Carl von Clausewitz, the Fog-of-War, and the AI Revolution: The Real World Is Not A Game Of Go

112 Pages·2018·2.33 MB·English

by  Rodrick Wallace (auth.)

SPRINGER BRIEFS IN APPLIED SCIENCES AND
TECHNOLOGY  COMPUTATIONAL INTELLIGENCE
Rodrick Wallace
Carl von
Clausewitz, the
Fog-of-War, and
the AI Revolution
The Real World Is
Not A Game Of Go
123
SpringerBriefs in Applied Sciences
and Technology
Computational Intelligence
Series editor
Janusz Kacprzyk, Polish Academy of Sciences, Systems Research Institute,
Warsaw, Poland
The series “Studies in Computational Intelligence” (SCI) publishes new develop-
mentsandadvancesinthevariousareasofcomputationalintelligence—quicklyand
with a high quality. The intent is to cover the theory, applications, and design
methods of computational intelligence, as embedded in the fields of engineering,
computer science, physics and life sciences, as well as the methodologies behind
them. The series contains monographs, lecture notes and edited volumes in
computational intelligence spanning the areas of neural networks, connectionist
systems, genetic algorithms, evolutionary computation, artificial intelligence,
cellular automata, self-organizing systems, soft computing, fuzzy systems, and
hybrid intelligent systems. Of particular value to both the contributors and the
readership are the short publication timeframe and the world-wide distribution,
which enable both wide and rapid dissemination of research output.
More information about this series at http://www.springer.com/series/10618
Rodrick Wallace
Carl von Clausewitz,
the Fog-of-War, and the AI
Revolution
The Real World Is Not A Game Of Go
123
RodrickWallace
Division of Epidemiology
TheNew YorkState Psychiatric Institute
NewYork,NY
USA
ISSN 2191-530X ISSN 2191-5318 (electronic)
SpringerBriefs inApplied SciencesandTechnology
ISSN 2520-8551 ISSN 2520-856X (electronic)
SpringerBriefs inComputational Intelligence
ISBN978-3-319-74632-6 ISBN978-3-319-74633-3 (eBook)
https://doi.org/10.1007/978-3-319-74633-3
LibraryofCongressControlNumber:2017964243
© The Author(s), under exclusive licence to Springer International Publishing AG, part of Springer
Nature2018
Thisworkissubjecttocopyright.AllrightsarereservedbythePublisher,whetherthewholeorpart
of the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations,
recitation, broadcasting, reproduction on microfilms or in any other physical way, and transmission
orinformationstorageandretrieval,electronicadaptation,computersoftware,orbysimilarordissimilar
methodologynowknownorhereafterdeveloped.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this
publicationdoesnotimply,evenintheabsenceofaspecificstatement,thatsuchnamesareexemptfrom
therelevantprotectivelawsandregulationsandthereforefreeforgeneraluse.
The publisher, the authors and the editors are safe to assume that the advice and information in this
book are believed to be true and accurate at the date of publication. Neither the publisher nor the
authorsortheeditorsgiveawarranty,expressorimplied,withrespecttothematerialcontainedhereinor
for any errors or omissions that may have been made. The publisher remains neutral with regard to
jurisdictionalclaimsinpublishedmapsandinstitutionalaffiliations.
Printedonacid-freepaper
ThisSpringerimprintispublishedbytheregisteredcompanySpringerInternationalPublishingAGpart
ofSpringerNature
Theregisteredcompanyaddressis:Gewerbestrasse11,6330Cham,Switzerland
Preface
Corporateinterests andtheiracademicclients nowclaim thatartificialintelligence,
via recent advances in deep learning and related technologies, is ready to take on
management of critical real-time processes ranging from driverless cars on intel-
ligentroadstotheconductofwar.Inthepast,corporateinterestshavealsoclaimed
thatsmokingisharmless,environmentalcontaminationunimportant,faultyairbags
are safe, polyvinyl chloride furnishings and finishings in fires are no more dan-
gerous than wood, and made any number of other assertions that, in the long run,
have caused massive human suffering. In many cases, aggressive marketing by
thoseinterestswasabletobuildedifices“toobigtofail,toobigtojail”,privatizing
profits while socializing costs. Corporate AI advocates for driverless cars and
autonomous weapons stand on the verge of creating such conditions for their
products. Absent intervention, others will follow.
Thecentralthesisofthismonographisthatcognitivealgorithmicentitiestasked
with the real-time management of critical processes under rapidly shifting “road-
way” conditions will face many of the same conundrums and constraints that
confront the conduct of warfare and other forms of conflict. As with conventional
trafficflow,suchroadwaysneednotbepassive,butmayengageoremployentities
having their own purposes, mechanisms, and cognitive abilities. These may range
across biological, individual, social, institutional, machine, and/or hybrid mani-
festationsanddynamics,fromcancer,murder,andneoliberalcapitalism,toCentaur
or autonomous battlespaces.
From the Somme and Passchendaele, to Blitzkrieg madness and Cold War
preparations for human extinction, Vietnam, and the current Middle Eastern
bludgeonings, the art and science of warfare has been singularly unable to cope
with what the military theorist Carl von Clausewitz characterized as the
“fog-of-war”and“friction”inevitabletohumanconflict.Weargueherethat,inthe
realworld,ArtificialIntelligencewillfacesimilarchallengeswithsimilarorgreater
ineptitude. The biblical injunction not to puttrust in thechariots of Egypt is likely
to take new meaning over the next century.
v
vi Preface
More specifically, the monograph’s first chapter shows how straightforward
arguments from control and information theories imply that emergence of the AI
revolution from games of Chess and Go into the real world will fatally encounter
the central matters of the Clausewitz analysis. Promises of graceful degradation
under stress for large numbers of driverless vehicles on intelligent roads, of pre-
cision targeting that avoids civilian collateral damage for autonomous or so-called
man/machine centaur weapons, of precision medicine under even normal living
condition, let alone during the current slow disasters of climate change and social
decay, of the ability to manage financial crises in real time with agent-based
models, and so on, are delusive groupthink or marketing hype that will be
beta-tested onhuman populations,agross contraventionoffundamentalmoraland
legal norms.
The second chapter extends the model to nonergodic cognitive systems, a par-
allel to the nonparametric extension of more familiar statistical models.
This requires some comment.
Cognition—biological, social, institutional, machine, or composite—most sin-
gularly involves choice that reduces uncertainty. Reduction of uncertainty implies
the existence of an information source dual to the cognitive process under study.
However, information source uncertainty for path-dependent nonergodic systems
cannotbedescribedasaconventionalShannonentropysincetimeaveragesarenot
ensemble averages. Nonetheless, the essential nature of information as a form of
freeenergyallowsstudyofnonergodiccognitivesystemshavingcomplexdynamic
topologies whose algebraic expression is in terms of directed homotopy groupoids
rather than groups. This permits a significant extension of the data rate theorem
linkingcontrolandinformationtheoriesviaananalogtothespontaneoussymmetry
breaking arguments fundamental to modern physics. In addition, the identification
of information as a form offree energy enables construction of dynamic empirical
Onsager models in the gradient of a classic entropy that can be built from the
Legendre transform of even path-dependent information source uncertainties. The
methodologyprovidesnewanalytictoolsthatshouldproveusefulinunderstanding
failuremodesandtheirdynamicsacrossabroadspectrumofcognitivephenomena,
ranging from physiological processes at different scales and levels of organization
to critical system automata and institutional economics.
The third chapter provides a worked-out example, making a schematic appli-
cationofthetheorytopassengercrowdingonvehicle-to-infrastructure(V2I)public
transitsystemsinwhichbusesorsubwaysbecomesocrowdedthattheyareordered
by a central control to begin a degraded “skip-stop” service. D. Wallace and R.
Wallace (1998) examine such “skip-stop” dynamics for fire service in New York
City, a policy called “fallback” in which increasing demand was met by a pro-
grammed decline in the dispatch of emergency equipment. The results, for the
Bronx, Central Harlem, and so on, were spectacularly catastrophic during the
1970s.
Preface vii
The fourth chapter provides another case history, examining how failure of the
dynamicsofcrosstalkbetween“tactical”and“strategic”levelsoforganizationwill
leadtoanotherversionoftheJohnBoydmechanismofcommandfailure:therules
of the game change faster than executive systems can respond.
The fifth chapter comes full circle, applying the theory explicitly to military
systems. Here, the powerful asymptotic limit theorems of control and information
theories particularly illuminate target discrimination failures afflicting autonomous
weapon, man/machine centaur or cockpit, and more traditional structures under
increasing fog-of-war and friction burdens. Argument indicates that degradation in
targeting precision by high-level cognitive entities under escalating uncertainty,
operational difficulty, attrition, and real-time demands will almost always involve
sudden collapse to an all too familiar pathological state in which “all possible
targetsareenemies”,otherwiseknownas“killeveryoneandletGodsortthemout”.
The sixth chapter examines real-time critical processes on a longer timescale,
through an evolutionary lens. The basic finding is that protracted conflict between
cognitiveentitiescantriggeraself-referential,coevolutionarybootstrapdynamic,in
essence a “language that speaks itself”. Such phenomena do not permit simple
command-loopinterventionsinJohnBoyd’ssenseandareveryhardtocontain.An
example might be found in the evolutionary transformation of the Soviet Union’s
military forces, tactics, and strategy in the face of German Bewegungskrieg from
the battles of Moscow to Stalingrad, and then Kursk, and in the “insurgency” that
followed the 2003 tactical successes of the US in Iraq. Another example can be
found in the systematic resistance of the defeated Confederate states after the US
Civil War that resulted in the withdrawal of US troops and the end of
Reconstruction in 1877, permitting imposition of the Jim Crow system of racial
segregationandvotersuppressionthatlastedwellintothelatterhalfofthetwentieth
century.
The final chapter sums up the argument: Caveat Venditor, Caveat Emptor.
Some explicit comment on methodology is in order. The basic approach is
through the asymptotic limit theorems of information and control theories, leading
tostatistical models that,likeregressionequations, aretobefittedtoobservational
or experimental data. The essential questions do not, then, revolve around the
pseudoscientific manipulation of metaphors abducted from “nonlinear science”, as
devastatingly critiqued by Lawson (2014), but rather on how well these statistical
models work in practice. Mathematical models that surround, or arise from, the
development of these tools should be viewed in the sense of the theoretical ecol-
ogist E. C. Pielou (1977) as generating conjectures that are to be tested by the
analysis of observational and experimental data: the word is never the thing.
The author thanks Barry Watts and a number of anonymous commentators for
suggestions and differences of opinion useful in revision.
New York, USA Rodrick Wallace
viii Preface
References
Lawson, S.,2014.Non-LinearScienceandWarfare:Chaos,complexityandtheUSmilitaryinthe
informationage.NewYork:Routledge.
Pielou, E.C.,1977.MathematicalEcology.NewYork:JohnWileyandSons.
Wallace, D.,Wallace,R.,1998.APlagueonYourHouses.NewYork:Verso.
Contents
1 AI in the Real World . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
1.2 The Data Rate Theorem . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
1.3 The ‘Clausewitz Temperature’. . . . . . . . . . . . . . . . . . . . . . . . . . 5
1.4 A Bottleneck Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
1.5 Dynamics of Control Failure . . . . . . . . . . . . . . . . . . . . . . . . . . . 7
1.6 The Failure of Cognition. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
1.7 No Free Lunch. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 16
1.8 The ‘Boyd Temperature’. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
1.9 Flash-Crash Market Pathologies. . . . . . . . . . . . . . . . . . . . . . . . . 25
1.10 Network Fragmentation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
1.11 The Ratchet . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
1.12 Operational and Strategic Failure. . . . . . . . . . . . . . . . . . . . . . . . 31
1.13 Failure, Remediation, and Culture . . . . . . . . . . . . . . . . . . . . . . . 36
1.14 The Synergism of Phase Transitions in Real-Time Critical
Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 38
1.15 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43
2 Extending the Model. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47
2.2 Generalizing the Data Rate Theorem . . . . . . . . . . . . . . . . . . . . . 50
2.3 The Transitive Cognitive Decomposition . . . . . . . . . . . . . . . . . . 52
2.4 Environmental Insult and Developmental Dysfunction . . . . . . . . 52
2.5 Other Complexity Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
2.6 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55
ix