Algorithms have become near mystical beings. Here on earth they make their users happy and their creators very rich. Every minute a few algorithms manipulate millions of Uber drivers and their customers. It turned Garrett Camp, the Canadian inventor of the Uber algorithm and co-founder of the company, into the transcendental choirmaster of the ride-sharing world and a billionaire to boot. Currently the third richest Canadian, Camp knows a thing or two about evolutionary algorithms. Uber is not the only algorithm playing a greater role in our daily lives. Self-improving AI machines diagnose cancer with far greater accuracy than your harried family physician. Almost twenty-five years ago IBM’s Deep Blue defeated world champion Garry Kasparov at chess. Today Go masters are no match for Google’s AlphaGo. EMI, not the record label, but the company Experiments in Musical Intelligence, composes Bach-like chorales that are, embarrassingly, extolled in blind reviews as far better than the real thing by classical connoisseurs. Most evolutionary biologists interpret organisms algorithmically. Neuroscientists study “thinking” in terms of patterns in the brain’s regional centres of operation. From medicine, to economics, to finance, to law, it is hard to find a discipline or profession today that does not increasingly employ algorithms. There are some laggards, especially in the humanities such as philosophy or literary criticism, but maybe their days are numbered as well.
The possibility of the sidelining of humanity by algorithmic entities, that may eventually assume legal-personhood status, has been popularized of late by the world historian, Professor of History at the Hebrew University of Jerusalem, and self-appointed philosopher Yuval Harari, author of Homodeus: A Brief History of Tomorrow (2015) and Sapiens: A Brief History of Humankind (2014). In Homodeus he forecasts a “Great Decoupling” of intelligence from consciousness. In this disenchanted future non-conscious intelligence reigns supreme while humans become superfluous adjuncts far more burdened than enlightened by conscious awareness. Our calculative wit will decline steadily as machines relieve us of the toil of doing “maths,” as our mothers used to say. In fact, given our room temperature IQ in relation to AI entities, humans are more likely to get in the way of algorithmic progress. There would be no accidents on the road if all cars were driverless, propelled by unconscious algorithms. The question is, what will humans do in this utopia/dystopia?
The growing algorithmic presence in modern civilization has dire consequences for some beloved ideals. The individual self of the Enlightenment and classical liberal philosophy, replete with inalienable rights and the inviolable borders of legal personhood, is portrayed by Harari as a myth that science can no longer tolerate. In fact, anything that comes out of the human imagination, any idea, concept, notion, universal, or generality, is a myth usually deployed for the purpose of eliciting human cooperation, but obviously devoid of objective reality. Modern secular humanism, a more stridently subjectivized variant of the ancient Protagorean “man is the measure of all things” humanism, is now being supplanted by a post-humanist technology that will algorithmically design “meaningful” individual experience and feeling. This will, of course, de-individualize the human pursuit of happiness and certainly impede our imaginative aspirations. The homogenization of meaning and value will make us all one of Voltaire’s bastards to invoke John Ralston Saul’s less than felicitous phrase.
Once Again the End of Philosophy
The ascent of the all-devouring algorithmic state will be littered with the carcass of just about every traditional philosophical absolute or disposition. Many of our modern philosophical orientations, especially in the idealistic tradition, are derived from the elevation of consciousness, or more preciously self-consciousness, a term Harari assiduously avoids, to primary referential status in the realm of learned discourse. A unitary self, free human agency (even if embodied), monotheisms, spirit, genius, authorship, composer, artist, idea, sovereignty, the general will, legal culpability, or just being something special, all go by the wayside. There are no univocals, no thoughts, no persons, in the algorithmic universe. Humans cease to be individuals, or moral beings, as the species becomes populated with Harari’s “dividuals.” All is unconscious process, though not necessarily Whiteheadian. Consciousness still has a place in his Process and Reality, though not as an absolute centrality.
The only possible unitary concept in Harari’s history of tomorrow is the end-purpose of the algorithm. Transportation-algorithms are just that, the process of efficiently moving people and things from A to B unimpeded by human interaction, except for human initiation of the algorithm. Eventually, machine learning will even replace the initiatory uniqueness assigned to human effort. Even the Garrett Camps of the world will be rendered obsolete by their own creations. This is more than Fukuyama’s end of history, it is the end of philosophy. Harari is by no means the first person to proclaim its demise. The end of philosophy is the inevitable result of the superceding of consciousness by intelligence. Homodeus undone.
Since Babylonian times an algorithm has been simply a set of rules for producing a particular result. The emphasis is on the particular. Your mom’s recipe for brownies is an algorithm. The Global Public Health Intelligence Network uses sophisticated algorithms to track infectious diseases and potential pandemics. No algorithm is definable without a specialized functional purpose. Algorithms are fundamentally teleological, which means they are designed for some end or a “telos.” Traditionally, philosophers have taken the teleological status of something as being either intrinsic or extrinsic. For Aristotle, an “entelechy” is something having an end in itself. For Immanuel Kant, a living being, since it is internally organized, has intrinsic purposiveness. Neither Aristotle nor Kant show up in Homodeus. Indeed, one is hard put to find any protracted engagement with a philosopher in Harari’s work, if one discounts Woody Allen or Chairman Mao. True, Jeremy Bentham, Confucius, Richard Dawkins, Daniel Dennett, René Descartes, Epicurus, Michel Foucault, Daniel Kahneman, Karl Marx, John Stuart Mill, Thomas Nagel, Steven Pinker, Jean-Jacques Rousseau, St. Augustine, Alan Turing, and Lorenzo Valla, all make cameo appearances, usually to prop up some well-established historical point. In Homodeus algorithms are defined as having their purposiveness externally determined either by humans or other algorithms. One looks in vain in Harari’s world-view for an engagement with the internal/external purposiveness issue. He just notes that algorithmic functionality is always externally determined (383). His portrait of humanity’s algorithmic future is much less persuasive because of this basic lapse in philosophical commission.
Functionality and Totality
Algorithms, their essence determined by an external functional instrumentality, can only admit of a relative totality, vis-à-vis the variables necessary to perfect the functionality of this or that algorithm. Philosophical totality, historically, the fundamental aspiration of many systems of thought, does not permit categories or thoughts to exist outside of the envisaged totality. This may be the unconditional totality, a non-determinative, unknowable, transcendental idea, of Kant’s Transcendental Dialectic, or the immanently knowable Absolute Idea of the Hegelian Absolute of Absolutes. In either case, the internal status of such an unconditional or comprehensive totality is always consciously represented by the human mind actively engaging with that totality. Harari’s unconscious algorithms require no engagement by us once they are initially produced. Strictly speaking, algorithms are not created, though they have the appearance of ex nihilo creation. They are produced because they are always relative to some external reality, to nature, to our built environment, or to filing systems.
Algorithms do not know themselves as algorithms. Self-awareness is not required for their functionality. Nor do they see themselves as standing alongside other algorithms. A transportation algorithm per se cannot adapt to an outside force that threatens the whole transportation system. It itself is an externality that is vulnerable to many other externalities. Dealing with these alien externalities necessitates new algorithms. Algorithms have no self-image. Nor do they have any will to acquire a self-image. This nisus is fundamental to human consciousness. Algorithmic instrumentality is powerless to eliminate it, we hope, as much as consciousness may at times interfere with its own functionality. Humans will always like to drive cars no matter how bad at it they may be. Calculative perfection is only one of many conflicting human ideals. Likewise, the same is true for theorizing about totality. The philosophical concept of totality does not require a signification of all contingent facts. That is not its purpose. A concept of totality is an attempt to signify reality as a structured whole, within which any particular fact (or any group or set of facts) can be comprehended.
Enter Plato’s Parmenides - An Algorithm’s Worst Nightmare
Harari’s prediction for a humanity devoid of humanity loses its argumentative and rhetorical appeal, albeit remaining Mephistophelean, when one considers philosophically the relation of the one to the many, or of part to whole, or the many other structural values on display in the high eidetic world of Plato’s great forms and kinds. If the author of Homodeus was familiar with Plato’s Parmenides he would have paid more attention to the negative and positive aspects of the ontological status of the one and the many with which no algorithm could possibly contend. The Parmenides, ever interpretatively elusive, is the anti-algorithm of the Western tradition. Neo-Platonic syncretism is one of the best exercises around in anti-algorithmics. As Hitler, after the invasion of the United States, noted that Charlie Chaplin would be one of the first to go, so will algorithms dictate the summary elimination of Plotinus and Damascius.
Harari is hard put to get rid of the ineffable One, or the run-of-the-mill one, or even ironically the many, no matter how hard he tries. If you hypothesize an uber-algorithm, or one algorithm controlling all other algorithms, perhaps thought of as totality of specialized algorithms, you must still reflect on the ontological relation between this one algorithm and its many part-algorithms. It could be argued that a calculative intelligence requires no second order thinking about its internal relations either to itself or other algorithms. This is debatable. Certainly consciousness naturally inclines towards such reflections. It knows that the one cannot be eliminated from reflective thinking, as much as the “dividuals” may attempt to do so. For example, Harari claims that humans, in an algorithmically dominated world, would not be able to control the “motivation of a system smarter than themselves” (381). This is nonsensical because motivation is itself based on a self-referential consciousness that must, at a minimal, invoke the ontological status of oneness. I may have multiple motivations to do something, but my consciousness is the unifier. Without that I have only psychopathologies of one sort or other, but even this requires a unified consciousness to make such a psychiatric assessment. Plato’s Parmenides surveys the structural values of the one/many relation while classical German Idealism incorporated these values into the categorical subtleties of the phenomenology of consciousness and the many shapes and bi-directionalities of its correlational grades of mind.
An external algorithm could never know me better than I know myself (383) simply because no human ever completely externalizes their thoughts, feelings, or dreams. To counter this obstacle to algorithmization, Harari and the neuro-scientists assume that because the algorithm knows all my internal processes, as biochemically and electro-chemically arrayed, then all my external actions can be algorithmisized. Given how much people hate Netflix deciding what movies they should watch, or Amazon orchestrating your book preferences, it is very doubtful that unruly consciousness will ever let algorithms run amok in their living rooms. Besides, who really knows his or her self? Even if it is an illusion, consciousness will always autonomously fantasize that it is living according to its own wishes and dictates. Yet another reason why algorithms could not obtain absolute status like consciousness since the latter makes mistakes as a part of its beingness, just like memory requires forgetfulness, Jorge Borges’ Funes the Memorius comes to mind, or the lessons of history are always ignored by our obliviousness to history, passim Barbara Tuchman’s The March of Folly. Algorithms are not, and cannot be, dialectical beings. Therein lies their fatal flaw.
Sovereignty, Agency, and the Unanimity Problem
Harari, when discussing Waze, a GPS-based navigational system, envisages that it could become an algorithm that acts like a sovereign agent (398). For algorithms to become sovereign agents, it is necessary that all humans, human drivers in this case, must use it. There is no room for erratic conscious beings when it comes to re-routing vehicles because of traffic jams or road construction. And it makes no sense to re-route everyone to the relatively open road because it will then become congested. Since Waze has more information than any individual human it will decide which drivers to re-route. Users of the application have to trust the system in order for it to make the re-routing decisions for each individual. Unanimous participation and complete trust are necessary for Waze, as a sovereign agent, to work flawlessly.
This transference of sovereignty from human to machine is unlikely and inherently self-contradictory. Sovereignty by definition cannot cancel itself. Sovereign human decision-makers, qua drivers, may willingly transfer their decision-making autonomy to the Waze-networked application. They can just as easily withdraw it, regardless of the reason for the withdrawal. Navigational efficiency may not be the only reason for participation in Waze. Would Waze refuse to let you disengage from the app? It would have to make such refusal an essential part of its algorithm if it is to be understood as a sovereign agent. No sovereign agent willingly surrenders its autonomy. The Enlightenment formulated a new concept of human nature based on the sovereignty of every human will. All humans would have to individually and voluntarily surrender their autonomy for the absolute algorithm to assume the status of sovereign agent. Unlike humans sovereignty is not something integral to the essence of Waze as a navigational algorithm. The concept of sovereign agency is in mortal conflict with unanimity. Algorithms, I am sorry, have no chance of attaining the status of autonomous, sovereign entities.
Karl Popper’s Historicism Makes a Comeback
Historicism reigns supreme in the algorithmic universe (399). Harari likes to envisage a virtual agent representing all of a human-user’s interests because it has a total historical record of those interests. We have a foretaste of how annoying this historicism can be when our phone, unprompted, notifies us of what we did a year ago, or that we ought to reserve a table at a restaurant. This is notification hell if all our personal history is completely memorialized. It is at least conceivable that a loved one might love us more for forgetting an anniversary or a birthday. Harari opines that the threat to individuality will, in the future, come more from within than, for instance, from a dictator using algorithms to take away our freedom (402). The individual will be nothing but a religious fantasy as greedy corporations break down our organs into electro-chemical subsystems in order to orchestrate our lived experiences.
One of the worst books around is Karl Popper’s The Poverty of Historicism (1957) for, among other things, its misrepresentations of great philosophers. The Poverty of Historicism continues its allegations against Plato and Hegel initially taken up in an earlier work called The Open Society and its Enemies (1945). These books, however, do say some things that Harari might entertain about the inexorable laws of historical destiny. Homodeus celebrates, or at least embraces as a unique historical trend, the destruction of the univocal and the individual. His vision of the algorithmic future is, inconsistently, historicist in the Popperian sense of the word. The remodelling of the human in the algorithmic dystopia appears inevitable. Harari offers no alternative to the instrumental totality of our future algorithmic society. Popper’s “piecemeal social engineering” is not presented even as a remote possibility.
Down With the Hippocratic Oath
Our prevailing philosophy of utility-maximization in economics and elsewhere feeds into possibilities for the algorithmic invasion of civilization. Kant’s categorical imperative could be the only defence. A good example is Harari’s portrayal of the development of algorithmic medicine (403-408). Instead of tending to the sick and sticking to their Hippocratic Oath, doctors in the future will be utilized to upgrade humans, i.e., to make already healthy people more healthy, i.e., more superior than they already are health-wise. Tending to the sick becomes a secondary matter. Health inequality will be greatly exacerbated. The goals of health restoration and maintenance become supplanted by the “health” needs of an instrumentalized superhuman performing outsized services for the system.
Why would a society that views health as a basic human right sanction such a wholesale abandonment of the Hippocratic Oath? Most of Harari’s algorithmic fantasies and elevations depend upon an excess concentration of resources in the hands of a few. As long as a society remains reasonably democratic and aspires to broad visions of equality in health and education, and believes in the pursuit of the common good, then the chances of algorithms becoming the Frankenstein weapons of the filthy rich are remote. Harari has a view of human nature, detailed in Sapiens, that has been fundamentally static for tens of thousands of years. In Homodeus, he stumbles into juxtaposing this view with the Darwinian view of human beings as purposeless products, decidedly not created, of the evolutionary process. Why at a certain point in the process, with the advent of the Agricultural Revolution (fraud?) around 12,000 years ago, human evolutionary development came to an end, is unclear. It is also unclear why the advance of algorithmic culture could somehow undo human nature, reinforced by Enlightenment philosophy that refuses to treat human beings as means to an end, and today writes our dignity into endless codes of ethics and conduct.
The Data Brotherhood
In the final chapter of Homodeus, called “The Data Religion,” Harari states that few people would want to be disconnected from the great dataflow of society. The assumption is that dataists believe that experiences are only meaningful if they are shared, i.e. externalized. The humanistic tradition since the Renaissance, buttressed by the Cartesian principle of subjectivity, explores the world of meaningful values within our interior landscapes. At the same time, historically, science replaced religion as the principal conduit for the human understanding of nature. A fundamental aspect of modernity is that these inward and outward directions to the human project go hand in hand. Dataists, and dataflow absolutists, seek to eliminate absolutely the inward direction of meaning-exploration. The problem is that the outward very much depends on the inward. Dialectic once again comes back to bite Harari. People will still write in full awareness that few may read. Giambattista Vico, and numerous other vanity publishers, know the routine. This is a wonderful thing. Wonder (θαũμα, θαυμάζω) and philosophy have a long relationship. In the dataflow universe there is no wonder, and no philosophy.
Know Thyself - Wasting Time
Harari might pretend that his best engagement with philosophy is the section “Know Thyself” towards the end of the chapter on the data religion. If love, Platonic or otherwise, is the presumed foil of the data cultists, would not all-knowing algorithms, Harari asks, figure out soon enough the hormonal matrix? He thinks the transition from a homo-centric to a data-centric universe is a philosophical revolution on the order of the scientific and humanist revolutions of three or four centuries ago (453-454). The comparison is most inapt. The humanist revolution reduced God to a figment of our imaginations or, if the idea was to be retained, to a useful social construct. Science, on the other hand, simply disengaged from the divine requirement. The modernity of the seventeenth century onwards certainly did not reduce humanity and its fragile consciousness to a function in an algorithmic process. Absolute consciousness had its strongest advocates in Johann Fichte, Karl Reinhold, and G.W.F. Hegel and their followers. Theirs was a grandiose effort to reconcile religion and science, subjectivity and objectivity, the infinite and the finite, the absolute and the relative, the Naturwissenschaften and the Geisteswissenschaften, the list is quite expansive. Functionality is only an aspect of their absolute theorizing, not itself the overall determinant. What is there in any possible infatuation with algorithms that would incite us to give up all of this theorizing in any philosophical, economic, or social imaginary?
It was not uncommon in the nineteenth century for political economists, like Henry George, to single out the capacity to produce as the pivotal distinction between humans and animals. Humans do not create ex nihilo, but they do produce things out of a material substratum. Nonetheless, we like to think of ourselves as creators in the world of art and ideas. We would rather say Beethoven created than produced great music. Humans, not animals, produce, or is it create (?), algorithms for tax systems and space travel. Do algorithmic entities, AI entities, have distinctive self-learning capacities such that it can be said that they are creators (or producers?) no different from their original human creators? Are algorithms disposed to create or produce as are humans to write novels or treatises, or produce art irrespective of whether we are better at it than non-conscious AI entities? Human production is often nebulously luminous, like James Joyce’s Finnegan’s Wake, rather than singularly purposive like an air traffic algorithm. Our human beingness, apart from the degree of consciousness or depth of intelligence, seems to require these “reflective judgments,” as Kant liked to style them, about things and states. Would an AI entity need any estimation, or calculation, of the sublime to accomplish any singular purpose? It is hard to imagine so. By the way, the sublime is defined as that which defies calculation or definition.
Algorithms are Not the Conclusion
Homodeus has the virtue of, at least, alerting us to possible horrors to come, as well as some that are already attendant upon us. It is too bad that Harari does not discuss Hegel’s master slave dialectic, or Kant’s transcendental ideas, or even Nietzsche’s Ecce Homo if one wants a dose of a hard to fathom melancholia. The humanist revolution does not allow for the demotion of human priorities, however vaguely defined. Science will not allow its scientists to be abandoned to the algorithmic juggernaut. That is their practical revolution. Life, as mind, spirit, or will resists, as its mystical mandate, a functional, calculative definition. It will always do so. Some may call this faith-based rather than evidence-based reasoning. It could equally be argued, in a sort of reverse Anselmianism, that it is understanding seeking faith, a faith understood as anything that is not an algorithm.
Yuval Noah Harari, Homodeus: A Brief History of Tomorrow, Toronto, McClelland & Stewart, 2015. Index, Notes, and Images.