Inhuman Intelligence: on “The Automatic Fetish” and “The MANIAC”
Beverley Best | The Automatic Fetish | Verso | May 2024 | 368 pages
Benjamín Labatut | The MANIAC | Penguin Press | October 2023 | 368 Pages
When artificial intelligence was first dramatically climbing in existential and financial salience and significance a couple of years ago, one of the buzzwords it brought with it to the public sphere was ‘alignment.’ Artificial intelligence, its boosters claimed, promised to dramatically reshape not just this and that industry, but the organization of the world as a whole. Humanity was on the cusp of a dramatic rupture: AI would not just make human labor more efficient but alter how we organized our society.
While boosters claimed this was mostly a good thing, they conceded that it was perhaps risky to put a non-human force at the steering wheel. This is the problem of alignment: AI-knowers claim sympathy for the worry that AI may work towards non-human ends. The problem is keeping AI ‘aligned’ with human goals and preventing misalignment, whereby the human creation may take on a mind of its own, and begin to engineer ends not aligned with, or actively harmful towards, its human inventor. In their most apocalyptic tone, the alignment-anxious speculate that AI could hack weapons systems, or destroy the world economy’s digital infrastructure, or unleash an engineered pandemic. Misalignment, researchers claim, is as existentially risky as nuclear war and pandemics.
These fears aren’t new: see Frankenstein, Terminator, and The Matrix. What AI has brought them is a new claim to realism. What will a non-human intelligence do to human history and the human world? Will the end of humanity be the human invention of an intelligence greater than itself?
•
The maniacal obsession at the heart of Benjamín Labatut’s new novel The MANIAC, is intelligence, human and artificial. The ‘maniac’ names both John von Neumann, the breathtakingly influential scientist whose destructive and creative genius is at the heart of the book, and the acronym for an early computer von Neumann helped create, the MANIAC (Mathematical Analyzer, Numerical Integrator, and Computer). Von Neumann, described in chapters written in the first person by various historically fictionalized real figures from his life, emerges as a sublime character: awe-inspiring and terrifying, von Neumann’s mind probed the limits of the thinkable, and obsessively pursued the foundations of order, mathematics, and intelligence itself.
Labatut’s book is far from a paean to a genius taken too young (von Neumann died at 53 from a cancer he probably contracted from his work on the nuclear bomb) but rather an illustration of the horror of the destructive power of intelligence. It is not even clear that von Neumann was a human before he was an intelligence, his generation of Hungarian physicists having once been dubbed “the Martians.” Rather, von Neumann’s is a mind invested in the replication of mind as such. He dreamt of intelligence as a self-replicating process, the shared medium of both biological life and information. Von Neumann wanted to design self-replicating space ships to seed the universe with intelligence. In the pursuit of a unified theory of intelligence, he tried to understand the mind as a kind of computer and the computer as a kind of biological system.. He dreamed of intelligence sloughing off its contingent mortal coil and realizing its true essence.
Von Neumann sought to make humanity superfluous by finding a new and more efficient form for intelligence. Without quite articulating his fear of von Neumann, Labatut suggests it, by ending his book with a vignette set decades after the Hungarian physicist’s death, the defeat of the best human Go player by AlphaGo’s AI. The future of intelligence is not human, this parable tells us: instead, intelligence will be itself, and the invention of this kind of nonhuman intelligence, in Labatut’s narrative, bears a genealogical relation to the invention of the hydrogen bomb, and the game theoretical justification of Mutually Assured Destruction, inventions von Neumann made key contributions to. Labatut’s von Neumann surveys the universe and finds human intelligence wanting. He wants a higher, purer form.
In Labatut’s hands, this possibility is both hauntingly existential and physically worrying. On the one hand, von Neumann’s approach to problems provided mankind with both the scientific capability to destroy the world in the H-Bomb, and the game theoretic justification for producing an arsenal of these weapons. On the other hand, the very quality of what it is like to be a human is demeaned by AI. Human games are mastered by a power greater than ourselves. Chess can no longer be played free of an anxiety that the opponent is a cheater using a powerful engine instead of their own ingenuity. AI boosters claim they will produce thousands of novels every year and put human authors out of business; AI ‘artists’ flood the internet with schlocky images; education administrators panic that students will never again write their own essays. What are humans to do in this world where AI does everything we do, but “better”? AI will have rendered the human mind superfluous. We would exist merely to enjoy and consume the products of a higher intelligence.
•
At the heart of the AI discourse of misalignment is a belief that the world as it exists today is meaningfully aligned with human ends. Surely there are signs of an intelligence organizing the world all around us. A whole slew of technical apparatuses lie ready and waiting to facilitate the human life process: roads to be driven and walked on, supermarkets and restaurants to get food from, appliances to smooth our social reproduction, political institutions to organize it, etc. etc. etc. But even as this world evinces rationality, is the rationality organizing it a human one?
In a sense, of course this rationality is human: the light switches in my house are all at a pleasingly human height. The couch I am writing this article on is similarly and pleasingly proportionate to my body. And I know, obviously, that the light-switch height and the couch’s suppleness are not coincidentally that way; they were made that way, by human minds.
But appearances may not be the whole truth. In Beverly Best’s latest contribution to Marxist scholarship, The Automatic Fetish, Best offers a new interpretation of the third volume of Karl Marx’s Capital. She argues that in this volume, Marx showed that the concrete world, the world as it appears to us every day, is actually organized by capital. Capital, though, does not appear as itself in that concrete world. Instead, capital is a relation between people that organizes how they work and the distribution of the products of their work. Capital is conceptually coherent, but this conceptual coherence can only appear in abstraction from the concrete world. Best pursues an understanding of this coherence by way of an interpretation of Marx’s book. What emerges is an understanding of capital as a kind of non-human intelligence which shapes the human world. Since the third volume of Capital was published by Friedrich Engels in 1894, ‘marxological’ debates have raged about where to place it in Marx’s body of thought, and whether it can even be taken as a legitimate reflection of Marx’s mature theory. Engels composed the volume from texts Marx (who died in 1883) had written thirty years prior in the mid-1860s. After writing the manuscript that Engels would later publish as volume three, Marx wrote and brought to publication volume one in 1867, which he then reworked in multiple editions during the 1870s. Across the 1860s and ‘70s, Marx constantly reimagined not just the particular volume he worked on, but what the overarching project of Capital itself was — and then never returned to the manuscript he had drafted in the mid ‘60s. Michael Heinrich has argued that the three volumes of Capital we have today are not so much an incomplete version of one project, but several fragments of separate projects, each of which, with the crucial exception of volume one, was left unpublished and considered incomplete by their author.
Best’s Automatic Fetish presents a bold argument in favor of the coherence of the work Engels published in 1894 as the third volume of Marx’s Capital, and interprets that text as though it was the final part of a cogent trilogy. This will be controversial in the world of marxological scholarship, but bracketing that important scholarly debate, there is something interpretatively constructive in Best’s insistence that we read volume three as a complete and cogent work. Positing volume three’s cogency gives her permission to interpret the book as a sustained argument. The Automatic Fetish helps us then see what that sustained argument would look like. For Best, Capital III is the critical moment in Marx’s oeuvre he moves from the conceptual to the concrete, where he analyzes how capital actually appears, as a multitude of different business interests competing with one another, in light of his theory of the essential nature of capital, which he had described in the prior volumes.
Best calls this shift, from capital as a concept to the concrete world, Marx’s ‘perceptual physics.’ With it, Best draws our attention to how, for Marx, capital appears differently depending on our perspective of it. A quotidian, everyday perspective on capital only gives us a partial glimpse of what capital is like, and cannot see how it works as a whole. These everyday perspectives of what capital is like are not false — in fact, they are essential to capital’s reproduction — but they cannot grasp how capital organizes its system as a whole. And, when we follow Marx as he moves back and forth between capital-as-concept and capital-as-it-appears, what we learn is that this whole has a rationality — only it is not a human rationality. And so, the force of Best’s interpretation of Marx’s argument is to claim that the world is already animated by a misaligned artificial intelligence.
One significant example of this perspectival dynamic is Marx’s treatment of profit. In Volume I, Marx showed that profit is a function of surplus value, how much time a worker works over and above the cost of their labor power. This implies that industries that depend more on human labor (“variable capital”) and less on machines, raw materials, and so on (“constant capital”) should be the most profitable. But this is plainly not the case. Instead, businesses make profit based on their size irrespective of what Marx called the ‘organic composition’ of their capital, the ratio of their investment between variable and constant capital. The point here for Marx is that Volume III, unlike Volume I, is concerned not with a single capital understood in abstract, but with the concrete world of many capitals. In that concrete world, capitals compete with each other over profit, and the competitive dynamic of the market pools and redistributes the total aggregate of profit between capitals. The size of the total magnitude of profit is still determined by the total amount of surplus value, but the distribution of profit is determined by the market.
Each individual capitalist, then, tries to cut their costs and sell their product for less than it costs to produce it. This means that each concrete, particular capitalist has an incentive to reduce the absolute sum of surplus value in the economy by reducing the amount they invest in human labor. This dynamic – where the essence of profit at the level of the whole appears for individual capitalists as something different, if not contradictory – is what Best calls Marx’s perceptual physics. In the particular case of profit, this dynamic gives rise to the (in)famous ‘tendency of the rate of profit to fall’ as capitalists are incentivized to shrink the size of their investment in variable capital even as that variable capital, at the level of the whole, is the source of profit. The individual capitalist thinks of what shrinks profit at the level of the social whole as the source of their own particular profit, and from their perspective, market crises driven by systemic overproduction appear as arbitrary events, even as they actively contribute to them.
Volume III proceeds on from the opening discussion of price, competition, and crisis, to an exploration of different kinds of capitals and capitalists: industrial, merchant, financial, and culminates in a long discussion of rent that will feel particularly abstruse and obscure to contemporary readers. In all these examples, the upshot of Marx’s perceptual physics is to show that while the particular kind of capital — merchant, financial, landed — must appear as in and of itself productive, at the level of the social whole, where capital is conceptually coherent, these particular capitals are actually just claims on a surplus produced by workers that capitalists compete for control over. That is, it appears as though land’s fertility is value-producing and so intrinsically rent producing; in the world of finance, it appears as though money creates more money. Buying low and selling high; money itself; and land: each of these are taken by the capitalist to be value producing, but the argumentative thrust of Volume III is to demonstrate that while their concrete appearance as intrinsically valuable is crucial to the way capital reproduces itself, at the level of capital’s conceptual core, they are instead rather claims on a surplus produced by workers that capitalist compete with each other for.
Read through Best’s heuristic of perceptual physics, then, Volume III shows how capital appears differently depending on the perspective we take. There is an everyday perspective, where we view capitalist reproduction from within capital’s totality. From this perspective, profit looks like a question of price, not surplus-value; money appears to grow; and so on. Marxist theory offers us up a second perspective upon capital as a social whole or totality. Roughly, the former ‘embedded’ perspective, is capital as it appears concretely; the latter, perspective upon capital is capital seen abstractly and conceptually. Marx’s analysis hopes to show these two levels interact, and how the latter, ‘in the last instance,’ predominates.
This isn’t to say capitalists are wrong to think of their profit in terms of cost and price. It is rational for them to think about it in this way! Capital’s appearance as profit is crucial for its functioning. If capital appeared to us all as value, individual capitalists would have no notion of how to make decisions about price. Best’s point is rather that this distinction between capital’s appearance and its essence is crucial to capital’s reproduction.
Dividing the analysis of capital into these two distinct levels of abstraction creates serious problems for Marx’s theory of capital. Most obviously, there is the difficulty of explaining how the abstract level interacts with the concrete level. The latter is the actual; the former is the concept that claims to somehow animate or shape the latter. But how do we actually see this virtuality that claims to animate the real? One iteration of this problem, notorious since the 1890s when Eugen von Böhm-Bahwerk first began the marginalist counteroffensive, is the ‘transformation problem,’ or the question of just how Marx’s theory of value interacts with — constrains? determines? influences? — the world of prices. Rudolf Hilferding, then a young theoretician of the German Social Democratic Party, responded that Marx’s theory is not meant to predict prices but rather to understand the movement of capital as a whole. Its incapacity to predict prices means it fails only for the capitalist aiming to buy low and sell high.
One payoff of the perceptual physics within Volume III is the “theory of the rate of profit to fall,” which I summarized earlier. In the wider Marxist literature, the (non)existence of the ‘tprf’ is widely debated: did Marx really leave his ultimate theory of capitalist crisis unpublished? Is this ‘tendency’ an ultimate theory of the end of capital or a theory of how capital reproduces itself through crisis? Best does not resolve these debates. Her chapter on the tendency is at pains to make clear that Marx enumerated countertendencies alongside the tendency and she is rather agnostic about whether the tendency means doom for capital. In fact, at the end of the chapter, Best pivots away from questions concerning the tendency itself, and instead argues that the really deleterious thing about capital is how it restricts the human capacity to enjoy abundance in non-capitalistic forms, and how that restriction is facilitated by capital’s perceptual capacity, its ability to make it appear as though life within capitalism is an open-ended entrepreneurial game where we compete for self-valorizing resources, rather than a system of value reproduction and expansion predicated on the extraction of surplus-value from the working class.
This sort of worry gets to the heart of what is valuable about Marx’s theory and its two distinct levels of abstraction, worries about the ‘transformation problem’ and the theory of crisis aside. If capital is the sort of thing which has a conceptual structure — if instead of an open-ended space for humans to perform bold entrepreneurial experiments, it instead biases their behaviors in directions that only become visible to them when they abstract away from their particular circumstances, and which proceed according to a particular logic — then humanity is being ruled by something that resembles an inhuman intelligence.
The second term that structures Best’s book besides ‘perceptual physics’ is ‘automatic fetish.’ The fetish, for Marx, is the appearance of a social relation as an object. An automatic fetish, then, is a social relation that through its very reproduction appears as an object. Marx uses the term specifically for interest-bearing capital, but we can see why Best believes it works as a stand-in symbol for the whole of Volume III. On Best’s reading, the crucial dynamic that animates Volume III is how capital appears as a set of more or less abstract objects – prices, businesses, goods, and the like – even as at the level of the whole, capital is a relation which follows a conceptual logic that determines society’s motion. Our projects are not really our own — they are instead small parts of a larger history, dictated by a conceptually coherent nonhuman subject, capital.
•
In the late 18th century, the philosophers of the Scottish Enlightenment tried to make sense of the emergence of the institutions that made their newly wealthy and (from their perspective) civilized society possible. No human agent, it seemed, had come in and sui generis produced the regime of property rights they believed had made their society so rich and so civilized. These institutions, instead, were in the Scotsman Adam Ferguson’s famous phrase “the result of human action, but not the execution of any human design.” As Smith, the other Scottish Adam, put it in his epoch-making Wealth of Nations, the regular application of the rule of law and property rights were produced by “the silent and insensible operation of foreign commerce and manufactures.” The world was organized by forces that acted through, but without the control of, humans, who in this reading of history were barely agents. In Ferguson’s and Smith’s view, this led, beautifully and naturally, to the flourishing of wealth and civilization.
For some of capitalism’s ideological supporters, its inhumanity is an explicit benefit. The deeply influential neoliberal and Nobel laureate Friedrich Hayek drew explicit analogies between computers, markets, and minds, and argued that the market is essentially the largest informational processor available and so should be charged with organizing society. History ought not be steered by the human mind, a comparatively small and inefficient information processor; like von Neumann, Hayek dreamed of a world organized by an inhuman intelligence.
Labatut clearly fears this possibility. He hints at the existential horror of humanity being smothered out of a universe transformed into a chessboard for an AI to master. What his novel misses is all the ways that the world in which we live is already organized by a nonhuman intelligence. The arrangement of our society is not random. But neither is it organized by a human mind. What Best and Marx show is its nonhuman coherence.
For Marx, as for Labatut, the world and history ought to be human. If we participate in organizing ourselves but do not make decisions about how we are organized, then we have simply failed to be fully truly human. History remains natural and not really historical insofar as we do not take responsibility for how we have made it. For a human history to come into existence, for us to exit this natural pre-history, we must conquer capital and become the subject of history. The threat to humanity being fully human, taking responsibility for history and its own freedom isn’t an AI invented in Silicon Valley, but the fact that we are on a rollercoaster ride of booms and bust, exploitation and immiseration, and ever greater accumulation of riches, that we did not design, and which does not work for our ends. Labatut and Best both fear this outcome. But Labatut does not realize we already live in a world governed by an alien intelligence. Politics should not attempt to defend the world as it was from a technological threat that first emerged from the mind of a Martian from Hungary — it should attempt to overturn the domination of humanity by capital.