Pate? 10 APR 1975 2139=PDT

From: LEDERBERG at USCIS]

Subject: ARPA RESEARCH FOLICY,

To3 LICKLIDER

ec?! FETGENRAUM at SUMEX=AIM, LEDERBERG at SUMEX@#AIM

DEAR LICK:

EL FETGENBAUM HAS SHARED YOUR 'EASTER MESSAGE' AND LATER APTICULATIONS
OF CONCERN WITH ME, I HAVE BEEN GIVING SERIOUS THOUGHT TO MANY OF THE
POINTS YOU RAISE, AND WILL BE DISCUSSING THEM FURTHER WITH ED AS A PRE@
LIMINARY AKL A VEHICLE TC REPLY TO YOU AND TO THE ISSUES YOU RAISE,

MEANWHILE T HAC ONE CONCRETE 6UT SMALL GUESTION, AND TWO METAe
GUESTIONS THAT IT wILL BE PURSUING FURTHER,

1, OBVIOUSLY THE IMPLICATIONS OF THE INSTANT HARDWARE CRITERION FOR
ARPA RESEARCH POLICY ARE TO CISCQUNT ANY SIGNIFICANT CONTRIBUTION OF ACA»
CEMICALLY RASED RESEARCH FOR QUR LONG RANGE DEFENSE NEEDS, MY PERSONAL
RESEARCH PROGRAM BENFFITS FROM APPA SUPPORT; BUT I WILL CERTAINLY FIND
SOME KAY TO SURVIVE WITHOUT IT, I 4M MUCH MORE PESSIMISTIC ABOUT THE
LONG RANGE SURVIVAL OF MY COUNTRY IN A WORLD OF HARSH TECHNOLOGICAL=
MILITARY COMPETITION IF THIS POLICY IS SYSTEMATICALLY PURSUED, JT AM
SURE YOU CAN FIND MANY MORE EXAMPLES THAN THOSE OF WHICH I AM PERSONALLY
AWARE THAT WOULD BEAR OuT THE VERY HIGH COSTS SF AN INSTANT HARDWARE
POLICY; AND PERHAPS IT IS FVEN OUT OF PLACE FOR ME TO MENTION SUCH
EMBARRASSMENTS, BUT IF I THINK OF THE HECTOMILLIONS THAT WERE WASTED

BY THE PURSUIT OF THIS PHILOSOPHY, E,G,, IN THE HI@®POWER LASER CATEGORY
THAT SHOULD RE SUFFICIENT REMINDER,

THIS IS NOT TO SAY THAT EVERYTHING AROUT PAST POLICIES was
IDEAL == MUCH GF THE PROCEDURAL RECOMMENDATIONS OF THE BERLEKAMP AMC
CUMMITTEE WAS DIRECTED TO IMPROVING THE SCIENTIFIC aNO TECHNICAL YIELD
FROM ARPA'TS INVESTMENTS, AND YOU WILL RECALL THAT I DID NOT BELIEVE
THAT APPA WAS THE APPROPRIATE AGENCY TO SUPPORT WORK IN FIELDS LIKE
NEUROBIOLOGY ALONG LINES THAT SEEMED TOO REMOTE FROM ARPA'S MISSION,
AND TO BE QUITE BLUNT. IT WAS RATHER UPSET BY LUKASIK'S ENTHUSIASM FOR
A RESEARCH PROGRAR (PINNEQ@SRI) THAT WAS ADVERTISED IN THE LOCAL PRESS
AS DIRECTED TQ THE READING OF A PERSON'S INNERMOST THOUGHTS; AND I GOT
NOWHERE [N MY CURRESPONDENCE WITH LUKASIK ON THIS POINT,

Qs WELL TRE PREVIOUS waS A MAJOR META@ISSUE, HERE NOW IS THE SMALL
ONES JUST WHAT IS THE STATUS OF THE TECHNOLOGY (AND OF THE LEVEL OF
THE MILITAERY REQUIREMENT) FOR THE PROBLEM YOU MENTIONED OF EXTRACTING
INFORMATION FROM MORSE CODE IN 4 NOTSY CONTEXT? JI SHOULD HAVE THOUGHT
THAT WAS CRACKED AT LEAST & MECADE &£G03 BUT INDEED I AM LUNAWARE OF

ANY SPECIFICS, TF IT IS STILL AN OPEN PROBLEM IT IS OBVIOUSLY A SIG
 SLEDERBENGOIL/LICK, TXTei SAT {2"APR=75 Bt 15am

NIFICANT STEPPING STONE THAT SHOLLD PRECEDE ANALOGOUS EFFORTS IN A)
SPEECH UNDERSTANDING, AND R)THE SONAR PROBLEM, DO YOU HAVE ANY REFERENCES
OR OTHER SCUPCES OF AMPLIFICATION?

3, M2 RATSES THE SECOND METS@ISSUEs PLAINLY WE NEED BETTER MACHINERY
THAN NOW EXISTS FOR BRINGING ABOUT A BETTER UNDERSTAND OF WHAT ARE REA.
MILITARY REQUIREMENTS IN RELATION TO TRE OPPORTUNITIES GF INFORMATION
SCIENCES RESEARCH, THE REQUIREMENTS AND OPPORTUNITIES CAN NOT REALLY HE
EXPLORED IN MUTUAL ISOLATION: ONE CAN HARDLY KNOW WHAT IS 'REQUIRED!
WITHOUT AN INFORMED PERSPECTIVE ON WHAT IS POSSIBLE,
$o

HOW DIO YOU GENERATE YOUR LIST? IS YOUR ALGORITHM AN OPTIMAL ONE
IN PROPORTIUN TQ THE SIGNIFICANCE OF THE CHALLENGE IT ADDRESSES?
4, AN AFTERTHOUGHT: I wILL BE SAYING THIS AGAIN: ONE OF THE REASONS
THAT I DECIDED TO COMMIT A SIGNIFICANT PART OF MY FURTHER CAREER TO COMe
PUTER INTELLIGENCE wAS MY ANYTETY WHETHER UNAIDED HUMAN INTELLIGENCE WAS
UP TO THE TASK OF ASSURING OUR SURVIVAL THROUGH THE NEXT SO YEARS OF
NUCLEAR@WEAPONS COMPETITION, THE PLACE THAT I FORESEE A,I, HAVING ITS
MOST CRUCIAL ROLE IN DEFENSE IS AT THE TOPMOST LEVEL =#TO HELP ASSURE
THAT THE FINGER THAT HAS 15 MINUTES TO DECIDE ABOUT THE RED BUTTON WILL
BE ADEQUATELY INFORMED ABOUT ALTERNATIVE DECISIONAL OPTIONS, I DON!T
HAVE TO TELL YOU BROw FAR WE ARE FROM A SITUATION WHERE WE WOULD PERMIT
ANY A,I, PROGRAM TO HAVE A SIGNIFICANT PART IN SUCH DECISIONS, BUT I
DO NOT THINK TT TS ARROGANT TO SUGGEST TKAT A TECHNOLOGY THAT CAN BRANCH
THROUGH A KNOKLEDGE# AND DATA#ORTENTED DECISIONAL TREE LIKE DENDRAL DOES
HELP CORROBORATE THE EVENTUAL FEASIBILITY OF SIGNIFICANT HELP TO TOP}
LEVEL DECISIGN=MAKERS FACING MUCH MORE COMPLEX PROBLEMS,

I AM DEPRESSED AT THE TROUGKT THAT THIS PROBLEM WILL BE NISMISSED
AS TOO VISTONARY OR TOM LONG RANGE TO BE OF INTEREST TO ARPA, AND I WELL
KNOW THAT WE SHARE COMMON OSSTACLES IN PERSUADING OTHERS OF THESE VIEWS,
BUT IF I NID NOT EXPRESS THEM TO YOU YOU WOULD BE LESS INFORMED AT
LEAST OF MY QWN CONVICTION ABOUT THEM,

YOURS

JOSH

(PLEASE REPLY PREFERABLY TO LEDERBERG@SUMEX: WE ARE DOWN FOR PM JUST
NOW, PS OUR IMP HOOKUP SEEMS TO BE WORKING VERY WELLESS1S34)