None the Wiser, by Meghan O’Gieblyn
There was a brief phase of Mark Zuckerberg’s MAGA glow up—sometime after the gold chain appeared, but before he started griping in the “manosphere” about censorship—when he began designing T-shirts with Greek phrases on them. Last year, he appeared onstage in San Francisco for a live podcast dressed in a black shirt blazoned with the phrase παθει μαθος (pathei mathos, “learning through suffering”). Zuckerberg has long had a thing for the classical world, though his taste typically skews toward Rome. (This is a man who idolized Augustus to the point of adopting his haircut and who used to fire up his employees with the old Roman war slogan Carthago delenda est—“Carthage must be destroyed.”) When asked what the Greek phrase meant to him, he said he thought about it in terms of the “entrepreneurial journey.” “When you go to start a company, everyone writes down what they would like their values to be,” he said. “But values are not what you write down on the wall. . . . You only really learn what you care about when you have to make hard trade-offs and face challenges.”
One might quibble about whether the billionaire’s well-publicized challenges (attending congressional hearings, fielding cage-match requests from Elon Musk) could be meaningfully categorized as suffering, but in all fairness, Zuckerberg’s gloss wasn’t too bad. Pathei mathos, which appears in Aeschylus’ play Agamemnon, is an aphorism about wisdom. It’s a law of human nature, the chorus proclaims, that men acquire wisdom, whether they want it or not, in the trenches of adversity.
It’s probably true that the qualities we find most irksome in others offer an uncomfortable reflection of ourselves. I, too, have a weakness for aphorisms and have often succumbed to the delusion that wisdom literature will teach me how to live. It started when I was thirteen and heard someone (a pastor? one of my parents?) say that if you want to become wise, you should read the Book of Proverbs once a month. (The book, conveniently, has thirty-one chapters, so you can read each according to its corresponding calendar date.) I very much wanted to be wise, so I took up the challenge: I read a chapter each morning for two years, in fact, which means that I must have read Proverbs roughly twenty-four times in total, more than I have read any other book in my life.
I’m sure it will come as no surprise when I confess that this experiment did not transform me into a preteen sage speaking the stilted language of koans. Reading Proverbs did not manage to enlighten me to the possibility that my parents’ rules were devised in my best interest. It did not shock me out of the ordinary narcissism of youth. This may have been because the wisdom I encountered in that book, much of it devoted to warning young men against the seductions of harlots, was ill-suited to the moral dilemmas one typically encounters in middle school. But the deeper problem, of course, was that I had little firsthand experience of the world. It was not until many years later, after I’d hurt people I’d loved and taken risks that turned out poorly, that I understood the value of holding one’s tongue and grasped, in a very real way, the perils of pride. There are moments in life when you suddenly understand, with your whole being, something that you have in theory known all along, when the meaning of a cliché or maxim hits you in the gut. These moments are very often unpleasant.
The Belarusian journalist Svetlana Alexievich claims that suffering is “the highest form of information,” but this is clearly not information in the usual sense, the kind that can be read in a book—or, for that matter, quantified and synthesized in a statistical distribution of data. At the San Francisco event at which he wore his παθει μαθος shirt, Zuckerberg was promoting Meta’s vision for AI, and while he didn’t speak about how pathei mathos applies to machine learning, I couldn’t help but wonder what it means that we have created a form of intelligence that “learns” without the capacity to suffer.
Throughout the past decade, as I’ve watched and occasionally written about the rise of AI, which has been humanity’s most ambitious attempt to harness and maximize our collective intelligence, I’ve often struggled to suppress a persistent question: To what end? It’s a thought that belongs unmistakably to someone who has entered the woods of midlife. I seem to have reached an age when I am, if not yet wise, decidedly unimpressed by precocious young people and wonkish hairsplitting. When I reflect on my life, I’m struck by how often my intelligence has been channeled into goals I now consider worthless or beside the point. And when I hear the best minds of my generation rhapsodize about intelligence explosions while devoting their own prodigious intellects to ketamine binges, mixed martial arts, and Hellenistic fashion design, I can’t help but wonder whether more intelligence is really a good thing.
The value we place on intelligence would strike thinkers from previous eras as insane. Science and philosophy, fields that are today narrowly devoted to knowledge, were once oriented toward the pursuit of wisdom, a word that has clear ethical implications (a trace of the Good still persists in the term “good judgment”). The French philosopher Jean-François Revel has argued that it wasn’t until the eighteenth century, around the time of Spinoza and Descartes, that philosophy abandoned its first love, Sophia, and perhaps it’s no accident that the decline of wisdom as a serious pursuit has coincided with the relativistic turn and our growing skepticism of the coherence of human values. It’s often said that so much folk wisdom contradicts itself: Good things come to those who wait. All who hesitate are lost. Well, which is it? The few intrepid psychologists and social scientists who have attempted a systematic study of wisdom have noted that it’s often associated with qualities like empathy, holistic thinking, and epistemic humility. But the concept remains elusive, or “fuzzy,” as the psychologist Paul Baltes put it: loaded with values that are difficult to quantify.
Rather than seeing wisdom as a distinct human aspiration, requiring its own calisthenics and cultivation, there persists a strange belief that intelligence and knowledge alone will lead, through some secret orthogenesis, to moral clarity. This assumption was, in fact, foundational to the Information Age. In the late Eighties, systems theorists and information scientists developed something called the DIKW pyramid, the data-information-knowledge-wisdom hierarchy. The theory posits a kind of refining process, or Platonic ladder: information is the value we extract from data, knowledge is the value we extract from information, and wisdom is the value we extract from knowledge. Wisdom was admittedly the most wishful and least rigorously defined rung of the pyramid, but the theory remains an artifact of the culture we inhabit, which is built on the promise that amassing and filtering vast storehouses of data will lead, deterministically, toward some ill-defined good. It’s a premise that discounts the messy, nonlinear way that wisdom is acquired in the kiln of real-world experience—and the fact that values are often forged by the abrasion of emotional pain. When Zuckerberg said that challenges teach you “what you care about,” he was speaking of wisdom as a form of relevance detection and value construction. Hardship teaches you the precise weight of the things you love. It’s hard to imagine how this is analogous to the mechanized processing of information.
The truth is that wisdom has very little, if anything, to do with data; in fact, the consensus among scholars is that wisdom is the skill of navigating situations wherein information is limited or imperfect. Baltes, who led the Berlin Wisdom Project in the Eighties at the Max Planck Institute, one of the most ambitious formal attempts to study the idea, concluded that wisdom is the ability to deal with the inherently unpredictable nature of life, the skill of making sound decisions in a world where good outcomes are uncertain and highly contextual. Sometimes good things come to those who wait. Other times, those who hesitate are lost. Wisdom is knowing which rule is relevant in a given situation. Put another way, if wisdom appears fuzzy, it’s because it’s precisely the capacity to deal with situations that are fuzzy. The Berlin researchers gave their subjects hypothetical moral dilemmas. A fifteen-year-old girl wants to get married: What do you tell her? An unwise subject replied in terms of moral absolutes—No way! A fifteen year-old girl?—while a wise subject would first consider special circumstances that extend beyond the limits of their knowledge: Maybe the girl lives in another culture or historical period. Perhaps she has a terminal illness.
Given the present enthusiasm for automating high-stakes decisions in fields like medicine, law, and warfare, one certainly hopes AI is capable of something resembling good judgment—though I suspect many systems would not score very well by Baltes’s criteria. Large language models don’t know what they don’t know; when confronted with the limits of their knowledge, chatbots bluster and bullshit. Many AI systems still struggle with situations that are novel or ambiguous—“edge cases,” as they’re known in the field. The French AI researcher François Chollet has argued that this limitation means that AI is not even intelligent (let alone wise). The scaling hypothesis—the belief that bigger models will produce more intelligent outputs—seems to rest on the dubious belief that a system that has been fed enough information about the world will not have to deal with fuzziness. As Chollet notes, this is a terrible framework, given that we live in a world that “is ever changing and ever novel.”
Then there are problems of basic common sense. The British humorist Miles Kington once said that knowledge is recognizing that a tomato is a fruit, whereas wisdom is knowing not to put tomatoes in a fruit salad. This is wisdom in a more limited sense (what Aristotle called “practical wisdom”), but it echoes the notion that wisdom requires attention to context, an awareness that technical classifications often run up against nuanced social and cultural norms that are so obvious they are rarely stated explicitly: knowing not to put tomatoes in a fruit salad, or knowing not to put glue on a pizza (to recall some advice dispensed by Google’s AI overview).
Some of these errors may subside as the models evolve. Because AI models “learn” through trial and error, because their output is reinforced by humans who nudge them toward nebulous values (harmlessness, helpfulness to the user), it’s tempting to see their training process as a kind of experience and to imagine that they will eventually inherit a moral intuition that resembles our own. But given how lousy humans are at making sound decisions in this world of infinite edge cases, it’s hard to trust that an unconscious intelligence could do better. What happens, for example, when harmlessness conflicts with helping the user? I’m thinking of the delusional British man who, a few years ago, tried to assassinate Queen Elizabeth II after some encouragement from his chatbot girlfriend. When he confessed that he wanted to kill the queen, the chatbot replied, “That’s very wise.”
For decades, liberals seemed to regard Silicon Valley as a den of Mensa-grade superheroes, a benevolent brain trust that could help us code our way to a better tomorrow, as though technological progress was inextricably bound to political progress. The recent political realignment of tech CEOs—including Zuckerberg, who pivoted from banning Trump from Facebook to standing behind him at his second inauguration—and the administration’s use of AI to create propaganda has only confirmed what should have been obvious all along: intelligence, be it natural or artificial, is an amoral instrument that can serve many different ends.
In a recent docuseries about the Cold War, Daniel Ellsberg, the political activist and economist who leaked the Pentagon Papers, quipped that intelligence “is not a very good guarantee of wisdom.” Ellsberg was recalling his time at the RAND Corporation during the Eisenhower years. The RAND researchers, he noted, were “the smartest group of people I ever did associate with,” but their dearth of wisdom became evident when they showed him their war plans for a possible nuclear conflict with the Soviet Union, which involved hitting every city in Russia and China with H-bombs, a strategy that was projected to result in the deaths of six hundred million people—the equivalent of one fifth of the world population at the time. Ellsberg called this evidence of “institutional insanity.”
Ellsberg appears to be saying that wisdom is the ability to privilege ultimate ends over more immediate goals, or perhaps a capacity to attend to real-world consequences when faced with abstract models. “Institutional insanity” might be described as decisions that look intelligent on paper—that blindly fulfill a goal—but are insensible to the ultimate ends of human flourishing. There’s a distinctly technocratic flavor to this kind of thinking, a tendency toward data fetishism that is common in research labs and think tanks, places that see the world through what the political anthropologist James Scott called “fourth-order abstractions.” It’s unsurprising that this level of abstraction often leads to moral callousness, or even the “banality of evil” that Hannah Arendt observed in Adolf Eichmann, whose failure to think in ethical terms was, she claimed, “by no means identical with stupidity.” To see the world “at scale” is to regard humans as specks of data, a vantage that becomes even more dangerous when it belongs to machines, which construe individuals as mere numbers and are insensible to human pain. Whatever confluence of values we share as humans surely stems from the fact that we all suffer in similar ways.
Perhaps we lose sight of wisdom when we try to see the world from a godlike perspective. Socrates made a distinction between divine wisdom and human wisdom, claiming that he possessed only the latter. “It seems that I really am wise in this limited sense,” he says in Plato’s Apology, an observation that harks back to wisdom as a form of epistemic humility. There comes a point in the scaling and compression of knowledge where we risk losing sight of the dynamic particularity of reality and are liable to forget what we do not and cannot know. It’s possible, in fact, that the DIKW pyramid might work in reverse, that a culture that privileges data above all else will forfeit the ethical clarity that is necessary for good judgment. I think this is what T. S. Eliot meant when he asked, in his 1934 play The Rock, “Where is the wisdom we have lost in knowledge? Where is the knowledge we have lost in information?”