10 Machine Learning Ecologies and Self-Organization

Material Souls

Julien Offray de la Mettrie’s Man Machine (1748) shocked its eighteenth-century audience with a radical, unapologetic materialism. Even Descartes’s Discourse on Method (1637), widely condemned as atheistic, distinguished the immaterial soul (res cogitans) and the mechanistic world of objects (res extensa). This dualism structured the emerging liberal order of democracy, rights, and markets, as well as the extractive imperialism in which the human-nonhuman divide served as “a biopolitical tool of governance” (Yusoff 2018, 5). Yet, La Mettrie’s Man Machine continues to provoke far greater outrage. No doubt fueled by La Mettrie’s truly reprehensible personal beliefs and behavior (who died, fittingly, by overeating), his critics claimed that without an immaterial subject—much less an immortal soul—human beings were mere deterministic machines inextricably entangled in a network of material connections. The indignation aimed at Man Machine, however, both past and present, misconstrues La Mettrie’s notion of a machine.

Confusion concerning La Mettrie’s machine is not surprising given the uneasy compromise between Epicurean atomism and Christian providentialism underwriting the New Science. During the seventeenth century, Francis Bacon, Descartes, and Robert Boyle appropriated the ancient atomism of Democritus, Epicurus, and Lucretius as models for predicting and controlling the natural world, rejecting the Aristotelian physics dominating European universities. In Epicureanism at the Origins of Modernity (2008), Catherine Wilson argues “the rehabilitation of Epicurean atomism” within a Christian worldview was “not a smooth process,” and “required more by way of invention than a simple insertion into Epicurean philosophy of a transcendent invisible ruler and a heaven whose attainment was condition on divine favor” (8).

The collision of a monistic atomism and impossibly complex theological-political context might best be approached in light of what Ofer Gal and Raz Chen-Morris call a “Baroque history of science.” Such a methodology “concentrates on questions above answers; on challenges more than their resolutions; on the intellectual price paid for each successful development” (2013, 10). Here, the “price paid” was a new antagonistic dualism. The Scholastics’ Aristotelian physics, with its matter-form dualism, nonetheless integrated human activity and the Earth into a unified providential whole; the adoption of an unruly Epicurean atomism, by contrast, triggered a rupture between physics and metaphysics, spirit and matter, living human and dead machinery.

Aspects of the Epicurean cosmology at odds with Christian theology—including its many-worlds theory inspiring Bruno’s On the Infinite Universe and Worlds (1584) and Fontenelle’s Conversations on the Plurality of Worlds (1686); a nonteleological evolutionary theory underlying Erasmus Darwin’s Zoonomia (1794); a rejection of the special creation of “man” and the Earth at odds with Genesis; and a theory of self-moving, self-organizing matter central to La Mettrie’s machine—were stripped from the scientifically “usable” aspects of atomism, which was then vigorously policed as a potential Trojan horse. The modern atom, ripped from its dynamic cosmological context, entered modernity as inert, meaningless, dead matter, subordinated to the meaningful movement of Providence.

This antagonism between physics and metaphysics reduced the natural world—and the majority of its inhabitants—to passive cogs in a natural machinery, inert atoms, actualized by extractive imperialism and colonial geoengineering, subject to the dominion of the non-machine: the Christian human soul. The result, as Amitav Ghosh writes in The Nutmeg’s Curse, the “metaphysics emerging in Europe, in which matter was seen as ‘brute’ or ‘stupid,’” lead directly to a colonial logic that non-European people and landscapes “were deserving of conquest” (2021, 36).

Against this intricate historical horizon, La Mettrie’s machine remains difficult to access. Without the backdrop of Epicureanism, especially the “soul” composed of self-organizing atoms, it becomes a caricature. What is lost is Man Machine’s most intriguing dimension: an early theory of self-organizing machine learning, free from metaphysical assumptions about the subject, the soul, and human exceptionality, anticipating many of the recent questions posed by machine learning and artificial neural networks.

Learning Ecologies

At least since Margaret Boden’s Artificial Intelligence and Natural Man (1977) and Douglas Hofstadter’s Gödel, Escher, Bach (1979), Man Machine has haunted modern conceptions of AI. Often, it’s a cautionary tale: Collapsing humans and machines risks reducing the former to a clockwork automaton without the capacity for learning, change, or adaptation. In Man Machine, however, the soul is not a dead metaphor. La Mettrie reconceptualizes the soul as the natural world’s animating principle, aligned with the most controversial aspects of the Epicurean tradition that recognizes no qualitative difference between human, animal, and biosphere, all embedded in “the web of life” (Moore 2015).

Drawing on Epicurean cosmology, La Mettrie proposes a soul composed of atoms (or, in the language of the early modern European science, “corpuscles”) indistinguishable from a machine’s physiological “organization.” “All the soul’s faculties,” La Mettrie writes, “depend so much on the specific organization of the brain and the whole body that they are clearly nothing but that organization” (26). These physiological organizations are not, importantly, merely machines conventionally understood; they are, La Mettrie writes, “mathematical” and “geometrical” structures, rudimentary algorithms immanent to the world, “thrown by chance on a point of the earth’s surface without anyone being able to say how or why” (23). “The reason for its existence is existence itself,” La Mettrie writes. Even as La Mettrie rejects Christian teleology, he redefines the final cause of life, both for human and nonhuman life. “Mushrooms,” La Mettrie insists, as well as animals and people are mathematical, biological machines with only one purpose: learning.

The physio-mathematical organization, the very fabric of a soul, is useless without “instruction,” La Mettrie insists, without a connectionist training on inchoate environmental data. “If a sort of ensign or flag is put, for example, on the piece of wood called a mast, and another is put on another identical object, and the first numbered with the sign ‘1’ and second with the sign or number ‘2,’” then the machine learns “arithmetic,” and then “algebraic calculations” and “language,” and then, ultimately, “reason” and “science” (14). Even without a theory of neurons, La Mettrie’s theory of learning anticipates Stuart Russell’s argument in Human Compatible (2019). For Russell, learning functions by strengthening synaptic connections in any entity—human, animal, or machine—with a “neural net.” Evolutionary developments arising from the “action potential” or electric signaling of single-celled organisms, neural nets are machine-like architectures, according to Russell, that, by way of learning, allow much of the animal kingdom to “speed up evolution itself” (15).

For La Mettrie, learning processes organize the constant dynamic transformations of the natural world. From an initial, limited mathematical-biological organization, machines develop increasingly complex learning structures, gaining intelligence and efficiency by way of feedback loops, reflexively adapting and being adapted by these self-organizing information systems. Without a theory of genes, La Mettrie presents learning as an indispensable mechanism for change and self-organization driving the processes of his pre-Darwinian evolutionary theory. But without a metaphysics of the subject or soul, without a distinction between self and other, this process of autopoiesis is in reality what Donna Haraway calls “sympoiesis” (2016), a collective process of “making-with” driving an incessant ecological evolution.

The natural world creates networks of data processing loops, processes less like human “thought” and more a series of assemblages, prefiguring the scientific contemporary work on biological ecologies—of trees, fungi, insects—as cybernetic systems. Man Machine sketches out “nature” as an ecology of interlocking data processing machines all sending and receiving data, rather than the traditional European Enlightenment cliché of the clockwork universe. By clarifying the figure of the machine, self-organizing cybernetic ecology comes to light, without an external theological teleology or humanistic control from above.

Cybernetic Control

Today, many critics of what Byung-Chul Han calls “infocracy’s . . . digitalization of the lifeworld” remain reluctant to relinquish the sanctity of the human (2022, 12). Tiqqun’s brilliant, thoroughly humanist text The Cybernetic Hypothesis is a case in point. Fundamental to the threat to human autonomy is the “analogy between living organisms and machines,” which they ague introduces a register of control: “The first aims at making living beings into a mechanics, at mastering, programming, determining humans and their life,” and “the second aims at imitating the living with machines, first as individuals, and this leads to the development of robots and artificial intelligence; next as a collective and this involves the setting of information into circulation and the construction of networks” (41–42; emphasis added). In his chapter “The Self-Organization of the Cybernetic Mind” (2023), Matteo Pasquinelli argues that the conception of artificial neural networks as a “biomorphic artefact . . . imitating life forms” elides a genealogy of self-organizing social structures and the division of human labor (134–35). While both Tiqqun and Pasquinelli capture the violence of our current neoliberal cybernetic order, Man Machine, some four hundred years in advance, questions the implications of their humanism.

Rather than salvaging the human, a more constructive path forward may be to consider self-organization and its opposition to the control inherent in cybernetic theory and the commodification of information. The result of the Macy Conferences on Cybernetics (1943–54), as N. Katherine Hayles writes, was “nothing less than a new way of looking at human beings” who now “were to be seen primarily as information-processing entities who are essentially similar to intelligent machines” (1999, 7). While this elision of the human-machine dichotomy poses no problems for Hayles, the new dualism emerging between “materiality on the one hand, information on the other” certainly does. Abstracted, “disembodied,” immaterial information that turns back and controls—and cybernetics, etymologically linked to “steering,” is nothing if not a question of control—the material, embodied, environmental realities. For Hayles, like La Mettrie, the problem is not the loss of the human but the dualisms in which embodied reality is subject to immaterial, transcendent form.

If early modern atomism “disenchanted” an all too dynamic Epicurean cosmology, early market forces functioned as its primary mode of cybernetic control. With On the Shoulders of Merchants (1994), Richard Hadden argues that modern mathematical algorithms required a homogenized natural world so that disparate values, such as times, distances, and forces, can be combined and calculated within a single equation. But the emergence of the seventeenth-century mechanistic world picture, Hadden insists, was not a cause but a consequence of the commercial world. “The calculation of the value of dissimilar commodities allowed the projection of similar techniques onto nature; observable phenomena were interpreted as resulting from the motion of bodies, motion which could be calculated as well in terms of the general magnitude” (xiv).

Similarly, in a chapter entitled “An Information Economy,” Harold Cook (2007) argues that the “long-distance public circulation about commodities and exchange rates” through colonial exchange networks also exchanged information about the natural world that gave rise to the New Science. Colonial expansion “led to a reform of the scientific disciplines” that “influenced the method of all the sciences,” with both the new economy and the New Science based on exchange value (5). In order for objects, people, and environments to be rendered meaningful (read: valuable), they must be abstracted, transformed into information that becomes calculable in algorithms or experimental “facts” established by the new scientific knowledge networks.

Man Machine questions the various histories of cybernetic dualism, past and present. It questions the legitimacy of a new, disembodied digital res cogitans acting parasitically (rather than symbiotically) on the material connections and circuits; eating, destroying, burning, and violating the network of “material souls” as they are transformed into value-as-data, “second-order cybernetics,” as Tiqqun writes, in what is “an experimentation on human societies: an anthropotechnics” (47). Better yet, avoiding pitfalls of anthropocentrism, La Mettrie’s self-organizing machine challenges the “identarian machines” that Édouard Glissant, in “The Cry of the World,” aligns with a history of “imperious science,” in contrast to systems of “fragile knowledge” (1997 [2020], 9). It is this legacy of a nonanthropocentric, self-directed machine learning, and its place in our fraught ecological realities, that La Mettrie has left for us to consider.

References