Artificial Idolatry: An Invitation to Not Worship

It gave him the greatest trouble, Nietzsche wrote in 1882, “to realize that what things are called is incomparably more important than what they are.” As rose-sniffers have long known, the scope of Nietzsche’s realization demands some restriction. Nevertheless, the determined effort to unify all machine learning under the single name “AI” corroborates Nietzsche’s sense that a thing’s “reputation, name, and appearance” has a way of growing “merely because people believe in it, until it gradually grows to be part of the very thing and turns into its very body” (Gay Science, sec. 58).

What gets unified by the name “AI”, as I will show, conforms rigorously to the memorable description of “a god” given by the Dominican priest and theologian Herbert McCabe. As McCabe argues in “The God of Truth” (chapter 3 of God Still Matters), a “god” is not God. A “god” is necessarily part of the universe, whereas the source of all being, truth and goodness is necessarily not part of the universe. “For whatever else the creator might be, he, she or it cannot be a part of what it creates. It would have to exist before it exists, which is a contradiction” (GSM, p. 30). Anything believed in as “a god” has the following characteristics: (1) It is an object of worship; (2) It has a certain “mystique,” since it is taken to be part of the world order and yet “not altogether subject to the natural laws of physics or other sciences” (GSM, p. 31); (3) It is thought to be quite powerful, yet manipulable by human beings; (4) The god’s power must be simultaneously stressed and strategically de-emphasized, lest the god appear demonic; (5) The god has its own class of promoters who gain in proportion to their capacity for attracting worshipers, and who disseminate propaganda as if it were the consensus of experts with access to the laws of history.

Considered with strict reference to these characteristics, is “AI” a god? The answer developed here is “Very much so.” In what follows, my primary aim is to demonstrate this. My secondary aim is to counsel non-belief in this god, particularly for those tasked with educating people amid mass idolatry.

1. “AI” as object of worship

What makes something an object of worship is not a property that inheres in the object, but the approach taken by the worshiper. The object of worship must somehow prompt a mixture of awe, fear, and respect on the part of the worshiper. Though worship is sometimes conscious, it need not be. A person who centers his life on the prospect of accumulating wealth might not consciously regard himself as worshipping money. Yet the relentlessness with he pursues wealth suggests that “worship” is in fact the right term for his practical attitude. As McCabe observes, people can worship “without even knowing that this is what they are doing” (GSM, p. 30).

One may therefore ask: How does the contemporary worship of “AI” display itself?

The first mark of constituting “AI” as a worship-object is the baptism of distinct forms of machine learning under the common name “AI.” The algorithm’s technical details are not important. What matters is its status as “AI”—or better, “an AI.” (Here the lexical similarity to “a god” is revealing.) As the quickest look at recent marketing campaigns can attest, the very name “AI” is totemic. It is intended to have a “mesmeric force,” functioning as the word “ought” functions in Kantian moral philosophy (G.E.M. Anscombe, “Modern Moral Philosophy,” p. 8). To speak of an operation as “powered by AI” or “AI-driven” is to distinguish it, to mark it out as something special.

The second important way that believers constitute “AI” as an object of worship is their determination not to suspend belief. Those skeptical about the value of introducing “AI” into the classroom will pose certain questions. These include: Does “enhancing” a process with “AI” necessarily improve that process? What are the costs of injecting “AI” into such processes? Do the benefits outweigh the costs? To see that “AI” is becoming an object of worship, one may note the way that among the “AI”-faithful, such questions are either superficially asked or not asked at all. In the extreme (but not infrequent) case, they are actively discouraged.

2. The mystique of “AI”

Whatever the processes of constituting an object of worship, nothing can remain such an object for long unless it bears a certain “mystique.” As McCabe remarks, “a god has a mystique about it” (GSM, p. 31). Promoters of “AI” cultivate an air of mystery about the precise ways that large-language models obtain their results. Tech companies strive to market “AI” in a way that stimulates demand, but they cannot let consumers into the secrets of power, the arcana imperii. Powers that would inspire worship must keep some important things hidden from worshipers, as Vico and Tacitus both knew. Reveal the secrets and the mystique will vanish.

To maintain the mystique proper to a god, the relationship between promise and reality must be obfuscated. Here McCabe’s description of playing state-run lotteries as “worshipping a familiar minor god” is helpful (GSM, p. 31) The lottery maintains its mystique by generating an air of mystery about who hits the jackpot and by what means. There is no actual mystery; those who grasp the facts of probability will not be tempted to buy a lottery ticket. But as long as there are people who look upon the lottery as a mysterious, unpredictable source, “to some extent outside the order of nature” (as McCabe says), it will remain a popular god. To ensure that their products retain the necessary mystique, tech marketers must publicize the instances where people are able to use “AI” in a way that garners massive winnings in exchange for little or no work. Structurally speaking, this is identical to the lottery; there are a few individual winners and many losers.

In our present moment, the most effective way for “AI” promoters to publicize the phenomenon of “massive wins in exchange for no work” is to create a large number of apparent winners, who will naturally get the word out to others who want to join the party. This explains why tech companies are particularly eager to sell “AI tools” to students. At the moment, students can use chatbots to win big. (Here “to win” means “receiving a high grade” or “gaining a credential with minimal effort.”) But this is a strictly temporary situation. Grades and credentials handed out without any growth in knowledge, skill, imagination, erudition, or competence will soon become worthless. But as long as companies can sell a product that appears to miraculously generate huge wins in exchange for little or no work, “AI” will have a mystique about it.

3. Powerful yet manipulable

In the great epics of Homer, the most powerful god is Zeus. Yet his awesome power does not place him beyond the reach of manipulation. Hera knows how to push her husband’s buttons, and Athena is no slouch at working her dad. Even humans can sometimes get to Zeus. Like Zeus and his fellow gods, “AI” is seen as being immensely more powerful than human beings yet capable of being manipulated by them.

“AI”’s power must be seen as godlike. It must seem to differ in kind (and not merely degree) from meagre technologies like the electric can opener. Accordingly, advertisements for “AI” stress how much it can do now, and how much more it will do in the future. The more things “AI” can do, the greater its claim to godlike power. Hence the desire of marketers to take any recognizable human enterprise and sell it anew with the tagline “AI-powered.”

However powerful, “AI” must also be manipulable. We must be able to bend it to our will. Just as the older gods had their favorites, “AI” will share its power with initiates who sign up for special programs.The dream is that of shifting the intellect’s work from contemplation of higher realities to the practical accomplishment of worldly tasks. Today “AI” is routinely praised in the terms of unexamined practical Cartesianism—the truly American ideology, as Tocqueville noticed. Anyone who bends his education to the accumulation of “real-world AI skills” will be able accomplish so many things in such little time. “Virtue has to come to consist of doing something in less time than someone else,” Nietzsche observed toward the end of the nineteenth century. Nietzsche’s observation is a lament, because he knows that the usual fate of those who combine “saving time” with the habit of “constantly chasing after gain” is to “expend their spirit to the point of exhaustion” (Gay Science, sec. 329). But those enchanted by the godlike power that “AI” promises to confer are bound to miss this point. They will be charmed by advertisements that speak of “mastering AI.” Once mastered, “AI” can be used to master nature and all that it contains.

But can nature be mastered? The ancients were skeptical. “You may drive out Nature with a pitchfork, yet she will ever hurry back,” says Horace (Epistles 1.10). The modern technological spirit from which “AI” springs tends to insulate itself from any such doubts. One day, Francis Bacon assures us, modern science will vindicate itself as the triumph of artifice over nature (The New Organon, aphorism 107). That no such thing had happened in his own time did not matter for Bacon. He could say (and rightly) that the artifices of modern science were in their infancy. But since the seventeenth century, the sophistication and power of the artificial have massively increased, much in the manner that Bacon projected. The “A” in “AI” signifies artifice on a new scale. It points to artificial devices that are powerful enough to make what Bacon had forecast into a new reality. Thus far, however, Nature remains more potent than the pitchfork.

4. Tool or weapon?

The arrival of new gods does not coincide with clear vision. In the case of “AI,” the lack of clarity is especially palpable. If machine learning turns out to be half as powerful as its boosters say, the odds that we will retain anything like sovereignty over it are vanishingly low. Successful power has a way of putting itself in the driver’s seat, sooner or later.

Those sanguine about the new god’s arrival do not like to admit the plausibility of this scenario. Accordingly, they find themselves obliged to engage in strategic deflation of the new god’s power. The deflation must be strategic rather than wholesale, because if “AI” gets weakened too much, its mystique will suffer. Nevertheless, “AI” proponents cannot afford exclusively to stress its power. They need a plausible story about how “AI” is powerful but not that powerful. Otherwise, the power of “AI” will begin to strike many as malevolent, or even demonic. The challenge for worshipers is to speak of their god in a way that both emphasizes its power and keeps it within certain bounds.

As it happens, the language for meeting this challenge already exists. It can be easily found in Bacon, who reliably associates the advancement of knowledge with “instruments” and “tools.” Advocates of “AI” like to say that “AI is a tool.” Here their Baconian inheritance appears to serve them well. If a “tool” is an instrument for getting a job done, then “AI” appears less like a god and more like a humble but useful implement. It seems comparable to a screwdriver or hammer or wrench or any other tool that increases our power as long as we use it, but then can be put safely back into its box.

In response to this apparent verdict of common sense, I offer a line from the Coen Brothers’ 2016 film Hail, Caesar!: “Would that ’twere so simple!” If “AI” were only like a screwdriver, with no godlike mystique, there would hardly be a question of worshipping it. But regardless of whatever analogies one might construct between “AI” and familiar tools, it is obviously unlike a screwdriver or hammer—or even digital tools like a spreadsheet or word-processor. Its purported power is far greater than that. Its proponents acknowledge this, of course, when they are in “selling mode,” appealing to the god’s power and our desire to share in it.

A more appropriate comparison is that between “AI” and a weapon. A weapon shares two primary features with a tool. First, a weapon can be used to get a quick result, whether or not the user possesses any understanding of how the weapon is made, or the means by which it achieves its effects. Second, any weapon can be used for good or ill. When pushed hard enough, some “AI” proponents acknowledge this problem. They do so in a way that recalls the teacher of rhetoric that Plato portrays in the Gorgias. The weapon that Gorgias is marketing is rhetoric, a speaker’s ability to persuade an audience of anything, in any domain—whether or not the speaker possesses any knowledge of how things are within that domain. Ancient Greeks were drawn to promises about rhetoric’s power as many are drawn today to “AI.” Both promise a quick result, bypassing the work required to understand something for oneself.

As a tool—or a weapon—for getting what one wants, rhetoric has an undeniable appeal. Teachers of rhetoric such as Gorgias would sell their ability to impart the knack of non-rational persuasion to others in much the same way that today’s “AI” marketers try to sell their product. Under pressure by Socrates, Gorgias reluctantly admits that rhetoric resembles a weapon that can be used either justly or unjustly. If the student who learns rhetoric from Gorgias has no knowledge of justice, and therefore cannot distinguish between rhetoric’s just and unjust use, where will he obtain this knowledge? To this question posed by Socrates, Gorgias replies that he will himself teach the student about justice (Gorgias, 460b). (For a small fee and in just three weeks, the London School of Economic and Political Science assures us, you can acquire “a conceptual toolkit to practice ethical reasoning.”) But Gorgias suddenly finds the ground shifting under his feet, since his claim to possess knowledge of justice has no credibility. (His whole schtick, after all, is to teach an art that can get results precisely when knowledge is absent.)

Sellers of “AI” find themselves in the same position as Gorgias. Under duress, they find themselves obliged to admit the existence of a question about the “responsible” use of “AI.” Like Gorgias, they will implausibly proclaim their own capacity to teach ethical uses of AI. Like Gorgias, they have little experience in reflecting on justice and its demands. This lack is evident both from their uncritical attitude toward propaganda and from their indifference (if not outright hostility) to a serious study of the long traditions that seek wisdom about ethical questions. It is not impossible, of course, that a particular “AI” enthusiast might decide to read for herself the books that lie at the core of these traditions, rather than rely upon electronically generated summaries. But she would also have to embrace the multiple demands that these books make on the mind. Most importantly, she would have to quit selling “AI” to students who are tempted to use “AI” as a way of avoiding intellectual exertion.

5. “AI in the classroom”: a call for resistance

What Plato calls “periagoge” (“turning around,” Republic 518c) and later traditions call metanoia (“change of mind”) has never been a mass event. My intention is not to suggest that the current “AI” industry will have a kind of collective metanoia. Like any other profit-driven entity, it will push as hard as it can, until it seems more profitable to direct its efforts elsewhere. A more practicable goal is to nourish skepticism among those connected to education—teachers, students, administration, support staff—about attempts to impose “AI” on classrooms and curricula.

“AI” marketers reliably appeal to two powerful passions—anxiety and the desire for ease. “AI won’t wait for you,” promises DataCamp. Better hop on the bandwagon now, before it’s too late. Uncertainty about the future, a primary cause of anxiety, yields to a new condition where you can “future-proof your skills with 580+ courses, AI career tracks, and certifications.” Life will be easy, if only you accept Carnegie Mellon’s offer to “get started with the essentials of deep learning for AI to get ahead in your career.” Again, we see the appeal to anxiety, accompanied by the prospect of a comfortable life where “AI” will “build the perfect, custom schedule.” Why? So you can “get more done in less time,” with the implicit (but lying) promise that time left over will translate into contented leisure.

To promote salutary skepticism about “AI,” one can observe the gap between promise and reality. As anyone with experience in the workforce knows, the predictable result of outsourcing jobs to machines is unemployment. If a worker does not lose his job, he will be reassigned to other tasks. Nothing, however, ensures that such tasks are more satisfying or less burdensome than the job that was destroyed by automation. In appealing to the desire for ease and increased leisure, “AI” advertisements promise a rosy scenario with no trade-offs. They conveniently fail to mention the other possibilities.

Marketers of “AI” are relentless in their efforts to convince us that we now live in the “age of AI.” One should not underestimate the impact of sheer repetition. Say enough times that the new god is in your midst and demands to be served, and people will start to believe it. A mystique will arise. Anyone who resists adapting to what gets promoted as “the way things are changing,” according to the “laws of history,” is called “reactionary.”

But haven’t we seen this movie before? The imperative to be “with it”—to swim with the cultural stream no matter what, on the ground that “whatever is newer is better”—is dangerous. In the extreme case, it produces trend-chasing intellectuals who support a regime that murders millions of its own citizens, as Leszek Kolakowski observed with respect to mid-twentieth-century Russia. (See his Modernity on Endless Trial, pp. 5-6 and p. 39.) A recent advertisement from the once-respectable Chronicle of Higher Education (now one of the most visible and least principled hawkers of “AI” in the educational sector) proclaims that “AI is changing the classroom—but many faculty remain cautious.” As a faculty member, I am not so much “cautious” as amused by this attempt to insult my intelligence with a made-up statistic: “In fact, 82% worry students may become too reliant on AI, weakening critical thinking skills.” The “in fact” is a nice touch, as is the invocation of “critical thinking skills”—a phrase that only 18% of faculty use without irony. (Two can play this game.) The Chronicle suggests that faculty like myself need to “learn” something we don’t know—namely, how “institutions are using AI as a pedagogical partner—helping faculty personalize instruction, redesign assessments, and guide students toward deeper learning.”

This language is oddly, strangely, surreally detached from the realities of any classroom in which I have taught. I have been privileged to witness moments of “deeper learning.” Each of them, without exception, has occurred in the course of discussing or writing about primary texts. These texts, and the spaces in which one can still read and talk about them, bear no relation to the fantasies of technocrats who would like to turn their ghoulish vision of “screens everywhere, all the time” into universal law.

Promotions of “AI” that rely on false dichotomies and manufactured statistics, while flying under the banner of education, must be understood as efforts to imbue the new god with a kind of mystique. They are driven by the desire for economic gain, a desire that has no relation to the good of students who deserve better. One suspects that universities trying to “brand” themselves as “AI-current” are engaged in a desperate attempt to address sharp losses of revenue due to declining enrollments. “Students are already using AI. Are you teaching them how?”—asks an outfit that calls itself “Learning.com.” My answer is “no.” I am persuading them to do their own reading, thinking, and writing. I am seeking to develop their natural ability to see through vapid slogans and fashionable buzzwords. I am helping them to form their power of judgment, which I take (with Montaigne) to be a primary aim of education. If, despite my skepticism about the phrase’s relentless and cynical deployment, this is indeed the “age of AI,” then I am teaching my students how to be “untimely” in Nietzsche’s sense. (See especially the third section of his essay “Schopenhauer as Educator.”)

An education in untimeliness teaches its recipients to regard themselves not as children of their age, but as their stepchildren. It is not a course that can be completed in three weeks. It is an adventure that, unfolding over the course of a life, affords some protection against technocratic tyranny.

Image Credit: Nicolas Poussin, “Adoration of the Golden Calf” (1634)