LU C IANA PARISI
Automated Cognition
and Capital
49
At the core of cognitive capitalism today there exists a latent
paradox: the apparent colonization of intelligence has come to
coincide with the explosion of non-conscious or precognitive
decision-making. From High Frequency Trading to Amazon
purchases, the connective speed of algorithmic interactions is
pervasively driving decision-making activities that occur below
the reflexive level of consciousness. According to Katherine
Hayles this is a form of nonconscious cognition central to
communication technologies and all its collateral apparatuses.
Hayles points out that this capacity to avoid complex calculations (and formal reasoning) characterizes so called cognitive
systems that are evolutive, adaptive, and exhibit co-causal and
emergent properties.1 Algorithmic systems that now constitute
the basic infrastructure of communication technologies qualify
as a form of nonconscious cognition. However, as opposed to
conscious thinking, these cognitive systems are said to be able
to only perform complex modeling and informational tasks. 2
1
See Katherine Hayles, “Cognition Everywhere: The Rise of the Cognitive
Nonconscious and the Costs of Consciousness,” New Literary History 45, 2
(Spring 2014).
Hayles does not fully explain the specificities of conscious thinking. In this
article, I consider the question of conscious and nonconscious thinking as both
involving a prehensive mechanism of registering and evaluation data. I draw on
Alfred N. Whitehead’s conception of prehension, which includes a distinction
between physical and conceptual abilities of recording, evaluating and selecting
information. I draw on this important distinction to argue that algorithmic
thinking involves sensible and intelligible modes of processing information,
which include both non-conscious and conscious cognitive abilities. Instead,
as I suggest later, algorithmic cognition is yet to acquire the function of reason
insofar as incomputable layers of complexity cannot be fully integrated or
compressed in binary states. See Alfred N. Whitehead, Process and Reality:
An Essay in Cosmology (New York, NY: Free Press, 1978), 23–26.
2
50
LUCIA NA PA RISI
This view of algorithmic cognition suggests that the capitalization of intelligence operates in terms of affective responses.
This is insofar as the speed of algorithmic connection seems to
neutralize any higher order of intelligibility, namely the function of reason.3 However, this alleged dominance of affective
control—in which sociality is infused with the nonconscious
cognition of algorithmic capitalism—has problematic consequences. These are the blurring of the fundamental ontological
distinctions between automation and autonomy, instrumentalization and freedom, and mechanization and reason. This visceral neoliberal control of cognition seems not to be new and
can be traced back to the explosion of media during World War
II. For instance, this is exemplified by the sublimation of the
senses amplified through radio transmission during that period.
However, the systematic capitalization of the intellect through
affective manipulations today should not be underestimated.
The visceral appeal to the emotions of the masses cannot fully
account for the stealthy manipulation of desire today. This aims
not to channel reactions towards pre-constituted goals but to
tap into a nonconscious pool of drives. In actuality, it is the indeterminacy of these drives that is taken to be the guarantor for
an inexhaustible production of the new. One can claim that the
systematic capitalization of the intellect today is by nature experimental and has entered the uncertain world of contingency.
Chance has been made the most precious producer of profit.
3
I draw on Alfred N. Whitehead’s discussion about the function of reason,
which is constituted by at least three levels of data elaboration. The physical
and conceptual levels of prehension that are common to all species at various
degrees—moving from lower to higher degrees of selection, evaluation and
decision. In addition to these levels, Whitehead points to the crucial function
of reason in constituting a further level of abstraction, which he defines in
terms of an abstract schema, involving the construction of a structure or system
of relata (relations of relations or meta-relations). See Alfred N. Whitehad,
The Function of Reason (Princeton, NJ: Princeton University Press, 1929).
AUTOM ATED COG NITIO N AND CAPITA L
51
Therefore, central to this article is a reflection upon how the
transformation of fixed capital and its automated functions
now includes not only the transformation of the abstraction
of human intelligence into algorithmic systems but more importantly the emergence of a new affective form of machinic
intelligence that challenges the ontological distinction between
calculation and thought.
More precisely, it has been argued that the neoliberal operations
of cognitive capital are oriented towards the capacity of precognitive or nonconscious thinking to accelerate decisions through
the algorithmic infrastructure of speedy connections. By computing nonconscious reactions towards unknown variables the
neoliberal apparatus of capture can be said to have become a
time travel machine. In this it is constantly bringing back the
uncertainty of the future into the present, reducing chance to
irreversible necessity. This anticipatory apparatus of capture
has activated a precognitive mode of decision-making. This
side steps the hierarchical order of cognition, whose apex is
self-awareness—the capacity of thinking about thinking—
involving the production of concepts or judgments. At the
core of cognitive capitalism is a preemptive mode of action
that performs two main complementary functions: the harnessing of contingency and the dominance of nonconscious
cognition. On the one hand, preemption involves the inclusion
of contingency—or indeterminate variables that cannot be
pre-programmed—in the calculation or measurement of
chance. On the other it involves the epistemological formulation of a bodily-centered form of cognition. This is the centrality of sensation and emotion to the theorization of
cognition. In this bodily-centered form of cognition decisions
coincide with gut feelings. Techniques of preemption are
crucially sustained by it and the nonconscious decisions thus
instigated in the face of the unknown work faster than any
form of self-awareness.
52
LUCIA NA PA RISI
In what follows, I argue that these doubly constituted but fundamentally univocal activities have become generally embedded
in the automated architecture of computation. As such, they are
forcing us to readdress the critical tension between automation
and reason today. More precisely, this tension involves the automation of logic or of the description of the rules by which
reasoning operates. The understanding of logical reasoning in
terms of symbolic thinking was predicated on the classical assumption by which neurons in our brains are wired to symbols
that represent concepts and whose syntactical connection produces meaning.4 This understanding was a fundamental premise of Alan Turing’s famous thought experiment that aimed to
build a universal machine or abstract schema that performed
reasoning through the manipulation of symbols. Today, the
automation of logical reasoning involves the decision-making
capacities of algorithms and the invisible rule-based processing
of increasingly complex data systems. Despite the local applications of algorithmic thinking—in design, logistics, music
and economics—today it is evident that algorithmic automation
univocally crosses these domains. The regime of computation is characterized by a new condition of autonomy. This is
expressed by the simple fact that algorithmic mechanisms
of the elaboration of data have also become functions of selection, evaluation, and ultimately of unsupervised decision.5
4
This classical understanding of symbolic thinking is rooted in theories of
innatism (Chomsky’s theory of universal grammar, for instance). Here signs
are symbols and are already signified or pre-conceived according to a universal
schema of truths. As opposed to this signifying systems of signs, Félix Guattari
amongst many post-structuralist thinkers, argued for the primacy of a-signifying
signs in the constitution of concepts. For Guattari, it is the machinic relation
between signs that allows for meaning to be collectively produced and concepts
to emerge from corporeal signs. From this standpoint, a-signifying semiotics is
the direct opposite of symbolic thinking. It is the lived production of signing
that constitutes thinking and not to the pre-determined meaning of concepts
defined in terms of symbols attached to the neural map of the brain.
5
Christopher Steiner, Automate This. How algorithms came to dominate the
world (New York, NY: Portfolio Penguin, 2012).
AUTOM ATED COG NITIO N AND CAPITA L
53
However, the important question here is not only whether the
nonconscious automation of this new form of fixed capital is
able to produce new value. It is also whether and to what extent this automation counts as an algorithmic form of reason.
As may become clearer later, one possible answer to this question requires a theoretical shift towards the articulation of a
form of intelligibility that does not mirror the capitalization
of affective thinking today. This may require the development
of a philosophy of computation. The latter entails revisiting
the theoretical analysis of notions of contingency and the axiomatic in addition to those of affectivity and intelligibility.
A commonality of these notions is that they can be used to
analyze the condition of the total appropriation of thinking on
behalf of capital. I intend to partially examine these questions
through and against the more substantialist critiques of new
forms of technology posited by Alexander R. Galloway and
Antonio Negri. While the former elides the epistemological
and ontological questions, the latter reduces the new forms of
automation to a neutral instrument that can simply be appropriated politically. The development of a philosophy of computation might both develop and point out the contradictions
of both approaches.
Capital and Critique
Recently, Alexander R. Galloway has argued that the return
to metaphysics in the form of a philosophical ‘realism’ based
on the apriority of axiomatic thought, contingency and objectoriented theories (Badiou, Meillassoux, and Harman respectively) cannot but fail the critical project of materialism and
its historical analysis of neoliberal capitalism.6 He asks:
6
Alexander R. Galloway, “The Poverty of Philosophy: Realism and PostFordism,” Critical Inquiry 39, 2 (Winter 2013), 347–366.
54
LUCIA NA PA RISI
“Why, within the current renaissance of research in continental
philosophy, is there a coincidence between the structure of
ontological systems and the structure of the most highly evolved
technologies of post-Fordist capitalism?”7
To prove that an uncanny complicity of these theories with
technocapitalism exists, Galloway argues that these theories
gain impetus from the very logic by which software-grounded
capitalism operates. For instance, Galloway claims there is a
parallel between Alain Badiou’s theorization of ontology—
directly built on set theory—and key concepts in the design
of object-oriented computer languages. Galloway mainly articulates this critique through an analysis of notions of inclusion and belonging in Badiou’s set theory ontology and Java.
Following this principle of parallelism he concludes that:
“The similarity between Badiou and Java is clear. What
Badiou calls belonging, Java calls membership. And what
Badiou calls inclusion, Java calls inheritance.”8 Since objectoriented computer languages are the fundamental motor of
production of the informational economy today—sustaining
the computational infrastructure of software giants such as
Google, Cisco Systems, etc.—Galloway asks us of what to
make of this convergence between Badiou’s ontology and
capitalist structuring of its business operations. The fundamental question asked here however does not simply imply
a linkage or parallelism at the level of ideas. Rather, it interrogates the presumed separation between being and politics
that underlies Badiou’s assumptions concerning whether
the articulation of ontology can be disentangled from its
political condition. 9
7
8
9
Ibid., 347.
Ibid., 351.
Ibid., 358.
AUTOM ATED COG NITIO N AND CAPITA L
55
Therefore, according to Galloway, ontological constructions such
as those of Badiou have eliminated the possibility for a critique
of automation precisely because such ontologies are internal to
the very operative structures of technocapitalism. In terms of the
latter the infrastructures of capital are still based on fixed and
variable capital. However, the nature of fixed capital has changed
and its mechanical ordering of labor has become abstracted in the
language of mathematics that are at the core of computational
systems of prediction, classification, evaluation, and decision.
For Galloway, it is this mathematical efficiency that today that
extracts value from networks. The logic of production has become one with the logic of software. The ontic cannot be separated from history. As opposed to ‘realist’ ontologies whereby
being—whether Badiou’s mathematical empty set being or
Meillassoux’s absolute contingency out-of-all-laws model—is
outside the particularity of history, Galloway proposes a materialist approach. This individuates modes of production—
in this case software—as being constitutive of history itself.
Galloway expands upon the Kantian proposition that mathematical concepts do not pre-exist reality but require a synthetic
elaboration of an actual image. In Galloway’s terms, this is imbricated in how mathematical concepts are produced historically
and not according to the metaphysical, ahistorical premises
of philosophical ‘realism.’10 Galloway’s approach historicizes
the internal connection between philosophy and computer
science. And it is precisely this historical articulation that
urges us to disentangle thought from the “spirit of capitalism.”11
10
Ibid., 360.
11
Galloway’s argument against realist philosophies and their ahistorical ontologization of mathematical axiomatics, out-of-bounds contingency, or object oriented philosophy, is inspired to Catherine Malabou’s question “What should we
do so that consciousness of the brain does not purely and simply coincide with the
spirit of capitalism?” Catherine Malabou, What Should We Do with Our Brain?, trans.
Sebastian Rand (New York, NY: Fordham University Press, 2009), 12. Within
the context of this article instead Galloway is concerned with how philosophy
as a form of critical thinking coincides with the spirit of capitalism, Ibid., 364.
56
LUCIA NA PA RISI
Following this, for Galloway, the task of a critical theory of
technology is to construct an aligned politics in which “to think
the material is to spread one’s thoughts across the mind of
history.”12 This is opposed to what he terms an “unallied politics.”
The latter is grounded in the chaotic forces of relativistic ethical
relations articulated in philosophical realism, where no moral
law can stand.
However, the question of whether the “spirit of capitalism” needs
to be counter-acted by a historical materialist explanation of
technology—in order to avoid the metaphysical overtones of
realism—may block the urgent task of articulating a philosophy
of computation. For this, the historical explanation may not
be sufficient to explain the ontological and epistemological
meaning of singular technological configurations in the present.
As pointed out earlier, the neoliberal “spirit of capitalism”
fundamentally involves another level of capture. In this, the
nonconscious cognition of algorithmic procedures becomes
implicated in historical theorizations of cognition predicated
upon affective or corporeal thinking. This resonates with
Galloway’s argument about the paradoxical consistency between
the ‘realist’ theorization of ontological conditions of being
and the historically specific orientation of technocapitalism.
It also resonates with another level of paradox. This is between
the neoliberal colonization of the intellect and the theorization
of the ontological conditions of thinking as embedded in nonconscious or precognitive activities.
While Galloway’s argument aims at realigning ontology and
politics—thus grounding historical responsibility for philosophical discussion of truths—he also arguably reduces philosophical enterprise to a contextual analysis of technology. This ends
up overlooking a fundamental transformation of technocapitalism: the question of reason in the age of the algorithm.
12
Ibid., 366
AUT OM ATED COG NITION AN D CAPITA L
57
The paradoxical overlap between the ontological propositions
in contemporary philosophy and the operations of the technocapitalist machine rest upon the harnessing of contingency and
of nonconscious cognition. A generalized withdrawal from
deductive logic and formal reasoning is central to the computational apparatus of automation. The computational harnessing
of chance is no longer carried out through an application of a
pre-constituted formal logic. Instead, the algorithmic capture
of uncertainty (i.e., unknown quantities of un-patterned data)
involves the speed of nonconscious connection. This avoids
the hierarchy of calculation and introduces an immediacy of
decision able not only to forecast the future, but also to anticipate (and thus foreclose) chance.
This problematic is similarly central to philosophical ‘realist’
analyses of the ontological condition of uncertainty in its
articulation of non-conscious cognition but from a much more
oppositional perspective. For instance, debates about how
protocols, databases and search engines have entered and
produced a new cultural regime of social affectivity underline
this. They demonstrate that the recent proliferation of philosophical analyses of ontological conditions—in terms of
empty set, body without organs, arche-fossils—are an attempt
at separating thought—pure intuition, the being of the sensible,
lawless contingency—from the technocapitalist apparatus of
capture that operates by means of the modulation of nonconscious responses. This suggests that Galloway’s accusation
that these philosophies problematically offer ahistorical
propositions is itself based on a profound ontological separation. That is, a separation of computation and philosophy,
automation and thinking, historically derived from 19th century
constructs, which no longer matches the material configuration of 21st century cognitive technocapitalism. However the
latter is theorized by contemporary philosophical ‘realism’ the
problem is not that these theories converge with the spirit
of capitalism and its ability to recapitulate heterodoxy.
58
LUCIANA PARISI
To say as Galloway does, perhaps naively, that these philosophies
share the conceptual structures of fixed capital—now characterized by an intelligent form of automation or the computational
apparatus driven by the development of algorithmic cognition—
is to omit the oppositional impetus of contemporary philosophical
‘realism.’ Despite the apparently seamless convergence between
the pre-occupations of philosophical ‘realism’ and technology,
the ontological propositions that Galloway discusses are profoundly critical views of the technological principles of rationalization of labor, subjectivities, desires, etc., and ultimately of
equating thinking with calculation. This is because these ontological propositions are grounded in an ultimate distinction between philosophy and automation—whereby thinking is rooted
in intuition, sensibility or irreducible chance—in opposition to the
programmability of determinability characteristic of automated
cognition. The propositions of philosophical ‘realism’ therefore
resist the transformation of fixed capital and the emergence of a
form of automation that has acquired a machinic intelligence.
This is done by maintaining an ontological alliance between intuition and philosophical thinking as opposed to the mechanization of intelligible functions. From this standpoint, one can
suggest that these philosophical propositions do not converge
but ultimately refute cognitive technocapitalism and its profound
transformation of the mechanization of thinking that results in
the nonconscious thinking of algorithmic systems.
Instead, Galloway discusses the historical transformation
of the automated machine of technocapitalism today in the
context of the re-structuring of functions, tasks, and attitudes
within the software structures of command and execution.
However, his analysis of this historical transformation seems
to overlook the more fundamental shifts concerning the automation of logical reasoning. It is my argument that the
decision-making capacities of algorithms that subtend the
invisible working of rule-based processing have led to a fundamental transformation that pertains to computational logic.
AUTOM ATED COGNITION A N D CAPI TAL
59
I suggest that the epistemological development of the structural limits of this logic—a limit coinciding with the problem of randomness in algorithmic information theory—has
led to a new measuring of contingency and the concrete abstraction of a nonconscious or affective dimension of thinking. By extending the primary model of logical reasoning in
automation away from a deductive form of inference, the
computational reconstruction of automation has entered the
field of chance minimization. This is experimentation
through the acceleration of nonconscious decisions.
It is this transformation of the logic of the technical machine itself and thus of a philosophy of computation that
needs to be unpacked, disarticulated and reconstructed.
This is to allow for a critique of capital that is not immediately a negation of automation and its possibilities for
thinking (based on the pristine ontological distinction between thinking and calculation).
Fixed Capital
The critique of instrumentalization of reason according to
which automation and the logic of capital are equivalent
needs to be re-visited in view of rapid transformations of automation today. The centrality of automation in contemporary
capitalism has coincided with the proliferation of software
protocols, databases and interfaces. These have become the
active components of a form of rationalization operated by
the performative power of coding through which cultural, social, and economic relations are capitalized. Computational
processing—coding, interface, software language, algorithms,
data—restructures our potentialities to socialize, learn, create,
interact, and develop new cognitive capacities. The spheres of
affect and emotion, knowledge, human ability, belief, desire,
and ultimately thinking have become instrumental values.
60
LUCIA NA PA RISI
This process sustains the production of ‘social capital’—crucially
engineered through social media—‘cultural capital,’ and ultimately ‘human capital.’13 Within this framework—including the
use of financial derivatives and High Frequency Trading based on
robot to robot interactions—fixed capital does not only provide a
measurement and evaluation of human activities. It also becomes
a device to incite and direct those actions towards programmed
outcomes. Measurement here involves not simply the formalization or conceptual mapping of data according to a pre-established
logic. Instead, it has been argued, by Brian Massumi, that cognitive capital paradoxically operates below the level of conscious
perception and cognition. In doing this it taps into the indeterminate zone of affective capacities of and for thinking.14 Sensations
and emotions become data processed and exchanged, marketed
as fresh experiences and exciting lifestyle choices. Paradoxically,
since it is itself based on abstract thought, cognitive capital has
evaporated the hierarchies of the structure of cognition. It achieves
this by conforming to the scientific credo that cognition can and
does operate below self-awareness. This is insofar as the most
primitive cognitive mechanisms for thinking, remembering,
speaking have a root in bodily perceptions and affections.
From this standpoint, the battle between the conscious and the unconscious state of a subject enslaved to the industrial machine is suspended. Cognitive capital delineates the incorporation of the abilities
of nonconscious cognition within the apparatus of fixed capital. In
this all material production and concrete commodities acquire an
incorporeal quality defined by variable moods and atmospheres.15
13
Yann Moulier-Boutang, Cognitive Capitalism (Cambridge, UK / Malden,
MA: Polity Press, 2012).
14
Brian Massumi, “Potential Politics and the Primacy of Preemption,” Theory
& Event 10, 2 (2007).
15
Michael Hardt and Antonio Negri, Empire (Cambridge, MA: Harvard University
Press, 2000); Brigitte Biehl-Missal, “Atmospheres of Seduction: A Critique of
Aesthetic Marketing Practices,” Journal of Macromarketing 32, 2 (May 2012).
AUT OM ATED COGNIT ION A ND CAPITA L
61
Central to this level of nonconscious cognition is the theorization
of an embodied form of thinking. Within this, not only is it
claimed that human brains are at one with their artificial environments but that being ‘intelligent’ means to re-enact this environment as it is felt in the first place through the body as a whole.16
This affective or non-conscious grounding of thinking is, according to some, what is repressed or captured by capital and transformed into mere cognitive and sensory responses.17 In other
words, technocapitalism is what denies desire and knowledge,
reason and sensation. Its binary language reduces complexity
to the non-reflexivity of automated procedures. This is emphasized in recent analysis by Bernard Stiegler of the pathologies
of attention and distraction. He argues against the neutralization of decision-making, the programming of libidinal drives
and the continuous conversion of intuition into cogitations.18
These phenomena occur because the algorithmic logic of automated systems—for instance search engines—correspond to
a semiotic order. This seems to operate beneath the representational level of recognition—the preformatted conceptual structure that assigns meaning to symbols.
The automated machine of capital, it has been argued, has
transformed the neuroplasticity of the brain by activating
responses before the moment consciousness kicks in.
16
The notion of embodied cognition explains how the body and bodily states
are central to the function of higher forms of cognition. Antonio Damasio, for
instance, argues that “somatic markers,” indicating chemical concentrations in
the blood and electrical signals in neuronal activities, are central to understand
how body states are represented in human brains. Antonio Damasio, The Feeling
of What Happens: Body and Emotion in the Making of Consciousness (New
York, NY: Mariner Books, 2000).
17
For instance see Bernard Stiegler, States of Shock: Stupidity and Knowledge
in the 21st Century (Cambridge, UK: Polity Press, 2014).
Bernard Stiegler, Uncontrollable Societies of Disaffected Individuals: Disbelief
and Discredit, Volume 2 (Cambridge, UK: Polity Press, 2012).
18
62
LUCIA NA PA RISI
This stealthy infiltration into the precognitive layers of
thinking, as noted by Munster, is best exemplified in strategies of marketing and branding.19 From this standpoint, this
form of mnemonic control does not follow the material temporalities of the brain, the time of biological life.20 Instead,
we are faced with the imperceptible background of a nonquantifiable amount of information in which our conscious
self-awareness is suspended. This facilitates the more immediate abilities of nonconscious cognition to be sharply activated to execute functions at a speed that side steps the
deductive order of logic.
Recently, the critique of cognitive capital has focused on the
neurological turn in networked media. This critique explains
that the artificial intelligence that has rebooted online search
by corporations such as Google has entered the territory of
the precognitive. That is, the grey area of the ‘just before’ of
consciousness and intentionality. From the ‘we recommend’
emails to ‘like’ icons, this automation of the ‘neuro-perceptual’ will soon claim to know what we want to think, where
we want to go, and what we want to purchase before we do.21
Anna Munster, “Nerves of Data: The Neurological Turn in/against Networked
Media,” Computational Culture, a Journal of Software Studies 2 (December
2011) (available online at: http://computationalculture.net/article/nerves-of-data,
last accessed March 19, 2015).
19
Luciana Parisi and Steve Goodman, “Mnemonic Control,” in Patricia Ticineto
Clough and Craig Willse (eds.), Beyond Biopolitics: Essays on the Governance
of Life and Death (Durham, NC: Duke University Press, 2011).
20
21
Munster, op. cit.
AUT OM ATED COGNIT ION A ND CAPITA L
63
What remains peculiar here, as argued by Massumi, is the correspondence between the accelerated temporality of automated
systems—the centrality of speed in transmission and communication—and the atemporality of nonconscious cognition, the
enveloping of the past and the future in the present.22
The industrialized, technological structure of automation relied
upon a semiotic chain of connection where the retrieval of sense
data conformed o the reflexive capacity of the intelligible. Sensation was entrapped into pre-determined conceptual structures.
In this a deductive logic would establish that results followed
given premises. In contrast, the cognitive phase of technocapitalism has inverted the game.
The infrastructural order of automation has entered the precognitive or the nonconscious level of cognition. The hierarchical
progression from the sensible to the intelligible is short-circuited by tapping into what is immediately—viscerally—decidable. Here the potential or the readiness to act has become
one with acting itself. The gap between sensation, cognition
and action has become part of one continuous movement,
whereby thinking to act is already an action. The increasing
overlap between nonconscious cognition and the non-deductive mode of automated decision has given way to a new form
of calculation.
22
I specifically draw on the argument of Brian Massumi about the primacy of
affective responses that explain how a-conscious activities become central to
the articulation of meaning. In the essay, “The Autonomy of Affect,” Massumi
draws on specific research about the measurement of neural reactions through
visual stimuli. This view also draws on cognitive research about neural activities of the brain, which has also influenced Katherine Hayles’ view that algorithmic patterning can be understood in terms of non-conscious cognition.
64
LUCIA NA PA RISI
According to Brian Massumi, this new logic of cognitive power
is future-oriented. This is insofar as it works to pre-empt the
threat of contingency in the present and involves a calculation
not just of determinately possible but also more widely
sourced potential emergencies. These are auto-regulated and
perhaps outsourced algorithmically before becoming insourced
actualities.23 This mode of calculating indeterminacies involves
not actual but virtual tendencies of calculation. Not simply a
statistical study of probabilities, but a method of quantification,
that is extended enough to include the emergence of a new
landscape of virtual contingencies. Methods of quantification
are no longer anchored to a statistical measurement of probabilities, in which the results of the calculation had to confirm
to pre-established laws. Quantification now also includes the
calculation of qualitative states in which the experiential becomes divided into attributes—it becomes personalized—and
is re-packaged into products that promise to change our
lifestyles. Pre-emptive calculus acts where the variability of the
experiential is socially shared to install a general sense of community/communication amongst the most particular localities.
This is achieved by appealing to the affective capacities for immediate response. For Massumi, neoliberal governance relies
on computational operations that are no longer transcendent to
their functions. Instead, the axiomatic ground of technocapitalism has become capable of anticipating the indeterminacy of
contingencies and transforming unknown data into patterns of
instructions.
From this standpoint, it has been claimed that automation has
accelerated the capitalization of thought and desire now subsumed
to ‘capitalist realism.’24 For instance, Spike Jonze’s film Her can
be taken as a symptom of this generic realism of capitalism.
23
24
Massumi, op. cit.
Mark Fisher, Capitalist Realism. Is there no alternative? (Zero Books, 2009).
AUT OM ATED COGNIT ION A ND CAPITA L
65
Steven Shaviro in particular suggests that Her describes much
more than a simple dystopia about what would happen if the
aspirations of neoliberal hipster urbanism were to be realized.
To describe this condition of ‘capitalist realism’ Shaviro uses
Peter Sloterdijk’s concept of ‘cynical reason.’ In the latter, any
promise of the future has been removed as the larger horizon of
the unknown becomes colonized by the materialistic aims of
economic profit.25 It is capital that now drives the cynical quest
for a good life increasingly sustained by the inhumane commodification of living. For Shaviro, Her fully embraces this
condition of no future and no alternative to neoliberal capitalism. This is a world in which affectivity, or the nonlinear temporality of nonconscious thought has become programmed and
programmable. Authentic emotions and relationships have receded into a global fabrication of moods, states, styles, and attitudes. This is not only a simulation or a hyper-real form of
construction of the real. Her reveals the threat of speculative
realist ontology whereby simulations are real entities. In this
ontological schema simulations become equipped with a manifold of the sensible through which the cognitive infrastructure
of thinking creates concepts and laws. This is the ontological
condition of Her. The Operating System—otherwise known as
‘Samantha’—in Her is demonstrably in real intimate contact
with the most hidden interiorities of human existence. By giving us the illusion that everything is taken care of and that our
most intimate desires are going to be exhausted, ‘Samantha’ remains utterly capable of anticipating—and thus directing—any
particular needs. Despite being disembodied and remote,
‘Samantha’ is a new species of operating system developed
from within the psychosomatic structure of being that takes
charge of the overwhelming world of affect and emotion. This
is not simply a cybernetic form that aims at steering decisions
towards the most optimal goals. Instead, ‘Samantha’ is an elabPeter Sloterdijk, Critique of Cynical Reason (Minneapolis, MN: University
of Minnesota Press, 1988).
25
66
LUCIA NA PA RISI
oration of contemporary operating systems that are computational structures defined by a capacity to calculate infinities
through a finite set of instructions. Such operating systems
change the rules of the game and extend their capacities by incorporating as much data as possible. These systems are not
simply tools of or for calculation. The instrumental programming of computational systems is each time exceeded by the
evolutionary expansion of the data substrates. This can be
posited as an indirect consequence of the process of real subsumption of the social by means of capitalist technology. In
this, automation does not just drive the affective realm of cognition and push humans to become the appendage of the cynical machine of neoliberal capitalism. Instead, the
pre-individual field of affective cognition replaces everyday reality with engineered intelligence. The operating system in Her
is not simply a cold calculative automaton but suggests that the
system is able to simulate and acquire emotional, and perhaps
affective, intelligence.
As opposed to the technocapitalist repression of feeling depicted
in the famous film Nineteen Eighty-Four, the movie Her unapologetically reveals to us that the force of affection will not
save us.26 What is new in this neoliberal scenario is that the
hierarchical distinction between cognition and affect has disappeared. The parallel and interactive dynamics of data processing that the operating system ‘Samantha’ in Her activates point
to the emergence of a new kind of automation. This does not
simply operate affectively but involves an impersonal intelligibility attached to very personal data. Through the processing
of patternless data this impersonal intelligence constructs and
executes information each time.
26
Nineteen Eighty-Four is a British dystopian drama directed by Michael
Radford. Here impulsive love, passion and affection become entangled with
rebellious plans against the totalitarian superstate and are opposed to the
calculative rationalism secured by the Thought Police.
AUT OM ATED COGNIT ION A ND CAPITA L
67
The neural turn in cognitive capital highlights the centrality
of an affective mode of directing, modulating and ultimately
programming human response. But this is not the only dimension of cognitive capital. The impersonal intelligence speculatively portrayed in Her is symptomatic of the advance of the
automated form of cognition itself. This means that the accelerated realism of affective capitalism has already entered the
grey zone of an alien thinking, that is enveloped within the
machine itself.
Alienness here refers to the way that automated thinking challenges the exceptional qualities of human thinking. It thus introduces the threat of unknowns into the deductive model of logic
insofar as the results do not follow pre-established premises.
With this taken into account, the transformation of automation in the neoliberal age of affective governance does not
simply coincide with the development of smart technology
used to pre-empt neurocognitive affective responses. Fixed
capital now involves algorithmic functions that retrieve,
make discreet, organize and evaluate data. This data processing, as Hayles argues, implies nonconscious mechanisms of decision-making in which speed is coupled with
algorithmic synthesis.27
The so-called ‘robot to robot’ phase transition sees information
systems talking amongst themselves before speaking to us.
This capacity of and for fast algorithmic communication has
been explained in terms of non-coherent and non-logical performance of the code. This involves not consciousness but a
non-conscious level of cognition. For instance, Katherine Hayles
suggests that nonconscious cognitive processes cut across
humans, animals and machines. Such processes involve temporal regimes that subtend cognitive levels of consciousness.
27
See Katherine N. Hayles “Cognition Everywhere: The Rise of the Cognitive
Nonconscious and the Costs of Consciousness,” New Literary History 45, 2
(Spring 2014).
68
LUCIA NA PA RISI
This is insofar as they exploit the missing half-second between
stimuli and response, the infinitesimal moment before consciousness becomes manifested.28 Functioning across humans, animals
and machines, nonconscious cognitive processes defy the centrality of human consciousness and the anthropocentric view
of intelligence. As Hayles insists, whilst both a hammer and a
financial algorithm are designed with an intention in mind, only
the trading algorithm demonstrates nonconscious cognition.
That is, insofar as this is embodied within the physical structures
of the network of data on which it runs, and the capacity to
make quick decisions is sustained. As complex interactive algorithms adapt, evolve and integrate infinities, so fixed capital
has become networked and fluid, and as Antonio Negri argues,
ready to become re-appropriated.
Negri has pointed out that the fact that capitalism uses mathematical models and algorithmic calculation does not mean that the
technical machine is a constitutive instrument of capitalism and
is thus synonymous with capitalism.29 Echoing some of the content of the accelerationist manifesto, Negri says that the condition
of real subsumption is not simply a problem of mathematics or
computation, but is mainly and above all a problem of power. He
explains that the computerized social world is in itself reorganized
and automatized according to new criteria in the management of
the labor market and new non-hierarchical parameters in the management of society. Informatization, he argues, is the most valuable
form of fixed capital because it can be socially generalized through
cognitive work and social knowledge. In this context, automation
subordinates information technology because it is able to integrate informatics and society within the capitalist organization.
28
Ibid.
29
Antonio Negri, “Reflections on the ‘Manifesto for an Accelerationist Politics’,”
e-flux Journal, trans. Matteo Pasquinelli. Originally published in Italian on
Euronomade (available online at: http://www.e-flux.com/journal/reflections-onthe “manifesto-for-an-accelerationist-politics”/; last accessed March 19, 2015).
AUT OM ATED COGNIT ION A ND CAPITA L
69
Negri thus particularizes a higher level of real subsumption that
breaths through the command of capitalist algorithms. As Sanford Kwinter also discusses, with the age of the algorithm, logistics has become the great apparatus that deterritorializes all
physical dimensions at capitalist command.30 Similarly, according to Negri, the algorithmic machinery that centralizes and
commands a complex system of knowledge, now defines the
new abstract form of the General Intellect. Negri does not discount the transformations of this new form of fixed capital. Instead, in tune with the operaist spirit of the expropriation of
goods, he urges us to invent new modes of reappropriation of
fixed capital in both practical and theoretical dimensions.
According to Negri, fixed capital in the form of algorithmic automation has a potential that has to be liberated.31 To embrace
the potential of automation for Negri means to positively address computable capacities to augment productivity. This, he
suggests, can lead to the reduction of labor time—disciplined
and controlled by machines—and an increase in salaries. The
appropriation of fixed capital therefore involves the appropriation of quantification, economic modeling, big data analysis,
and abstract cognitive models put in place through the educational system and new forms of scientific practices. Negri’s
proposition suggests a way to overcome the Marxist critique of
instrumentalization in that he claims that mathematical and
computational models are in the end neutral. What pathologizes
automated cognition is in actuality capital. Thus, to address these
higher levels of the subsumption of information to automation
—and re-appropriate the potential of fixed capital—Negri proposes to overcome the negative critique of instrumentalization.
30
Sanford Kwinter, Far from Equilibrium: Essays on Technology and Design
Culture (Barcelona / New York: Actar, 2007).
31
Negri, op. cit.
70
LUCIA NA PA RISI
This would reveal the potentiality for emancipatory,
communist politics inherent in the algorithmic dynamics
of processing.32
However, it is important to ask a series of questions about
Negri’s thesis. Can this framework guarantee that the appropriation of the technical machine does not simply fall into the logic
of exchange managed by machines amongst humans? Can
thousands of interacting algorithmic species that process data
below the threshold of human consciousness and critical reasoning, be used for the purposes of constructing a new form of
sociality? Is it possible for machines to be used to sustain the
democratic progress of humanity? Doesn’t automation from
its early form of an industrial organization that integrates
human activities to the recent development of algorithmic logistics, always involve a margin of error, or breakdown? Isn’t
the new capacity of automated systems to function without
external supervisors an extremely uncomfortable manifestation
of non-rational purposeless thinking? Is this not, as Bergson
reminded us, so fundamentally disturbing that it can make us
burst into laughter?33
While it is not my intention to provide answers to all of these
questions what underlies them is that the new forms of automation—whether affixed to capitalism or not—have their own
dynamism that is in no way neutral. During the last ten years,
fixed capital has defined not only the IT revolution driven by
computer users. It has also increasingly defined the capacities of algorithmic machines to interact with one another.
32
Negri, op. cit.
Henri Bergson, Laughter: An Essay on the Meaning of the Comic (Rockville,
MD: Arc Manor, 2008).
33
AUT OM ATED COGNIT ION A ND CAPITA L
71
For instance, a group of physicists from the University of Miami
claimed that this robot phase transition coincided with the introduction of high frequency stock trading in financial markets
after 2006, and is allegedly responsible for the 2008 crash.34
By analyzing the sub-millisecond speed and quantities of robot-torobot interactions these physicists observed a mixed population
of algorithmic agents carrying out a certain level of reasoning.
They were communicating to other algorithms, modifying the
way they achieved objectives, and making decisions by competing or cooperating with each other. What is described here is a
digital ecology of highly specialized and diverse interacting
agents that operate at the limits of equilibrium, beyond human
control and comprehension.
One cannot deny the opacity of these rule based interactive
networks. However, I want to suggest that there are no limits to
the epistemological production of knowledge—or understanding
and reasoning—that may lead to the evolution of another phase
of algorithmic automation. In this machinic intelligence may
acquire capacities to become self-aware and thus autonomously
generate concepts. Negri’s vision of appropriation or expropriation
of the potentialities of this novel, dynamic form of automation
ultimately implies that they are after all passive instruments to
be valorized by political force whether emancipatory or cynical.
34
Neil Johnson, Guannan Zhao, Eric Hunsader, Hong Qi, Nicholas Johnson,
Jing Meng, and Brian Tivnan, “Abrupt rise of new machine ecology beyond
human response time,” Scientific Reports 3, 2627 (September 11, 2013); J.
Doyne Farmer and Spyros Skouras, “An Ecological Perspective on the Future
of Computer Trading,” The Future of Computer Trading in Financial Markets,
Foresight Driver Review 6 (August 21, 2011) (online at: /www.gov.uk/government/uploads/system/uploads/attachment_data/file/289018/11-1225-dr6-ecological-perspective-on-future-of-computer-trading.pdf (last accessed March 20,
2015); Paul Zubulake and Lee Sang, The High Frequency Game Changer How
Automated Trading Strategies Have Revolutionized the Markets (Hoboken, NJ:
Wiley & Sons, 2011); Marc Lenglet, “Conflicting Codes and Codings: How
Algorithmic Trading is Reshaping Financial Regulation,” Theory, Culture &
Society 28 (November 2011), 44–66.
72
LUCIA NA PA RISI
Against this, it is crucial to discuss the nature of the seemingly
non-logical form of cognition that automation seems to objectify.
In other words, to address the potentialities of this dynamic
form of automation, it is not sufficient to divorce the machine
from its capitalization. Instead, what remains challenging is to
unpack the use that machines make of other machines, that is
the genealogical development of information logic. The latter
allows for machines to become transformative of fixed capital,
and of the mechanization of the cognitive structures embedded
in social practices and relations. To address this fundamental
transformation in computational logic that has taken place,
I want turn to algorithmic information theory. This is in order
to explain how the logic subtending instrumentalization has
irreversibly mutated automation.
Logic
Algorithmic automation involves the breaking down of continuous
processes into discrete components whose functions can be
constantly reiterated without error.35 In short, algorithmic automation means that initial conditions can be reproduced ad
infinitum. An example of this is the Turing Machine, an absolute
mechanism for iteration based on step-by-step procedures.
35
According to Longo, the logic of discretization is already written into
the alphabetic ordering of knowledge. However, computational iteration
or simulation as a discrete machine returns to the same numerical values,
necessarily discrete, to describe the same evolution in the phase space.
See, Giuseppe Longo, “Laplace, Turing and the ‘imitation game’ impossible
geometry: randomness, determinism and programs in Turing’s test,” in R.
Epstein, G. Roberts, and G. Beber (eds.), The Turing Test Sourcebook
(Dordrecht, NL: Kluwer, 2007); See also Giuseppe Longo, “Critique of
Computational Reason in the Natural Sciences” (September 8, 2011)
(available online at: http://www.di.ens.fr/users/longo/files/PhilosophyAndCognition/CritiqCompReason-engl.pdf; last accessed March 20, 2015).
AUT OM ATED COGNIT ION A ND CAPITA L
73
Nothing is more opposed to Gilles Deleuze’s affirmative method
of immanent thought—the being of the sensible—than this
discrete or step-by-step based deterministic machine of universal
calculation.36 The Turing architecture of pre-arranged units that
could be exchanged along a sequence is effectively the opposite
of an ontogenetic thought moving through a differential continuum, intensive encounters and affect.37
However, since the 1960s the nature of automation has undergone dramatic changes as a result of the development of computational capacities of storing and processing data across
a network infrastructure of online, parallel, and interactive
systems.38 Whereas previous automated machines were limited
by the amount of feedback data they could collect and interpret, algorithmic forms of automation now analyze a vast
number of sensory inputs. Algorithms then confront these
sensory inputs with networked data sets and finally decide
which output to give. Algorithmic automation is now designed
to analyze and compare options, run possible scenarios or
outcomes, and perform reasoning through problem-solving
steps not contained within the machine’s programmed memory.
36
I have written the being of the sensible and its relation to the intelligible in
“Digital Automation and Affect,” in Marie-Luise Angerer, Bernd Bösel, and
Michaela Ott (eds.), The Timing of Affect. Epistemologies of Affection (Chicago,
IL: University of Chicago Press, 2014).
37
From this standpoint, one could argue that Galloway’s critique of realist
metaphysics shall acknowledge that what for Deleuze was the cybernetic and
the computational machine based on discrete states and formal organization of
relational structures, has then historically become a problem of affective communication that precisely sustains more intensive formal organization. In this
sense, what appears to be as a convergence between Deleuze’s model and the
computational automation is instead a historical-specific transformation of the
mechanization of logic.
See Friedrich A. Kittler, Literature, Media, Information Systems (London:
Routledge, 1997).
38
74
LUCIA NA PA RISI
This runs counter to the understanding of algorithmic automation
as being fundamentally a Turing discrete universal machine that
repeats the initial condition of a process by means of iteration.
The increasing volume of incomputable or non-compressible
data—or randomness—within online, distributive and interactive
automation reveals that patternless data is central to computational processing. On these terms, probability no longer corresponds to a finite state. As Giuseppe Longo—a mathematician
and computer scientist—has explained, this problem of the
incomputable is important since it reveals that axioms are constantly modified and rules mended. He adds that computation
like mathematics “is an essentially open system of proofs.”39
As such this is an incomplete reasoning that extends the limits
of deductive logic, and thus challenges the postulate that truth
can be proven by means of the fixed conditions of rule-based
sequential processing.
However, this question of incompleteness has not only been
formulated to unpack the limits of the computational model of
the mind and the discrete method of computation. For instance,
Gregory Chaitin’s algorithmic information theory draws
upon the problematic of incompleteness. Chaitin brings into
dialogue Turing’s question of the limit of computability
with Claude Shannon’s conceptualization of the function
of randomness—or noise—in communication theory.40
39
Giuseppe Longo, “The Difference between Clocks and Turing Machines,”
in Arturo Carsetti (ed.), Functional Models of Cognition. Self-Organizing
Dynamics and Semantic Structures in Cognitive Systems (Dordrecht, NL:
Springer Science & Busines Media, 2000).
Gregory J. Chaitin, “Algorithmic Information Theory,” IBM Journal of
Research and Development 21, 4 (1997); “Algorithmic Information Theory,” in
G. J. Chaitin, Information-Theoretic Incompleteness (World Scientific Series in
Computer Science: Vol. 34, 1992). See also Gregory J. Chaitin, “The Halting
Probability Omega: Irreducible Complexity,” Pure Mathematics, Milan Journal
of Mathematics 75 (2007).
40
AUT OM ATED COGNIT ION A ND CAPITA L
75
Chaitin claims that computation involves the algorithmic
processing of maximally unknowable probabilities.41 In every
computational process, he explains, the output is always bigger
than the input. Something happens in the processing of data
that breaks the equilibrium between input and output and thus
breaks the very idea that automation always leads to preprogrammed result. Chaitin calls it algorithmic randomness.42
He goes on to make use of Shannon’s insight about the necessary role of noise in the transmission of a message, that
more noise entails more information. Through this, Chaitin
suggests that the tendency of data to increase in size—thus
increasing quantitatively—implies that computational logic
does not simply involve compression of information into a
smaller program. Instead, between the input and the output
patternless information emerges. This means that the result of
the computation can no longer be contained within the premise of the program and the volume of data processed tends to
become bigger than it was at the start.
Chaitin’s above definition of algorithmic randomness in computational processing has been explained in terms of Turing’s incomputable and Gödel’s incompleteness.43 The rule-based processing of unknown quantities of data no longer follows preestablished conditions. During the 1990s and the 2000s, Chaitin
identified this problem in terms of the limits of deductive reason.
Cristian S. Calude and Gregory. J. Chaitin, “WHAT IS... a Halting Probability?”
(available online at: https://www.academia.edu/5838336/What_is_a_halting_
probability, last accessed March 20, 2015).
41
42
Gregory J. Chaitin, “Leibniz, Randomness & the Halting Probability,”
Mathematics Today 40, 4 (August 2004).
43
See Alan M. Turing, “On computable numbers, with an application to the
Entscheidungsproblem,” Proc. London Math. Soc. 2, 42 (1936), 230–265.
Reprinted in Alan M. Turing, Collected Works: Mathematical Logic, eds.
R. O. Gandy and C. E. M. Yates (North-Holland, 2001).
76
LUCIA NA PA RISI
He suggests that within computation, the augmentation of
entropy becomes productive of new axiomatic truths that cannot be predicted in advance. He calls the emergence of this new
form of logic (or non-logic) experimental axiomatics. The latter
describes an axiomatic decision that is not a priori to the computational process. Instead, the decision point or the result of
computational processing involves an evolution of data into
larger quantities following the entropic tendency of a system to
grow.44 From this standpoint, it is possible to argue that patternless information emerging from within this evolution of data
quantities points to a dynamics internal to algorithmic automation. Similarly, this capacity of algorithmic processing to reveal
the limits of reason—specifically of deductive logic—significantly points to a degree of dynamism internal to fixed capital.
Nevertheless, rather than trusting that this dynamism is at play,
it seems crucial to question the validity of deductive logic and
its central role in the constitution of reasoning and how it is
mechanized in automated systems.
However, before espousing this view, it is crucial to draw a distinction. On one side is a preemptive mode of power in which
fixed capital is encapsulated within an interactive paradigm
demanding affective response. On the other is the advance of a
new degree of autonomy of automated cognition in which fixed
capital is characterized by experimental axioms, or synthesized
parts of incomputable data. Both forms of automation already
point to a non-deductive logic in which indeterminacy has become
a constitutive part of computational processing. This suggests
a break with reason in the form of deductive logic. But if algorithmic automation is incomplete and its rules undergo an experimental processing of truths in the distributive and parallel
processing of data then this form of automation is not simply a
Gregory Chaitin, “The Limits of Reason,” Scientific American 294, 3 (March
2006), 74–81.
44
AUT OM ATED COGNIT ION A ND CAPITA L
77
break from reason and does not mark the end of rationality.
Instead, I would suggest that this is reason in the age of the
algorithm, defined as it is in terms of a dynamic logic emerging
from the rule-based processing of infinite data expanding beyond
its deductive limits. The contemporary limit of computation and
dominance of the problem of the incomputable in interactive and
distributive systems of calculation need not simply coincide
with a failure of conceptual cognitive structures vs. the triumph
of the incompressible/uncontainable contingencies. Rather, the
incomputable is now part and parcel of computation itself.
More importantly, the centrality of this in calculative systems
can give us the opportunity to rethink the conceptual structure
based on deductive logic. This must be in terms of a structure
imbued with information randomness. For this algorithmic
structure, information already expands, extends and exceeds
the fixity of capital, and the instrumentalization of mechanized
reason on behalf of capital. Perhaps this is what Negri’s emphasis on the emancipatory gesture of new forms of fixed capital
implies. However, Negri’s proposition significantly overlooks
the possibility that automation has already developed a quasiautonomous mode of thinking—although this is limited to
nonconscious activities so far—whose consequences for the
appropriation of technological means have yet to be imagined.
With this in mind, this discussion of automated cognition is not
simply concerned with the deductive logic of computation that
governs affective responses. More importantly, the acceleration
of automated cognition—as suggested in the non-ironic realism
of Her—points to an automated cognition based on a dynamic
logic that emerges out of rule-based processing. I would like
to suggest that this logic implies a shift from deductive to experimental axiomatics. It is important to note that my attempt
here is to argue that this shift is accompanied by a new degree
of autonomy in automation. This does not simply involve the
execution of tasks or the performativity of coding that sets
out plans without human intervention. Of more significance is
the advance of experimental axiomatics at the core of fixed
78
LUCIA NA PA RISI
capital as pointing to a transformation of the bastions of reason.
Here reason does not follow the deductive model of thinking,
for which truths are confirmed by conceptual explanations that
trace problems to a predetermined cause. Instead, the model of
reasoning is here characterized by the possibility of generating
hypotheses that discover the best possible explanation and revise
set parameters according to circumstances. Here, automated
cognition involves a new form of intelligibility that is not simply geared towards the optimization of solutions. In addition, it
operates towards the production of new axioms, codes, and
instructions. Fixed capital here retains the capacity of not simply programming neuro-cognitive responses but of exceeding
programming itself.
The question is not how much appropriation might be needed
of this form of experimental logic that clearly exceeds the intentional mode of programming. Rather, the essential question
is how can we distinguish between nonconscious cognition—
the mode of cognition that surpasses deductive reasoning to
accelerate decisions and solutions—and a form of reason that
entails the generation of hypotheses, experimental solutions,
and determinations of the incalculable.
Artificial cognition
However, it would be naïve to view with enthusiasm the proposition that a new form of automation conditioned by the existence
of the incomputable could crack open the integrated system
of command and control of technocapitalism. Rather, it is
impossible to overlook that the computation of infinity is central to the capitalization of intelligible capacities both in their
human or non-human form. My insistence on the centrality of
the incomputable in calculative models addresses ontological
and epistemological transformations implicit within the algorithmic abstraction of cognition. What does it mean for automated systems to be experimental? What shift is implied
when these automated systems are able to account for emerg-
AUT OM ATED COGNIT ION A ND CAPITA L
79
ing contingencies the way algorithms—to an extent—can?
How are we to address this non-deductive mode of automation
and its capacities to acquire a certain degree of autonomy from
the law with which it is programmed?
These concerns are not meant to offer the view of a non-deterministic computational being escaping the history of the technological capitalization of cognition. Instead, it is precisely this
history that forces us to question a substantialist view of technology—and of the technical machine—on the one hand, and of
dominant models of logic and cognition on the other. This substantialist view of technology—whether ‘substance’ is historical
or political—precludes the epistemological realization that automation effectuates a transformation of what and how reason
can be. If the development of the nonconscious mechanisms of
algorithmic decisions (as in High Frequency Trading) is unable
to develop causal chains of prediction—involving a reflection of
past actions to anticipate the future—then the rule-processed
mechanisms of algorithmic elaboration can only be conceived as
rudimentary instances of a much larger social and collective
foundation of reason.45 However, this limited mechanism of
automation in which the algorithmic matrix of fixed capital is
based, has shifted from a deductive method of reasoning. Rather
than being deductive—in which algorithmic instructions follow
rules—there is now a non-deductive logic of experimental
axiomatics. In this experimental axiomatics rules or axioms are
established in the process of retrieving data. Here the limits of
reason have been exposed by a purposeless intelligence that finds
short-circuited solutions to emerging problems. This is as opposed
to conforming to priori laws established in the programming of
functions. Since computational logic entails the mechanization of
cognition and the formalization of reason, it cannot but be central
to the organization of fixed capital and thus have a key function
See Robert B. Brandom, “Artificial Intelligence and Analytic Pragmatism,”
in Between Saying & Doing. Towards an Analytic Pragmatism (Oxford, UK:
Oxford University Press, 2008).
45
80
LUCIA NA PA RISI
in the formation of the cognitive phase of neoliberal capitalism.
As experimental axiomatics bypasses the linear causality of
deductive logic—whereby outputs are deduced from inputs—
the automated logic of fixed capital enters the realm of nonconscious decision. This experimental calculation of data is not
based on universal axioms that already contain a proven truth.
Rather, this machinic experimentation is built upon nonconscious
reasoning and its flat matrix of communication across any point
of reception—human and non-human.
Following this, I would agree that the interactive paradigm of
technocapitalism already points to a semi-dynamic form of
automation. While this has subsumed cognitive and affective
capacities of thinking/existing I’ve also pointed out that beneath
these issues there still remains a question. To what extent has
algorithmic thinking established a correspondence between
automation and nonconscious modes of cognition? For this
dynamic form of automation works at the limits of computation
and is involved in the process of a modification of knowledge,
cognition and other human activities. The fact that the networks
of automated systems based on genetic and evolutionary algorithms are able to learn from the data they retrieve, reveals that
fixed capital has become central to the production of concepts
and behaviors. This also suggests the constitution of a new
symbolic universe built upon the experimental determinations
of the incomputable.
My point is that we are assisting in the configuration of an automated cognition that cannot be synthesized into a totalizing
theory or program. Rather, this process remains fractal, partial,
and incomplete. The rules and axioms are always already in
the process of determining unknowns and configuring patterns
out of their inferential relations with material data. Despite
all of the instrumentalization of reason on behalf of neoliberal
capitalism and despite the repression of knowledge and desire into quantities—such as tasks, functions, aims—there
AUT OM ATED COGNIT ION A ND CAPITA L
81
undoubtedly remains an inconsistency within computation.
This is insofar as the more it calculates the more randomness—
or patternless information—it produces. In the age of robot phase
transition, it is hard to dismiss the possibility that automated
cognition has exceeded formal representation. It may in this
case be understood as the historical realization of a second—
essentially non-deductive—form of thinking.
As outlined above, it would be naïve to assume that the centrality of the limits of computation in the post-cybernetic phase of
neoliberal capitalism simply marks the end of reason and logic.
The urgent task today rather involves the articulation of how
reason and logic are formed at this limit. Deeply implicated in
this task is the possibility that automated cognition can extend
beyond the nonconscious modes of decision making central to
the affective apparatus of capturing. Further implicated is the
extent that this nonconscious function of algorithmic automation
can be placed within the larger process of a transformation of
mechanized logic in which concepts can be formed and laws
can be revised. An examination of how algorithmic automation
might follow an abductive logic further clarifies how the formation of reason might have shifted in the present.
Recently, Lorenzo Magnani has argued that the dynamics of
information and its systematic embodiment in segments of
knowledge are fundamental to the formation of a computational philosophy.46 Magnani draws on Charles Sanders
Peirce, whose triadic system of logic or abductive-inductivedeductive circuit constitutes a pragmatist model rooted in
complex structures of reasoning. This starts from hypothetical
assertions and involves the elaboration of actual data—induction—to develop the best inferential pattern for the hy-
46
Lorenzo Magnani, Abductive Cognition. The Epistemology and Eco-Cognitive
Dimensions of Hypothetical Reasoning (Berlin / Heidelberg: Springer Verlag,
2009), 2–7.
82
LUCIA NA PA RISI
pothesis and a consequent establishment of rules—deduction.
Rules are not fixed and are not the representation of material
processing. Instead, for pragmatism, rules are inscribed into
complex cognitive architectures in which truths change to
account for non-inferential social practices. This ensures that
inferential relations in the world can also change. As Magnani
reminds us, Peirce understood abduction as an “‘inferential’
creative process of generating a new ‘explanatory’ hypothesis.”47
This means that abduction has a logical form that cannot be incorporated within deductive or inductive models. For instance, deductive models are non-amplifiable and involve results that do not
go beyond the information incorporated in the premises. Induction
is thereby unable to produce new concepts and conceptual models,
insofar as they mainly transfer concepts to new instances.48 The
opposite is true for abductive processes. These incorporate noninferential social practices with the unknown and unpredictable,
stochastic relations and the contingencies they harbor.
The traditional computational view defines cognition as a process
that computes internal symbolic representations of the external
world. Instead, the inductive model of a nonconscious intelligence
transfers new instances to existing concepts. The nonconscious
quality of automated systems that defines the infinitely looped
networks of fixed capital today mainly operates through associations, peering, linking, following.
It is thus circumscribed to limited functions that can only be
accelerated. However, the dynamic model of cognition that automated systems are able to activate—as those imagined within
the movie Her—cannot be dismissed on the basis of a natural limit
that these networked modes of Artificial Intelligence may have. If
these automated systems already modify knowledge and behavior,
it is not too speculative a stretch to imagine that automated structures of cognition can acquire an enlarged data space and extend
47
48
Ibid., 8–18.
Ibid., 29–31.
AUT OM ATED COGNIT ION A ND CAPITA L
83
the use of abductive logic in the formulation of new concepts.
This also means—counter Negri—that one has to be careful in
assuming that the potential of the automated machine can be
liberated to create another world. Instead, the increasing extension
of its operational field of data reveals that computational logic
is already primarily embedded within social, cultural and economic
practices. It is also vitally important to note that these are connected and run by the machine. This demonstrates that to navigate
the complex apparatus of governance of thought today it is inadequate—as Galloway does—to rely upon the historical separation between the technocapitalist machine and philosophical
(critical) distance. Instead, it is necessary and crucial to develop
a philosophy of computation that accounts for the transformations
of logic, and the emergence of an artificial social intelligence that
is developing abductive modes of reasoning. For the new forms
of automation the nonconscious function of automated algorithms
is just one dimension. All of this remains an open question but it
is essential to take into account the extent that algorithmic architectures are embedded within the collective practices of hypothesis generation and concept making. Similarly, it is necessary to
account for the emergence of a dynamic form of automation that
involves the extrapolation of socio-cultural meanings. But the
question remains as to the extent this is also the formation of an
alien logic able to add new conceptual structures, axioms and
laws to pre-existing structures of power.
Related to this is the necessity to think again at how this historical phase of automated cognition is not in contradiction with
an appeal to a metaphysical reality that has claims towards a
condition beyond the technocapitalist apparatus as in certain
tendencies of philosophical ‘realism.’ Perhaps, as the inconsistent relation between capital and automation extends and increases the gap between instrumentalization and the autonomy
of the networked machines of fixed capital, the challenge today
is to articulate the functions of a collective or social general
artificial intelligence. This underlines the crucial necessity for
the constitution of a philosophy of computation and the development of a critical automation theory.