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
THE IVIEA.NII~JG OF ARTIFICIAL INT i ELLIGEI~TCE
o THERE ARE AS 1V1.~1N~ DE~'II'~IITI~ONS OF "ARTIFICIAL ~-
INTELLIGENCE" AS THERE ARE PRACTITIONERS --
AND NO ONE ACCEPTS ANYONE ELSE'S
WHAT IS Il~/IPORTANT TO KNOW APOUT AI AR]E
THE FEATURES THAT DISTINGUISH IT FROIUI
COI~TVENTIONAL COI'vIPUTER SCIENCE
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
WHAT IS NEW ABOUT AR7'IF~CIAL INTE~~IG~NCE?
~ IT USES COMPUTERS FOP. SY1~iBOLIC PROCESSES,
AI,IN TO MODES OF HUMAN TI30UGHT, RATI-IER
THAN FOR NUMERIC IUJANIPULATION . - . -- :-
* IT SEEKS TO REPRESENT KNOWLEDGE IN THE
COMPUTER IN A I1/iANNER THAT FACILITATES
THESE S~'MBOLIC PROCESSES AND USES KNOW
LEDGE THAT IS OFTEN.-INCOMPLETE AND
SUBJECTIVE
IT STRESSES INTERACTIONS BETWEEN MAN AND
MACHINES THAT ARE COMFORTABLE AND THAT
CREATE CONFIDENCE
~ IT EMPLOYS THESE PP,OCESSES TO PERT ORM_ _
AUTOMATED REASONING
- DRA~'~~ING INFERENCES
- PRUNING DECISION TREES
- SELECTING OPTIONS
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
TYPICAL E LES ?F ARTIFICIAL
' INTELLIGEI~TCE APPLICATIONS
SOLVE INTEGRAL CALCULUS PROBLEMS ~CTSING
HEURISTICS TECHNIQUES
PERFORM MEDIC~IL DIAGNOSIS AND MAKE
PRESCRIPTIONS
- PULMONARY DISEASE (PUFFS)
-INFECTIOUS BLOOD DISEASE (I4~II'CIN)
- INTERNAL MEDICINE (INTERNIST)
INTERPRET GEOLOGICAL SURVEY DATA FOR
NIINERAL AND OII. EXPLORATION
(PROSPECTOfl~) -
ANALYSIS OF MOLECULAR. STRUCTURES BASED
ON NMR, MASS SPECTROSCOPY, AND CHENIICAL
DATA (DEIVDR.~i.)
DESIGN OF COMPUTER_SYSTEMS CONFIGURA-
TIONS TO MEET GIVEN CUSTOMER REQUIRE-
MENTS (R1, X~?Nj
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
I-IOW CAN ARTIFICIAE IT~TTELLIGET~~CE
A. DIFFERENCE TO TI3E N1-~~TY?
APPLICATIONS
A) EXPERT SYSTEMS
DECISION SUPPORT SYSTEMS
TARGET CLASSIFICATION
VISION PROCESSING
B) NATURAL LANGUAGE
- MESSAGE AUTOMATION
- QUERY/RESPONSE FOR LARGE DATA. BASE
- MESSAGE ENTRY SYSTEM
- SPEECH UNDERSTANDING
C) DISTRIBUTED PROBLEM SOLVING - --.
- MULTISENSOR INFORMATION INTEGRATION
FOR:
P3C
DDG 51
SUBACS . '
-AUTONOMOUS VEIIICLES
- Si1~AR.T WEAPONS
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
1E~~~.1 S~S'T'E1S
- CAPTURE KN~VVLEDGE A.Ie~.D _. _.
REASONING PROCESSES USED B~
hC[7MANS II~T COlVIiLEX PROBLEli/~
SOLVING.
C~CTER,I~~D BY KNOV~TLEDGE-
IIVTENS~VE S~OLIC
COIi~iP~TTATI0I~TS.
- APPLIED TO COIi~LEX PR0~3L~MS
NOT WELL-CTITR.~D ~3Y
IvI~.THEM~TICA~ MODELS.
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
EXPERT SYSTEM POR MAINTENANCE AND
TROUBLESHOOTING
PURPOSE -USE AI, ATE AND MAN MACHINE
INTERACTIONS TO GUIDE A
TECHNICIAN IN TROUBLESHOOTING
ELECTRONIC EQUIPMENT
- USE AI EXPERT SYSTEM
TECHNOLOGY TO GENERATE ATE
CODE AUTOMATICALLY
APPROACH -DEVELOP GENERAL AI
TOOLS TO BE APPLIED
TO DIVERSE MILITARY
ELECTRONIC SYSTEMS
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
MOTNATION:
LACI~# OF SKILLED TECHNICIANS
LIlVIITATIONS OF CURRENT ATE
- TIME REQUIRED
- COST OF ACQ~TISIZ'ION
- SPEED OF FAULT ISOLATION
-- RESOLLJTI?N Off' FAULT ISOLATION
- RIG~ITY
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
I~TATU~,AL LANGUAGE SYSTEMS
- CAPTURE THE COl~'IPLEX KNOWLEDGE AND_ __
TECHNIQUES THAT A HUlVIAI~T USES TO
UNDERSTAND SPOKEN OR WRITTEN LANGUAGE
BY:
~ PARSING
~ PERFORMING SYNTACTIC AND SEMANTIC
ANALYSIS
~ UNDERSTANDING THE CONTEXT
- CHARACTERIZED BY THE USE OF LINGUISTIC
KN011~~LEDGE AND COM~'UTATION~ LINGUISTICS
COUPLED WITH WORLD KN01~'~TLEDGE
- APPLIED TO THE .AREAS OF COMPUTER
UNDERSTANDING OF WRITTEN TEXT OR
SPOKEN LANGUAGE
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 REGULARIZE
PARSE
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
~nfoar~aT~nn
f10~ESSAGE E~lTRY SYSTEM
FORMAT
USER
APPLICATION
APPLICATION2
PRO-FOR11/IA EIVTRIf
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
CURRENT SYSTEM STRUCTURE
M ESSAG E
M ESSAG E
ANALYSIS
NARRATIVE
PRO-FORMA
INFORMATION
INFORMMATION
FORMATTING
APPLICATION
~APPLICATION2
FORMATTED
NARRATIVE
INFORMATION FORMATTING
= CONVERSION OF NARRATIVE INTO A STRUCTURED
DATA SASE
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
DISTRIBUTED PROBLEM SOLVING S~'STE~/IS
- CAPTURE T~ KI~TOLEDGE AND REA.SOI~TING ~ ._ --~--
POWERS THAT ARE P,E~UIRED TO I~~~KE
AUTONOMOUS OR SEl~-AUTONOMOUS DECISIONS
- C~[ARACTERI~ED I3Y DISTP,IBUTED NET~~TORKS OF
SEMI-AUTONOMOUS PR,OY~LEI~/1 SOLVING IoIODES THAT
ARE CAPABLE OF COOPERATING WIT~I OTHER
NODS TO SOLVE A SINGLE I'R.OBLEi~
- APPLIED TO AREAS SUCH ~S MULTI-SENSOR FUSION
AND/OR, SOPHISTICATED A.UTOI~TOI~1I?US VEHICLES
S~~7CH AS SMAPLT ~'~EAPOI~TS AND INTELLIGENT
ROBOTS _ ~~ - .
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4 .
CrROUi~ MRf@I&A~Ef\fiE191T !'F~~~LEM
CROSSLINK
~__
BATTLE GROUP.
WITH
CORRELATION
CENTER ~ ,
PROCESSING
SATELLITE
RELAY
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
RECCE
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
NAVY AI PROGRAII~IS ARE AIl'~~ED AT SOLVING PROBLEMS
1) INCREASE READINESS AND REDUCE COST
- EXPERT SYSTEM TO MAINTAIN ELECTRONIC
EQUIPMENT AND TO GENERATE ATE CODE
AUTOMATICALLY (NAVAIR)
- DECISION SUPPORT SYSTEMS FOR COMBAT
MANAGEMENT (MAR.INE CORPS)
- TRAINING (JDL)
2~ REDUCE ~iANPO~ER REQUIREIvIENTS AND OPERATOR
OVERLOAD
- NATURAL LANGUAGE PROCESSING TO AiJTO~~iATE
MESSAGE IiANDLING AND TO REDUCE ER.P.OJR.S
(3M)
- MULTISENSOR INFORMATION INTEGRATION, DISPLAY
AND INTERPRETATION
- EXPERT SYSTEM FOR OPERATIONAL PLANNING
- SPEECH UNDERSTANDING (NAVAIR-DARPA
STII.ATEGIC COMPUTING INITIATIVE)
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
CRITICAL AI TECHNOLOGIES
EXPERT SYSTEMS
A) AUTOMATED KNOWLEDGE ACQUISITION
SYSTEMS
B) ADAPTIVE LEARNING SYSTEMS
C) INFERENCE METHODS FOR TIME CRITICAL
SITUATIONS
NATURAL LANGUAGE
A) HIGH SPEED PARSER
B) VERSATILE G
C) IMPROVED SPEECH UNDERSTANDING
SYSTEMS
DISTRIBUTED PROBLEM SOLVING
A) AUTONOMOUS PROBLEM SOLVING
B) ADAPTIVE SEARCH TECHNIQUES
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
I~~lAAT~lI~lp4INT:4,UGENCB '
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
OUTLINE
1) OVERVIEW OF NCAI-CAI
- PP~OJECTS
2) TECHNICAL DISCUSSION OF PROJECTS
- MAINTENANCE AND TROUBLESHOOTING
- vIULTISENSOR INFORMATION INTEGRATION AND ISAR CLASSIFIER
3) POTENTIAL NAVELEX/NRL PROGRAMS
- DIAGNOSTICS FOR MAINTENANCE AND TROUBLESHOOTING
- SURVEILLANCE
- TARGET CLASSIFICATION (ACOUSTICS AND NON ACOUSTICS)
- BATTLE GROUP LEVEL MANAGEMENT AND ASSESSMENT
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
C~~~S~S ?~ / U Cv"t~A~ ~R?4J I.oCT`L~
NATURAL
DISTRIBUTED PROBLEM
E3;PERT S~'STEMS
LANGUAGE
SOLVING
MAINTENANCE AND
MESSAGE. ~
MULTISENSOR
TROUBLES(~OOTING
AUTOMATION
FUSION
COMBAT
ADAPTIVE
IV(ANAGEiViENT
CONTROL
TARGET
CL~-lSS1FiCATIOiU
~ ~.
OPERB~1'#ONAL
PLAN(~ING
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
ti -''t
..t~eie~
~6g1111(Ill(Illfl(1(111((111111((((i~(I(I
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
L~~'EI-L'.~ ~~ST~' l~~ I''~I~. I~11I.t`~:~I'~T'~`~T~- ~I`~T ~" ~ .A~
TI~0~1~3LES~#~~1'INC
PUP.P?SE - .USE , ~.TE;!~.~ I~v~N?i~~~I.~CI-II~1~E
~~ ~ ~i VT~ 3.,1 ~.`ir~ ai.~~.~ T ~' ~-~Ti3~3~~~'.' ~.
TECI~NTCI~.1~T Il'~ TI~~~~LES~IC~'T'INC
]ELECT~CI~C E~UIPVIENT
- USE ~1[ E~PE~,T S~STEI~~I
TECI-IIOT~L?C~ T~ GI~~NE~~T]~ A.TE
C~1~E ~~JT~TIC~~.LL~
~.PP~.?~CI-I - LE~-~L?P CEI~E~,~L ~
T??LS '~'~ ~E APPLIED
T? DI~TEP.SE II~~LIT~E',Y
ELECT~~i'~IC SYSTE~~IS
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
MOTNATI~N:
- LACK ?F S~[LLED TFC~NICIANS
-~..L1iV~~'~4.i~~~~ ~~' C V R.~~'N'~~ ~~-.T~
- T~VIE RE~~TIRED
- C?ST ?F ACQ~J~S~T~CI~I
- SFEED ?F FA~TLT IS ~3L~TIC~N
- RCS?L~JT~~l'~T ?~' F~IJL'~C~ ISCLA.T~CN
- RIC~ITY
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
~EC~I~ICAI.~ ISS~J~S:
- ]EIJ?I~ ~ I~~~I~I~T~T
TE~~I~T?~~~Y
- AVOID Kl~TO~,ED~GE ~CQU~~I~IOl~I
B~JTTLENECI{
- DEL?IT E~~TIT~~ l-~.~~ ~~
- PF~.O~DE ~I~H Q~I~I.,I~~-USER
INTERFACE
~.IJT?I~~~IC~I.~L~
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
~~ ~,~~E COfVif-1LER
RULE
~EI~Ei~ATOR
PRIOR NfO?EL OF UNIT
Ufa?ER TEST
(M/~NUAL~~IE~PERTS)
~RlITIAL R~JLE BASE
(E~~'E"RT)
1~(~OV~ILED~E
6~~E
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
~P~e~~a~?~oc ~v~rEn_n
TEST
RESULT
ICIVOWLEDGE
BASE
INFERENCE
ENGINE
UPDATES iVIODEL 1
CONDITION OF UNIT'
' EVIDENCE OFD
HONING FOUND
A FAULTY
~COIVtI'ONEN1',
WHAT AND HOW
TO TESTIREPAiR
TEST
TECHNICIAN
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 1~1UX IN
Of4P .VIDEO IN ._.~
OOE SELECT ~-~
~,q00 }Iz
V AC
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
SEIf TEST
AND
BIT
SYNC SEP
ANO
RASTER
GEN
EI.1I
FRTER
LIGHTING
PAi~~EL
DISPLAY
ELECT(t0~~}ICS
ASSCIIUL.Y
X, Y, Z, ~IUPS, CRT
OUTPUT SEPISE
AND DISPLAY ELECTRONICS
ASSES-1BLY
Figure 6. Converter-pisplay - Alock Diagram
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
FEATURES OF MAI\'TEi~?Al\TCE A1VD TROUBLESI-i00TiNG
EXPERT SYSTEM
.1) 1-fL~D I1~ITIATIVE
2) AI\'ALOG OR DIGITAL
3) COST OF TEST, SET-Ut AI~TD RISK ARE TAKEN
INTO ACCOUNT
4) DYNAV1IC N10DEL OF BELIEFS
- EVIDENTIAL REASOI\~ING
- NO\T-VIONOTOI~~IC REASOl~'ING
- ADf1PTIVE TROUBLESHOOTING TREE
- FORWARD A1VD B~-1CICtiVARD REA~OI\~ING
5) HEURISTIC SEARCH
G) NEXT BEST TEST
7) FUNCTIONALITY BEING ADDED
8) LEARNING FEATURE TO BE ADD ~ D
- STATISTICS
- AUTOVIAI'IC DEDUCTION OF ~?UA.LITATIVE RULES
- CAD CAM
0) AUTOYIATIC MULE GENERATION
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
A TWO-DIMENSIONAL RADAR IMAGE CONSTRUCTED FROM
DOPPLER RETURNS CAUSED BY THE iiri0TI0N 01=' THE TARGET
POORER QUALITY THAN A VISUAL IMAGE
? NOISY
~ REFLECTION INTENSITIES VARY GREATLY,
EVEN FROtVI SIMILAR VIEWING ANGLES
m SHIP ORIENTATION AND CROSS-RANGE SCALE
IN THE INTAGE ARE UNKNOWN
THIS IMAGE RECOGNITION PROBLEM IS NOT WELL-SUITED FOR
TRADITIONAL STATISTICAL PATTERN RECOGNI'T'ION TECHNIQUES
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
UNCLASSIFIED
Rd~iVGE/DO~PL~R PITCH CONCEPT
~~
`a
DOPPLER
DOPPLER MEASURED BY RADAR
PROPORTIONAL TO HEIGHT
UNCLASSIFIED
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
~IP~i~?~ ~~ ~IE~~~t~
ALLOW FOR AUTOMATIC PROCESSING W{~EN PRACTICAL
AN? APPROPRIATE
? M,AI~E USE OF TWE ?PERATOR'S JUDGEMEf~TS AND
PAT~'EI~N RECOG~lITION CAP,~6ILITIES
~ CAPTURE ThIE KNOWLEDGE AND REASONING /ROUT
ALTERNATIVE CLASSiI=ICATI?NS T~tAT IS Ci?IARACTERISTiC
OE ~,~1 EXPERT ,ANALYST
~ Si-rOW PROMISE 01=NEAR-TERM TRANSITION TO Ti~E FLEET
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
TESTED ONLY ON I-IiGI~ C~UALITY, h11GW RESOLUTION
! f~/i ~ G ES
USES ONLY A SINGLE STATIC FRAME OF II~~AGERY
0 CAN IT DlSTlNGUISW AMONG SIMILAR SWIR CLASSES?
? DOES NOT ~4LL01lV FOR USER QUERIES, OBSERVATIONS,
EYC.
IS 11' l~RACTICAL 1=?R TtrIEO~ERA1'IOI~oIAL N,~,VY?
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
~.~ ~~~~,~~ ~li,~~~o~~~~~~~rv s~r~~~~~
i ~,~ ~
i~~AGC
f~EA~Ui~Ei'~iEf~T
COi~RELATi~~I
Li~T OF
~~ ~~ CAIV D I DATE
chili' CLA~~ES
~ t~~
~i U i~ ~ ~
E~ ~ E ~A~"E~ i~
E~PEi~T
~~~Tl~ ~ 4
~U~JECTi~JE
?E~Ci~ii~T'i?i~i~~
iii?~1' Lii`ELY
CLA~~iF1CATi?~
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
~,'~e~e~'?~s ~'a-o~ ~~ ~~~~~,~ ~~~.~s~~~a.tzarl trial
islraid -Thu Dec 1 13:16:57 Ifl83
~ti'elcome to the ISr'~. image interpretation aid
(NCf1t~.AI)
- ---T::I~4:C'r %o:nm7cd:.,,,.~gri+~L.
>EntE:r the image name or t: imaoa8~
>Eutc:r the class name or t: * '~
Do you wish to see the results of each question ? ~
fo u;hr,:f degree do you believe that
(+ profr'le-has-isolaled-directors fwd-of bridge-with-a-li~aear-taper) ? -3
I neithf:r suspect uor doubt the possibility of ForestSherman, Bainbridge,
Sverdlov, Coontz, Leahy, Belknap, Truxtun, California or Virginia
I doubt. that imaoe84 could be LongBeach
To what degree do you believe that
(~ proj'le-leas-a-gap-between-aJt-mast-and-aft-superstructure-block ? 5
I strongl}? suspect that image84 is Coontz
I doubt that irrlage&~ could be ForestSherman, Sverdlov, Leahy, F3elknap,
'Truxtun, California or Virginia
I strongly doubt that image34 is Bainbridge or LongBeach
To what dcgre.e do you bel:eve that
(3 profile-Iris-isolrled-di; ecfors-on-aft-superstructure-block ? 5
-----------------------
I strongly suspect that irnaGeS~ is Coontz
I sli~ictl~? doubt that ima~cS~1 could be California
I drnibL that ima~c5t could he Leahy, Belknap, 'Truxtun or Virgil-pia
I stront;l}~ dr~ul,t tl-:c? 1i71~t~;C~~ IS horestSlierrnan, Bainbridge, S~-crdlo~:
or l,c~ri~I;cacli
-----------------------
7?o a:/cat degree rlo you belie2~e that
(r ;,~alile->ia.st-heights-core-crtuai~ ? -~
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
I strongly suspect that image84 is Coontz
I slightly doubt that image84 could be California
I doubt that image84 could be Leahy
I strongly doubt that image84 is ForestSherman, Bainbridge, Sverdlov,
LongBeach, Belknap, Truxtun or Virginia
To what degree do you believe that
(~ profile-aJt-mast-is-on-top-of-aft-superstructure-block) ? -b
i S;trL`Rgi~..SLL'SS cCi tha? SRri.'nTE'.$4 L5 ~:oO71tT,
I doubt that image84 could be Leahy
I strongly doubt that image84 is ForestSherman, Bainbridge, Sverdlov,
LongBeach, Belknap, Truxtun, California or Virginia
To what degree do you believe that
(+~ profzle-deck-is-level-with-no-o,~J'sets) ;' ~
I :strongly suspect that image84 is Coontz
I :strongly doubt that image84 is ForestSherman, Bainbrido~, Sverdlov,
LongBeach, Leahy, Belknap, Truxtun, California or Virginia
No more ask able questions about image84
~~* Summary of hypotheses about *~X
image84
Coontz
4.999935
California
-4.841409
Leahy
-5.000000
Truxtun
-5.000000
Belr,:nap
-5.000000
Virginia
-5.000000
Sverdlov
-5.000000
Bai~abridgc
-5.000000
ForestSherman
-5.000000
LongBeach
-5.000000
I st.mngly suspect that imageS~t is Coontz
I .strongly doubt that image&~ is ForestSherman, Bainbridge, Sverdlov,
I.,ongl3eac}r, I.,eahy, Belkn:r.p, Truxtun, California. or Virginia
-----------------------
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
~~~'~~~~
~ ~d~~Ll'U~IV~~ T 1 ~w~ 1 ~IJ
101 TOTAL If~~AGES OF THE CHOSEN SWIP CLASSES
52 PLAN VIEWS
49 PROFILE VIEWS
18 it~11AGES OF SWIPS NOT iN THE DATA BASE
lid 8~ OF TF~E 101 TRIALS, TFiE SI~iP CL~,SS RANICED 1st
BV TWE EXPERT SYSTEIZfi WAS TWE CORRECT' CLA SSIFiCATION
? ~0~~~~~TE~7'L~ ~l~G~l~L L~I~CR1~1~1AT'i0~;1~
CORRECT TRIALS:
AVG CONFIDENCE IN TOP RANKED CLASS 1,G92
AVG CONFIDENCE iN 2nd ~~+NI~ED CLASS --=2.383
T'i-IIS f~/iEANS THE SYSTEiUr USUALLY COIEfIES UP'1PifITH
Of~LV ONE P LAUSiBLE CLASSIFICATION
INCORRECT TRIALS:
AVG CONFIDENCE 1N TOP RAi~IKED CLASS -0.1G1
AVG CONFIDENCE IN 2nd RANKED CLASS --~Q.990
A FEW CLASSES, USUALLY iNCLUD[~~tG THE CORRECT
ONE, .ARE CLUSTERED TOGETFIER AS ALTERNATIVE
CLASSI Fi CATI O NS
Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
 Approved For Release 2008/09/04 :CIA-RDP86M00886R000500040011-4
PtJf~PO~E -
INVESTIGATE AND DEMONSTRATE HOW ARTIFICIAL INTELLIGENCE tAl)
TECHNIQUES CAN BE USED TO IMPROVE THE INTEGRATION OF INFORMA-
TION FROM DIVERSE SENSORS IN SUPPORT OF NAVY COM~/IAND Ft CON-
TROLOPERATIONS IN ELECTRONIC WARFARE (60TH AT THE PLATFORM AND
THE BATTLE GROUP LEVEL)
P~~~~/A~1T AI TECWI~iC~UC~ -
EXPERT SYSTEMS DEVELOPMENT -CAPTURING AND EMULATING
DECISION-MAI